Modals in the Languages of Europe: A Reference Work 9783110219210, 9783110219203

This book is the first comprehensive survey of modals and modal constructions in the languages of Europe. It is a collab

183 90 3MB

English Pages 590 [588] Year 2009

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Frontmatter
Contents
List of contributors
Abbreviations used in glossing
1. Introduction
2. Modals in the Germanic languages
3. Modals in Irish
4. Modals in the Romance languages
5. Modals in Greek
6. Modals in the Slavonic languages
7. Modals in Baltic
8. Modals in Albanian
9. Modality in Romani
10. The grammaticalisation of modal auxiliaries in Maltese and Arabic vernaculars of the Mediterranean area
11. Modal verbs in Balto-Finnic
12. Modals in Hungarian
13. Mood and modality in Berber
14. Modality in Basque
15. Modals in Turkic
16. Concluding chapter: modal constructions in the languages of Europe
Backmatter
Recommend Papers

Modals in the Languages of Europe: A Reference Work
 9783110219210, 9783110219203

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Modals in the Languages of Europe



Empirical Approaches to Language Typology 44

Editors Georg Bossong Bernard Comrie Yaron Matras

Mouton de Gruyter Berlin · New York

Modals in the Languages of Europe A Reference Work Edited by Björn Hansen Ferdinand de Haan

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines of the 앪 ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Modals in the languages of Europe : a reference work / edited by Björn Hansen, Ferdinand de Haan. p. cm. ⫺ (Empirical approaches to language typology ; 44) Includes bibliographical references and index. ISBN 978-3-11-021920-3 (hardcover : alk. paper) 1. Europe ⫺ Languages ⫺ Modality. I. Hansen, Björn, 1964⫺ II. Haan, Ferdinand de, 1965⫺ P299.M6M65 2009 4151.6⫺dc22 2009027078

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at http://dnb.d-nb.de.

ISBN 978-3-11-021920-3 ISSN 0933-761X © Copyright 2009 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany.

Contents

List of contributors…………………………………………………. Abbreviations used in glossing……………………………………...

vii xii

1. Introduction ................................................................................... Ferdinand de Haan and Björn Hansen

1

A. Modals in Indo-European languages (Western branch) 2. Modals in the Germanic languages ………………………...……. Tanja Mortelmans, Kasper Boye and Johan van der Auwera

11

3. Modals in Irish …………………………………………..……… Peter McQuillan

71

4. Modals in the Romance languages …………………………..…. Bert Cornillie, Walter De Mulder, Tine Van Hecke and Dieter Vermandere

107

5. Modals in Greek .......................................................................... Anastasios Tsangalidis

139

B. Modals in Indo-European languages (Eastern branch) 6. Modals in the Slavonic languages ................................................ Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

167

7. Modals in Baltic ............................................................................ Axel Holvoet

199

8. Modals in Albanian ....................................................................... Walter Breu

229

vi

Contents

9. Modality in Romani ..................................................................... Viktor Elšík and Yaron Matras

267

C. Modals in Non-Indo-European languages 10. The grammaticalisation of modal auxiliaries in Maltese and Arabic vernaculars of the Mediterranean area ………………... Martine Vanhove, Catherine Miller and Dominique Caubet

325

11. Modal verbs in Balto-Finnic ....................................................... Petar Kehayov and Reeli Torn-Leesik

363

12. Modals in Hungarian ................................................................... Erika Körtvély

403

13. Mood and modality in Berber ..................................................... Amina Mettouchi

431

14. Modality in Basque ..................................................................... Alan R. King

457

15. Modals in Turkic ………………………………………………. Lars Johanson

487

16. Concluding chapter: modal constructions in the languages of Europe .................................................................................... Björn Hansen and Ferdinand de Haan

511

Subject index ……………………………………………………... Languages and language varieties index …………………………. Names index ………………………………………………………

561 569 573

List of contributors Johan van der Auwera Center for Grammar, Cognition and Typology University of Antwerp Belgium [email protected] Juliane Besters-Dilger Slavisches Seminar Universität Freiburg Germany [email protected] Kasper Boye Department of Scandinavian Studies and Linguistics University of Copenhagen Denmark [email protected] Walter Breu Sprachwissenschaft/Slavistik Universität Konstanz Germany [email protected] Dominique Caubet INALCO Paris France [email protected] Bert Cornillie Department of Linguistics University of Leuven Belgium [email protected]

viii

List of contributors

Walter Demulder Department of Linguistics University of Antwerp Belgium [email protected] Ana Drobnjakoviü Department of Linguistics University of Leuven Belgium [email protected] Viktor Elšík Institute of Linguistics and Finno-Ugric Studies Charles University Prague Czech Republic [email protected] Ferdinand de Haan Department of Linguistics University of Arizona Tucson USA [email protected] Björn Hansen Institut für Slavistik Universität Regensburg Germany [email protected] Tine Van Hecke Department of Linguistics University of Antwerp Belgium [email protected]

List of contributors

Axel Holvoet Lietuviǐ kalbos institutas Vilnius Lithuania [email protected] Lars Johanson Seminar für Orientkunde Universität Mainz Germany [email protected] Petar Dimitrov Kehayov Institute of Estonian and General Linguistics University of Tartu Estonia [email protected] Alan King Trinidade kalea 1, 3 eskuina 20800 Zarautz Gipuzkoa Euskal Herria Spain [email protected] Erika Körtvély Department of Uralic Linguistics University of Szeged Szeged Hungary [email protected] Yaron Matras School of Languages, Linguistics & Cultures University of Manchester United Kingdom [email protected]

ix

x

List of contributors

Peter McQuillan Department of Irish Language and Literature University of Notre Dame USA [email protected] Amina Mettouchi University of Nantes and Institut Universitaire de France Nantes France [email protected] Catherine Miller Centre Jacques Berque Rabat Morocco [email protected] Tanja Mortelmans Center for Grammar, Cognition and Typology University of Antwerp Belgium [email protected] Reeli Torn-Leesik The Department of English Language and Literature University of Tartu Estonia [email protected] Anastasios Tsangalidis School of English Dept of Theoretical and Applied Linguistics Aristotle University of Thessaloniki Greece [email protected]

List of contributors

Martine Vanhove LLACAN - UMR 8135 du CNRS Centre Georges Haudricourt Villejuif France [email protected] Dieter Vermandere Department of Linguistics University of Antwerp Belgium [email protected]

xi

Abbreviations used in glossing

1 2 3 ABL ABS ACC ACT ADE ADJ ADV AGR ALL AOR ART AS AUT AUX CM COMP COND CONJ COORD COP CVB CXN DAT DEB DEF DET DFC DU ELA EMPH ERG ESS F FS FUT GEN

first person second person third person ablative absolutive accusative active adessive adjective adverb(ial) agreement allative aorist article annexed state autonomous form auxiliary challengeable marker complementizer conditional conjunctive coordination copula converb construction dative debitive definite determiner definite conjugation dual elative emphasis ergative essive feminine free state future genitive

habitual hypothetic i -ik medial conjugation (Hungarian) ILLAT illative IMP imperative IMPF.PART participium imperfectum IMPRF imperfect IMPS impersonal IND indicative INDF indefinite INE inessive INF infinitive INFL inflection INS instrumental INTRA intraterminal IPFV imperfective IPP infinitivus pro participio IRR irrealis JUNCT junctor LK linker LOC locative LP linking particle LVC light verb construction M masculine MIDDLE middle voice MOD modal MOD.AFFIX modal affix MP modal particle N neuter Nnon- (e.g. NSG nonsingular, NPST nonpast) NEC necessitative NEG negation, negative, negative coverb NFPRS non-focal present NMLZ nominalizer, nominalization NOM nominative HAB HYP

Introduction xiii OBJ OBL OPT PASS PFUT PFV PFX PL PLPF POSS POS POT PP PREF PREP PRF PROG PROX PRS PRT PST PTL PTCP PURP

object oblique optative passive perfective future perfective prefix plural pluperfect possessive possibility potential mood pragmatic particle prefix prepositional case perfect progressive proximal/proximate present partitive past particle participle purposive

PV Q REAL REFL REL REM SBJ SBJV SG SPPL SUBLAT SUP SUPER TAM TERM TRANSL V VB VN VOC VOL

preverb question particle/ marker realis reflexive relative remote past subject subjunctive singular suppletive sublative supine superlative tense-aspect-mood terminative nominal case translative verb verbal verbal noun, nomen actionis vocative voluntative

1. Introduction Ferdinand de Haan and Björn Hansen

1. The aim of the book The aim of this book is to describe the properties of modals or to be precise modal constructions in the European area and to compare the systems in individual languages or language families from an areal and genetic perspective. For the sake of contrast, the book also looks at the expression of modality in some languages just outside of Europe. The book tries both to give a description of the modals in the individual languages and to account for the nature and status of modals in general. There are diverging traditions in the treatment of modals as a separate category in the linguistic descriptions of individual European languages; this ranges from applying morphological and morphosyntactic criteria in English (the well-known “NICE-properties”) 1 to descriptions in other languages that take a fully semantic approach (essentially only those elements are modal that express modal notions), without investigating whether such elements have different formal properties as well. The English modals have been described from different perspectives and within a whole series of different theoretical frameworks and can probably be considered as one of the best studied language categories overall. In comparison to the Germanic languages, modals are much less studied for other languages and language families. Many descriptive linguists do not consider the expression of necessity, possibility and volition part of grammar and, consequently, do not recognise modals as a category in their own right. Frequently, modals are treated as purely lexical elements and are, therefore, excluded from grammatical descriptions. These diverging descriptive traditions notwithstanding, in the last decade there has been considerable progress in research on the universal and typological features of the semantic space of modality (for the state of the art see de Haan 2006 and Nuyts 2006). The hub for the typological research on modality has been the Center for Grammar, Cognition and Typology located at the University of Antwerp. In 1998 van der Auwera and Plungian presented a universal semantic map covering the semantic space of possibility and necessity. This highly influential paper contains all cross-linguistically relevant synchronic and diachronic connections

2

Ferdinand de Haan and Björn Hansen

between modal, pre-modal and post-modal meanings and has been developed further in recent years (e.g. van der Auwera and Ammann 2005, van der Auwera 2008). Another contribution to the cross-linguistic study of the semantic side of modality is de Haan (1997) on the interaction of modality and negation. Building on these semantic works we look at modals from a holistic perspective, taking into consideration both the semantic and the morphosyntactic nature of modals and examine how grammaticalised the notional category of modality is in the languages of Europe. The development of modals is generally considered a typical case of a grammaticalisation process, i.e. a change in which constructions shed their lexical status and acquire a more or less grammatical status. Grammaticalisation comes in degrees: within one language not all modals are grammaticalised to the same extent, and languages do not grammaticalise corresponding modals to the same degree either. We take as our starting point the known grammaticalisation parameters of Lehmann (2002, first edition 1985), as they are not language-specific (unlike the NICE properties). The basic idea behind Lehmann’s parameter approach is that grammaticalisation is understood as a complex macroprocess which is grasped in terms of six aspects located at the paradigmatic and the syntagmatic level: “The weight of a sign, viewed paradigmatically, is its integrity, its substantial size, both on the semantic and phonological sides. Viewed syntagmatically, it is its structural scope, that is, the extent of the construction which it enters or helps to form. The cohesion of a sign with other signs in a paradigm will be called its paradigmaticity, that is, the degree to which it enters a paradigm, is integrated into it and dependent on it. The cohesion of a sign with other signs in a syntagm will be called its bondedness; this is the degree to which it depends on, or attaches to, such other signs. The paradigmatic variability of a sign is the possibility of using other signs in its stead or of omitting it altogether. The syntagmatic variability of a sign is the possibility of shifting it around in its construction.” (Lehmann 2002: 110)

For some languages detailed work on the grammaticalisation of modals has already been done, partly on the basis of Lehmann’s work; e.g. Goossens (1987) and Traugott (1989) for English, Diewald (1999) for German, and Hansen (2001, 2004, 2006) for Slavonic. However, there is a considerable lack of work comparing the degree of the grammaticalisation of modal elements in different languages. The goal of this book is to present

Introduction

3

similarly inspired work on other languages of Europe, thus allowing for a cross-linguistic perspective. Table 1. Parameters of grammaticalisation (Lehmann 2002: 110) Paradigmatic

syntagmatic

weight

Integrity

structural scope

cohesion

Paradigmaticity

bondedness

variability

paradigmatic variability

syntagmatic variability

We focus on the modal elements (and modal affixes) and account for their status in terms of Lehmann’s (2002 [1982]) parameters of grammaticalisation, shown in Table 1. In line with the recent approaches mentioned above we understand modality in a narrow sense as comprising exclusively the meanings ‘necessity’, ‘obligation’, ‘possibility’, ‘permission’, and ‘volition’. We will, therefore, exclude the category of mood which is defined as a morphological verbal category which in contrast to modals is an obligatory category in those languages that have it (de Haan 2006: 33). We will not be concerned with related notions such as conditional, subjunctive, optative, and imperative/hortative, although in the case of modal affixes, the distinction might be hard to draw. We shall likewise exclude all types of evidential markers. Whereas modal meanings like possibility and necessity refer to the commitment of the speaker to the truth of what he/she is saying, evidentiality refers to the source of evidence the speaker has for his statement (see de Haan 2001). As a first working definition offered to the individual contributors of the present volume we proposed to define modals as word-like elements which are polyfunctional in the sense that they express no less than two types of modality (see van der Auwera 1999, van der Auwera and Ammann 2005). One frequently distinguishes between three types of modality: dynamic, deontic, and epistemic modality. Polyfunctionality can be seen here as a process of semantic bleaching and, thus, as the result of semantic shifts that are typically encountered in the grammaticalisation parameter of integrity. This book tries to address the following questions: How can we describe modals as a cross-linguistic category? Where is the borderline between modals and fully lexical expressions of modality? What are the relevant parameters for a typology of modal systems? How can one detect

4

Ferdinand de Haan and Björn Hansen

different degrees of grammaticalisation? Which features of the modal systems are typical for Standard Average European? Can we spot areas of convergence and of divergence in the modal systems of the languages of Europe and immediate surroundings?

2. The languages studied The languages addressed in this book are shown on Map 1 and Table 2 below, with the exception of the various Romani languages, which are essentially found all over Europe.

Map 1. Map of the languages included in the study

Some language families are more or less fully represented (Baltic, Slavonic, Finno-Ugric, plus those consisting of only one language, such as Basque, Albanian and Modern Greek) while others are represented with only a small number of languages. Germanic is represented by five languages which detail the geographical spread (English, two Continental West Germanic languages and two North Germanic languages). Romance

Introduction

5

is represented by French, Spanish, Italian and Romanian, again for reasons of geography, and Celtic is represented by Irish, with appropriate remarks on other Celtic languages. Arabic is covered by Moroccan, Egyptian and Levantine Arabic, as well as Maltese, which accounts for a good spread of languages adjoining the European linguistic area. The same is true for Turkic. Because our focus is on the European linguistic situation we did not seek articles from specialists in the Indo-Iranian and Indo-Aryan language families. Table 2. Genetic affiliation of the languages PHYLUM INDO-EUROPEAN (Western branch) Germanic Celtic Romance Greek INDO-EUROPEAN (Eastern branch) Slavonic

Baltic Albanian Indo-Iranian AFRO-ASIATIC Semitic Berber URALIC Balto-Finnic Ugric BASQUE TURKIC

languages English, Dutch, German, Icelandic, Danish Irish French, Italian, Spanish, Romanian Greek Russian, Ukrainian, Belorussian, Polish, Czech, Slovak, (Upper and Lower) Sorbian, Slovene, Serbian/Croatian, Bulgarian Latvian, Lithuanian Albanian Romani Egyptian Arabic, Moroccan Arabic, Levantine Arabic, Maltese Kabyle Finnish, Estonian, Livonian, Votic, Ingrian, Karelian, Veps Hungarian Basque Turkish, Azeri, Uyghur, Uzbek, Gagauz and others

6

Ferdinand de Haan and Björn Hansen

3. Outlook over the individual chapters Given the fact that the book consists of a unified theme, we will forego the usual listing of individual chapters that one finds in a collection of papers such as this one. Such a listing would be highly repetitious, given that all of them deal with “modals in language family X” (with the exception of the final, concluding, chapter). Instead, we will say a few words on the differences in emphasis between contributions. Although we suggested a framework for the structure of the individual chapters (see section 1 above), individual contributors were free to choose how closely they followed the framework, if they felt it was less relevant for their data. Therefore, the chapters on Basque, Berber and Turkic focus more on the modal constructions than on the grammaticalisation framework while others (such as Germanic, Slavonic and Finno-Ugric) utilised the framework to a much higher degree. In addition, given the constraints of time and space, some chapters did not address their entire language family but imposed additional restrictions. We have already mentioned the restrictions on languages in Germanic, Romance and Celtic. Furthermore, the scholars working on Romance limited themselves mainly to modals of necessity, with remarks on other modals when appropriate. Finally, given the nature of the Romani family, it is fitting that the emphasis there lies on the borrowing of modals and modal constructions into the various Romani dialects and the syntactic typology of modal constructions. There is a wealth of data and an important fount of information for anyone working in the areas of language contact and grammaticalisation theory. It goes without saying that a book of this scope is the result of the work of many hands. First and foremost we would like to thank Johan van der Auwera, the inspiration behind the whole project, who had the idea to organise a one-day workshop on ‚Modals in the languages of Europe’ at the 38. Conference of the Societas Linguistica Europaea held in Valencia, September 2005. This workshop brought the authors together and was the kick-off for the current book project. Apart from that, Johan has helped us during the initial phase of editorial work and has given us useful comments on the concluding chapter. We would also like to express our gratitude to all contributors for their commitment and patience. A special thank goes to Sheila Dooley for proofreading the extensive manuscript. Last but not least we would like to thank Arek Danszczyk and Stefano Bertaina for their competent technical assistance during the process of bringing this volume to press.

Introduction

7

Notes 1.

i.e. Negation: modals allow negation contraction (mustn’t vs *hopen’t), Inversion without do-periphrasis (Can we go? vs *Hope we to go?), Code (John can swim, so can Bill) and Emphasis (*Yes, I DÒ can come. vs Yes, I DÒ hope to come.) (Quirk et al. 1985: 137).

References de Haan, Ferdinand 1997 The interaction of negation and modality. A typological study. New York: Garland. de Haan, Ferdinand 2001 The relation between modality and evidentiality. In Modalität und Modalverben im Deutschen, Reimar Müller and Marga Reis (eds.), 201–216. (Linguistische Berichte Sonderheft 9) Hamburg: Helmut Buske Verlag. de Haan, Ferdinand 2006 Typological approaches to modality. In The Expression of Modality, William Frawley (ed.), 27-69. Berlin: Mouton de Gruyter. Diewald, Gabriele 1999 Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalität. (Germanistische Linguistik 208) Tübingen: Niemeyer. Goossens, Louis 1987 The auxiliarization of the English modals: a Functional Grammar view. In Historical development of auxiliaries, Martin Harris and Paolo Ramat (eds.), 111-143. Berlin: Mouton de Gruyter. Hansen, Björn 2001 Das Modalauxiliar im Slavischen. Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen. München: Otto Sagner. Hansen, Björn 2004 The boundaries of grammaticalization. The case of modals in Russian, Polish and Serbian/Croatian. In What makes grammaticalization? A look from its fringes and its components, Bisang Walter, Nikolaus P. Himmelmann, and Björn Wiemer (eds.), 245271. (Current trends in linguistics Studies and Monographs 158) Berlin: Mouton de Gruyter.

8

Ferdinand de Haan and Björn Hansen

Hansen, Björn 2006 Na polputi ot slovarja k grammatike: modal’nye vspomogatel’nye slova v slavjanskich jazykach [Halfway from lexicon to grammar: modals in the Slavonic languages]. Voprosy jazykoznanija 2006/2, 68-84 Lehmann, Christian 2002 Thoughts on Grammaticalization. A programmatic Sketch. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Nuyts, Jan 2006 Modality: Overview and linguistic issues. In The Expression of Modality, William Frawley (ed.), 1-27. Berlin: Mouton de Gruyter. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A comprehensive grammar of the English language. London Traugott, Elizabeth Closs 1989 On the rise of epistemic meaning: An example of subjectification in semantic change. Language 65, 31-55 van der Auwera, Johan 1999 On the semantic and pragmatic polyfunctionality of modal verbs. In The semantics/pragmatics interface from different points of view, Ken Turner (ed.), 49-64. Oxford: Elsevier. van der Auwera, Johan 2008 In defence of classical semantic maps. Theoretical Linguistics, 34/1, 39–46. van der Auwera, Johan, and Vladimir Plungian 1998 Modality’s semantic map. Linguistic Typology 2, 79–124. van der Auwera, Johan and Andreas Ammann 2005 Overlap between situational and epistemic modal marking. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gil, and Bernard Comrie (eds.), 310–314. Oxford: Oxford University Press. van der Auwera, Johan and Andreas Ammann with Saskia Kindt 2005 Modal polyfunctionality and Standard Average European. In Modality. Studies in Form and Function, Alex Klinge, and Henrik Høeg Müller (eds.), 247–272. London: Equinox.

A. Modals in Indo-European languages (Western branch)

2. Modals in the Germanic languages1 Tanja Mortelmans, Kasper Boye and Johan van der Auwera 1. Introduction: the Germanic modals In contrast to e.g. Romance (see Cornillie et al., this volume) and Slavonic (see Hansen 2004, Besters-Dilger et al., this volume) modals, the Germanic modals can be claimed to stand out – morphologically, syntactically and semantically – such that they can be said to build a grammatical paradigm. From a morphosyntactic point of view, the central members of the modal verb class are preterite-presents and are followed by a bare infinitive. Semantically, the Germanic modals can be characterised by the fact that they express both root (i.e. non-epistemic) and epistemic modal meanings. 2 This implies neither that the class of modal verbs can be easily delineated – there are peripheral candidates striving for admittance (e.g. brauchen in German, behøve in Danish) – nor that the class of modal auxiliaries or expressions is diachronically stable (cf. Krug’s (2000) discussion of emerging modals in English). Moreover, it is well-known that there are differences between the modals in the different Germanic languages, both with respect to individual modal items within the paradigm of a particular language (the German modal wollen, for instance, seems to be less grammaticalised than the other modals of its paradigm, see section 3) and crosslinguistically with respect to entire paradigms (such that the entire modal verb paradigm in one language can be said to be grammaticalised to a larger extent than in another; the English modal verb paradigm is a case in point). In the following, we will compare the various core or central modals in the Germanic languages by means of Lehmann’s parameters of grammaticalisation. As these parameters concern three aspects of grammaticalisation (weight, cohesion and variability), which are further differentiated along a syntagmatic and paradigmatic axis, we want to establish whether the grammaticalisation of the Germanic modals involves the various processes to the same extent. “Consequently, if we want to measure the degree to which a sign is grammaticalized, we will determine its degree of autonomy. This has three

12

Tanja Mortelmans, Kasper Boye and Johan van der Auwera principal aspects. First, in order to be autonomous, a sign must have a certain weight, a property which renders it distinct from the members of its class and endows it with prominence in the syntagm. Second, autonomy decreases to the extent that a sign systematically contracts certain relations with other signs; the factor inherent in such relations which detracts from autonomy will be called cohesion. Third, a sign is the more autonomous the more variability it enjoys; this means a momentary mobility or shiftability with respect to other signs” (Lehmann 2002: 108-9).

Whereas a sign’s weight (both in terms of its paradigmatic weight (integrity) and its syntagmatic weight (structural scope) and the degree of paradigmatic and syntagmatic variability are assumed to decrease with increasing grammaticalisation, a sign’s paradigmaticity (its degree of paradigmatic cohesion) and its degree of bondedness (its syntagmatic cohesion) are expected to rise with grammaticalisation. In the following table 1, an overview is given of the six processes leading to an increased degree of grammaticalisation. In our study of the Germanic languages, we will focus on the central modals in five Germanic languages 3 only: English, Dutch, German, Danish and Icelandic. The highly grammaticalised English modals – the grammatical properties of which have been described in great detail in numerous studies (for instance, Quirk et al. 1985, Biber et al. 1999) – will function as a starting point and standard of comparison; we will offer more detailed accounts of the central modals in Dutch and German (as representatives of the West Germanic languages) and of Danish and Icelandic (representing the North Germanic or Scandinavian languages), an inventory of which is given in table 2. As stated already, we consider as central modals those that are preterite-presents (including will and its counterparts) and combine with a bare infinitive. As high text frequency is a typical characteristic of grammatical items (see Bybee 2003: 603), we also assume that central modals – being the most strongly grammaticalised ones – are frequent. Not only will we mainly restrict ourselves to five Germanic languages, we will also leave out the parameters of (formal) integrity, bondedness and syntagmatic variability. Formal integrity 4 and bondedness (syntagmatic cohesion) 5 seem to be less relevant for the modals in the languages under account.

Modals in the Germanic languages

13

Table 1. Lehmann’s parameters and grammaticalisation processes (on the basis of Lehmann 2002: 110) Parameter

Weak grammaticalisation

Strong grammaticalisation

Process

Integrity

bundle of semantic features; possibly polysyllabic

attrition

paradigmaticity

item participate loosely in semantic field free choice of items according to communicative intentions item relates to constituent of arbitrary complexity item is independently juxtaposed item can be shifted around freely

paradigmatization

paradigmatic variability structural scope

bondedness syntagmatic variability

obligatorification

condensation

coalescence

few semantic features; oligo- or monosegmental small, tightly integrated paradigm choice systematically constrained, use largely obligatory item modifies word or stem item is affix or even phonological feature of carrier item occupies fixed slot

fixation

Table 2. The central modals in five Germanic languages English

Dutch

German

Danish

Icelandic

can shall may will 7 must / / / (ought)

kunnen zullen mogen willen moeten / / /

können sollen mögen wollen müssen dürfen / /

kunne skulle måtte ville / (turde) 8 (burde) /

(kunna) 6 skulu mega (vilja) / (þurfa) munu (eiga)

14

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

Decrease of syntagmatic variability, which manifests itself through increasing fixation, can be traced back to general constraints regarding word order in the various languages rather than to individual properties of the modal verbs themselves. 9 Our focus will therefore lie on the parameters of paradigmatic variability, paradigmaticity, and structural scope. As far as these remaining parameters are concerned, it has been observed that English does not formally distinguish between non-epistemic and epistemic meanings. The English central modals show a very high degree of paradigmaticity, on the one hand, and a low degree of paradigmatic variability and structural scope, on the other, and this irrespective of whether they express non-epistemic or epistemic meanings (see Goossens 1985 for a similar claim). On the semantic side, the central modals do not seem to be able to express lexical or premodal meanings like ‘to like’, ‘to want’ or ‘to know’ anymore, which points to the fact that there has been a lot of semantic attrition. Our starting point will therefore be the high degree of grammaticalisation of the entire set of English modals. In fact, four morphosyntactic features can be distinguished that set the central modals clearly apart from other auxiliary verbs and lexical verbs (Krug 2000: 44). The English modals have no -s form in the present tense (*he cans), i.e. they make no person or number distinctions, they do not have non-finite forms (*canning, *to can), they show abnormal time reference to the extent that past tense forms do not normally express past time and they can only be combined with a bare infinitive (He must be there, *He must it). These features are indicative of the decrease of intraparadigmatic variability (as each modal has either only one (must) or two (may/might) forms at its disposal, its formal variability is low) and structural scope (as the English modals can only modify another infinitive) and the increase of paradigmaticity (the set of English modals is limited and formally highly similar). These features, however, do not hold for the German, Dutch or Danish modals: they all have non-finite forms, for instance, their past tense forms do refer to past time, and they allow for complementation patterns other than the bare infinitive. However, a closer look at the modals in other Germanic languages reveals that non-finite uses are not always sanctioned; more particularly, it seems that the non-English Germanic modals – in clear contrast to the English ones – formally distinguish between non-epistemic and epistemic modality, whereby the latter is generally more strongly grammaticalised than the former. 10

Modals in the Germanic languages

15

In the following sections, we will address the relevant parameters for the Germanic languages under consideration. We will first present the inventory of the central modals in the respective languages, have a look at the semantic range of the central modals (whether or not they still express non-modal 11 meanings, for instance) and then apply the various parameters to the members of the modal paradigm, in order to establish differences between individual members of the paradigm. We will thereby focus on three questions: (a) if and to what extent is the distinction between non-epistemic and epistemic meanings relevant for the grammaticalisation degree of the various modals in the languages under consideration? (b) are the central modals in the Germanic languages under consideration as a group grammaticalised to the same extent, that is: are there languages in which the central modals overall seem to have acquired a higher degree of grammaticalisation in comparison to the central modals in another language? (c) are there differences between the individual items of a language’s central modal paradigm regarding degree of grammaticalisation?

Before embarking on our analysis of the modals in the other Germanic languages, we will offer a brief description of the range of meanings expressed by the modals, the aim of which is to clarify a number of semantic concepts. Modal meanings can be classified in various ways. A crucial distinction for this chapter is the one between non-epistemic and epistemic modality, whereby the latter “concerns an indication of the estimation, typically, but not necessarily by the speaker, of the chances that the state of affairs expressed in the clause applies in the world” (Nuyts 2006: 6), as in (1)

Someone is knocking at the door. That will be John.

As counterparts of will prominently figure in table 2, we will include verbs of volition in our account, whereby it should be clear that these counterparts have not reached the high degree of grammaticalisation of English will, which tends to express epistemic meanings (prediction and predictability) considerably more often than volitional ones (see Coates 1983: 171). An adequate classification might distinguish between dynamic and deontic meanings (in the non-epistemic realm of modality). Dynamic modality is typically taken to include “capacities/abilities/potentials and

16

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

needs/necessities […] inherent to the first argument participant” (Nuyts 2006). Nuyts, in accordance with Palmer (2001), also includes cases in which the potential or necessity must be ascribed to (i.e. is inherent in) the situation as such (and is not dependent on a particular participant), as in (2). (2)

a. The book need not be in the library. It can also be on my desk. b. John cannot be in Spain. I’ve just seen him in the grocery store.

As will appear in the discussion to come, this type of dynamic modality (termed circumstantial dynamic (Palmer 2001), situational dynamic (Nuyts 2006), non-deontic participant-external (van der Auwera and Plungian 1998)) is hard to distinguish from genuine epistemic modality. Finally, deontic modality offers an indication of the degree of moral desirability of the state of affairs, as in (3). (3)

We should be thankful for what he has done for us, so we must find a way to show our gratitude to him.

Whereas deontic modality will not figure prominently in our discussion (in fact, most modals seem to be able to express deontic meanings, without this having important repercussions on their grammatical behaviour), we will have to introduce the category of evidential modality or evidentiality, which refers to “the source of evidence the speaker has for his or her statement” (de Haan 2006: 57). Evidentiality concerns the expression of direct and indirect sources of information, and in the latter category another subdivision is made between inferentials and hearsays or reportatives. Interestingly, it will appear that hearsay evidentiality (“the speaker was not a witness but obtained knowledge about the action from another person”, de Haan 2006: 57) is generally expressed by a member of the core modals in the Germanic languages under discussion (e.g. sollen in German, zou (and moeten, see de Haan 2001) in Dutch, skulle in Danish and munu in Icelandic) with the English modals as an important exception (Old English sculan has been attested with evidential meaning (Traugott 1989: 41), which it has lost, however). In the following sections, we will present the central modals’ paradigms for Dutch (section 2), German (section 3), Danish (section 4) and Icelandic (section 5). For each section, a general introductory overview will be given, followed by a detailed discussion of Lehmann’s parameters of

Modals in the Germanic languages

17

paradigmaticity, paradigmatic variability and structural scope for the modals in each language. 2. The Dutch modals The group of Dutch core modals has only five members: kunnen (‘can’), zullen (‘shall’), mogen (‘may’), moeten (‘must’) and willen (‘will’). Often, the verb hoeven – a negative polarity item the meaning of which can be compared to that of need in English or brauchen in German – is mentioned as a sixth modal verb (see for instance the list of modal verbs in the Dutch reference grammar Algemene Nederlandse Spraakkunst (Haeseryn et al. 1997: 984)). In contrast to the other ones, however, hoeven is not a preterite-present and it is followed by a te-infinitive instead of a bare one (although hoeven sometimes takes a bare infinitive as well, see Haeseryn et al. 1997: 970-973). It should also be noted that the use of (niet) hoeven (expressing lack of necessity) is typical of Northern Dutch, whereas in Southern Dutch (Flemish), moeten in combination with the negation marker niet is more frequently used instead of niet hoeven (Diepeveen et al. 2006: 14ff.). (4)

Dat hoef je that need.PRS you ‘You needn’t do that.’

niet te doen. not to do.INF

The verb durven, meaning ‘to dare, to have the courage’, has left the group of modal verbs (unlike its cognate in German and Danish, for instance, where dürfen viz. turde expresses permission and probability). Note, however, that Dutch durven has developed an auxilary use, typically expressing a frequentative meaning. This use is particularly common in Southern Dutch (Ronny Boogaart, p.c.). Typically, the verb combines with a te-infinitive, but uses without te are attested as well (Haeseryn et al. 1997: 1010). (5)

Dan durft het wel eens fout then dare.PRS it well once wrong ‘Then it sometimes goes wrong.’

(te) lopen. (to) go.INF

In contrast to the English modals, the Dutch modals have a full present and past tense paradigm, and they allow non-finite forms (infinitive, past

18

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

participle (except for zullen), and present participles for kunnen and willen (kunnend, willend)). Imperatives and passives, however, are generally ruled out (Haeseryn et al. 1997: 94-100), although an imperative for willen is mentioned (wil-wilt; (Haeseryn et al. 1997: 98). More details will be given in section 2.1. (6)

a. Hij/Zij moet/moeten he/they must.PRS.SG/PL ‘He/they must do it.’

het doen. it do.INF

b. Hij moest het doen. he must.PST it do.INF ‘He had to do it.’ c. Hij heeft het moeten doen. He have.PRS it must.INF do.INF ‘He’s had to do it.’ d. Hij heeft het gemoeten/gekund/gemogen/gewild. must.PTCP/can.PTCP/may.PTCP/will.PTCP he have.PRS it ‘He’s had to/he’s been able to/he has been allowed to/he has wanted it.’ One of the striking features of the Dutch modals is the fact that they express a wide range of non-modal meanings. These can be found not only for willen ‘to want’ and kunnen ‘to know’, the cognates of which in other languages also often have premodal meanings, but also for mogen and moeten, both of which express (dis)like, sympathy and antipathy (Barbiers 2002). (7)

a. Hij moet haar niet. he must.PRS her not ‘He doesn’t like her.’ b. Ik mag geen soep. 12 I may.PRS no soup ‘I don’t like soup.’

Modals in the Germanic languages

c. Ik mag hem I may.PRS him ‘I like him.’

19

wel. well

Moreover, in some (southern) varieties of Dutch, moeten can also mean ‘to owe’ (Diepeveen et al. 2006: 100f.). (8)

Hoeveel moet ik u? you how.much must.PRS I ‘How much do I owe you?’ In combination with a nominal object designating a language, the verb kunnen can occur with the meaning ‘to know’. 13 (9)

Hij kan Frans. he can.PRS French ‘He knows French.’

The verb kunnen in combination with the preposition tegen can also express (physical or psychological) aversion. (10) Ik kan niet tegen katten. I can.PRS not against cats ‘I can’t stand cats.’ As far as epistemic meanings are concerned, all core modals are claimed to have developed them (Haeseryn et al. 1997: 984). It should be noted, however, that a number of these epistemic meanings are not as strongly developed as in English. Nuyts (2001: 187; cp. also van der Auwera 1999: 60), for instance, claims that mogen has lost its epistemic meaning in present-day Dutch and that the epistemic use of kunnen is highly instable (most cases of epistemic modality are in fact vague between dynamic and epistemic modality). Moreover, Nuyts doesn’t mention an epistemic meaning for Dutch willen. The only two central modals that in Nuyts’s view can express epistemic or evidential modality are moeten (11a)) and zullen (whose past tense has reportative meaning, see (11c)). (11) a. Ze moet oponthoud she must.PRS delay

gehad hebben. have.PTCP have.INF

20

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

‘She must have been delayed.’ (otherwise she wouldn’t be an hour late) b. Ze zal wel oponthoud gehad hebben. she shall.PRS well delay have.PTCP have.INF ‘She will have been delayed.’(I gather) c. Ze zou oponthoud gehad hebben. she shall.PST delay have.PTCP have.INF ‘She’s said to have been delayed.’ In the following subsections, we will apply the parameters of paradigmaticity (2.1), paradigmatic variability (2.2) and structural scope (2.3) to the Dutch modals. 2.1. Paradigmaticity On the whole, i.e. irrespective of whether the Dutch modals invite an epistemic or a non-epistemic reading, the degree of paradigmaticity has increased over time (see Birkmann 1987: 375). Important elements in this respect are the size of the paradigm on the one hand, and its formal homogeneity on the other. Like English, Dutch has a relatively small paradigm of core modals whose members (which can be characterised as preterite-presents) can all take bare infinitives. Still, one could claim that paradigmaticity (in particular: the formal similarity between the members of the modal paradigm) is much more outspoken in English than in Dutch. In English, the individual members of the modal paradigm have fewer forms at their disposal (in the present tense there is only one form (can, no *cans), and formal variation is restricted to vowel change and addition of a dental suffix in the case of the ‘past’ forms of the modals (will-would, can-could). For must, there is no alternation between a present and a past form anymore. (12) He said he must / had to go back. In Dutch, however, there is clearly more formal variation. A number of the Dutch modals (kunnen, mogen, zullen) show vowel alternation between

Modals in the Germanic languages

21

singular and plural (e.g. ik kan – wij kunnen), which is one of the typical characteristics of the preterite-presents (Birkmann 1987: 55). Table 3. Formal variation in the present-tense paradigm of the Dutch modals

ik (1st sg) jij (2nd sg) hij/zij (3rd sg) wij/jullie/zij (pl)

kunnen present

willen present

mogen present

zullen present

moeten present

kan kan / kunt kan kunnen

wil wil / wilt

mag mag

zal zal / zult

moet moet

wil willen

mag mogen

zal zullen

moet moeten

Moreover, in the singular paradigm of kunnen, willen and zullen, the first and third persons are identical, whereas the second person has two options at its disposal. 14 Note that the use of the forms without -t increases the homogeneity of the singular paradigm, there’s only one remaining singular form. This points to an ongoing tendency towards simplification or regularisation of the present tense paradigm. Moreover, the loss of the forms with -t makes the modal verbs ‘different’ from ordinary lexical verbs, which have -t in both the second and the third person singular (jij drinkt, hij drinkt vs. jij kan, hij kan). As far as the past tense is concerned, we observe that the modal willen has two past tense forms at its disposal: wilde and wou, with the latter being the more recent one (see table 4). The other Dutch modals only have one preterite form (moest, kon, zou, mocht). Table 4. Formal variation in the past tense paradigm of Dutch willen willen past ik (1st sg) jij (2nd sg) hij/zij (3rd sg) wij/zij/jullie (pl)

wilde (more frequent)/ wou 15 wilde / wou wilde /wou wilden / (substandard) wou(d)en

Another factor, this time contributing to the modals’ paradigmaticity, is the fact that in combination with an infinitive, the expected past participle of the modal (in perfect and pluperfect constructions) is “replaced” or “substituted” by the infinitive form (the so-called infinitivus pro participio

22

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

or IPP-construction 16 ). The use of the past participle (gekund, gemoeten, etc.) is therefore restricted to cases in which the modal verb has a nonverbal complement. It should be noted, however, that Dutch IPPconstructions occur with a relatively wide range of verbs, i.e. they are not restricted to the modal verbs, but also occur with inchoatives and durative verbs, for instance (het is beginnen te regenen, lit. it is start (IPP) to rain (INF), ‘it has started to rain’) (see Ponten 1973 for more details). In German, as we will see, there has been a narrowing down of verbs with IPP to the extent that in present-day German, this category is more or less restricted to the modal verbs (together with brauchen) and verbs of perception. As the Dutch modals share the IPP-parameter with a wider range of verbs, this factor contributes less to the Dutch modals’ paradigmaticity than in German, where IPP is much more restricted and in a way typical of the modals. (13) a. Hij heeft het moeten (*gemoeten) zeggen. must.INF (*PTCP) say.INF he have.PRS it ‘He’s had to say it.’ b. Dat heeft hij nooit that have.PRS he never ‘He’s never had to do this.’

gemoeten. must.PTCP

Another characteristic of Dutch modals is the fact that in compound past tenses, either the auxiliary selected by the modal (hebben) or the auxiliary selected by the infinitive complement (zijn) can be used. The latter case is indicative of higher grammaticalisation, as the modal verb ‘loses’ its typical properties, as it no longer controls the selection of hebben as a perfect auxiliary (for a discussion of the parameters that favour either hebben or zijn, see Hofmans 1980a-b). So, modals in combination with infinitives typically taking zijn (like gaan ‘to go’ in example (14a)), can also take zijn (next to hebben) as a perfect auxiliary. Note that this phenomenon is typical of Dutch: German, Danish and Icelandic allow only of a have-cognate in the perfect (14b) (but see Cornillie et al. (this volume) for a similar observation pertaining to modal verbs in Italian). (14) a. Hij is naar huis he be.PRS to-home ‘He’s had to go home.’

moeten gaan. must.INF go.INF

Modals in the Germanic languages

b. Er hat/*ist nach Hause he have.PRS/*be.PRS to.home ‘He’s had to go home.’

23

gehen müssen. go.INF must.INF

Strangely enough, the perfect auxiliary zijn can also be used when the modal (without an infinitive) is combined with a directional phrase (as in example (15) naar de garage ‘to the garage’). This is difficult to account for in grammaticalisation terms, as one would expect less grammaticalised behaviour in this (clearly non-epistemic, dynamic) environment. (15) Onze auto is/ heeft vaak naar de garage gemoeten. must.PTCP our car be.PRS/ have.PRS often to the garage ‘Our car has had to be taken to the garage often.’ 2.2. Paradigmatic variability Paradigmatic variability concerns the freedom with which the language user chooses a sign. It is high when the speaker’s freedom is barely restricted, so that s/he can pick out forms of the paradigm according to his/her own communicative needs; paradigmatic variability is restricted to the extent that the speaker’s options are narrowed down. As far as this parameter is concerned, Dutch seems to make a clear distinction between non-epistemic and epistemic modality, to the extent that non-finites are (more or less) excluded in epistemic uses. This affects the degree of intraparadigmatic variability, i.e. “the selection of alternatives which are in opposition” (Lehmann 2002: 138): with epistemic uses, only finite forms can be chosen. Let’s have a look at the following examples. In the IPP-constructions (16a) and (16b), the infinitives kunnen and moeten, respectively, only have a non-epistemic reading. The same holds for the past participles gemoeten and gekund in (16c), which get a deontic reading. (16) a. Hij heeft het kunnen zeggen. he have.PRS it can.INF say.INF ‘He has been able to say it.’ b. Ze heeft het moeten lezen in de krant. she have.PRS it must.INF read.INF in the newspaper ‘She had to read it in the newspaper.’

24

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

c. Dit had beter gekund en gemoeten. this have.PST better can.PTCP and must.PTCP ‘It could and should have been better.’ One important counterexample has to be mentioned, however. In combination with zou (formally a past tense of zullen), a formal infinitive kunnen appears that can get a ‘conditional’ epistemic reading, as in (17a) (see Van Ostaeyen and Nuyts 2004: 108). Moreover, the complement infinitive after kunnen can be omitted with zou so that we get an epistemic meaning expressed by a nonfinite form of a modal (kunnen) and complemented by a that-clause (17b). It should be stressed that this seems to be possible only with kunnen. (17) a. Het zou kunnen zijn dat Jan thuis is. it shall.PST can.INF be.INF that Jan at.home be.PRS ‘Jan could be at home.’ b. Het zou kunnen dat it shall.PST can.INF that ‘Jan could be at home.’

Jan Jan

thuis is. at.home be.PRS

2.3. Structural scope As far as structural scope is concerned (involving the loss of syntagmatic weight), Dutch modals show an increased preference for infinitival complements in the epistemic realm. In their dynamic and deontic uses, however, they often occur without infinitive complements. The Dutch core modals have a clear liking for intransitive uses and allow for a wide range of non-infinitive arguments, which clearly distinguishes them from German and Danish modals. In Dutch, non-epistemic modals are still compatible with directional prepositional phrases (18a) and adverbs (18b), which used to be the case for English modals up to the Early Modern English period as well (see Plank 1984: 325). Moreover, the modal verb willen (in its non-epistemic use) is compatible with a that-clause (in object-function).

Modals in the Germanic languages

(18) a. Hij mag/kan/moet/wil he may.PRS/can.PRS/must.PRS/will.PRS ‘He may/can/must/wants to go home.’

25

naar huis. to home

b. Alles moet (mag, kan) weg. everything must.PRS (may, can) away ‘Everything must/may/can be sold.’ c. Ze wil dat hij naar huis komt. she will.PRS that he to home come.PRS ‘She wants him to come home.’ Dutch non-epistemic modals can also be combined with past participles to express passive meanings. Note that the passive auxiliary (worden) can be added without difference in meaning. (19) a. Dat moet gezegd! that must.PRS say.PTCP ‘That must be said!’ b. De hele gang moest geschilderd. the entire corridor must.PST paint.PTCP ‘The entire corridor had to be painted.’ Dutch modals also allow for nominal objects and adjectives. As shown in Barbiers (2002), combinations with adjectives are only possible when the adjective denotes a value on a bounded scale, i.e. a value that is either zero or one. 17 This is the reason why a Dutch modal like moeten can be combined with adjectives like vol ‘full’, leeg ‘empty’ and dood ‘dead’, but not with adjectives like lang ‘long’ or oud ‘old’. (20) a. Hij moet en zal een snoepje. he must.PRS and shall.PRS a candy ‘He insists on having a candy.’ b. Deze fles moet vol/leeg. this bottle must.PRS full/empty ‘This bottle must be filled/be emptied.’

26

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

c. Moet dat nu echt kapot? must.PRS that now really broke ‘Do you really want to destroy it?’ Finally, the Dutch modals can also occur in ‘intransitive’ structures. 18 Note that these sentences are not really felt to be elliptical, as it is often very difficult to decide what has been left out. (21) a. Moet dat nu echt? must.PRS that now really ‘Is this really necessary?’ b. Kan dat? can.PRS it ‘Can it be?’ Van Ostaeyen and Nuyts (2004: 43-46) observe that for kunnen, the frequency of such main-verb or main-verb like uses has increased in present-day Dutch. In early modern Dutch, these uses were rare. Van Ostaeyen and Nuyts (2004: 46) hint at the idea that a process of formal degrammaticalisation (at least as far as this particular parameter is concerned) is going on here. It should also be stressed that in this mainverb like use, the meaning of the verb is not at all restricted to dynamic modality; in example (22), the meaning of kunnen is clearly deontic. (22) […] wat u doet kan helemaal niet, een klooster bouwen op het grootste Joodse kerkhof ter wereld. (Van Ostaeyen and Nuyts 2004: 45) ‘what you’re doing is not done (lit. ‘can not’ i.e. ‘is not allowed’), building a monastery at the spot of the biggest Jewish cemetery in the world.’ The fact that Dutch allows this main-verb like deontic use is also in line with the observation that in Dutch the modal source (i.e. the entity in which the modal relation originates) can appear as an onstage participant introduced by the preposition van.

Modals in the Germanic languages

27

(23) Hij mag/moet dat niet van mij. He may/must.PRS that not from me ‘He is not allowed to do that/he needn’t do that, as far as I’m concerned.’ Note also that the epistemic use of zou kunnen (without zijn) as in example (17b) above seems to suggest that intransitive uses of the modals can express epistemic modality as well. 2.4. Summary: the Dutch modals The Dutch modals do not seem to score particularly high with respect to the three grammaticalisation parameters. As far as paradigmaticity is concerned, we have observed that there is considerable variation within the paradigm, i.e. the Dutch modals do not behave in a particularly homogeneous way. There are double forms in the present paradigm of zullen, willen and kunnen; three modals show up vowel alternation between singular and plural (kunnen, mogen, zullen), two others don’t (moeten, willen). The past tense is built irregularly as well (especially past tense moest stands out) and has double forms in the case of willen (wilde/wou). The past participle either ends in a dental suffix (gekund, gewild) or in -en (gemoeten, gemogen). Paradigmaticity seems to increase for the epistemic modals, as only three modals seem to have developed a well-established epistemic reading (moeten, zullen and to a lesser extent kunnen). Moreover, participles of the modals are not compatible with an epistemic reading, which increases paradigmaticity as well. Typical of the Dutch modals is the fact that the perfect auxiliary zijn can occur next to hebben in composed past tenses, i.e. zijn can be selected by the infinitive complement of the modal. The distinction between non-epistemic and epistemic meanings is particularly relevant if we consider the parameters of paradigmatic variability and structural scope. Both are on the decrease when the modals are used epistemically, as the epistemic use tends to rule out nonfinite forms of the modals (with some exceptions) and epistemic modals normally require an infinitive complement, thus restricting their structural scope.

28

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

Table 5. Dutch modals: paradigmaticity paradigmaticity [+] – Class of central modals consists of only five members. – Dutch epistemic modals build an even smaller paradigm (no epistemic meanings for mogen and willen). – IPP-constructions restrict the use of the past participle. – Perfect auxiliary zijn (next to hebben) in perfect constructions with modal verb and infinitive taking zijn.

[-] – For each modal, there are present, past, infinitive and participle (both present and past) forms (only zullen lacks participle forms). – Irregular formation of the various forms (past tense, past participle) decreases formal homogeneity of the paradigm. – Formal variation (double forms) in the present/past tense paradigm of willen, zullen and kunnen indicates lack of formal stability.

It should be stressed that non-epistemic uses of the Dutch modals are compatible with a particularly wide range of arguments, i.e. the Dutch modals generally possess a high degree of structural scope, which distinguishes them – as we will see – from the modals in German and Danish. Dutch modals are compatible with past participles, adjectives, noun phrases, complement clauses and pronouns. Table 6. Dutch modals: paradigmatic variability and structural scope paradigmatic variability [+] [-] Non-epistemic modals occur in finite Epistemic modals favour finite and nonfinite forms (participles, forms (with some exceptions). infinitives). structural scope Non-epistemic modals occur in Epistemic modals clearly prefer ‘intransitive’ structures and can be infinitive complements. Their combined with nouns, prepositional structural scope is rather narrow. phrases, adjectives, participles and infinitives. Their structural scope is rather wide.

Modals in the Germanic languages

29

In some cases, there seems to be an imperfect match between epistemic meanings and the expected increased degree of grammaticalisation, on the one hand, and non-epistemic meanings and the expected lesser degree of grammaticalisation, on the other. We find epistemic meanings with ‘intransitive’ kunnen, as in (24), repeated from (17b). (24) Het zou kunnen dat it shall.PST can.INF that ‘Jan could be at home.’

Jan Jan

thuis is. at.home be.PRS

Conversely, main verb uses allow for zijn as a perfect auxiliary (as if an ellipted infinitive (verb of movement) governed the selection of the auxiliary), as in (25). (25) Hij is naar het ziekenhuis gemoeten. he be.PRS to the hospital must.PTCP ‘He had to go to the hospital.’ From an intralinguistic perspective, moeten and zullen seem to be the most strongly grammaticalised modals in Dutch, whereas willen and (to a somewhat lesser extent) kunnen are at the other end of the grammaticalisation spectrum, i.e. they behave still more strongly like full verbs. The latter verbs exhibit an almost complete conjugation paradigm (with present participles and even imperative forms – for willen), show formal instability (a range of alternatives) and allow for the widest range of complementation patterns. 3. The German modals The paradigm of German modals consists of six preterite-present 19 members: können, sollen, müssen, mögen, dürfen and wollen. They can occur in the present and past indicative, in the present and past conjunctive (labelled Konjunktiv I and Konjunktiv II, respectively), and in perfect tense forms (in which case the so-called Ersatzinfinitiv is used; see also section 2.1 on IPP-constructions in Dutch). (26) a. Er he

muss/musste/müsse/müsste nach Hause. must.PRS.IND/PST.IND/ PRS.CONJ/ PST.CONJ to.home

30

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

b. Er hat nach Hause fahren müssen. he have.PRS to.home drive.INF must.INF ‘He’s had to drive home.’ c. Er hat nach Hause gemusst. he have.PRS to.home must.PTCP ‘He had to go home.’ Strictly speaking, the core modals’ paradigm is more elaborate than in the case of the Dutch modals, which have no synthetic conjunctive forms. This, however, seems to be due to the general morphological complexity of the German verb system, in which the category of mood is still productive. Formally and functionally, the negative polarity verb brauchen is “attracted” towards the category of German core modals (Duden 2005: 465, Askedal 1998). Note that the verb does not belong to the group of preteritepresents and is normally combined with a to-infinitive. First, its past conjunctive often gets Umlaut (bräuchte instead of the regular form brauchte, which is analogous to the past conjunctive forms müsste, könnte, möchte and dürfte); second, a present tense form brauch (without -e, resp. t, thus resembling the other -e and -t-less forms muss, kann, darf etc.) is used both in the first and third person singular. Note that this form is not accepted in standard German. A third indication of brauchen’s integration in the modal verb paradigm is the loss of the infinitival marker zu, illustrated in (27): (27) Er braucht es nicht (zu) tun. he need.PRS it not (to) do.INF ‘He needn’t do it.’ Another verb whose membership in the category of German modals is somewhat unclear is the future auxiliary werden. In contrast to English (will/shall), Dutch (zullen) or Danish (skulle), none of the German core modals expresses futurity. The verb werden has stepped in here. It does not belong to the group of preterite-presents, but patterns with a bare infinitive in present-day German. Originally meaning something like ‘turn’, werden developed inchoative uses (mainly in combination with a present participle) in Old High German. The grammaticalisation of werden with a bare infinitive into a future auxiliary took place in the Middle High German period, in which it came to replace earlier ‘future’ uses expressed by the predecessors of the modals sollen, wollen and müssen (see for an account

Modals in the Germanic languages

31

on the history of werden, Kotin 2003: 151-163 and 166-175). Apart from syntactic arguments, werden’s inclusion in the group of German modals can also be argued for with functional arguments, as it has developed an epistemic use as well, as in (28) – for a discussion of the grammaticalisation status of werden in present-day German, see e.g. Mortelmans (2004). (28) Er wird jetzt wohl he become.PRS now well ‘He will be there by now.’

da sein. there be.INF

As far as the semantics of the German modals is concerned, it can be noted that the modals express a wide range of meanings, from lexical to dynamic, deontic and epistemic ones. The German modals mögen (both indicative mag ‘like’, ‘want’ and past conjunctive möchte ‘would like to have’), wollen ‘want’ and können ‘know’ express non-modal meanings: (29) a. Ich mag keine I may.PRS no ‘I don’t like soup.’

Suppe. soup

b. Ich möchte ein Kind. I may.CONJ.PST a child ‘I would like to have a child.’ c. Ich will ein Kind. I will.PRS a child ‘I want a child.’ d. Er kann das Gedicht auswendig. he can.PRS the poem by heart ‘He knows the poem by heart.’ No other lexical meanings can be discerned for the German modals, though, which means that their range of non-modal meanings is a bit lower than in Dutch, where moeten can both mean ‘like’ (see example (30a) repeated from (7a) above) and ‘owe’ (see example (30b) repeated from (8) above). Note that the German counterpart of Dutch moeten meaning ‘to owe’ is the verb sollen rather than müssen. This sollen is found in main verb like use until the 18th century (Diewald 1999: 322), as exemplified in

32

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

example (30c). In present-day German, however, the infinitive seems to be obligatory. (30) a. Ik moet haar niet. I must.PRS her not ‘I don’t like her.’ b. Hoeveel moet ik je? how much must.PRS I you ‘What do I owe you?’ c. Was soll ich Ihnen für den what shall.PRS I you for the ‘What do I owe you for the tee?’

Tee? (Goethe) tea

Every German modal has developed an epistemic meaning; in the case of sollen and wollen, this is of the evidential kind, expressing reportative modality (comparable to zou in Dutch, skulle in Danish and munu in Icelandic). The frequency of these epistemic meanings, however, is rather low. Diewald’s admittedly rather restricted corpus of present day German material (Diewald 1999: 217) yields frequencies between 15.8 % (for dürfte) and 0.0 % (for reportative wollen). This might be indicative of the fact that the epistemic variant is not as well-entrenched in German as it is in English (where the epistemic use of the modals is much more frequent, compare e.g. Abraham 2002: 8). Due to a lack of comparable analyses, it is rather difficult to judge German and Dutch in this respect. 3.1. Paradigmaticity The German core modals form a relatively concise paradigm with a rather strong formal homogeneity. All members are preterite-presents (or, in the case of wollen, are modelled on them); five out of six verbs show the original vowel alternation between singular and plural in the present tense paradigm. Only sollen has everywhere. No double forms as in Dutch seem to exist. The past tense is built by means of the dental suffix -e, which is attached to the stem. The past participle ends in -t.

Modals in the Germanic languages

33

Table 7. Present, past and past participle forms of the central modals in German present tense ich, er muss ich, er kann ich, er will ich, er mag ich, er darf ich, er soll

wir müssen wir können wir wollen wir mögen wir dürfen wir sollen

past tense

past participle

musste konnte wollte mochte durfte sollte

gemusst gekonnt gewollt gemocht gedurft gesollt

Unlike Dutch, no tendencies towards further simplification or regularisation of the paradigm can be observed; the present-day German paradigm seems to be more stable than the Dutch one. With some modals, epistemic readings are only compatible with particular forms; this holds for past conjunctive dürfte and present indicative mag (this formal restriction could also be considered as an example of the decrease of paradigmatic variability, see section 3.2.). (31) a. Sie dürfte (*darf) nicht mehr ganz nüchtern sein. she may.PST.CONJ not anymore completely sober be.INF ‘She’s probably not completely sober anymore.’ b. Mag sein, dass das in der Schweiz wohl may.PRS be.INF COMP that in Switzerland indeed funktioniert. function.PRS ‘It’s very well possible that this functions in Switzerland.’ It has already been noted that both German and Dutch have the IPPconstruction. In German, however, IPP mainly occurs with the core modals and brauchen ‘need’, with the causative verb lassen ‘make’, and with the perception verbs sehen ‘see’, hören ‘hear’, fühlen ‘feel’ and spüren ‘feel’ (Duden 2005: 473). The verbs of perception can take a past participle as well. In Dutch, the category of verbs taking IPP is much larger, so that the IPP-parameter adds to the paradigmaticity of the modals in German to a larger extent than in Dutch, where IPP does not single out the central modals.

34

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

3.2. Paradigmatic variability As is the case for Dutch, epistemic uses of the modals have to be finite, i.e. “EMV [epistemic modal verbs] cannot occur in non-finite environments – i.e. no epistemic reading emerges in embedded constructions” (Abraham 2001: 11). So, the perfect tense of müssen in the following example can only express dynamic modality and does not invite an epistemic interpretation. (32) Er hat das Buch bis Dienstag lesen müssen. he have.PRS the book till Tuesday read.INF must.INF ‘He had to read the book by Tuesday.’ (example taken from Diewald 1999) In Dutch, zou (past tense form of zullen) can be combined with an infinitive of a modal to express epistemic meaning. The combination of German würde with the infinitive of a modal (*? würde sein können in (33b)), however, is awkward, especially when the modal has epistemic meaning. (33) a. Het zou kunnen dat hij it shall.PST can.INF COMP he

komt. come.PRS

b. Es könnte sein, dass er kommt. (*Es können, …) it can.PST.CONJ be.INF COMP he come.PRS ‘It could be that he’ll come.’

würde

sein

However, some authors state that there are exceptions to this rule. Reis (2001: 295), for instance, claims that non-finite forms do allow an epistemic interpretation in irrealis conditional constructions and in attributive infinitival complements like the ones in (34); according to Reis, therefore, the non-finite gap is not absolute. For Reis, semantic factors are responsible for the fact that non-finites are generally not compatible with epistemic meaning, but the general incompatibility is not a grammatical characteristic of the German epistemic modals as such.

Modals in the Germanic languages

35

(34) a. Nach allem, was ich weiß, hätte er dann after everything that I know have.PST.CONJ he then zu Hause sein müssen. home be.INF must.INF ‘On the basis of what I know, he should have been home by then.’ b. Nach allem, was ich weiß, hätte er da after everything that I know have.PST.CONJ he then noch in Prag sein können. still in Prague be.INF can.INF ‘On the basis of what I know, he could have been in Prague at the time. c. Der Gedanke, sich täuschen zu müssen, drängte sich auf. the thought be.mistaken.INF to must.INF forced itself on ‘The idea that he was wrong forced itself upon him’ 3.3. Structural scope In their non-epistemic use, the structural scope of the German modals is considerably higher than in English, as the German modals (just like the Dutch ones) can still be combined with directional prepositional phrases 20 and adverbs. (35) a. Er kann nach Hause. he can.PRS to.home ‘He can go home.’ b. Alles muss weg. everything must.PRS gone ‘Everything must go.’ Still, there has been considerably more condensation in German than in Dutch, as a number of structures without a complement infinitive are ruled out in German. Put differently, the complexity of constituents a German modal can combine with is considerably lower than in Dutch, so that its structural scope is smaller than the structural scope of a corresponding Dutch modal. Intransitive-uses (i.e. without an infinitive, see (36a-b), uses

36

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

with a past participle (36c) and uses with adjectival complements (36d) are clearly out for the German modals. (36) a. ?*Darf (kann) das? (But: Darf (kann) das sein?) Mag (kan) dat? May (can) that ‘Is it permitted/possible?’ b. ?*Muss das nun wirklich? (But: Muss das nun wirklich sein?) Moet dat nu echt? must.PRS it now really ‘Is this really necessary?’ c. *Es muss gesagt, dass...) Het moet gezegd, it must.PRS say.PTCP ‘It must be said.’

dass (but: Es muss gesagt werden, dat … that

d. *Der Stuhl darf weiß. De stoel mag wit. the chair may.PRS white ‘The chair may be painted white.’ German modals can only be combined with nominal objects in the case of können, wollen and mögen (see examples under (29) above), and with complement clauses introduced by dass in the case of wollen and mögen (see example (37a), taken from Diewald 1999: 55). Contrary to Dutch, uses of müssen and sollen with an NP (as in (37b)) are generally impossible. (37) a. Er mag, dass du ihm hilfst. he may.PRS COMP you him help.PRS ‘He wants you to help him.’ b.*Er muss und soll einen Bonbon. he must.PRS and shall.PRS a candy ‘He insists on having a candy.’

Modals in the Germanic languages

37

The German modals can be combined with pronominal objects like das and es, as in the following example (see for discussion also Öhlschläger 1989: 65-68). (38) Er kann es, er muss es, er darf es. he can.PRS it, he must.PRS it he may.PRS it ‘He’s able to do it, he must do it, he may do it.’ 3.4. Summary: the German modals It seems that the German modals score higher with respect to the grammaticalisation parameters than the Dutch ones, especially with regard to the parameter of structural scope. The German modals more often take an infinitive – also in their non-epistemic uses – than the Dutch ones (compare German das kann sein with Dutch dat kan). Uses in which the modal is combined with a past participle (*das muss gesagt) or an adjective (*er muss tot) are ruled out in German, and German modals resist the combination with nominal objects to a greater extent than Dutch ones (German *er muss einen Bonbon vs. Dutch hij moet een snoepje). Table 8. German modals: structural scope structural scope [+]

[-]

– German modals can be combined with noun phrases and prepositional phrases in nonepistemic use only.

– German modals show a strong tendency towards infinitival complements; combinations with past participles and adjectives are ruled out, intransitive structures do not seem to occur either. – German epistemic modals share an outspoken preference for infinitive complements.

As far as paradigmaticity is concerned, the German paradigm is more complex than the Dutch one (allowing for synthetic conjunctive forms, for instance), on the one hand, and formally more stable and regular than the Dutch one, on the other. These two observations point into different directions as far as grammaticalisation is concerned: for German, the higher

38

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

complexity (i.e. the wide array of formal realisations of a particular modal like muss, müsste, müsse, müssen, gemusst etc.) could be interpreted as signalling low paradigmaticity (which would be symptomatic of low grammaticalisation), whereas the lack of formal stability in Dutch leads to an increase of formal variants (e.g. wou/wilde), hence to a more complex paradigm, which again can be taken to signal low grammaticalisation. As the formal complexity of the German modal paradigm is completely in line with the greater inflectional possibilities of the German verb in general, we are inclined to think that it is the second element, i.e. the Dutch lack of formal stability and internal regularity in contrast to the stability and regularity of the German modal paradigm, which should be stressed as more significant in the comparison of both languages. Table 9. German modals: paradigmaticity paradigmaticity [+]

[-]

– German modals build a relatively small paradigm. – The German paradigm is homogeneous and fairly regular (past tense = stem + -te, all past participles end in -t). – There is no formal variation (double forms) in German. –IPP-constructions are obligatory. The category of verbs that obligatorily take IPP is restricted to the modal verbs (with brauchen) and heißen, lassen and sehen.

The paradigmatic variability of the German modals clearly decreases when they are used epistemically: with only a few exceptions (see Reis 2001), the epistemic use is restricted to finite forms of the modals.

Modals in the Germanic languages

39

Table 10. German modals: paradigmatic variability paradigmatic variability [+]

[-]

In German (as in Dutch), nonepistemic modals occur in finite and non-finite forms (participles, infinitives).

In German (as in Dutch) epistemic modals favour finite forms of the modals (with a few exceptions).

From a semantic point of view, all of the German modals have developed an epistemic or evidential reading. In contrast, the existence of strongly entrenched epistemic uses for every modal in Dutch can be doubted (see Nuyts 2001). 4. The Danish modals The Danish paradigm of modal preterite-presents consists of six members: ville ‘will’ (cf. however footnote 5), skulle ‘must’, måtte 21 ‘may’/‘must’, kunne ‘can’, burde 22 ‘ought to’ and turde ‘dare’. 23 All members can be constructed with a bare infinitive. (39) Hun vil/skal/må/kan/bør/tør svømme. she will/must/may/can/ought.to/dare.to.PRS swim.INF ‘She wants to/must/may/can/ought to/dares to swim.’ One member, turde, can also be constructed with an infinitive marked by at (‘to’). (40) Han tør ikke (at) svømme. he dare.PRS not (to) swim.INF ‘He does not dare to swim.’ In this respect, turde is similar to the verbs behøve ‘need’ and gide ‘feel like’, ‘be bothered (to)’ that are normally included among the Danish modals, although they are not preterite-presents (e.g. Skyum-Nielsen 1971, Brandt 1999, Boye 2001, Jensen 2005).

40

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

(41) a. Butleren behøver ikke (at) være morderen. butler.DEF need.PRS not (to) be.INF killer.DEF ‘The butler doesn’t have to be the murderer.’ b. Jeg gider ikke (at) læse I is.bothered.PRS not (to) read.INF ‘I can’t be bothered to read that book.’

den that

bog. book

Among the six modal preterite-presents, kunne, skulle, ville and måtte may be taken to constitute a core paradigm. First, they are the only modals that cannot be represented by gøre ‘do’ (see Brandt 1999: 17). Compare: (42) a. Jeg tør /behøver ikke se den I dare /need.PRS not watch.INF that *gør du? do.PRS you ‘I dare/need not watch that movie – do you?’

film – movie

b. Jeg kan/vil/skal/må ikke se den film – I can/will/must/may.PRS not watch.INF that movie *gør du? do.PRS you ‘I cannot/will not/must not/may not watch that movie – do you?’ Second, kunne, skulle, ville and måtte are by far the most frequent of the modal verbs (Brandt 1999: 27). Some of the modals have additional non-modal meanings. Among the preterite-presents, kunne like its German cognate can still mean ‘know how to’, as in (43). (43) Han kan en sang. he can.PRS a song ‘He knows a song.’ The modals kunne, burde and skulle are all regularly used epistemically. So is måtte with necessity meaning, in contrast to måtte with possibility meaning (e.g. Boye 2001: 51-54). Note that epistemic skulle (44b) has an evidential interpretation, just like its cognates in German (sollen) and Dutch (zou).

Modals in the Germanic languages

41

(44) a. Butleren kan/ bør/må være morderen. butler.DEF can/ought/must.PRS be.INF killer.DEF ‘The butler may/should/must be the murderer.’ b. Butleren skal være morderen. butler.DEF must/shall.PRS be.INF killer.DEF ‘The butler is the murderer, it is said.’ The modal turde also seems to have epistemic uses. However, these are found mainly in idiomatised expressions such as (45). (45) Det tør være overflødigt at nævne at. it dare.PRS be.INF superfluous to mention.INF that ‘It should be unnecessary to mention that…’ As for the modal ville, it is disputed whether it has an epistemic variant or not (e.g. Davidsen-Nielsen 1990: 161-163, Brandt 1999: 60-62, Boye 2001: 58-64). In any case, epistemic uses of ville as in (46) would be very rare. (46) Hun vil være i Danmark she will.PRS be.INF in Denmark ‘She will be in Denmark right now.’

lige nu. right now

As has already been noted, måtte can be used epistemically only with necessity meaning, not with possibility meaning. As it is the case with other Germanic modals, the past tense forms of the Danish modals are regularly used as past-tense hypotheticals instead of pure tense forms. So, in (47), past tense kunne weakens the strength of the epistemic judgement. (47) Butleren kunne være morderen. butler.DEF can.PST be.INF killer.DEF ‘The butler could be the murderer.’ 4.1. Paradigmaticity The most salient feature of the Danish modals (preterite-presents as well as non-preterite-presents) is the ability to co-occur with a bare infinitive. This

42

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

feature is what defines the paradigm of modals, and it has defined the paradigm all the way back to Common Nordic (Old Norse). For some individual modals, however, a diachronic change in paradigmaticity can be registered, which manifests itself in the gradual loss and acquisition, respectively, of the infinitival marker at (‘to’) before the infinitives accompanying the modals. Thus, while it seems that turde is currently moving out of the paradigm (i.e. it seems that it is gradually acquiring at before its accompanying infinitive (see example (48)), behøve seems to be entering the paradigm (i.e. it seems that behøve is losing at before its accompanying infinitive (see example (49)) (Hansen 1977). This mirrors the behaviour of brauchen in present-day German, which is said to increasingly lose the infinitival marker zu (see section 3.1.). (48) a. Han tør he dare.PRS

ikke danse. (old construction) not dance.INF

b. Han tør ikke at danse (new construction) he dare.PRS not to dance.INF ‘He doesn’t dare to dance.’ (49) a. Han behøver he need.PRS

ikke danse. not dance

(new construction)

b. Han behøver ikke at danse. (old construction) he need.PRS not to dance ‘He doesn’t need to dance.’ While the modal paradigm can be defined in terms of the infinitival marker, the Danish modals show a few other peculiarities as well. Notably, the inflection of Danish modals is reduced in comparison to non-modal verbs such as e.g. svømme (‘swim’). In the words of Brandt (1999: 69), modals “practically only occur in four forms” (here illustrated with kunne): infinitive (kunne), present (kan), past (kunne), and past participle (kunnet) (in which case they always select have (‘have’) rather than være (‘be’) as an auxiliary; cf. section 2.1. on Dutch). Among the other verb forms found in Danish, present participles (kunnende) and passives (kunnes) of some of the modals 24 can be found, but all modals lack an imperative form. In fact, it might be argued that the inventory of modal forms is even smaller. First, the infinitives of preterite-present modals are formally

Modals in the Germanic languages

43

identical to the pasts, both in spoken and written language. Second, for some speakers, the past participles of preterite-present modals are formally identical to the infinitives (cf. section 2.1. on the infinitivus pro participio or IPP-construction) and thus the pasts. This latter convergence of forms can be registered even in the written language. Thus, rather than the ‘correct’ (50), (51) is frequently encountered. 25 (50) Han har kunnet svømme. he has can.PTCP swim ‘He has been able to swim.’ (51) Han har kunne svømme. he has can.INF swim ‘He has been able to swim.’ This means that morphologically the preterite-present modals might be taken to move diachronically towards having only two regularly used forms: present (kan/må/skal/vil/tør/bør) and non-present/non-finite (kunne/måtte/skulle/ville/turde/burde). 4.2. Paradigmatic variability While all the Danish modals co-occur with non-modal verbs (in the form of bare infinitives) in the same clause, the use of non-modal verbs is not linked with a systematic choice between modals. Thus, Danish modals – just like the Dutch and German ones – are not obligatory in any useful sense of the term. As mentioned in the previous section, Danish modals do not have as many forms at their disposal as regular main verbs. They show a defective conjugation pattern. For epistemic modals, this pattern is even more defective than for non-epistemic modals. Notably, epistemic modals lack participles (Brandt 1999: 96), and – as observed for Dutch and German – infinitive modals can only rarely be used epistemically. The following example, however, would be a case in point. (52) Han synes at måtte være morderen. he seem.PRS to must.INF be.INF murderer.DEF ‘It seems he must be the murderer.’

44

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

Accordingly, a modal infinitive may occur with an infinitival periphrastic blive-passive in its scope. Whereas the morphological -s-passive is compatible only with a non-epistemic reading of the modals in the scope of which it occurs (53b), the blive-passive typically goes with an epistemic reading (Lauridsen and Lauridsen 1989: 248). (53) a. Han synes at kunne blive valgt i morgen. he seem.PRS to can.INF be.INF elected tomorrow (blive-passive) ‘It seems that he may (epistemically) be elected tomorrow.’ b. Han synes at kunne vælges i morgen. he seem.PRS to can.INF elect.PASS tomorrow (-s-passive) ‘It seems he has the (dynamic or deontic) possibility to be elected tomorrow.’ 4.3. Structural scope In their non-epistemic uses some of the modals do not require an infinitive (see Brandt 1999 and Boye 2001 and 2005 for details). Among the preterite-presents, only kunne readily occurs with a noun phrase. (54) Han kan en sang. he can.PRS a song ‘He knows a song.’ A similar construction is found with ville. With ville, however, the construction has a ceremonial, archaic, or, as Brandt (1999: 74) puts it, “slogan”-like flavour. (55) Vi vil frihed! (Brandt 1999: 74) we will.PRS freedom ‘We want freedom!’ With a pronoun or a general noun phrase which can be interpreted as having “infinitival meaning” (Brandt 199: 70-71), the construction is possible with other modals as well (cf. Davidsen-Nielsen 1990: 21).

Modals in the Germanic languages

45

(56) a. Hun skal/må/tør ikke noget. she must/may/dare.PRS not anything ‘She has no plans.’/‘She is not allowed to do anything.’/’She does not dare to do anything.’ b. Hun kan/vil/må/skal en masse. a lot of things. she can/will/may/must.PRS ‘She can/wants to/may/must do a lot of things.’ In their non-epistemic uses, all Danish modals except turde, kunne and måtte with possibility meaning readily co-occur with directional adverbials (but not adjectives) (cf. Boye 2001 and 2005 for a cognitive analysis which captures the distribution of directional adverbials with modals). (57) Hun skal hjem. she must.PRS home ‘She has to go home.’ With ville (‘will’) even finite complement clauses may appear. This behaviour unites ville with its German (wollen) and Dutch (willen) cognates. (58) Kongen vil at den sorte prins forlader king.DEF will.PRS COMP the black prince leave.PRS landet. country.DEF ‘The king wants that the black prince leaves the country.’ With epistemic modals, however, all these constructions are impossible. Epistemic modals always require a bare infinitive. (59) a. Det kan være hun er gået. it can.PRS be.INF she is go.PTCP ‘It may be she has gone.’ b. Det kan være hun ikke it can.PRS be.INF she not ‘She may not have gone.’

er gået. is go.PTCP

46

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

c. Hun kan være gået. she can.PRS be.INF go.PTCP ‘She may have gone.’ d. Hun kan ikke være gået. she can.PRS not be.INF go.PTCP ‘She cannot have gone.’ Omission of the infinitive is impossible (in contrast to the Dutch construction (60b)): (60) a. *Det kan hun it can.PRS she

er gået. is go.PTCP

b. Het kan dat ze it can.PRS COMP she ‘She may have gone.’

weg is. go.PTCP is

Thus, while in their non-epistemic uses the Danish modals have a quite wide structural scope, in their epistemic uses their structural scope is narrow. It should be mentioned, finally, that both non-epistemic (61a-b) and epistemic (61c) modals may co-occur with a subject only. In that case, however, a complement is represented by ellipsis, or the meaning of a complement is retrievable from the context. (61) a. Hun skal gå nu. – Nej, hun skal ikke. she must.PRS go.INF now. no she must.PRS not ‘She must leave now. – No, she mustn’t/needn’t’. b. Jeg kann ikke længere. I can.PRS no longer ‘I cannot do that (obvious from the context) any more.’ c. Butleren må være morderen. – Må han? butler.DEF must.PRS be.INF murderer.DEF must.PRS he ‘The butler must be the murderer.’ – ‘Really?’

Modals in the Germanic languages

47

4.4. The Danish modals: summary The epistemic and the non-epistemic uses of Danish modal verbs do not show differences with respect to degree of paradigmaticity. However, in their epistemic use the modals have a much more restricted structural scope than in their non-epistemic use. As mentioned, this may be taken to suggest that epistemic modals are more grammaticalised than their non-epistemic counterparts (alternatively, one might speculate that the explanation is semantic, but unrelated to grammaticalisation; it is generally accepted that epistemic modals in contrast to non-epistemic ones take a proposition – i.e. something with a truth value – in their scope). In support of this, two facts may be noted: First, unlike non-epistemic Danish modals, epistemic modals do not occur as participles (Brandt 1999: 96) – which affects their paradigmatic variability – and second, unlike infinitives of non-epistemic Danish modals, infinitives of epistemic modals cannot occur as complements of other verbs (Brandt 1999: 136-137); however, they can occur in complements of other verbs. The main findings for Danish are summarised in table 11. Table 11. Danish modals: paradigmaticity, paradigmatic variability and structural scope paradigmaticity [+] – The Danish paradigm of core modals is relatively small (only 4 core members: kunne, skulle, ville and måtte). – The inventory of inflectional forms each modal has at its disposal is quite restricted as well: the past tense form is identical to the infinitive and, for some speakers, even to the past participle.

[-]

48

Tanja Mortelmans, Kasper Boye and Johan van der Auwera paradigmatic variability [+]

[-]

Non-epistemic modals occur in finite and non-finite forms (participles, infinitives). Non-epistemic modals allow passive forms.

– In general, Danish modals have less forms at their disposal than the modals in both German and Dutch (which hangs together with the strongly reduced inflectional paradigm in Danish) – Epistemic modals prefer finite forms of the modals (with some exceptions). – Epistemic modals do not allow passive forms.

structural scope [+]

[-]

Structural scope is wide with nonepistemic modals. – Non-epistemic modals combine with noun-phrases and directional adverbs, and ville even with finite complement clauses. – The co-occurrence with directional phrases seems to be somewhat restricted.

Epistemic modals always take infinitive complements, the structural scope is therefore narrow.

5. The Icelandic modals Icelandic has seven modal preterite-presents: vilja ‘will’, skulu ‘shall’, ‘will’, mega ‘may’, ‘must’, munu ‘will’, eiga ‘ought (to)’, kunna ‘can’, ‘may’ and þurfa ‘need’ (Einarsson 1945: 164-167, Thráinsson and Vikner 1995, Kress 1982: 245). In contrast to three non-modal preterite-presents, vita ‘know’, unna ‘love’ and muna ‘remember’, the modal preteritepresents all express meanings within the necessity-possibility range, most, if not all, of them have both epistemic and non-epistemic uses (Thráinsson and Vikner 1995), and they lack an imperative (Einarsson 1945: 103). Interestingly, only four preterite-present modals combine with a bare infinitive: vilja, skulu, munu and mega. Among these, mega is unique in

Modals in the Germanic languages

49

being capable of expressing both possibility (see 62c) and necessity (in 62d). In this sense, mega is comparable to Danish måtte (see section 4). (62) a. Hann vill lera málvísindi. 26 he will.PRS learn.INF linguistics ‘He wants to study linguistics.’ b. Þú skalt fara. you shall.PRS go.INF ‘You have to go.’ c. Hún má alveg taka she may.PRS well take.INF ‘It is fine that she takes my car.’

minn bíl. my car

d. Han mátti lesa upp og læra betur. he must.PST read.INF up and learn better ‘He had to read everything once more and learn it better.’ e. Þetta mun vera rétt. this shall/will.PRS be.INF right ‘This will be right’/‘This is probably right.’ The three remaining modal preterite-presents, eiga ‘ought (to)’, kunna ‘can’, ‘may’, and þurfa ‘need’, all combine with an að-marked infinitive; they can express both non-epistemic and epistemic meanings. (63) a. Það ætti að it ought.PST.SBJV to ‘It ought to be enough.’

vera be.INF

b. Han kan ekki að synda. he can.PRS not to swim.INF ‘He cannot swim.’ c. Við þurfum ekki að fara. we need.1.PL. not to go.INF ‘We need not go.’

nóg. enough

50

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

As skulu, vilja, mega and munu are the only preterite-present modals that combine with a bare infinitive, they may be considered the central modals of Icelandic. It should be noted that the morphological status of vilja is somewhat unclear in present-day Icelandic. Birkmann (1987: 377) states that vilja is a weak verb with irregularities in the 2nd and 3rd person singular of the present indicative and therefore excludes it from his list of preteritepresent modals (which thus contains six instead of seven members). However, vilja is undoubtedly one of the more frequent modals which – in combination with the fact that it combines with a bare infinitive – warrants its inclusion in this list of central modals. As in Dutch, German and Danish, some of the Icelandic modals have additional non-modal meanings. Like its Danish cognate kunne, for instance, kunna has the meaning ‘know’, ‘know how to’ besides its modal meaning of ‘possibility’. This meaning is expressed when the verb takes nominal complements as in (64a-b) – note that in (64b) ensku is a feminine accusative noun, rather than an adjective: as an adjective English translates into the adjective enskur. Similar to its German, Dutch and Danish cognates, vilja can also express the non-modal meaning ‘want’ (in 64c (provided to us by Johanna Barðdal)). (64) a. Hann kann fyrsta boðorðið. he can.PRS first commandment ‘He knows the first commandment.’ b. Hann kunni ensku. he can.PST English ‘He knew English.’ c. Hann vill þessa bók. he will.PRS this book ‘He wants this book.’ As for the central modals mega, munu and skulu, they all have clearcut epistemic uses. Notably, the modal mega (unlike its Danish cognate måtte; cf. section 4) can be used epistemically both with its possibility (65a) and its necessity (65b) meaning, and munu seems to be capable of expressing both epistemic necessity (65c) and a distinct evidential meaning of report (65d).

Modals in the Germanic languages

51

(65) a. Það má vel vera. . that may.PRS well be.INF ‘That is quite possible.’ b. Það má þá rigna mikið. it must.PRS then rain.INF much ‘Then it must be raining a lot.’ c. Þetta mun vera rétt. this shall/will.PRS be.INF right ‘This is probably right.’ d. Það mun hafa verið mús í there shall/will.PRS have.INF be.PTCP mouse in baðkerinu. bathtub.DEF ‘There is said to have been a mouse in the bathtub.’ e. Það skal hafa verið mús í baðkerinu. there shall.PRS have.INF be.PTCP mouse in bathtub.DEF ‘I am sure there has been a mouse in the bathtub.’ As for the fourth central modal, vilja, Thráinsson and Vikner (1995) seem to think that it has an epistemic use as well, and they give (66) as an example. (66) Mig vill sennilega vanta peninga. me.ACC will.PRS probably need.INF money ‘I will probably need money.’ However, it is not obvious to us that vilja in (66) is indeed epistemic. In fact, (66) is judged ‘odd’ by Ellert Thor Jóhannson (p.c.), and in a sentence like the following (provided by Ellert Thor Jóhannson) where vilja cannot be read with future meaning, only a non-epistemic reading is possible: (67) Hann vill vera á skrifstofunni sinni núna. he will be.INF in office his now ‘He wants to be in his office now.’ (* ‘It must be the case that he is in his office now.’)

52

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

Epistemic Icelandic modals differ from non-epistemic modals in some interesting respects. Epistemic modals allow of non-argument subjects (see examples 68a-b), and they cannot be pseudo-clefted. According to Thráinsson and Vikner, the former explains why only epistemic modals can occur with quirky (accusative or dative) subjects, as in (68a) where vanta requires an accusative subject (Harald rather than nominative Haraldur) and in (68b) where lika requires a dative subject (Haraldi) (examples are taken from Thráinsson and Vikner 1995: 60). (68) a. Harald vill oft vanta peninga. Harald.ACC will.PRS frequently lack money ‘Harald frequently tends to lack money.’ b. Haraldi ætlar að líka vel í Stuttgart. Harald.DAT intend.PRS to like well in Stuttgart ‘It looks like Harald will like it in Stuttgart.’ Note that the non-epistemic readings ‘Harald frequently wants...’ (for 68a) and ‘Harald intends…’ (for 68b) are excluded. 5.1. Paradigmaticity As mentioned above, there are seven modal preterite-present modals in Icelandic: kunna, þurfa, munu, skulu, mega, eiga, and vilja. Within this group, a morphosyntactic paradigm of four modals, munu, skulu, mega, and vilja, can be delimited by the fact that these verbs occur with a bare infinitive rather than with a marked infinitive. In fact, one might subsequently distinguish a core paradigm consisting of munu and skulu only. In contrast to the rest of the modal preterite-presents these two modals lack preterite indicatives 27 and present as well as past participles. Moreover, in contrast to the other modal preterite-presents, as well as in contrast to all other Icelandic verbs, they have infinitives ending in -u, which are in fact preterite infinitives (a preterite form of the infinitive vilja (vilju) is rarely found as well) (Einarsson 1945: 95, 103, 160, Hammerich 1960: 53). Apart from these properties, the Icelandic central modals are similar to other regular verbs: They are conjugated for number and person, as well as for tense (present vs. past) and mood (indicative vs. subjunctive), and they

Modals in the Germanic languages

53

have infinitives and (apart from skulu and munu) present and past participles. (69) Það hefði mátt rigna mjög mikið til þess að it had.CONJ may.PTCP rain.INF very much to it to stíflan brysti. dam break ‘It would have to rain a lot for the dam to break.’ On the whole, therefore, the Icelandic modal verbs still show a fairly complete conjugation paradigm. They conform to Thráinsson and Vikner’s generalisation over Scandinavian modal verbs: “Modal verbs show subject-verb agreement in those Scandinavian languages that have subject-verb agreement in general, i.e. in Faroese [...] and Icelandic, and not in those where the finite verb never shows any kind of agreement with the subject (nor with anything else, i.e. Danish, Norwegian, and Swedish).” (Thráinsson and Vikner 1995: 53)

It should also be noted that a number of non-preterite-presents complement the category of modal verbs in Icelandic: geta ‘can’, verða ‘must’, hljóta ‘must’ and ætla ‘intend’. 28 Whereas geta is combined with a past participle, the other ones take an að-marked infinitive. Interestingly, these verbs express both non-epistemic (a-c) and epistemic meanings (b-d), as the following examples with verða (70a-b) and geta (70c-d) show. (70) a. Hann verður he must.PRS ‘He must go.’

að fara. to go.INF

b. Það varð að vera. that must.PST to be.INF ‘That must have been so.’/‘That must have been the reason.’ c. Hann getur komið. he can.PRS come.PTCP ‘He is able to come.’

54

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

d. Það getur it can.PRS ‘It may be.’

verið. be.PTCP

We can therefore conclude that the process whereby the original preteritepresents develop into a modal paradigm which is clearly distinct from all other verbs has not taken place to a high extent in Icelandic. The ‘survival’ of the non-modal preterite-presents supports this conclusion as well as the fact that the modal preterite-presents behave in a less systematic way as far as their complementation patterns are concerned (allowing for both bare and að-marked infinitives in epistemic and non-epistemic uses). Moreover, other verbs – with different morphological and syntactic characteristics – have entered into the modal verb category of Icelandic. 29 We will return to this issue in section 5.4. 5.2. Paradigmatic variability Icelandic modal verbs still show a fairly complete conjugation paradigm (cf. above). If we disregard the lack of modal imperatives, which probably has semantic reasons, the Icelandic modals do not in general show a decrease of paradigmatic variability in the sense that there are less verb forms at disposal – the only exceptions here are munu and skulu. If we concentrate on the verbs expressing epistemic modality, however, we observe again that epistemic forms are typically finite. As a rule, epistemic modal infinitives cannot be combined with non-epistemic modal finites (example from Thráinsson and Vikner 1995: 78). The following example, in which the first occurrence of að kunna has an intended epistemic reading, is therefore unacceptable. (71) *Hann he

verður must.PRS

að to

kunna can.INF

að kunna synda. to can.INF swim.INF

Nevertheless, Thráinsson and Vikner argue that in both Danish and Icelandic an epistemic modal finite can combine with an epistemic modal infinitive; they provide the following examples for Danish (72a-b) and Icelandic (73a-b) (Thráinsson and Vikner 1995: 76-77):

Modals in the Germanic languages

55

(72) a. Det må kunne stå på en side. it must.PRS can.INF stand on one page ‘It must be possible to fit it onto one page.’ b. Der vil let kunne gå noget galt. there will.PRS easily can.INF go.INF something wrong ‘It will easily be possible that something goes wrong.’ (73) a. Það mun vilja rigni meðan þið eruð þar. it will.PRS want.INF rain.INF while you are there ‘It will tend to rain while you are there.’ b. Strákana ætlað i að vilja reka á land. boys.DEF intended to want.INF drift to land ‘It looked like the boys tended to drift ashore.’ As for Danish, however, Brandt (1999: 130-132) argues convincingly that both (72a) and (72b) involve non-epistemic modal infinitives (of the situational kind). Double epistemic modals are not possible in Danish. As for Icelandic, the examples seem dubious as well. For both (73a) and (73b), it can be claimed that vilja has a future rather than an epistemic reading. While double epistemic modals may be excluded in both Danish and Icelandic, however, epistemic modal infinitives are found (although only rarely) in both languages (cf. section 4.2.). The requirement is that the finite verb is epistemic (an epistemic raising verb, in fact) but non-modal. Thráinsson and Vikner provide Icelandic examples of this, but they are not quite convincing. (74a) and (74b) (provided by Ellert Johansson, p.c.) are better examples, and (74c) (repeated from section 4.2) is a good Danish example. (74) a. Mér virðist kunna að hafa rignt. me.DAT seem.PRS can.INF to have.INF rain.PTCP ‘It seems to me that it may have been raining.’ b. Hann hlýtur að verða að vera á Islandi. he seem.PRS to must.INF to be.INF on Iceland ‘It seems that he must be in Iceland.’

56

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

c. Han synes at måtte være he seem.PRS to must.INF be.INF ‘It seems that he must be the murderer.’

morderen. murderer.DEF

Both finite and non-finite modals are used epistemically and nonepistemically, then. 5.3. Structural scope Non-epistemic Icelandic modals are found combining with nouns and directional adverbials, rather than with infinitives. (75) a. Hann kann fyrsta boðorðið. he can.PRS first commandment ‘He knows the first commandment.’ b. Ég vill heim. I want.PRS home ‘I want to go home.’ Just like its cognates in Danish, Dutch and German, the modal vilja even combines with a finite complement clause. (76) Hann vill að þu komir. he will.PRS COMP you come.PRS ‘He wants that you come.’ With epistemic modals, however, all these constructions seem to be impossible. Epistemic modals require an infinitive (which may however be marked by að) or a past participle (in the case of the non-preterite-present geta).

5.4. The Icelandic modals: summary It could be argued that there is a highly grammaticalised category of four core modals (munu, skulu, mega and vilja) in Icelandic that share all the relevant characteristics of the core modals that we have used as delimiting

Modals in the Germanic languages

57

criteria so far: they are preterite-present polyfunctional modals taking a bare infinitive. One could even, as was argued in section 5.1., go one step further and claim that only two modals (munu and skulu) are at the heart of this category, on account of the fact that their formal inventory is heavily reduced. Following this line of thinking, the degree of grammaticalisation of the Icelandic core modal category would be rather high as it consists of only two members. However, the category of Icelandic core modals (whether it consists of either two or four members) seems to distinguish itself crucially from the core modal categories in the other languages to the extent that the differences with respect to the non-core modals are of a more gradual nature. Table 12. Icelandic modals: paradigmaticity paradigmaticity [+] Small ‘paradigm’ consisting of either two (munu, skulu) or four core modals (munu, skulu, mega, vilja).

[-] - Core modals share features with other modal and main verbs: (1) preterite-presents modals with bare infinitive: munu, skulu, mega, vilja (2) preterite-presents modals with að-infinitive: eiga, kunna, þurfa (3) non-preterite-presents modals with að-infinitive: verða, hljóta, ætla (4) non-preterite-present modal with past participle: geta (5) other modals: taka, fara, … (6) preterite-present main verbs: vita, unna, muna

Thus, we do not only find four modal polyfunctional preterite-presents taking a bare infinitive, but also three modal polyfunctional preteritepresents taking an að-marked infinitive 30 (eiga ‘ought (to)’, kunna ‘can’, ‘may’, and þurfa ‘need’), and a number of modal polyfunctional auxiliaries that are non-preterite-presents (e.g. geta ‘can’, verða ‘must’, hljóta ‘must’), most of which are combined with an að-marked infinitive (geta needs a past participle). Moreover, the Icelandic core modals share the preteritepresent feature with three main verbs vita ‘know’, unna ‘love’ and muna ‘remember’. Thus, we would suggest that the delimiting criteria in Icelandic do not have the same degree of saliency as in the other Germanic

58

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

languages, as the criteria are shared by a considerable number of other auxiliary (and even main) verbs as well. So, although a clearly delimitable group of highly grammaticalised core modals can be isolated, it remains doubtful whether this ‘paradigm’ is as highly paradigmatised as in the other Germanic languages, as it shares its main (formal as well as semantic) features with a considerable number of other verbs. As far as paradigmatic variability is concerned, we observe the same tendencies as in the other Germanic languages: with epistemic readings, finite forms of the modals are clearly preferred, although epistemic uses with infinitives (in the scope of finite epistemic raising verbs) can be found. It could be argued that paradigmatic variability is thus a bit lower than with the non-epistemic modals, as they can freely occur in all kinds of nonfinite constructions. Finally, epistemic modals have a much more restricted structural scope than non-epistemic ones, since they only combine with verbal complements (which, if one also takes the non core modals into account) are considerable more varied than in the other Germanic languages). As was the case for the other Germanic languages, this may be taken to suggest that epistemic modals are more grammaticalised than non-epistemic ones. Table 13. Icelandic modals: paradigmatic variability and structural scope paradigmatic variability [+]

[-]

Non-epistemic modals occur in finite and nonfinite forms (participles, infinitives).

Epistemic modals prefer finite forms of the modals, but do not exclude nonfinite forms.

structural scope Non-epistemic modals have a wide structural scope.

Epistemic modals only take verbal complements.

Modals in the Germanic languages

59

6. General conclusions In so far as grammaticalisation is conceived of in accordance with Lehmann’s parameters, we can set up a grammaticalisation cline, starting from the highly grammaticalised English modals and ending up with the Icelandic ones (due to their lower degree of paradigmaticity). The Danish modals with their rather restricted inflectional paradigm (mainly two forms, present-finite and non-present non-finite, in an extreme formulation) share many similarities with the English modals, whereas the German and Dutch modals hover somewhere in between. For the latter, the structural scope is remarkably wide, as the Dutch modals allow for a very broad range of complements and often occur in what we have labelled ‘intransitive’ use, i.e. without a complement (which however does not necessarily give rise to an elliptic analysis). So, although the grammar of Dutch can often be seen as intermediate between those of German and English (Van Haeringen 1956, Hüning et al. 2006), this does not seem to be the case for the grammar of the modals. The distinction between epistemic and non-epistemic modality is relevant for all four languages studied here: the epistemic modals are clearly more grammaticalised than the non-epistemic ones, although the often cited reluctance against nonfinite modals in epistemic use (“the epistemic non-finiteness gap”, see e.g. Abraham (2001)) had to be relativised for every language. In fact, we would suggest that this reluctance against non-finites in epistemic use is more linked to semantic reasons than to purely grammatical ones. As far as the question regarding the usefulness of applying Lehmann’s parameters to this particular group of verbs is concerned, we have to provide a two-sided answer. First of all, the parameters can be said to function as eye-openers and classification tools at the same time. For this clearly delimitable group of verbs, the parameters function as searchlights, leading one to observations that might otherwise pass unnoticed. Moreover, seemingly arbitrary facts concerning the form and behaviour of the core modals can be integrated in a larger framework and understood accordingly. This immediately brings us to the second point we would like to raise: these parameters should not be applied in isolation, i.e. without taking the entire structure of the verbal system of a particular language in account. In languages with a strongly developed inflectional system, the modal verbs can be expected to reflect this, i.e. they can be expected to have more inflectional categories at their disposal than modals in languages

60

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

where these categories do not exist (anymore). Whether this automatically leads to a lesser degree of grammaticalisation, however, is doubtful and can only be decided after carefully comparing the item with ordinary verbs in its own language with respect to the other grammaticalisation parameters. By the same token, the fact that a particular language doesn’t generally mark subject-verb agreement (e.g. Danish) will have its bearing on the formal make-up of the Danish class of core modals as well: it can be expected that Danish modals won’t be marked for subject-verb agreement either, and this could be interpreted as an indication of a higher degree of paradigmaticy (as the number of formal options within the paradigm decreases). From a cross-linguistic perspective, however, it would be wrong to claim on the basis of this observation alone that Danish modals score higher with respect to paradigmaticity than their German or Dutch counterparts, which do mark subject-verb agreement (and hence have more forms at their disposal). The former also implies that Lehmann’s conception of grammaticalisation does not in itself constitute an ideal means for measuring the degree of grammaticalisation: what is a sign of grammaticalisation according to Lehmann’s conception may in fact occasionally be due to factors unrelated to grammaticalisation as such. General semantic factors, for instance, or the overall structure of a particular language can highly influence the linguistic appearance of particular meanings and meaning categories without however being immediately tied to grammaticalisation per se. Notes 1. Thanks are due to the University of Antwerp and to the Belgian Federal Government (IAP grant P6/44) for financial help to Tanja Mortelmans and Johan van der Auwera (‘GOA’ grant 2003-2006). Kasper Boye wishes to thank The Carlsberg Foundation for a grant which made his work on the chapter possible. Finally, we thank Jóhanna Barðdal and Ellert Thor Jóhannsson for expert help on Icelandic, Karen Margrethe Pedersen for expert help on Danish dialects and Björn Hansen and Ferdinand de Haan for their editorial patience. 2. Many languages use different markers to express non-epistemic and epistemic possibility and necessity (van der Auwera and Ammann 2005: 310-314). 3. For more information on the modals in Norwegian, see Eide (2005); for Swedish, see Telemann et al. (1999); for Faroese, see Thráinsson et al. (2004); for Afrikaans, see Donaldson (1993), for Pennsylvania German, see Burridge

Modals in the Germanic languages

4.

5.

6.

7.

8.

9.

10.

61

(1997, 1998), for Yiddish, see Hansen/Birzer (in prep), for Low German, see Mortelmans (2007). The realization of will as ‘ll counts as an instance of attrition (decrease of formal integrity). It should be noted that the Danish modals ville and skulle do show attrition, as their pronunciation involves deletion of the final consonant (and schwa). The bondedness of a sign is defined as “the intimacy with which it is connected with another sign to which it bears a syntagmatic relation” (Lehmann 2002: 131). The English modals fuse with negation markers (can’t, mustn’t) and pronouns (I’ll), which obviously increases their degree of bondedness. In general, this parameter does not seem to be very relevant for the Germanic modals, as they all function as syntagmatically independent items, at least in written language (but see note 18). The Icelandic modals kunna, eiga and þurfa are added between brackets, as they combine with an að-marked infinitive only (which incidentally also holds for ought in the English group, which doesn’t combine with a bare infinitive either). Vilja is added between brackets as well, as it does not conform to the characteristics of a preterite-present. Following Birkmann (1987: 377), modal vilja is a weak verb with some formal irregularities in the 2nd and 3rd person of the present indicative (syntactically, vilja behaves as a central modal, combining with a bare infinitive though). The modal will and its cognates in the other Germanic languages are not real preterite-presents, as they go back to original optative forms with present meaning (Birkmann 1987: 116). In many Germanic languages, the willcognates have been formally integrated in the preterite-present paradigm, (with the exception of Icelandic). The Danish modals turde and burde are both considered non-central modals, the former because of the fact that it can also take an at-marked infinitive (see section 2.3.1), the latter because of its rather low frequency. Note also that in spite of the fact that they formally behave like preterite-presents in present-day Danish, their ancestors did not belong to this category (Birkmann 1987: 381). This does not imply that the central modals are thought to be completely regular or even uninteresting as far as word order is concerned. German and Dutch, for instance, differ with respect to the position of the modal in constructions with double infinitives (see also section 2.1). Compare Dutch hij heeft het niet kunnen zeggen (VAUX INFMOD INF) with Er hat es nicht sagen können (VAUX INF INFMOD) ‘He hasn’t been able to say it’. Moreover, there are interesting word order differences in subclauses. For the following example with epistemic moeten, Dutch allows three possibilities dat hij gezegd moet hebben/moet gezegd hebben/moet hebben gezegd), whereas German only allows the word order [PP INF VMOD] dass er gesagt haben soll, with the modal verb in final position. For the German modals, this is claimed by Diewald (1997: 19-27).

62

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

11. Note that it is not always that easy to distinguish between non-modal and modal (non-epistemic) meanings. It is quite unclear for instance whether German können meaning ‘know’ should be considered to express a non-modal (lexical) meaning or a modal (dynamic) one in the sentence Sie kann Deutsch (sprechen). 12. The use of mogen for edible entities is restricted to Southern Dutch (Diepeveen et al. 2006: 98). More frequently, the lexical verb lusten is used in these contexts. Conversely, the use of mogen with persons (expressing feelings of sympathy) is restricted to Northern Dutch (Diepeveen et al. 2006). 13. Although this is generally frowned upon and characterized as Southern Dutch, the modal kunnen pops up both in Southern and Northern Dutch corpora with a comparable frequency (Diepeveen et al. 2006: 39-40). 14. The forms without -t (which are the younger ones) are considered to be informal in Northern Dutch, which generally prefers the forms with -t. In Southern Dutch, however, no such distinction is made; here the forms without -t are typical of spoken language (Haeseryn et al. 1997: 94-100). 15. Whereas wou is mainly used in spoken language, the form wilde is more typical of written language (Renkema 1995: 103); the plural form ‘wou(d)en’ is considered as highly informal spoken language (Smets and Van Belle 1997: 138). 16. IPP-constructions are mainly characteristic of the West Germanic languages (but see section 2.3.1 for an interesting exception in Danish); they can be found in German, Afrikaans, Swiss German and Dutch (Schmid 2005:12). 17. The condition that the modal has to denote a value on a bounded scale when it combines with a nominal or adjective phrase does not hold for cases in which the modal has a lexical reading (see ex. (7a-b-c) above). 18. In fact, the possibility of such ‘intransitive’ uses can also be interpreted as an instance of less syntagmatic cohesion (see Cornillie et al., this volume). 19. The preterite present wissen governs an infinitive with zu, in which case it expresses an ability-like meaning. Epistemic meanings of wissen do not exist, though. 20. Following Öhlschläger (1989: 64), Diewald (1999) considers this use as ‘elliptic’. 21. The modal måtte (which is related to English may and Dutch mogen) may be regarded as multifunctional, as it expresses both possibility and necessity. Note, however, that epistemic måtte can only express (epistemic) necessity. 22. Counterparts of burde – which is absent in the West-Germanic languages – are also found among the Swedish (böra) and the Norwegian (burde) modals. 23. The modal turde has its origin in a fusion between two distinct verbs: one which is cognate with German dürfen and which meant ‘need, must’, and another which meant ‘dare’. In table 2, turde is represented as a cognate of dürfen.

Modals in the Germanic languages

63

24. Synthetic passives are “least uncommon” for ville (villes) and kunne (kunnes) (Brandt 1999: 69). Also skulles and måttes are marginally acceptable. 25. As for the formal identity between past participles and infinitives (and pasts) of preterite-present modals, it might be claimed that it is phonetically (and orthographically) conditioned only, residing in the omission of final -t from the conservative participles (e.g. kunnet) to yield forms identical to the infinitives (e.g. kunne). Such omission is occasionally found also with regular main verbs. However, the convergence of past participles and infinitives (and pasts) of preterite-present modals is found in several Danish dialects (e.g. Seelandic) where such a simple phonetic explanation is inadequate. In the case in point, the difference between the conservative participles and the infinitives (and pasts) is marked by presence vs. absence of ‘stød’ (the special Danish glottal stop) in addition to presence vs. absence of final -t. This suggests that the convergence resides in the disappearance of the past participle as a unique form, rather than in a phonetic omission (Karen Margrethe Pedersen, p.c.). 26. Unless it is noted, examples come from either Thráinsson and Vikner 1995: 55-57, or from Einarsson 1945: 161 and 164-167. 27. For munu there is the form mundi, which is however not used with past-time meaning. It is accordingly classified in Einarsson (1945: 103) as a preterite subjunctive side by side with myndi. 28. Birkmann (1987: 377) also mentions taka and fara. The verb geta has counterparts in English and Danish, too; for a study of these ‘acquisitive modals’, modals deriving from a lexical meaning ‘to acquire’, see van der Auwera, Kehayov and Vittrant (in print). 29. On the basis of a small-scale investigation (a comparison of a small parallel corpus), Hrafnbjargarson (2007) shows that the other Scandinavian languages (Danish, Norwegian, Swedish and Faroese) make use of a limited and much more uniform inventory of modals than Icelandic. So, Norwegian kunne translates as kunne or fá in Faroese, kunne or ville in Danish or kunna and få in Swedish. The Icelandic translation contains no less than four alternatives (geta, kunna, mega, skulu). Similarly, whereas Norwegian skulle is rendered by skula or kunna in Faroese, ville or skulle in Danish and skola in Swedish, Icelandic translates the occurrences by means of skulu, munu, eiga and Þurfa. The same holds for the translation of Norwegian måtte and ville: Icelandic has four alternatives for the former (hljóta, verða, mega and þurfa) and three for the latter (munu, vilja, ætla) whereas none of the other languages lists more than two alternatives for either verb. 30. Note that at-marked infinitives are found with some Danish modals as well, but only on a non-epistemic reading. In Icelandic, however, these complements occur with both non-epistemic and epistemic uses of the modals.

64

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

References Abraham, Werner 2001 Modals: toward explaining the ‚epistemic non-finiteness gap’. In Modalität und Modalverben im Deutschen, Reimar Müller and Marga Reis (eds.), 7-36. (Linguistische Berichte Sonderheft 9) Hamburg: Helmut Buske Verlag. Abraham, Werner 2002 Modal verbs: Epistemics in German and English. In Modality and its Interaction with the Verbal System, Sjef Barbiers, Frits Beukema and Wim van der Wurff (eds.), 19-50. (Linguistik Aktuell/Linguistics Today 47) Amsterdam/Philadelphia: John Benjamins. Askedal, John Ole 1998 brauchen mit Infinitiv. Aspekte der Auxiliarisierung. In Jahrbuch der ungarischen Germanistik 1997, Antal Mádl and Gunther Dietz (eds.), 53-68. Budapest: Gesellschaft Ungarischer Germanisten; Bonn: Deutscher Akademischer Austauschdienst. Barbiers, Sjef 2002 Modality and Polarity. In Modality and its Interaction with the Verbal System, Sjef Barbiers, Frits Beukema and Wim van der Wurff (eds.), 51-73. (Linguistik Aktuell/Linguistics Today 47) Amsterdam/Philadelphia: John Benjamins. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan 1999 The Longman grammar of spoken and written English. London: Longman. Birkmann, Thomas 1987 Präteritopräsentia. Morphologische Entwicklungen einer Sonderklasse in den altgermanischen Sprachen. (Linguistische Arbeiten 188) Tübingen: Niemeyer. Boye, Kasper 2001 The force-dynamic core meaning of Danish modal verbs. Acta Linguistica Hafniensia 33: 19-66. Boye, Kasper 2005 Modality and the concept of force-dynamic potential. In Modality. Studies in form and function, Alex Klinge and Henrik Høeg Müller (eds.), 49-80. London: Equinox. Brandt, Søren 1999 Modal Verbs in Danish. (Travaux du Cercle Linguistique de Copenhague XXX) Copenhagen: The Linguistic Circle of Copenhagen. Burridge, Kate 1997 On the trail of the Conestoga modal: recent movements of modal auxiliaries in Pennsylvania German. In Languages and lives. Essays

Modals in the Germanic languages

65

in honor of Werner Enninger, James R. Dow and Michele Wolff (eds.), 7-28. New York: Peter Lang. Burridge, Kate 1998 From Modal Auxiliary to Lexical Verb: The Curious Case of Pennsylvania German Wotte. In Historical Linguistics 1995. Vol II: Germanic Linguistics, Richard M. Hogg and Linda van Bergen (eds.), 19-33. (Current Issues in Linguistic Theory 162) Amsterdam: Benjamins. Bybee, Joan 2003 Mechanisms of Change in Grammaticization: The Role of Frequency. In The Handbook of Historical Linguistics, Brian D. Joseph and Richard D. Janda (eds.), 602-623. Malden MA, Oxford: Blackwell. Coates, Jennifer 1983 The semantics of the modal auxiliaries. London: Croom Helm. Davidsen-Nielsen, Niels 1990 Tense and Mood in English. A Comparison with Danish. Berlin: Mouton de Gruyter. de Haan, Ferdinand 2001 The relation between modality and evidentiality. In Modalität und Modalverben im Deutschen, Reimar Müller and Marga Reis (eds.), 201-216. (Linguistische Berichte Sonderheft 9) Hamburg: Helmut Buske Verlag. de Haan, Ferdinand 2006 Typological approaches to modality. In The expression of modality, William Frawley (ed.), 27-69. Berlin/New York: Mouton de Gruyter. Diepeveen, Janneke, Ronny Boogaart, Jenneke Brantjes, Pieter Byloo, Theo Janssen and Jan Nuyts 2006 Modale uitdrukkingen in Belgisch-Nederlands en NederlandsNederlands: corpusonderzoek en enquête. [Modal expressions in Belgian-Dutch and Dutch-Dutch] Amsterdam, Münster: Stichting Neerlandistiek VU Amsterdam, Nodus Publikationen. Diewald, Gabriele 1997 Grammatikalisierung. Eine Einführung in Sein und Werden grammatischer Formen. Tübingen: Niemeyer. Diewald, Gabriele 1999 Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalität. (Germanistische Linguistik 208) Tübingen: Niemeyer. Donaldson, Bruce 1993 A grammar of Afrikaans. (Mouton Grammar Library 8) Berlin: de Gruyter.

66

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

Duden 2005

Die Grammatik. 7., völlig neu erarbeitete und erweiterte Auflage. Herausgegeben von der Dudenredaktion. Mannheim, Leipzig, Wien, Zürich: Dudenverlag. Eide, Kristin Melum 2005 Norwegian modals. (Studies in generative grammar 74) Berlin: Mouton de Gruyter. Einarsson, Stefán 1945 Icelandic – Grammar, Texts, Glossary. Baltimore: The John Hopkins Press. Goossens, Louis 1985 Differentiating the English modals in Functional Grammar. In Antwerp Studies in Functional Grammar, Jan Nuyts (ed.), 49-71. Antwerp: University of Antwerp. Haeseryn, Walter, K. Romijn, G. Geerts, J. de Rooij, and M.C. van den Toorn 1997 Algemene Nederlandse Spraakkunst. [Grammar of Dutch] 2nd edition. Groningen: Nijhoff. Hammerich, L.L. 1960 Über die Modalverben der neugermanischen Sprachen. Zeitschrift für deutsche Wortforschung 16: 47-70. Hansen, Björn 2004 Modals and the boundaries of grammaticalization: The case of Russian, Polish and Serbian-Croatian. In What makes Grammaticalization? A look from its Fringes and its Components, Walter Bisang, Nikolaus P. Himmelmann and Björn Wiemer (eds.), 245-270. Berlin, New York: Mouton de Gruyter. Hansen, Björn and Sandra Birzer In prep. The Yiddish modal system between Germanic and Slavonic. A case study on the borrowability of modals. In Grammaticalization and language contact, Björn Wiemer, Bernhard Wälchli and Björn Hansen (eds.). Berlin, New York: Mouton de Gruyter. Hansen, Erik 1977 Behøver vi at? [Need we to?] Nyt fra Sprognævnet 19: 1-4. Copenhagen: The Danish Language Council. Hofmans, Mark 1980a Hebben of zijn: Een enquete naar het gebruik van ‘hebben’ of ‘zijn’ in de konstruktie Thww + Mhww + HWW in Nederland en Vlaanderen. [Hebben or zijn: a survey of the use of ‘hebben’ or ‘zijn’ in the construction Taux + Maux + MainV in The Netherlands and Flanders] Rapport d’activités de l’institut de phonetique 14: 83-129. Hofmans, Mark 1980b Hebben of zijn en de deverbalisering van de modale werkwoorden in het Nederlands. [Hebben or zijn and the deverbalising of modal

Modals in the Germanic languages

67

verbs in Dutch] In Linguistics in Belgium/Linguïstiek in België/Linguistique en Belgique 5, Marc Dominicy (ed.), 81-109. Brussel: Didier Hatier. Hrafnbjargarson, Gunnar Hrafn 2007 Generating a lexicon of Scandinavian modals. Nordlyd 34: 183-199. Hüning, Matthias, Ulrike Vogl, Ton van der Wouden and Arie Verhagen (eds.) 2006 Nederlands tussen Duits en Engels. Handelingen van de workshop op 30-9 en 1-10-2005 aan de Freie Universität Berlin. [Dutch between German and English. Proceedings of the workshop on 30-9 en 1-10-2005, Freie Universität Berlin] Leiden: SNL. Jensen, Torben J. 2005 Irrealitetsmarkører i dansk talesprog. En korpuslingvistisk undersøgelse af hvis-konstruktioner og modalverber i medarbejdersamtaler og sociolingvistiske interview [Irreality markers in spoken Danish. A corpus-linguistic study of hvis-constructions and modal verbs in appraisal and sociolinguistic interviews] Ph.D. Thesis. Copenhagen: University of Copenhagen. Kotin, Michail L. 2003 Die werden-Perspektive und die werden-Periphrasen im Deutschen. (Danziger Beiträge zur Germanistik 6) Frankfurt/M.: Peter Lang. Kress, Bruno 1982 Isländische Grammatik. Munich: Max Hueber Verlag. Krug, Manfred 2000 Emerging English Modals. A Corpus-based Study of Grammaticalization. Berlin, New York: Mouton de Gruyter. Lauridsen, Karen M. and Ole Lauridsen 1989 Modalverber med passiv infinitiv i moderne dansk. [Modal verbs with infinitive passive in Modern Danish] 2. Møde om Udforskningen af Dansk Sprog, 239-249. Lehmann, Christian 1995 Synsemantika. In Syntax. Ein internationales Handbuch zeitgenössischer Forschung. 2. Halbbd, Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), 12511266. Berlin, New York: de Gruyter Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Mortelmans, Tanja 2004 Grammatikalisierung und Subjektivierung: Traugott und Langacker revisited. Zeitschrift für germanistische Linguistik: deutsche Sprache in Gegenwart und Geschichte 32: 188-209.

68

Tanja Mortelmans, Kasper Boye and Johan van der Auwera

Mortelmans, Tanja 2007 Modalverben im Niederdeutschen: Ansatz zu einem Vergleich mit dem Modalverbbestand im Deutschen und im Niederländischen. Niederdeutsches Wort: Beiträge zur niederdeutschen Philologie, 47/48: 135-148. Nuyts, Jan 2001 Epistemic modality, Language and Conceptualization. Amsterdam: John Benjamins. Nuyts, Jan 2006 Modality: Overview and linguistic issues. In The expression of modality, William Frawley (ed.), 1-26. Berlin/New York: Mouton de Gruyter. Nuyts, Jan, Pieter Byloo and Janneke Diepeveen 2005 On deontic modality, directivity, and mood. A case study of Dutch mogen and moeten. Antwerp Papers In Linguistics 110. Antwerp: University of Antwerp. Öhlschläger, Günther 1989 Zur Syntax und Semantik der Modalverben des Deutschen. (Linguistische Arbeiten 144) Tübingen: Niemeyer Palmer, Frank R. 2001 Mood and modality. 2nd edition. Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986 Plank, Frans 1984 The modals story retold. Studies in language 8.3: 305-364. Ponten, Peter Jan 1973 Der Ersatz- oder Scheininfinitiv. Ein Problem aus der deutschen und niederländischen Syntax. Wirkendes Wort 23.2: 73-85. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1984 A comprehensive grammar of the English language. London: Longman. Reis, Marga 2001 Bilden Modalverben im Deutschen eine syntaktische Klasse? In Modalität und Modalverben im Deutschen, Reimar Müller and Marga Reis (eds.), 287-318. (Linguistische Berichte Sonderheft 9) Hamburg: Helmut Buske Verlag. Renkema, Jan 1995 Schrijfwijzer (3rd ed.). [Stylebook] Den Haag: Sdu. Schmid, Tanja 2005 Infinitival Syntax. Infinitivus Pro Participio as a repair strategy. (Linguistik Aktuell/Linguistics Today 79) Amsterdam: John Benjamins.

Modals in the Germanic languages

69

Skyum-Nielsen, Peder 1971 Modalverberne i Nudansk. [Modal verbs in Modern Danish] Unpublished prize essay. Smedts, Willy and William Van Belle 1997 Taalboek Nederlands. [Language book Dutch] Kapellen: Pelckmans. Telemann, Ulf, Staffan Hellberg and Erik Andersson 1999 Svenska Akademiens Grammatik 1-4. [Swedish Academy Grammar] Stockholm: Norstedts Ordbok. Thráinsson, Höskuldur and Sten Vikner 1995 Modals and Double Modals in the Scandinavian Languages. Working Papers in Scandinavian Syntax 55:51-88. Dept. of Scandinavian Linguistics, Lund University. Thráinsson, Höskuldur, Hjalmar P. Petersen, Jógvan í Lon Jacobsen and Zakaris Svabo Hansen 2004 Faroese. An Overview and Reference Grammar. Tórshavn: Føroya Fródkaparfelag. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65,1:31-55. van der Auwera, Johan 1999 On the semantic and pragmatic polyfunctionality of modal verbs. In The semantics/pragmatics interface from different points of view, Ken Turner (ed.), 49-64. Oxford: Elsevier. van der Auwera, Johan and Andreas Ammann 2005 Overlap between situational and epistemic modal marking. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gil and Bernard Comrie (eds.), 310-314. Oxford: Oxford University Press. van der Auwera, Johan and Vladimir Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79-124. van der Auwera, Johan, Petar Kehayov and Alice Vittrant In print Acquisitive modals. In Cross-linguistic Studies of Tense, Aspect, and Modality, Lotte Hogeweg, Helen De Hoop and Andrej Malchukov (eds.). Amsterdam: Benjamins. Van Haeringen, C.B. 1956 Nederlands tussen Duits en Engels. [Dutch between German and English] Den Haag: Servire. Van Ostaeyen, Gert and Jan Nuyts 2004 De diachronie van kunnen. [The diachrony of kunnen] Antwerp Papers in Linguistics 109. Antwerp: University of Antwerp.

3. Modals in Irish Peter McQuillan

1. Introduction The aim of this study is to describe how lexical and grammatical resources in Irish cover the two modal sub-domains of possibility and necessity, both of which are further sub-divided into epistemic and non-epistemic modality (van der Auwera and Plungian 1998). Epistemic modality concerns the speaker’s judgment on the truth-value of a proposition in terms of relative certainty. Non-epistemic modality is, on the other hand, ‘situational’ (van der Auwera, Ammann and Kindt 2005: 250) or circumstantial, possibility or necessity being a component of the situation itself rather than reflecting the speaker’s judgment of the situation from without. Irish is a member of the Celtic family of Indo-European. Along with the other Celtic languages its basic word order is Verb-Subject-Object and it shows the typical characteristics of VSO languages (Greenberg 1963). 1 The following is an outline sketch of the Irish verbal system: –



The following finite forms occur: – Past, present and future tenses. As we shall see below, modal auxiliaries are often defective with respect to these categories – Imperative, conditional and subjunctive moods. All modals are defective with respect to the imperative – An aspectual distinction between habitual and non-habitual in the past; only in the substantive verb is this distinction made formally in the present – There are two verbs ‘to be’: a copula linking the subject to a predicate noun or pronoun and a substantive verb connecting the subject to an adjective, prepositional or adverbial phrase. Both of these verbs participate in modal structures as described below The following non-finite forms occur: – Verbal adjective. This combines with the construction ‘to have’ to form a passive perfective (type: ‘I have it done’ / I’ve done it’)

72

Peter McQuillan



Verbal noun. This form supplies the equivalent of both a gerund / participle (as in the progressive construction ‘I’m doing it’) and an infinitive (‘I want to do it’). While the verbal noun, like the English infinitive, carries no markers of tense, mood, voice, or person, it continues to exhibit noun-like morphology — for example, where it is followed by a direct object, as in the progressive construction, the object is in the genitive case if a noun, and a possessive (preceding the verbal noun) if a pronoun. On the other hand, verbal nouns are modified by adverbs in non-finite clauses. In accordance with traditional practice, I will refer to the infinitival use of verbal nouns as verbal noun phrases.

For the purposes of the present discussion, there are two main dialects areas of Irish: Northern (Ulster primarily) and Southern (Munster and South Connacht [Conamara]). As regards periodization, Old Irish refers to texts written before c. 900; Middle Irish to the subsequent period up to c.1200; and Early Modern Irish to the succeeding period up to the seventeenth century (at least for present purposes). 2. The category of modals in Irish: semantics, syntax and morphology Although non-epistemic and epistemic modality are distinct semantic domains, it is a fact that in many languages, particularly, it seems, European ones, modal expressions are polyfunctional in that the same item can cover both domains (van der Auwera, Ammann and Kindt 2005: 255– 8). 2 All items discussed in this section are polyfunctional. 2.1. Semantics: Polyfunctionality The data is given the same basic arrangement as will be found in Ó Siadhail (1989: 289ff): verbs (2.1.1) and impersonal verb phrases (2.1.2). 2.1.1. Verbs Here we deal with originally full lexical verbs which have started to develop as grammatical markers of modality (Heine 1993: 70).

Modals in Irish

73

In the domain of necessity, the most important verb is caith ‘must’, ‘has to’ which derives from the full verb caith ‘use’, ‘consume’, ‘spend’, ‘wear’; ‘throw’. As a modal, caith occurs in all Irish dialects. However, it also continues to function as a full verb in present-day Irish in all dialects. For the area of possibility, the most significant verb is féad ‘may’, ‘can’, ‘be able’ (Scottish faod), which, in Ireland at least, survives only as a modal. It is derived from an Old Irish full verb meaning ‘get’, ‘obtain’, ‘procure’. The verb faigh also means ‘get’ and is still fully functional as a main verb. However, it has also started to develop as a modal where it occurs mostly in the conditional mood and is largely confined to the dialects of SE Munster. Here we may note that the development of faigh replicates that of féad. Thus we can speak here of a grammaticalization path to possibility in Irish (see 4.1 below). In addition, there are some verbs where the semantic subject is expressed by a prepositional phrase. For necessity we have dialectal (Donegal, NW Ulster) féad do ‘could (have)’; ‘should (have)’. This is a further development of the possibility modal féad above. So here we have a development from possibility to (weak) deontic necessity. See 4.2 below for further discussion. For possibility, northern dialects (Ulster; N Connacht) have the form tig le or thig le ‘may’, ‘can’, ‘be able’. This is derived from the full verb tig / thig ‘comes’ with the addition of a prepositional phrase as the semantic subject, here governed by the preposition le ‘with’, ‘to’. At an earlier stage of the language the preposition used was do ‘to’, ‘for’ or de ‘of’, ‘from’ rather than le. In Munster we have the historically related ráinigh ‘reach’, ‘arrive’ (variant past form ráinig) which has the meaning of possibility when used impersonally with the preposition le (see further 4.1.3 below). 2.1.2. Impersonal verb phrases Such expressions have the structure: Verb: Is

Predicate nucleus: X

Complement: Finite (epistemic) Non-finite (non-epistemic)

74

Peter McQuillan

In Irish, X is either a noun or adjective expressing a modal value, while the verb is the copula is ‘is’. A semantic subject can be introduced by a preposition, usually either do ‘to’, ‘for’ or le ‘with’, ‘to’, between the predicate and the complement. The syntactic structure of the complement will be discussed in 2.4 below. For necessity there is the noun éigean ‘force’, ‘violence’; ‘compulsion’; ‘necessity’. This is most commonly found in its past tense form (1) which used suppletively to the defective caith (2.1.1 and see 4.2 below): 3 (1)

B’ éigean be-PST necessity ‘I had to.’

dom. for me

For Munster especially we have the noun gá (Old Irish gád ‘danger’, ‘stress’, ‘need’) which is usually found in the negative as in (2): (2)



gá dom. necessity for me ‘I don’t have to / needn’t.’ NEG-PRS-be

This expression is typically used to negate the modal caith (4.2 below). Some Connacht dialects use the noun call (< English ‘call’) ‘need’; ‘claim’, ‘right’, instead of gá in this construction. A further couple of items particular to Munster should be considered here. An original noun foláir (Old Irish foróil ‘abundance’, ‘excess’) is now generally only found used modally with the negative copula (3a), while the adjective mór ‘big’, ‘great’ (i.e. ‘it is no great thing for me’ > ‘I must’) is also used with the negated copula (3b): (3)

a. Ní NEG-PRS-be

b. Ní

foláir dom. excess for me

mór great ‘I must / have to.’ NEG-PRS-be

dom. for me

With both (3a) and (3b) the idea is that if it is not too much to do something, then it should or must be done. Expressions of weaker obligation use the lexical items ceart and cóir, both of which can be taken as adjectives ‘right’, ‘just’, ‘fitting’, or as the

Modals in Irish

75

equivalent abstract nouns. In practice these are most commonly found preceded by the conditional copula ba as in (4): (4)

Ba

cheart / chóir right

COND-be

dom. for me’

‘I should.’ In the domain of possibility, most common is the item féidir, which, like foláir above, is only attested with the copula (5). The lexical source here is somewhat obscure — probably an Old Irish noun séitir ‘vigour’, ‘energy’ (see further 4.1 below): (5)

Is PRS-be

féidir liom. vigour to me

‘I can.’ 2.1.3. The syntax of polyfunctionality in Irish As indicated above, there is a general syntactic distinction between nonepistemic (non-finite) and epistemic (finite) complements. 4 For necessity, this can be exemplified for caith by (6a) non-epistemic and (6b) epistemic: (6)

a. Caithfidh mé imeacht anois. leave-VN now caith-FUT I ‘I must / have to leave now.’ (GGBC 236) b. Caithfidh (sé) go bhfuil an cluiche caith-FUT it that be-PRS the game ‘The game must be over.’ (GGBC 237)

thart. over

It may be noted that in (6a) and (6b) the verb caith is used in its morphological future indicative form as a general non-past (see 2.3 below). The proleptic pronoun sé ‘it’ in (7b) is optional in some dialects. In the case of féad, there is a tendency in many dialects to favour the morphologically future tense form as a general non-past (Ó Siadhail 1989: 292), as in (7a), while the conditional form is very common with this verb in epistemic use (7b; again the proleptic pronoun is optional):

76 (7)

Peter McQuillan

a. Féadfaidh sé teacht lá ar bith. féad-FUT he come-VN day any ‘He can / may come any day.’ (GGBC 234) b. D’fhéadfadh (sé) go raibh siad ann. féad-COND (it) that be-PST they there ‘They could / might have been there.’ (GGBC 235)

Note that (7a) is ambivalent between a deontic and a dynamic reading. As was noted in 2.1.1 above, a further semantic development has occurred in some dialects from possibility towards weak deontic necessity. This will be discussed further in 4.2 below. However, the non-finite construction can also be used epistemically. Note here that sé is no longer a ‘dummy’ subject: (8)

D’fhéadfadh sé imeacht le taom féad-COND he go-VN with bout tinnis. GEN-sickness ‘He could die of a bout of illness.’ (CD 116)

2.2. Syntax The syntax of modals shows certain differences from that of some modal content expressions (see 3 below). Compare the modals féad and féidir ‘can’ in (9) with the content expression (a)bheith ábalta ‘to be able’ in (10). Note here that the verbal noun is preceded by a particle (PTL.) a: (9)

a. Ní NEG

b. Ní

fhéadfaidh can-FUT

sin a that PTL

féidir sin a possible that PTL ‘That cannot be so.’

NEG

bheith amhlaidh. be-VN so

bheith amhlaidh be-VN so

(10) *Níl sin ábalta a NEG-PRS-be that able PTL *‘That cannot be so.’

bheith amhlaidh. be-VN so

Modals in Irish

77

Here the modal content word ábalta cannot be used impersonally as the two genuine modals can. The semantic specificity of ábalta means that it requires a willful agent as its first argument (subject). In contrast, the semantic generalization that modals typically undergo (leading to polyfunctionality, see 2.1 above and further 4.1 below) means that they are compatible with a greater variety of verb complements and can therefore assume the first arguments of those complements. However, modal content expressions differ from one another to the extent that they can participate in impersonal constructions. This is especially so where their nucleus is an overt lexical item. Compare (11a) and (11b) here (again the verbal noun is preceded by a particle a, here because it takes a direct object which precedes the particle): (11) a. Tá feidhm orm teach a be-PRS need on me house PTL ‘I need to build a house.’ b. Tá feidhm (le) teach a be-PRS need (to) house PTL ‘It is necessary to build a house.’

thógáil. build-VN thógáil. build-VN

As Hansen (2004: 248) puts it, modals “form a verbal complex which takes over the argument structure of the main verb”. Modals, therefore, have more affix-like traits than lexical expressions of modality. Thus, in the nonepistemic construction described above, the complement cannot have a separate subject (Ó Siadhail 1985: 288). 5 The modal assumes the first argument of the main verb while the verbal noun infinitive of the main verb is extraposed. McCloskey (1980) points to an apparent violation of this restriction on separate complement subjects for modals. In Irish many physical and mental states are expressed using a construction where the semantic subject is again governed by a preposition, as in (12): (12) Tá foighid againn. be-PRS patience at us ‘We have patience / are patient.’ When such a construction forms the verbal noun complement to a modal, foighid is apparently a separate complement subject, although the result is still grammatical (13):

78

Peter McQuillan

(13) Caithfimid foighid a bheith againn. must-FUT-1PL patience PTL be-VN at us ‘We must have patience / be patient.’ (McCloskey 1980: 354) As McCloskey points out, however, the fact that the sentence is perfectly well-formed shows that semantic rather than purely syntactic considerations are at work: the prepositional phrase againn ‘at us’ is felt to be the semantic or psychological subject of the ‘have’-construction, although it is grammatically indirect object and is, of course, co-referential with the matrix modal subject (‘we’). 2.3. Morphology – surface unit As will be clear from the examples already given, modal expressions in Irish appear as independent morphological surface units that occur with verbal noun and finite verb complements. A few additional observations may be made here. 2.3.1. The morphology of modal auxiliaries As is characteristic of auxiliaries in general, these typically show some defectiveness in their paradigms when compared with full lexical verbs (Heine 1993: 23). We can make the following generalizations in terms of the categories that they allow (defectiveness in the case of modals is italicized): Full a. b. c. d. e. f.

Imperative Verbal noun Conditional Future Present Past

Modal a. No imperative b. Generally no verbal noun c. Conditional d/e General non-past (either future or present form) e. Dialectally-determined past

In other words, most of the categories associated with full verbs are not entirely realized in the case of modal auxiliaries. There is a significant

Modals in Irish

79

dialectal dimension to this issue of defectiveness which will be discussed in some detail in 4.2 below. 2.3.2. The morphology of impersonal constructions As has been seen above, modal expressions of the impersonal type (2.1.2 and especially 2.2 above), use the prepositions le ‘with’, ‘to’ or do ‘for’, ‘to’ to introduce a semantic subject to the construction. One characteristic use of le is with a nominal predicate (noun or adjective) to indicate the possession of an opinion, state of mind, feeling, desire or impulse on the part of the semantic subject. On the other hand, the preposition do introduces a semantic subject for whom/which a state or condition applies. We can compare (14a) and (14b) with the predicate fearr ‘better’: (14) a. B’ fhearr liom imeacht. be-COND better with me go-VN ‘I would prefer to go.’ b. B’ fhearr dom imeacht. be-COND better for me go-VN ‘It would be preferable for me to go.’ In some cases, the distinction between the prepositions le and do can be used to distinguish between participant-external (15a) and participantinternal (15b) versions of the same modality: (15) a. Níorbh fholáir dó NEG-be-PST excess for him ‘He had to get up.’

éirí. get.up-VN

b. Níorbh fholáir leis éirí. NEG-be-PST excess with him get.up-VN ‘He had to get up.’ (GGBC 235) =‘He felt he had to get up.’ Furthermore, it may be observed that for modal predicates denoting necessity that take an oblique semantic subject the preposition of choice is do (external source) whereas for ability / possibility it is generally le (internal source). 6

80

Peter McQuillan

3. Other means of expression of possibility and necessity 3.1. Modal content expressions (Hansen 2004: 246) for Slavonic distinguishes between “typical modals” that express at least two types of modality (see 2.3 above) and “modal content” expressions that express only one. In Irish, such modal content expressions characteristically employ the substantive verb tá ‘is’. Here, as in 2.1.2 above, the modality is typically expressed lexically, although in a couple of cases the modality is a function of the entire expression. For necessity, we have a number of such lexical items: feidhm ‘function’, ‘use, ‘duty’, ‘need’ (see further 4.2 below); dualgas ‘duty’; oibleagáid ‘obligation’; fiacha ‘debts’ (pl); iallach or iachall ‘constraint’, ‘compulsion’ (GGBC 235–6). These frequently occur as a prepositional phrase governed by the partitive de ‘of’ with the semantic subject governed by the preposition ar ‘on’, as in (16): 7 (16) Tá sé de dhualgas orm é be-PRS it of duty on me it ‘I am (duty) bound to do it.’

a PTL

dhéanamh. do-VN

As can be seen from most of the glosses above, these constructions relate mostly to participant-external necessity. Extracted from such constructions (presumably) we have the common expression (17): (17) Tá orm é be-PRS on me it ‘I have to do it.’

a PTL

dhéanamh. do-VN

Here the verb phrase itself, devoid of any further lexical expression, conveys the modality. 8 The following, which is found especially in Northern Irish, appears to be borrowed from English I have to … (18): (18) Tá agam le hobair a dhéanamh. be-PRS at me to work PTL do-VN ‘I have to do (some) work.’ (GGBC 260) For participant-internal possibility there are a variety of options. The substantive verb is used with an adjective or prepositional phrase as in (19):

Modals in Irish

(19) Tá mé be-PRS I ‘I’m able.’

81

ábalta. able

The prepositional phrases in ann and in innimh / in inmhe (inmhe ‘wealth’, position’, ‘condition’; ‘strength’) are used with the same meaning. Other lexical items used here include cumas ‘power’, ‘capability’; caoi ‘means’, ‘opportunity’; gléas ‘means’, ‘facilities’; acmhainn ‘capacity’, ‘means’, ‘resources’. As was said above, these expressions differ from typically modal ones in that they express only one type of modality — in particular, there are no epistemic uses for any of these forms. However, it should be pointed out that, just as in 2.1.2 above, some of these expressions contribute to the grammatical paradigm of modality in that they can be used suppletively to modal auxiliaries (see 4.2 below). There is one verb to consider here: glac ‘should’, ‘ought to’, ‘had better’. This occurs in the conditional mood as a weak necessity modal in NW Ulster (Donegal) e.g. ghlacfá ‘you’d better’, ‘you ought to’ (Ó Siadhail 1989: 292) and in the future tense (Dónall Ó Baoill, personal communication). It is derived from the full verb glac ‘accept’, ‘obtain’, ‘procure’, which is found in all dialects. 3.2. Inflected and derived forms While modality in Irish is predominantly conveyed by free-standing lexical-morphological units, there are still certain residual forms that use inflectional and/or derivational morphology (GGBC 265–7). Old Irish has a form known as the verbal of necessity (NEC) used predicatively after the copula (Thurneysen 1946: 443) and this form is still common in Early Modern Irish as in (20) with the verb goin ‘wound’, ‘kill’ and the ending – ta (Greene 1979: 135, whence the following example): (20) Is gonta Tadhg. be-PRS kill-NEC Tadhg ‘Tadhg should / must be killed.’

82

Peter McQuillan

This construction survives in contemporary Irish only in stereotyped phrases following the negated copula such as (21) with the verb gearáin ‘complain’ (GGBC 265): (21) Ní

gearánta dom. complain-NEC to me ‘I can’t complain / mustn’t grumble.’ NEG-be-PRS

The verbal of necessity comes to be augmented by the prefix ion- / in- in the Middle Irish period. This formation continues in Modern Irish, although synchronically in- is now prefixed to the verbal adjective (VB ADJ) as with the form indéanta in (22a) from the verb déan ‘do’; or to the genitive singular of a noun, in this case inchomórtais from comórtas ‘competition’, ‘comparison’ (22b, see GGBC 267): 9 (22) a. Tá

sé in- déan - ta. PRS-be it PFX do VB.ADJ ‘It can be done.’

b. Níl

sé in - chomórtais léi. he PFX comparison-GEN with her ‘He is not comparable with her.’ / ‘He can’t be compared to her’

NEG.PRS-be

This formation expresses impersonal participant-external possibility. However, a semantic subject may be introduced by using the preposition ag ‘at’ (23): (23) Ní

raibh sé in - déan - ta aige. NEG PST-be it PFX do VB.ADJ at/by him ‘He shouldn’t have done it.’ (Ó Dónaill 1977: s.v. indéanta)

4. The grammaticalization of the modals We will now look at the development and status of modals in Irish in terms of Lehmann’s (2002) six parameters of grammaticalization. The parameters encompass two sets of relations: paradigmatic (4.1 to 4.3) and syntagmatic (4.4 to 4.6).

Modals in Irish

83

4.1. Parameter Integrity (Lehmann 2002: 112) As féad develops from the specific lexical concept of ‘getting’, ‘obtaining’ to the more generalized one of possibility, we have seen that it becomes polyfunctional in that it can express various types of possibility. To the extent that this semantic process involves a loss of specificity it is often called ‘bleaching’ or ‘desemanticization’; it is clear, however, that other processes such as metaphorical extension of meaning (from dynamic / deontic to epistemic, for example) or the conventionalization of implicature may also be involved (Hopper and Traugott 1993: 63ff). The syntactic correlates of these developments with this verb are the loss of its own argument positions and its increasing compatibility with a wider range of main verbs (see 4.3 and 4.4 below). Modals in Irish are not subject to particular formal or phonological erosion. The first step in establishing what Hansen (2004: 252) has called ‘the semantic space of modality’ is to describe the pre-modal or lexical space out of which modality evolves diachronically. Semantic sources for, or ‘paths to’, modality in Irish are presented immediately below.

4.1.1. Paths to necessity (i) ‘is used > needed’ (caith full verb, ‘use’, ‘consume’, ‘spend’, ‘wear’; ‘throw’): Greene (1975: 50) gives evidence from a number of Celtic languages for a semantic development of ‘use’, ‘function’ to ‘need’. This is borne out by consideration of the etymon feidhm (see 3 above) in Northern Irish (24): 10 (24) (A)tá feidhm ar X be-PRS use on X ‘There is a use/function for X ’ > ‘There is a need for X > ‘X needs (to)’ Southern Irish provides a further parallel to this general development with the expression oireann X do Y ‘X helps, serves, fits, suits Y’ (Greene 1975: 54). Thus in the Irish dialect survey (LASID) for item 1046 in the questionnaire, ‘you needn’t’, we find in West Cork ([Munster] 25):

84

Peter McQuillan

(25) Ní

oireann duit (etc.) suit/fit-PRS to you (It doesn’t fit / suit / serve you [etc.]) > ‘You needn’t …’ NEG

Old Irish has a construction for ‘need’ which consists of the verb ro-icc ‘reaches’ followed by the noun less ‘benefit’ (i.e. advantageous use) followed by the genitive case of the object needed if a noun, or preceded by a possessive if a pronoun (26): riccim for less. (Wb. 12d 33) NEG reach-PRS-1SG. your-PL benefit/ use (I don’t reach your use) ‘I don’t need you.’

(26) Ni

In current Irish two formations based on this expression survive: an abstract noun riachtanas ‘need’, ‘necessity’, derived from the verbal noun (Middle Irish ríachtain), and a corresponding adjective riachtanach ‘necessary’. It will be noted that the original object noun less (modern leas) has now been lost from the construction. We first find examples of caith in a possible modal sense around the twelfth century. Irish has a subjectless or autonomous form (AUT) of the verb in every tense and mood (often translating into English as a passive). In the present tense the ending is –t(e)ar and the relevant form here is caite(a)r ( Modern Irish féidir): An important source for possibility is the nominal predicate féidir (< Old Irish séitir via Middle Irish éitir). As has already been mentioned above, the lexical source here is somewhat obscure — an Old Irish noun séitir ‘vigour’, ‘energy’ is poorly attested but is already found with the copula is séitir la X ‘X can / is able’ in Old and early Middle Irish. For more on this etymon, see 4.1.3 below. (ii) ‘gets’ > ‘gets to’ > ‘can’ For possibility, the clearest path of development is that from lexical ‘get’, ‘find’, ‘obtain’, which is shown initially by the verb féad (Old Irish forms ad-cota ‘gets’ and (ní) éta ‘(doesn’t) get’, whence Middle Irish (f)ét, Modern Irish féad) and then later in the language, in the dialect of East Munster, by the verb faigh ‘get’, ‘find’. The syntax of the verb féad will be considered in more detail in 4.4 below. (iii) ‘movement in space’ > ‘comes from / to / with’; ‘reaches’, ‘arrives’ (especially compounds of Old Irish –icc): Verbs based on the stem –icc have a long ‘modal’ history in Irish. In Old Irish we have the verb con-icc ‘is able’, ‘has (the) power’ (for con- cf. Latin cum ‘with’) for which the form current today, the collocation of the verb t(h)ig (‘comes’) with the preposition le (‘with’), provides a parallel (see 2.1.1). In Middle and early Modern Irish, we also find the expression do-icc / ticc de ‘comes from’, or ‘arises from’ as in (30): (30) Tháinig díom a fhortacht. come-PST from me his helping ‘I was able to help him.’ (Ó Dónaill 1977: s.v. tar de) We may note also the idiom do-m-icc X in Old Irish (‘X comes to me’, -mbeing the 1SG. infixed object pronoun) which, when taking a noun subject (as semantic object), is often translated as ‘I get X’. However, when the nominal subject is a verbal noun we have (31):

Modals in Irish

87

(31) Cuman lim amal dom-áinic labrad. memorable to me how to me-come-PST speech-VN ‘I remember how I learned (got = became able) to speak.’ (DIL s.v. do-icc, 298) Here we at least approach the force of a modal auxiliary, even though this particular construction does not become productive as a modal auxiliary type. Another compound of –icc is ro-icc ‘reaches’, ‘succeeds’, ‘accomplishes’ which also develops into expressions of possibility. 12 Usually a preposition is added to express the logical subject, ó ‘from’ or le ‘with’, ‘to’ as in (32a), although the verb can also be used with a personal subject followed directly by a verbal noun clause (32b): 13 (32) a. Nach ráinig leó a rochtain. that-NEG reach-PST with/to them its reach-VN ‘that they could not reach.’ b. Ní

rainic nech cét-guin muice. reach-PST anyone first-wounding pig-GEN ‘No one could kill the first pig.’

NEG

The outcome of ro-icc (le) survives in Modern Irish (see 2.1.1 above). 4.2. Parameter Paradigmaticity (Lehmann 2002: 118) Grammaticalization posits a gradual distinction between ‘major’ linguistic categories (nouns, verbs, adjectives) and ‘minor’ categories (pronouns, auxiliaries, prepositions and conjunctions). Major categories are characteristically more ‘open’, or open-ended, minor categories are more ‘closed’ in terms of their membership. There are typically fewer grammatical than lexical expressions of modality. ‘Paradigmaticity’ refers to the extent to which a particular category or subcategory is clearly defined and tightly integrated. ‘Paradigmaticization’ refers to the evolution from ‘open’ to ‘closed’, from lexical to grammatical, from major to minor categories. For this reason it is also called ‘decategorialization’ (Hopper and Traugott 1993: 103).

88

Peter McQuillan

Under the rubric of paradigmaticity we are looking for evidence that helps us define modals formally, as a grammatical set. 14 In particular, modals often lose features normally associated with the full verbs from which they are derived, typically showing no imperative or non-finite forms. For example. Irish modal verbs show varying degrees of defectiveness, much of it dialectally based: 15 – No imperatives – 4.2.1:No non-finite forms (except marginally) – Idiosyncrasies with regard to – 4.2.2: past tense forms – 4.2.3: non-past forms – 4.2.4: negation The first of the above bullet points can be summarily accounted for: there are no imperative forms. 4.2.1. No non-finite forms For caith, the verbal noun caitheamh can be used as a noun in verb phrases of the type discussed in 2.1.2 above, see also (33a) and (33b) below, while a verbal noun is attested in the progressive construction for féad in Conamara (34): (33) a. Níl

caitheamh ar bith ort. obligation any on you ‘You don’t have to at all.’ (Béal. 1989: 38)

NEG-PRS-be

b. Níl

aon chaitheachtáil ann. any compulsion there ‘There is no compulsion’ (Ó Siadhail 1985: 291)

NEG-PRS-be

(34) Ní raibh mé ag féadachtáil … NEG-PST-be I PROG-féad-VN ‘I wasn't managing/ able to …’ (Ó Siadhail 1985: 292) These non-finite forms are clearly marginal in the verbal system as a whole.

Modals in Irish

89

4.2.2. Past Tense Forms Already in early modern Irish, a grammatical tract (BST, 44) proscribes all finite forms of modal caith not formed on the future stem (future indicative and conditional mood). This is still the position in the Irish of Ulster and largely of Connacht. However in southern Irish dialects present habitual (35a), past habitual (35b) and past non-habitual (35c) forms are found, as in its full verb counterpart: (35) a. Caitheann Donncha éisteacht. caith-PRS Donncha listen-VN ‘Donncha has to listen.’ (Ó Siadhail 1989: 291) b. Chaitheadh báid dul leis. caith-HAB-PST boats go-VN with him ‘Boats used to have to go with him.’ (An tOil. 65) c. Chaith daoine é a íoc. caith-PST people it to pay-VN ‘People had to pay it.’ (FFG s.v. caith) Otherwise, obligation reported as existing in the past is supplied suppletively by various nominal expressions based on the copula and substantive verb. Representative of the type of form used in the past is b’éigean do, which in fact occurs in all the dialects (see 2.1.2 above). 16 Outside of Ulster the morphological past tense of féad still expresses ability in the past ((36a) for Munster; (36b) for Connacht; note the dialect variation in the pre-verbal noun particles d’ and a): (36) a. Níor fhéadas suíochán d’ fháil. b. Níor fhéad mé suíochán a fháil. NEG-PST féad-PST-1SG seat PTL get ‘I couldn’t (= wasn’t able to) find a seat.’ (LASID item 184) In Ulster dialects, however, the past d’fhéad moves along a path from the report of ability in the past (could = was able) via hypothetical/ counterfactual (could have) to weak necessity (should have). 17 In (37) a

90

Peter McQuillan

young woman has been taken to task for striking a man who would not leave her alone and she responds: (37) D’fhéad seisean ligean domh. féad-PST he-EMPH leave to.me ‘He should have left me alone.’ (CO 50) (not ‘He was able to …’) The obligative sense appears to arise here as a conversational implicature from the non-implicative nature of modals: ‘he was able to leave me alone’ (but didn’t) > ‘he could / should have …’. 18 D’fhéad itself develops polyfunctionality and occurs also with epistemic force as in the sentence: Phós Seán i Sasana agus … ‘Seán married in England and …’ which is completed by (38): (38) D’fhéad clann a chloinne a bheith ann anois. féad-PST children of his children to be there now ‘his grandchildren could / should be there now.’ (Ó Siadhail 1989: 293) Ó Siadhail (1989: 293) reports on a further development with the modal féad in northern Irish where (39b) has developed from (39a): (39) a. D’fhéad féad-PST

tú é you it

a ráit to say-VN

leis. to him

b. D’fhéad duit é a ráit féad-PST to/for you it to say-VN ‘You should tell him / have told him.’

leis. to him

This development sees the normal subject pronoun that follows the verb replaced by a prepositional pronoun (inflected prepositional) form. The preposition involved here, do ‘to’, ‘for’, we have met above where it characteristically expresses the semantic subject of modal verb phrase predicates. This depersonalization has spread to the present tense so that féadann duit now occurs alongside féadann tú. 19

Modals in Irish

91

4.2.3. Non-past forms There are two separate issues to be considered briefly here: (i) neutralization of the present-future distinction in favour of a general nonpast; (ii) use of the conditional mood for past tense. (i) In the case of the necessity modal caith, we have LASID items 377 ‘I have to comb my hair before school’ and 379 ‘I have to shave every day’. Although the context is semantically habitual present, the response in all dialects is overwhelmingly the future form caithfidh mé (with the synthetic variant caithfead); however, some Southern dialects give an alternative present form caithim. 20 For possibility modals, we have LASID item 626: ‘We cannot live (without potatoes)’. Southern Irish here uses the modal féad and favours the future form ní fhéadfaimid (with the Munster variant ní fhéadfam) ‘we cannot’ (Ó Siadhail 1989: 292). Northern Irish, on the other hand, employs the modal thig le and has the present ní thig linn (or with the variant negation cha dtig linn) ‘we cannot’. As a modal, thig does not have a morphological future tense form and so the present must be used. (ii) The conditional mood form of a modal can be found with the sense of a simple past as in (40) from an Ulster dialect: (40) Chaithfinn imeacht mar bhí mé mall. caith-COND-1SG. leave-VN as PST-be I late ‘I had to go because I was late.’ (Ó Siadhail 1989: 292) This is one way, apart from suppletion (see 2.4.2 above) in which such dialects can compensate for the lack of a morphological past tense form. However, the same type of development can be found even where such a consideration does not arise. Thus, in the case of the verb féad, for LASID item 184 ‘I couldn’t find a seat in the room’ (see 34 above for the past tense here), most points record a conditional form either ní fhéadfainn (southern) or ní thiocfadh liom (verb thig le, northern, with the variant cha dtiocfadh liom). 21

92

Peter McQuillan

4.2.4. Negation Palmer (1995, 453–5) points out that there are two forms of irregularity that can affect the negation of modals and both of these apply to our Irish data: (i) There is suppletion of negative forms in the domain of necessity but the negation is placed ideally to negate the modality (i.e. not necessary) as in (41): (41) a. Ní NEG-PRS-be

b. Níl

gá necessity

duit. for.you

feidhm / féichiú need / necessity ‘You needn’t …’

NEG-PRS-be

ort. on.you

Another way is to indicate the inverse possible not, using the verb féad as in (42; LASID item 1046, point 75): (42) Féadaidh tú gan mórán oibre a dhéanamh. féad-PRS you NEG much work-GEN to do-VN (You can [not do any more work]) ‘You needn’t do any more work.’ Here the negation is again ideally placed since ‘possibility / permission not to’ is the equivalent of ‘not necessary to’. Note that the negative form gan indicates unambiguously that negation belongs to the non-finite verbal noun clause (as opposed to the finite clause negative ní). (ii) There is suppletion but also a lack of one-to-one correlation between the semantics and the syntax of negation: the negative is misplaced syntactically. 22 This occurs in Ulster dialects. Compare (43) below with (42) above, both of which are responses to item 1046 in LASID: (43) Ní

fhéadann tú níos mó oibre a dhéanamh. féad-PRS you more work-GEN to do-VN ‘You needn’t do any more work.’

NEG

Modals in Irish

93

Ideally this sentence as constructed should mean ‘you can’t / may not do any more work’, but in fact as already indicated it means ‘you needn’t …’. In other words, the negative which belongs semantically to the complement clause where it negates the proposition is syntactically displaced, or ‘raised’, to the main clause where ideally it should negate the modality of possibility but does not. 23 The ideal situation in language is one where the placement of the negative indicates the semantic scope of the negation. The Ulster use of ní fhéadann is strikingly irregular in this regard. I take this striking irregularity as further evidence for the greater degree of grammaticalization of modals in Ulster Irish than in the other dialects, since the negation of féadann in these cases is, in terms of the verbal system as a whole, thoroughly idiosyncratic. 24 4.3. Parameter Paradigmatic Variability (Lehmann 2002: 123) In the previous two sections we described some of the conceptual and formal dimensions of the grammaticalization process: loss of semantic specificity and certain morpho-syntactic properties. A further concomitant of these developments concerns the degree to which the speaker is free to choose a particular item according to his/her communicative intentions. Grammaticalization tends to diminish this freedom of choice and to make the use of an item largely obligatory in its context. This process is therefore sometimes called ‘obligatorification’ (Lehmann 2002: 162). Obligatorification here would involve a constraint on the speaker’s freedom to leave the general category of modality unspecified. Typically, however, modality is not ‘obligatory’ in the same way that tense, mood and number are in the finite verb. Hansen (2004: 263) points out that in Slavonic modals lose restrictions on the kinds on subjects with which they may combine. In Early Irish the verbs féad and caith take human, or at least animate, subjects. Over time, however, they lose this selection restriction as a consequence of the desemanticization discussed in 4.1 above, and can now appear with any verb, just as a tense marker can.

94

Peter McQuillan

4.4. Parameter Scope (Lehmann 2002: 128) As was discussed in section 4.1 above, the development of grammatical forms from lexical items starts at the conceptual level and can be accounted for, partly at least, by a loss of semantic specificity (Heine 1993: 54). Now I will look at some of the syntactic correlates of this development in the case of féad. In Old Irish the ‘independent’ form of this verb in the present indicative is ad-cota, the ‘dependent’ form –éta (as after the negative particle ní for example). 25 It characteristically has a human subject, expresses a lexical concept ‘gets’, ‘obtains’ and takes a concrete object as its complement (see Heine 1993: 54). A third argument, from which the thing is obtained, can be introduced by the preposition for ‘on’ ((44), note that the preverb atcontains the infixed object ‘it’): (44) At

-chotat -sat PV-OBJ ‘it’ get-PRS-3PL EMPH ‘They obtain it from God.’

for dia. (Ml 54a9) on God

The Old Irish dependent form -éta develops into Middle Irish fét- (modern féad). 26 Simple object nouns still occur (45a) but a verbal noun complement is becoming the norm (45b). The meaning is now ‘be able’, ‘can’ and is most frequent in negative and conditional (i.e. irrealis) contexts: 27 (45) a. O nar fhétsat bás ele since that-NEG-PST get-PST-3PL death other dó. (PH 2555) for.him (because they didn’t get another death for him) = ‘because they couldn’t kill him any other way.’ b. Ní

étaim get PRS-1SG (I don’t get a going) = ‘I cannot go.’

NEG

dano moreover

techt. (TBC 791) go-VN

If the verbal noun governs an object there are two possibilities. First, the object of the verbal noun itself is marked for possessive if pronominal, or genitive if it is nominal (46a). This can lead to a certain indeterminacy as to

Modals in Irish

95

the nominal or infinitival status of the clause. Second, the verbal noun is preceded by both its object and the preposition do ‘to’ (46b). In this case, either a noun or pronoun object will be accusative: (46) a. Níro fét a n-acallaim. (SC 81) NEG-PST get-PST-3SG their speak (to)-VN (He didn’t get their speaking to) = ‘He was not able to speak to them.’ b. Ní

fétar in chned do lesugud. (PH 7645) NEG get PRS-AUT the wound to heal-VN (One can’t [get] the wound to heal) = ‘The wound cannot be healed.’

Irish thus has two types of verbal noun complement: a more noun-like and a more verb-like variant. Disterheft (1997: 126) argues that these are in fact in complementary distribution in Middle Irish: the more nominal construction being used when the object is pronominal, the more verbal with a noun object. Synchronically, therefore, both types belong to the same ‘infinitival’ construction. By Early Modern Irish (47), however, the more clearly ‘verbal’ prepositional infinitive type becomes predominant: (47) Ní

fhéadaim romhar do dhéanamh. (TSh. 7546) NEG get PRS-1SG digging to do-VN (I don’t get to dig) = ‘I can’t dig.’

For the noun rómhar one can now substitute an object pronominal form, e.g. é ‘it’. Finally, the subject of féad is no longer necessarily identified with a willful human subject, as shown in by the use of a ‘dummy’ subject (48): (48) Féadfaidh sé sneachta a dhéanamh. can-FUT it snow PTL make / do-VN ‘It may snow.’ (Ó Dónaill 1977, s.v. féad) This gives us the argument structure of an auxiliary where the surface subject position of the matrix verb is the first argument of the non-finite verb (see 2.2 above).

96

Peter McQuillan

Thus the verb féad goes from its full lexical status in Old Irish where it takes the same kinds of nominal arguments as any other transitive verb to a situation where it characteristically takes a non-finite verb as complement. Historically féad is the verb that shifts earliest towards modal auxiliary status in Irish. 28 While we can account for these developments in terms of Lehmann’s parameter of condensation of structure, for epistemic use there is a clear tendency to widen structural scope from the verbal action itself to the proposition, which is reflected iconically in the use of a finite subordinate clause as in (6b) and (7b) above (see section 2.1.3). 4.5. Parameter Bondedness (Lehmann 2002: 131) Here an item goes from being independently juxtaposed, via the process of ‘coalescence’, to where it becomes an affix or even a phonological feature of its carrier. Irish modals show no evidence for this. 4.6. Parameter Syntagmatic Variability (Lehmann 2002: 140) Here an item evolves from being able to be shifted around freely, via a process of ‘fixation’, to where it occupies fixed syntactic slot. Since complements in Irish are invariably extraposed, the modal must precede its main verb. So syntagmatic variability does not exist here. But this is a feature of Irish complements in general. However, in epistemic usage, where modal expressions develop, or show signs of developing into adverbs, greater variability exists. Compare the following seventeenth-century examples, using the conditional autonomous form féadfaidhe (49a) and (49b): (49) a. Féadfaidhe nár sgríobhadar. (Desid. 8) féad-COND-AUT that not write-PST-3PL ‘It could be that / Maybe they didn’t write.’ b. Ní

fhuiléanga Dia, féadfaidhe. (Desid. 4427) suffer-FUT God féad-COND-AUT God will not allow, maybe, ...

NEG

Modals in Irish

97

We may almost be dealing with quasi-adverbial function here. The same flexibility of word order applies to the adverbial b’fhéidir ‘maybe’ which is only used epistemically and also to the necessity modal ní foláir (50a and b): (50) a. Ní NEG-be

foláir foláir

nó tá that PRS-be

sé tinn. he ill

b. Tá

sé tinn, ní foláir. he ill NEG-be foláir ‘He must be ill.’ (GGBC 236–7)

PRS-be

Such a development relates to other parameters above: decategorialization (4.2) and scope (4.4). Note in this regard that the use of invariant caithfidh for epistemic necessity approaches adverbial force except that its syntactic position is fixed in that it must precede its complement. 5. Conclusion In his discussion of the Slavonic modals, Hansen (2004) proposes that modals display a cluster of formal and semantic characteristics that distinguish them from lexical expressions of modality. These are arranged as a modal prototype into ‘central’ and ‘peripheral’ features, all of which have been incorporated into our discussion above: 1. Central features – Semantic – modals are polyfunctional (2.1) – Syntactic – modals form a verbal complex in which they assume the argument of the main (complement) verb (2.2) – Morphological – modals are superficially independent free-standing units (2.3) 2. Peripheral features – Semantic: modals lack fully lexical meanings beyond the modal ones (4.1) – Morpho-syntactic: modals show various kinds of defectiveness (4.2)

98

Peter McQuillan

In keeping with a prototype analysis, membership of the category modality is seen as gradual rather than “either/or”. In such terms, Hansen identifies (i) an internal core whose members show all five characteristics (in Irish féad, féidir, ní foláir — it may be noted in passing here that possibility grammaticalizes earlier and, in the terms of the present discussion, more fully than necessity in Irish); (ii) an external core whose members show all the central features (caith; most modal impersonal expressions); and finally (iii) a periphery showing some of the central features (modal content words, 3.1). As was noted in 3.1 above, however, even the more purely lexical expressions of modality can possess a grammatical function in that they are used suppletively, especially negatively and in the past tense, to compensate for the defectiveness of some of the more grammaticalized modals. Modals in Irish show: – A tendency to paradigmacity but still compete with / complement a large set of lexical modal expressions (4.2). The fact that modals cannot co-occur is a feature of this tendency; – Compatibility with any verb type but no obligatorification (4.3); – A condensation of syntactic scope that is relatively highly grammaticalized (4.4); – No signs of bondedness (4.5); – Very limited syntactic variability (4.6); – No postmodal developments, unlike Slavonic (e.g. no future of conditional forms developing from modals) with the exception of the adverbialization of b’fhéidir (4.6 above). Grammaticality of modals is decided by the extent to which syntactic restrictions apply and to which the items concerned can be described as constituting a limited rather than open-ended system (Palmer 1986: 5). In this respect, as in English, the defectiveness of modals shows a relatively high degree of grammaticalization (4.2), especially in Northern Irish. In Hansen’s (2004) discussion, it is not entirely clear to me whether the five parameters adduced are intended to apply to Slavonic only, or whether they are envisaged as having a more universal validity. Since defectiveness represents modality in Irish at its most grammaticalized stage, should it be relegated to ‘peripheral’ feature? It seems counterintuitive, for example, to regard foláir as more of a core modal than caith, for example, when one considers the greater frequency of the latter.

Modals in Irish

99

In order to situate this question of defectivness in the context of grammaticalization, we may look at Heine’s (1993) analysis of the developments of auxiliaries. On a cline of grammaticalization, auxiliaries tend to occupy an intermediate position between the two poles of “lexical” and “grammatical” (Heine 1993: 54ff). Heine identifies a succession of stages in the evolution from full verb to auxiliary: A. The verb expresses a lexical concept with a characteristically human subject and the complement is a concrete object B. Rather than a concrete object, the complement expresses a dynamic situation which can be formally expressed in various ways: (i) gerundial / participial / infinitival or (ii) clausal. For Irish at all periods option (i) means a verbal noun clause C. The verb acquires grammatical function and the subject is no longer especially identified with human / willful referents; the verb is now more and more associated with non-finite complements. There also develops the requirement that the subject of the modal and the complement should be identical. This stage marks the transition from full verb to semi- or quasi-auxiliary status. It is at this stage that sentences like (28) above are possible (the modal co-occurs with its full lexical etymon) D. The verb may be characterized by defectiveness. This represents the first move towards ‘decategorialization’ (see 4.2 above) In subsequent stages (E through G), items become subject to phonological erosion, cliticization and affixation; eventually they become purely functional, grammatical elements. 29 These final three stages are not a factor for Irish modals. Heine (1993: 65) makes the important point, however, that various of these seven stages of grammaticalization may be combined in the characterization of individual items. Thus a verb that is evolving toward a more grammatical status may still retain some full, lexical characteristics. In Irish, the verb caith would be an “A / D item”, where “A” accounts for its full and “D” its modal verb status. On the other hand, féad has “D”, or purely auxiliary status. In general terms, therefore, we would locate Irish modals at an intermediate stage of the grammaticalization chain and characterize their overall degree of grammaticalization as medium. 30 The overall conclusion of medium grammaticalization appears to square with more general studies, as well as with Hansen’s on Slavonic (2004).

100 Peter McQuillan Bybee, Perkins and Pagliuca (1994: 242) in a survey of over sixty languages, conclude that almost 80% of those language express what they call agent-oriented (non-epistemic) modality through independent, freestanding grammatical units (for epistemic modality the proportions are much more even, supporting the conclusion that epistemic modality is likely to be more grammaticalized). This applies even to a language like English which otherwise shows a relatively high degree of grammaticalization of modals. Notes 1. Except for the fact that Irish does not allow SVO order as an alternative. See McCloskey (1983: 12). 2. In the course of their analysis of modal polyfunctionality, van der Auwera, Ammann and Kindt (2005: 248) place Irish in a brief discussion of “Standard Average European” (SAE) based on five linguistic criteria. By these criteria, Irish emerges as “less SAE” than Hungarian, Russian, Spanish, Icelandic, French and German (no indefinite article; no ‘have’-perfect; no relative pronoun; having pro-drop — the only criterion met was the participial passive). 3. Unless otherwise stated, examples are my own. 4. For semantic explanations, see Noonan (1988: 90); Givón (1995: 125-6). For Irish see McQuillan, (2000). 5. This restriction applies to ‘control’ verbs in general, such as féach le ‘try’, the aspectual ba ghnách le ‘used to’, of which modals are thus a subset (McCloskey 1980: 354). As Hansen puts it (2004: 248): “Modals … form a verbal complex which takes over the argument structure of the main verb”. 6. With the predicate féidir the normal preposition in Modern Irish is le, but do is also found in the earlier language and still survives today. In this connection it should be noted that the Scottish Gaelic cognate phrase is is fheudar dhomh ‘I must’, where participant-external possibility develops into necessity. 7. “Duty” is mentioned by van der Auwera and Plungian (1998: 96) as one of the lexical sources for deontic necessity. 8. For some speakers, tá orm “I have to” conveys a greater sense of urgency than the modal verb caith “must” (Tara Macleod, personal communication). This appears to cut across any distinction between participant-internal and -external modality. 9. For a succinct account of the diachronic development of this construction see Greene (1966). 10. This has led further in Scottish Gaelic to the evolution of a modal verb feumaidh ‘must’.

Modals in Irish 11. 12. 13. 14. 15. 16. 17. 18. 19.

20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30.

101

Much more remains to be done on this specific development. As does the semantically related ro-saig ‘reaches’, ‘arrives’. For (32a) and (32b), see DIL under ro-icc. See also Palmer (1986: 33–8) for a comparison between English, German and French in this respect. For further discussion and examples, see McQuillan (1999). Palmer (1983, 207ff.) distinguishes between the modal and propositional (semi-modal) content of a clause to explain such defectiveness in the past of English modals. See Bybee (1995) for a discussion of the relation between past tense and modal sense. “Past combined with modality leaves open the possibility that certain conditions on the completion of the predicate have not been met” (Bybee 1995: 514). According to Ó Siadhail, the newer construction with do is felt to be ‘stronger’ or more emphatic than the original one with the simple pronoun. However, a native speaker of this dialect, Dónall Ó Baoill (p.c.), sees the distinction as basically generational: the newer construction emerges with speakers born after the mid-1950s or so. Mostly Munster, points (28, both items), (17), (22); one in Connacht, point (34). Although note point 56 with the past tense tháinig liom in the affirmative. According to Palmer this type of misplacement is most likely to occur across languages with the equivalent of must not, as exemplified by English needn’t where the modality rather than the propositional content is negated. It occurs at points (74), (78), (79), (81), (83) and (86a). For a comprehensive discussion of the relationship between modality and negation, see de Haan (1997). Known to scholars of Old Irish as, respectively, ‘deuterotonic’ (the second element of the verbal complex receives the stress: ad-cota) and ‘prototonic’ (the first element is stressed: -éta) forms. See McCone (1987: 2ff). In the Middle Irish period compound verbs become marginalized and new simple verbs develop, often based on the old prototonic stem as in the current case. See McCone (1987: 206ff). According to DIL. See ad-cota and fétaid; see also McCone (1987: 127) for likely examples as early as the eighth century. In Scotland, however, the verbal noun faotainn can still be used suppletively to the main verb faigh ‘get’, ‘find’. See for example English ’ll from will. Much as Hansen (2004) does for Slavonic.

102 Peter McQuillan References Aisl.MC 1990 An tOil. 1980 Béal. 1927BST 1944

Aislinge Meic Conglinne, Kenneth Jackson (ed.). Dublin: Dublin Institute for Advanced Studies. An tOileánach, Tomás Ó Criomhthainn; Pádraig Ó Maoileoin (ed.). Dublin: Helicon. Béaloideas: The Journal of the Folklore of Ireland Society. The Society. Bardic Syntactical Tracts, Lambert McKenna (ed.). Dublin: Dublin Institute for Advanced Studies.

CD 1976

Cith is Dealán, Séamas Ó Grianna (‘Máire’). Cork: Mercier Press.

CO 1976 Desid. 1975 DIL 1983 FFG 1985 GGBC 1960 LASID 1958 LL Táin 1967

Caisleáin Óir, Séamas Ó Grianna (‘Máire’). Cork: Mercier Press. Desiderius: otherwise called Sgáthán an Chrábhuidh, T.F. O’Rahilly (ed.). Dublin: Dublin Institute for Advanced Studies. A Compact Dictionary of the Irish Language. Dublin: Royal Irish Academy. Foirisiún Focal as Gaillimh, Tomás deBhaldraithe (ed.). Dublin: Royal Irish Academy. Graiméar Gaeilge na mBráthar Críostaí [Christian Brothers’ Irish Grammar]. Baile Átha Cliath [Dublin]: Mac an Ghoill. Linguistic Atlas and Survey of Irish Dialects, Heinrich Wagner (ed.). Vols. 1–4. Dublin: Dublin Institute for Advanced Studies. Táin Bó Cúalnge from the Book of Leinster, Cecile O’Rahilly (ed.). Dublin: Dublin Institute for Advanced Studies.

Ml 1975

Reprint ‘The Milan Glosses on the Psalms’, Whitley Stokes and John Strachan (eds). In Thesaurus Palaeohibernicus, Vol.1, pp.7– 483. Dublin: Dublin Institute for Advanced Studies.

1887

The Passions and Homilies from the Leabhar Breac, Robert Atkinson (ed.). Dublin: Royal Irish Academy

PH

Modals in Irish

103

SC 1975 TBC 1976 TSh 1931

Serglige Con Culainn, Myles Dillon (ed). Dublin: Dublin Institute for Advanced Studies. Táin Bó Cúailnge Recension 1, Cecile O’Rahilly (ed.). Dublin: Dublin Institute for Advanced Studies. Trí Bior-ghaoithe an Bháis [The Three Shafts of Death], Osborn Bergin (ed.). Dublin: Hodges Figgis.

Wb. Reprint ‘Codex Paulinus Wirzburgiensis’, Whitley Stokes and John Strachan (eds). In Thesaurus Palaeohibernicus, Vol. 1, 499–714. Dublin: Dublin Institute for Advanced Studies. Bybee, Joan 1995 The semantic development of past tense modals in English. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 503–517. Amsterdam/Philadelphia: Benjamins. Bybee, Joan, Revere Perkins and William Pagliuca 1994 The Evolution of Grammar. Tense, Aspect and Modality in the Languages of the World. Chicago/London: The University of Chicago Press. de Haan, Ferdinand 1997 The Interaction of Modality and Negation: A Typological Study. New York: Garland. Disterheft, Dorothy 1997 Syntactic innovation in Early Irish. In Dán do Oide. Essays in Memory of Conn R. Ó Cléirigh, Anders Ahlqvist and VČra ýapková (eds.), 123–134. Dublin: Institiúid Teangeolaíochta Éireann. Givón, Talmy 1995 Functionalism and Grammar. Amsterdam/Philadelphia: Benjamins. Greenberg, Joseph H. 1963 Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Language, Joseph H. Greenberg (ed.), 73–113. Cambridge: MIT Press. Greene, David 1966 The prefix in-. Ériu 20, 82–86. Greene, David 1975 A recent semantic shift in Insular Celtic. Zeitschrift für Celtische Philologie 34, 43–59. Greene, David 1979 Perfects and perfectives in modern Irish. Ériu 30, 122–141.

104 Peter McQuillan Hansen, Björn 2004 Modals and the boundaries of grammaticalization: the case of Russian, Polish and Serbian-Croatian. In What Makes Grammaticalization? A Look from its Fringes and its Components, Walter Bisang, Nikolaus P. Himmelmann, and Björn Wiemer (eds.), 245–271. (Current Trends in Linguistics 158) Berlin / New York: Mouton de Gruyter. Heine, Bernd 1993 Auxiliaries. Cognitive Forces and Grammaticalization. New York/Oxford: Oxford University Press. Hopper , Paul and Elizabeth Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. McCloskey, James 1980 A note on modern Irish verbal nouns and the VP-complement analysis. Linguistic Analysis 6 (4), 345–357. McCloskey, James 1983 A VP in a VSO language? In Order, Concord and Constituency, Gerald Gazdar, Ewan Klein and Geoffrey K. Pullam (eds.), 9–55. Dordrecht/Cinnaminson: Foris. McCone, Kim 1987 The Early Irish Verb. Maynooth: An Sagart. McQuillan, Peter 1999 On the modal verbs of Modern Irish: aspects of caith and féad. Éigse. A Journal of Irish Studies 31, 37–59. McQuillan, Peter 2000 Complementation and the subjunctive in early Irish. Ériu 50, 87– 132. Noonan, Michael 1985 Complementation. In Language Typology and Syntactic Description, Vol. 2, Timothy Shopen (ed.), 42–140. Cambridge: Cambridge University Press. Ó Dónaill, Niall (ed) 1977 Foclóir Gaeilge-Béarla. Baile Átha Cliath [Dublin]: Oifig an tSoláthair. Ó Siadhail, Mícheál 1989 Modern Irish. Grammatical Structure and Dialectal Variation. Cambridge: Cambridge University Press.

Modals in Irish

105

Palmer, Frank R. 1983 Semantic explanations for the syntax of the English modals. In Linguistic Categories. Auxiliaries and Related Puzzles, Frank Heny, and Barry Richards (eds.), 205–217. Dordrecht/Boston: D. Reidel. Palmer, Frank R. 1986 Mood and Modality. Cambridge: Cambridge University Press. Palmer, Frank R. 1995 Negation and the modals of possibility and necessity. In Modality in Grammar and Discourse, Joan Bybee, and Suzanne Fleischman (eds.), 453–471. Amsterdam/Philadelphia: Benjamins. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2, 79–124. van der Auwera Johan, Andreas Ammann, and Saskia Kindt 2005 Modal polyfunctionality and Standard Average European. In Modality. Studies in Form and Function, Alex Klinge, and Henrik Høeg Müller (eds.), 247–272. London/Oakville CT: Equinox.

4. Modals in the Romance languages Bert Cornillie, Walter De Mulder, Tine Van Hecke and Dieter Vermandere

1. Introduction: aims and restrictions In accordance with the objectives of this volume, the aim of this contribution is to determine the degrees of grammaticalisation of modal verbs in Romance by using Lehmann’s (2002) parameters. The results obtained will be used to answer the central questions this volume addresses, such as: (i) what is to be considered a modal? (ii) where is the borderline between modals and fully lexical expressions of modality? (iii) what are the relevant parameters for a typology of modal systems? Space, however, prohibits us from discussing all modal verbs in the Romance languages. We therefore agreed on the following restrictions: (i) we will only look into modal verbs of four Romance languages: French, Spanish, Italian and Romanian; (ii) we will restrict ourselves to verbs expressing “obligation”, i.e. the Romance equivalents of must and its synonyms; (iii) however, in addition to the deontic readings of these verbs, we will also look at their epistemic interpretation, when they express probability or certainty.1 The paper is structured as follows: in Section 2, we present the verbs to be discussed. Section 3 deals with the various parameters proposed by Lehmann (2002). The conclusions are summed up in Section 4. 2. The verbs of our sample For clarity’s sake, we will start by listing the verbs and constructions analyzed in this contribution; as we already pointed out, they are all expressions of “obligation” or “necessity” (van der Auwera and Plungian 1998): Italian: dovere, avere da, essere da, andare + passive (past) participle, bisogna, occorre, conviene

108

Bert Cornillie et al.

Spanish: deber, haber de, haber que, tener que French: devoir, avoir à, falloir, être à Romanian: a trebui, a avea de, a fi de 3. Lehmann’s parameters applied to Romance modals In this section we will present the results of the application of Lehmann’s parameters to the above-mentioned Romance verbs. As will become clear, we will pay more attention to the parameters along the paradigmatic axis than to those along the syntagmatic axis. Our main goal is to try to pin down the properties that characterise the modal verbs as a specific ‘modal’ paradigm in the selected Romance languages. 3.1. Weight/integrity The parameter of weight/integrity concerns two well-known processes of grammaticalisation: (i) desemanticisation or bleaching; (ii) attrition or the gradual loss of phonological substance. Unlike desemanticisation, attrition does not seem to play a role in the evolution of the Romance modals (cf. 3.1.2). 3.1.1. Desemanticisation A series of desemanticisation processes are observed in the Romance modals, some of which already took place in Latin. We will briefly present the desemanticisation processes involved as they developed from the Latin verbs debere (1) and habere (2), the Romance verbs meaning ‘to need’(3) and, finally, several other sources (4). 1) Debere Italian dovere, Spanish deber and French devoir all evolved from Latin debere ‘must’; Romanian, on the contrary, has no such verb. In Latin, the verb originally meant ‘I owe to/ I am indebted to’ and, when it was followed by an infinitive, ‘duty obliges me to’. However, already in Latin, debere was also a modal expressing obligation and necessity. So desemanticisation already took place in Latin where debere also had an

Modals in Romance languages

109

epistemic meaning, expressing probability (cf. Olbertz 1998: 246; Bolkestein 1980: 123-127).2 According to Bybee, Perkins and Pagliuca (1994: 200-201), the epistemic meaning of must is due to a metaphorical transfer of the basic deontic meaning,3 since “the contexts in which must has an obligation reading and the contexts in which it has an epistemic reading are mutually exclusive”. The “metaphorical change implies a shift to another domain”.4 However, Traugott and Dasher (2002: 131) point out that the metaphor account is based on the actual use of must, and that what appear to be metaphorical relations in synchronic analyses of contemporary senses, are frequently the result of invited inferences in semantic evolution.5 Moreover, even in present-day English, the deontic and epistemic senses can overlap, as in example (1): (1)

And anyway I think mental health is a very relative thing – I mean, mental health must be related to the sort of general mentality or whatever other word you use of the community you’re living in. (Coates 1983: 47, cited in Traugott and Dasher 2002: 112).

The same goes for Modern French (cf. Desclés 2003: 51). Moreover, several authors have pointed out that the interpretation of devoir is influenced by its context of use.6 Huot (1974: 41) systematically discusses the influence of several grammatical factors and points out that in (2a), devoir has a deontic interpretation (“obligation”), whereas in (2b), it gets an epistemic interpretation, because of the perfective form of the infinitive: (2)

a. On doit one/we must-IND.PRS.3SG ‘We must wait.’

attendre. wait-INF

b. On doit être venu. one/we must-IND.PRS.3SG be-INF come-PTCP.PRF-M.SG ‘Someone must have come.’ Such examples of overlap between deontic and epistemic meaning and of the influence of contextual elements make an account in terms of invited inferences more plausible. Following Traugott and Dasher (2002), we thus propose for debere and the modal verbs derived from debere an evolution involving schematically the following stages:

110

Bert Cornillie et al.

í to owe to; í obligation / necessity; í epistemic meaning: probability / certainty. Even though this representation is too schematic, it shows that at least two desemanticisation processes are needed to explain the evolution of debere and the verbs derived from it.7 It also shows that it is quite likely (though not necessary) that the epistemic readings of polyfunctional modals derive from the non-epistemic one (van der Auwera and Ammann 2005a,b,c). Semantic bleaching between (pre-)modal interpretations does not seem to function in isolation, but generally accompanies changes in other paradigmatic and syntagmatic parameters. As a clearer case of such combined (i.e. semantic and syntagmatic) shifts one could cite the Italian verb bisogna (see Benincà and Poletto 1994, 1996; Guéron 2000). Bisogna has an original, lexical meaning of ‘need’ and a specific case-grid in (3), but changes into an impersonal verb that takes a finite or infinite subject clause as in (4), (5): (3)

Per lo quale mi bisogn-ano fiorini for the:M.SG which:SG I:DAT need-IND.PRS.3PL florins:PL dugento d’oro. (Boccaccio, Il . Decameron, giornata VIII, novella 1) 200 of gold:M.SG ‘For which I need 200 gold florins.’

(4)

Bisogn-a che Carlo chiam-i must-IND.PRS.3SG that Carlo call-SBJV-PRS.SG medico. doctor:M.SG ‘It's required that Carlo calls a doctor.’

(5)

Bisogn-a chiam-are un medico. must-IND.PRS.3SG call-INF a:M.SG doctor:M.SG ‘Someone has to call a doctor.’

un a:M.SG

In Modern Italian (examples (4) and (5)), bisogna only has an impersonal 'obligation' interpretation. The personal 'need' reading is impossible, as is the more elaborate syntactic frame (3). Similar semantic shifts hold for dovere (and for avere da + infinitive) with the exception that, unlike bisognano in (3), their different readings are still possible in contemporary

Modals in Romance languages

111

Italian. As is typical of grammaticalisation processes, the development of the new meaning of debere has not led to the elimination of its older values: there is ‘layering’ of different readings (Hopper 1991: 22; Traugott and Dasher 2002). In Italian, French and Spanish, dovere, devoir and deber have (at least) three values, respectively: í it is a main, lexical ditransitive verb; í it can be a deontic modal, when it is followed by a bare infinitive; í it can be an epistemic modal, in which case it is also followed by a bare infinitive. In our view, this makes the Romance equivalents of debere polysemic verbs, whose meanings form a network.8 2) Habere The Latin verb habere meaning ‘to have’ or ‘to hold’ is the second major source giving rise to verbs expressing the modal meaning of obligation, and is attested in the four languages studied here. Yet, unlike debere, which itself expresses obligation, it seems more adequate to say that a whole construction rather than a verb has grammaticalised to some degree. Good examples are Italian avere da + infinitive, French avoir à + infinitive, Spanish haber de or Romanian a avea de + supine: the meaning of these constructions still seems largely transparent and can be obtained compositionally, since the prepositions da (Italian), à (French) and de (Spanish and Romanian) evoke the idea of an objective to be attained.9 This concords well with the infinitive — in French, Spanish and Italian — which is frequently used, according to Bybee, Perkins and Pagliuca (1994: 184185), to present the denoted activity as a projected one. The same holds for the Romanian supine. On the other hand, the presence of the verb have can be explained if one accepts that the construction comes from (6), from Gougenheim (1929:205): (6)

Il a France he have-IND.PRS.3SG France:F.SG ‘He has France to protect.’

à garder. to protect-INF

When the object10 is something that does not (yet) exist (as in “a letter to write”), the sense of possession may be bleached out and the pragmatic

112

Bert Cornillie et al.

inference of obligation can be stressed (cf Traugott and Dasher 2002: 139). A reading that gives more prominence to the obligation is found in the grammaticalised reformulation in (7): (7)

Il a à garder la France. he have-IND.PRS.3SG to protect-INF the:F.SG France:F.SG ‘He has to protect France.’

In Italian and Romanian too, both constructions coexist: in (8b) and (9b) the infinitive/supine appears as the complement of the direct object of avere, while in (8c) and (9c) the construction is embedded directly under avere / a avea. (8)

a. Non ho not have-IND.PRS.1SG ‘I have nothing.’

niente. nothing

b. Non ho niente da fa-re. not have-IND.PRS.1SG nothing to do-INF ‘I don't have anything to do.’ c. Non ho da fa-re niente. Not have-IND.PRS.1SG to do-INF nothing ‘I don't have to do anything.’ (9)

a. Are cinci copii de cresc-ut. have-IND.PRS.3SG five child:PL to raise-SUP ‘He has five children to raise.’ b. Are de cresc-ut cinci copii. have-IND.PRS.3SG to raise-SUP five child:PL ‘He has to raise five children.’ c. Mai ai de învăĠa-t! still have-IND.PRS.2SG to learn-SUP ‘You still have (a lot) to learn.’

In the (c) constructions, the verbs avere and a avea acquire a more abstract sense: they serve to indicate that the agent is associated with the projected

Modals in Romance languages

113

activity expressed by the infinitive and thus, “has” this activity present or accessible in a certain sense. Furthermore, Bybee, Perkins and Pagliuca (1994: 184) point out that the verb ‘to be’ is used in other languages to form deontic verbs or constructions. This is the case of the Italian essere da, the French être à and the Romanian a fi de. Spanish does not have the ser a/de construction, but the auxiliary haber seems to be more grammaticalised than its equivalents in Italian, French, and Romanian since its possessive meaning has completely disappeared. In (10) haber is followed by the preposition de. Such a haber de construction can receive both a future reading and a deontic reading. In both cases it presents a general rule to obey. (10) Las indicaciones de las notas en el the:F.PL indication:F.PL of the:F.PL note:F.PL in the:M.SG cuerpo del texto han de body:M.SG of the:M.SG text:M.SG have-IND.PRS.3SG to hacerse mediante números arábigos, make-INF.REFL by means of number:M.PL arabic:M.PL en superíndice o entre paréntesis. in superscript:M.SG or between bracket:M.PL ‘The indication of the notes in the body of the text has to be done by means of Arabic numbers, in superscript or between brackets.’ In other contexts the verb haber is not used as a possessive verb either. In present-day Spanish it is exclusively used with past participles as a temporal auxiliary, e.g. lo he dicho ‘I have said it’. Tener not only conveys the possession reading that haber had before but has also replaced the verb in the deontic readings. Whereas originally tener de was used, tener is attested in combination with the complementiser que from the end of the 15th century on. Once again, a possessive meaning leads to a grammaticalised deontic modal, as in (11): (11) No como dize sant isidoro dela not as say:IND.PRS.3SG saint isidoro:M.SG of the:F.SG orden del a b c. que la a order:F.SG of the:M.SG abc:M.SG that the:F.SG a:F.SG es primera. La b segunda la c be-IND.PRS.3SG first:F.SG the:F.SG b second:F.SG the c:F.SG tercera por que desta orden no

114

Bert Cornillie et al.

third:F.SG because of this:F.SG order:F.SG not tiene que hazer el gramático. have-IND.PRS.3SG that-COMP do-INF the:M.SG grammarian:M.SG (Antonio de Nebrija. Gramática castellana. 1492) ‘Not like Saint Isidoro in alphabetical order, where the a is the first, the b the second, and the c the third because the grammarian should not work in this order.’ The case of Spanish haber de and tener que illustrates that desemanticisation clearly can take place to different degrees not only between different languages, but also between different elements of the same language – which is, of course, no surprise. Interestingly, some constructions that have evolved from constructions implying habere not only express obligation, but also have a future reading and get an epistemic interpretation. Diachronic corpus data from the on-line Corpus del español (compiled by M. Davies) show that in Old Spanish haber de yields also futural (12a) and epistemic/inferential readings (12b) (see Cornillie 2007: chapter 7).11 (12) a. Dize que los moros todos tell-IND.PRS.3SG COMP the maur:M.PL all:M.PL han de yr a parayso & have-IND.PRS.3PL to go-INF to paradise:M.SG and han de comer mjel & leche & have-IND.PRS.3PL to eat-INF honey:F.SG and milk:F.SG and manteca & buñuelos & han de butter:F.SG and cookie:M.PL and have-IND.PRS.3PL to auer muchas moças a lot:F.PL girl:F.PL have-INF (Sancho IV. Castigos y documentos para bien vivir. 13th century) ‘One says that all Muslims will go to paradise and will eat honey and milk and butter and cookies and will have lots of young girls.’ b. Sy es cordura o if be-IND.PRS.3SG common sense:F.SG or se me posieron enel REFL I:DAT put-IND.PRET.3PL in the:M.SG creo que han

locura […] madness:F.SG coraçon & heart:M.3SG and de ser

Modals in Romance languages

115

think:IND.PRES.1SG COMP have-IND.PRES.3PL to be-INF verdaderas ca todo es enel because all:M.SG be-IND.PRES.3SG in the:M.SG real:F.PL poder de dios del pobre fazer power:M.SG of god:M.SG of the:M.SG poor:M.SG do-INF Rico & del rrico pobre. rich:M.SG and of the:M.SG rich:M.SG poor:M.SG (Anonymous. Libro del caballero Zifar. 14th C) ‘If it is common sense or madness that were put in my heart and I think they must be real, since God has the power to make the poor rich and the rich poor.’ Unlike haber de, the impersonal construction hay que is restricted to a deontic reading. This may again suggest that the grammaticalisation process does not concern the verb as such, but rather the whole construction of verb and preposition/complementiser, an idea that is more and more accepted nowadays. 3) Verbs meaning ‘to need’ In Romanian, the most common verb expressing deontic modality is a trebui, which has evolved from the Slavic verb trƟbovati ‘to need’, ‘to make use of’, ‘to make a sacrifice’ (Avram 1999: 220, Ciorănescu 2001: 802). When a trebui is followed by a subjunctive, it can have both a deontic (13) or an epistemic (14) reading. (13) Trebui-e must-IND.PRS.3SG ‘I have to leave.’

să plec. that leave-SBJV.PRS.1SG

(14) Dunărea trebui-e să Danube:ART must-IND.PRS.3SG that ‘The Danube must be near.’

fi-e aproape. be-SBJV.PRS.3SG near

It can also be followed by a past participle (15) or a supine (16), but only in its deontic use: (15) Trebui-e căuta-t-ă altă must-IND.PRS.3SG look_for-PST.PTCP other ‘We have to look for another explanation.’

explicaĠie. explanation-F.SG

116

Bert Cornillie et al.

(16) Trebui-e reacĠiona-t must-IND.PRS.3SG react-SUP ‘We must react calmly.’

cu calm. with calm

Trebuie + subjunctive can be compared to French il faut que, which comes from Classical Latin fallere ‘deceive, be lacking’. In Old French, the verb had two different uses, according to Herslund (2003). With an agentive subject, it meant ‘to miss, not to reach’, as in (17).12 (17) Et tex prendre qui and such take-INF that ‘And take what is missed.’

faut. (Erec 2939) be_lacking-IND.PRS.3SG

With an inanimate, non agentive subject (and without an object), it meant ‘to end, to run out’, as in (18). (18) A tant faillirent les paroles. (Erec 694) alors be_lacking-IND.AOR.3PL the:F.PL word:F.PL ‘With that many [words], the words stopped.’ However, when the subject was not an entity that runs out or has a natural endpoint by itself, the meaning may shift to ‘not to be available’. In this sense, the verb can be followed by a dative complement: (19) Garde qu’ il ne li (the horse) take care that it not him faille riens. (Erec 454) be_lacking:SBJV.PRS.3SG anything ‘Make that it (the horse) does not lack anything.’ When such a construction of faillir with a dative complement is used in a negative context, the verb can be followed by an infinitive (cf. Herslund 2003: 69): (20) Il ne vos faut it only (=ne … que) you be_lacking-IND.PRES.3SG que commander. (Meraugis, cf Kjellman 1913: 90). that command-INF ‘The only thing that remains for you [to be done] is commanding.’

Modals in Romance languages

117

The literal meaning of this sentence can easily be reinterpreted as expressing obligation, as in contemporary French: (21) a. Il ne faut pas it not (=ne…pas) be_necessary-IND.PRS.3SG la bouche pleine. the:F.SG mouth:F.SG full:F.SG ‘One should not speak with a full mouth.’ b. Il faut faire vite, it be_necessary-IND.PRS.3SG do quick des cas pareils. DET.INDF.M.PL cas:M.PL similar:M.PL ‘One should be quick, in similar cases.’

parler speak

dans in

Moreover, just as is the case for a trebui, it can be argued that il faut develops an epistemic interpretation, as in (22): (22) Dieu sait que les voitures God know-IND.PRS.3SG that the:F.PL car:F.PL ne manquaient pas à Luchon! Il not (=ne…pas) be_lacking-IND.IPF.3PL at Luchon! It fallait être une Fondaudège be_necessary-IND.IPFV.3SG be-INF a:F.SG Fondaudège pour y avoir amené to there have-INF bring_along-PTCP.PRF.M.SG son équipage. his:M.SG coach:M.SG (Mauriac, Nœud de vipères. 1932: 44) ‘God knows that Luchon had a lot of cars! One had to be a Fondaudège to bring one’s coach along.’ 4) Other sources Other verbs expressing deontic modality are even more restricted since they can only be used in a limited list of forms or contexts. We give the example of Italian impersonal verbs bisogna ‘must’, occorre ‘must’ and conviene ‘ought’ followed by a bare infinitive in (23), andare + past (passive) participle as in (24), and Spanish caber (25).

118

Bert Cornillie et al.

(23) a. Bisogn-a parl-ar-ne subito. must-IND.PRS.3SG talk-INF-of_it immediately ‘One must talk about it immediately.’ b. Occorr-e parl-ar-ne subito. must-IND.PRS.3SG talk-INF-of_it immediately ‘One has to talk about it immediately.’ c. Convien-e parl-ar-ne subito. fit-IND.PRS.3SG talk-INF-of_it immediately ‘One ought to talk about it immediately.’ (24) La casa v-a distrutt-a. the:F.SG house:F.SG go-IND.PRS.3SG destroy:PRF:PTCP-F.SG ‘The house has to be destroyed.’ (25) Cabe destruir la casa. fit-IND.PRS.3SG destroy-INF the:F.SG house:F.SG ‘It is necessary/convenient to destroy the house/ the house has to be destroyed.’ In the construction in (24), the verb andare loses its primary directional meaning. Rather, this meaning is interpreted in a more abstract way, namely to express an abstract future that yields a deontic meaning when it is combined with the passive (past) participle. Although standard Italian generally disallows andare a + infinitive as a periphrastic future (perhaps due to purist interventions in the 18th century, cf. Migliorini 1966), there are clear cases where it does have this futural reading, as in lo spettacolo va a cominciare ('the show is about to start') and andrà a finire che si faranno male ('in the end they will hurt each other'). Hence, there is a cline from movement over future to deontic modality. The case of Spanish caber is somewhat different, in that the most frequent reading of this verb refers to a process of fitting in a volume of space. In the deontic modal reading in (25), the volume reading is completely desemanticised and replaced by a general necessity reading. Note that this reading only appears in third person impersonal constructions. Again, these cases raise the question whether for these verbs the deontic interpretation is a separate meaning or whether it is an interpretation induced by the context. This in turn poses the problem of the limitation of

Modals in Romance languages

119

the group of verbs studied here: one could perhaps take into account French il convient que/de, Spanish conviene + infinitive and Romanian se cuvine/cade ‘to be proper’ + subjunctive. 5) Conclusions for desemanticisation All forms and constructions have undergone desemanticisation. There are in fact two desemanticisation processes. First, there is the development of a more abstract meaning from the initial, pre-modal, more concrete meaning; most deontic modals in Romance languages seem to come from six lexical sources:13 (i) debere ‘to owe’, (ii) habere and verbs meaning ‘have’ or ‘hold’, (iii) essere ‘to be’, (iv) a trebui, falloir and bisognare meaning ‘need’, (v) andare, occorrere and convenire which express movement, (vi) caber ‘fit into a volume of space’. Second, there is the development of the epistemic meaning from the deontic one. The meaning changes involved can be described using known grammaticalisation paths (as described e.g. by Heine 1993; Bybee, Perkins and Pagliuca 1994; Kuteva 2001) and invited inferences. They give rise to polysemic verbs, since the original, more concrete meanings, are still available alongside the grammaticalised, more abstract ones. As pointed out by Lehmann (2002), this does not mean that they might not have a Gesamtbedeutung; in fact, their meaning can probably best be described using a Langackerian network model, combining the possibility of one basic meaning with the existence of different, more precise, conventional meanings (see also Kronning 1996: 30, 90-121).14 Yet, more recent modal constructions lacking a whole paradigm, such as the ones with Italian andare and Spanish caber, indicate that the network representation need not apply to all verbs, and, more generally, that this representation requires a detailed description of the origins of each construction. Both processes of meaning change can have taken place to variable degrees: (i) the degree to which the desemanticisation has already proceeded is variable from language to language and from form to form and (ii) not all forms develop an epistemic meaning. This can probably be explained if one accepts the idea that it is not the verb alone that undergoes desemanticisation, but the verb in a particular construction. Moreover, change takes place in a particular context, as is proposed by Traugott and Dasher’s (2004) analysis of semantic change in terms of invited inferences. Finally, it has been shown that the degree of desemanticisation gives an

120

Bert Cornillie et al.

indication as to the degree to which a form or a construction has grammaticalised. An example in point would be the difference between Italian avere da + infinitive and bisogna + infinitive: in the former, we have shown that desemanticisation interacts with the syntagmatic axis (constructional frames), yet for avere da + infinitive the original ‘have’ construction is also allowing a possessive interpretation, while the original personal ‘need’ construction for bisognare has been completely lost. Bisogna can therefore be said to be more desemanticised than the avere + da construction. 3.1.2. Attrition Interestingly, none of the above-discussed forms seems to have undergone attrition to an important degree. This may point to the need for more detailed analyses of the pronunciation of these forms in real discourse, but it may also suggest that these modal verbs are still more autonomous elements than e.g. temporal auxiliaries such as English going to that can be shortened into gonna (cf Krug 2000). It may also be taken as an indication that desemanticisation and attrition do not necessarily go hand in hand. 3.2. Cohesion/paradigmaticity According to this parameter, grammaticalised expressions tend to form small, tightly integrated paradigms. In most traditional grammars, modal verbs are characterised as auxiliaries or semi-auxiliaries. In this section, we will examine whether the modal forms or constructions discussed here function as auxiliaries, and also whether they form paradigms. Thus, we will first describe the criteria relevant to determine the auxiliary status of verbs in Romance languages; these criteria are different from those used to determine the auxiliary status of verbs in Germanic languages. 3.2.1. Cliticisation Cliticisation can be considered a measure of grammaticalisation, because the movement of a clitic over an entire verbal complex shows cohesion of the modal plus main verb.

Modals in Romance languages

121

According to Jones (1996: 80-87), examples such as those under (26) show that the clitic pronoun cannot precede pouvoir and devoir in French, contrary to what is the case with être and avoir: (26) a. Jean l’ a résolu John it have-IND.PRS.3SG solve-.PTCP.PRF.M.SG (* a le résolu) ( has it solve-PTCP.PRF.M.SG ‘John has solved it.’ b. Pierre y est arrivé. Peter there be-IND.PRS.3SG arrive-PTCP.PRF.M.SG (* est y arrivé) (* be-IND.PRS.3SG there arrive-PTCP.PRF.M.SG) ‘Peter has arrived there’ / ‘Peter managed.’ c. Pierre peut/doit le résoudre. Peter can/must-IND.PRS.3SG it solve-INF ‘Peter can/must solve it.’ He concludes that devoir and pouvoir are not auxiliaries to the same extent as être and avoir. The fact that it is impossible to put the pronoun before the modal verb indeed seems to imply, amongst other things, that there really are two clauses in sentences containing devoir. In that sense, Italian dovere and Spanish deber would be somewhat further on the grammaticalisation/auxiliation chain, since they allow the clitic pronoun to be preposed. However, this conclusion should not be drawn too hastily: as Jones (1996: 82) also points out, forms like Pierre le peut résoudre were perfectly acceptable in Old and classical French. By contrast, Spanish had debe lo hacer before, but nowadays only admits lo debe hacer or debe hacerlo ‘(s)he has to do it’. Although more detailed research is clearly needed, this might suggest that French has been further down the line on the grammaticalisation chain than Italian and Spanish, but that it now climbs up the cline – a case of de-auxiliation, degrammaticalisation or, perhaps, lexicalisation. In any case, the fact that Italian dovere permits clitic-climbing suggests that it is more comparable to auxiliaries such as essere ‘to be’ and avere ‘to have’ than its actual French equivalent. However, this is less so for other modal expressions expressing obligation, such as those derived from

122

Bert Cornillie et al.

habere. For a start, in Italian, clitic elements that depend on the infinitive cannot climb towards avere and clitics that depend on avere cannot be lowered onto the infinitive, as shown in (27): (27) a. Ne

ho

ancora

tre

da correggere.

[from: ho ancora tre compiti da correggere]

of_it have-IND.PRS.1SG still ‘I still have to correct three of them.’ b. Ho

ancora

three to

correct-INF

da correggerne

tre.

[from: ho ancora da correggere tre compiti]

have-IND.PRS.1SG still c. *Ne of_it

to

correct-INF-of_it

ho ancora have-IND.PRS.1SG still

d. *Ho have-IND.PRS.1SG

ancora still

tre three

three

da correggere tre. to correct-INF three da correggerne.15 to correct-INF-of_it

With the impersonal verbs bisogna, occorre and conviene, no clitic climbing seems to be allowed. Neither is clitic climbing possible with the French avoir à construction (28). (28) a. *J’en ai encore I.of_it have-IND.PRS.A.SG still

à to

corriger correct

trois three

[from j'ai encore à corriger trois examens]

b. *Je l’ai I him.have-IND.PRS.A.SG

encore still

à convaincre to convince

[vs. OK J’ai encore à le convaincre]

By contrast, Spanish haber de and tener que readily allows for the two positions of the clitic (29). In combination with the clitic climbing with deber, this points to a clear increase of paradigmaticity or “auxiliarity”. In Romanian as well, clitics climb up to a avea (30). (29) a. Y no lo {han de/tienen que} and not it have-IND.PRS.3PL to/that-COMP por altruismo o ética sino because altruism:M.SG or ethics:F.SG but

hacer do-INF por because

Modals in Romance languages

123

supervivencia. survival:F.SG ‘And they don’t have to do it for altruistic or ethical reasons but for surviving.’ b. Y no {han de/tienen que} hacerlo por and not have-IND.PRS.3PL to/that-COMP do-INF-it because altruismo…. altruism:M.SG (30) Astea sunt problemele pe care le these:F.PL be-IND.PRS.3PL problems:ART that REL ACC.F.PL mai avem de rezolva-t. still have-IND.PRS.1PL to solve-PST.PTCP ‘These are the problems that we still have to solve.’ However, clitic climbing may not be the best means to detect (modal) auxiliaries, since it also applies to other verbs, e.g. the Spanish causatives dejar ‘let’ and hacer ‘make’, which lack coreference between the subject of the finite verb and the subject of the infinitive. 3.2.2. Loss of subcategorisation frames Lamiroy (1999), following Heine (1993), points out that, in processes of grammaticalisation, verbs tend to lose their original subcategorisation frames and become more associated with more abstract complements referring to situations, expressed by non tensed verb forms, such as infinitives. This idea is confirmed by most verbs studied here: debere was originally a ditransitive verb, avoir has a NP complement when it is not used to express deontic modality, etc. More in general, when verbs grammaticalise, the verbal complement tends to inherit the major syntactic properties that the first verb has lost. This can be seen in Italian in the choice of auxiliary. In (31) the compound past tenses in the dovere + infinitive construction can have the auxiliary of dovere itself (avere), or the auxiliary selected by the infinitive. (31) a. Dopo after

24 ore l’ex Presidente er-a 24 hour:F:PL the:SG ex president:SG be-PST.IPFV.3SG

124

Bert Cornillie et al.

dovut-o

and-are in ospedale

must-PFV:PTCP-M:SG go-INF

per sottopo-r-si

in hospital:SG to

submit-INF-REFL

ad un intervento chirurgico piuttosto to a:M:SG operation:M:SG surgical:M:SG rather delicato. delicate:M:SG ‘After 24 hours, the ex president had to go to hospital for a rather delicate operation.’ b. Un’escursione che av-rebb-e dovut-o a:F:SG excursion:SG that have-COND.3SG must:PFV:PTCP-M:SG conclud-er-si in giornata. conclude-INF-REFL in day:F:SG ‘An excursion that should have ended that day.’ c. Un compromesso con Maranghi che sa-rebb-e a:M:SG compromise:M:SG with Maranghi that be-COND-3SG dovut-o culmin-are in una presidenza must-PFV.PTCP.M:SG culminate-INF in a:F:SG presidency:F:SG di Francesco C. of Francesco C. ‘A compromise with M. that should have culminated in the presidency of F.C.’ In French, Spanish and Romanian, this possibility does not exist. In Spanish and Romanian, the auxiliary used for forming compound past tenses is always haber and a avea, respectively. Hence, it is only for French that this parameter may be taken to suggest that the modal verbs are less grammaticalised than the Italian ones. The other verbs that can take a bare infinitival complement always retain their proper auxiliary (as in their use as main verb). The fact that the Italian modal verbs (dovere, potere) do admit the auxiliary selected by the infinitive explains that they are generally considered as modal auxiliaries, rather than merely periphrastic constructions. This indicates a greater degree of paradigmaticity. Lamiroy (1999: 38) considers the selection of an infinitival complement rather than que P to be the distinguishing element par excellence of Romance auxiliary verbs. From this point of view, one could argue that devoir is an auxiliary verb, as are Spanish deber, haber de, tener que and

Modals in Romance languages

125

Italian dovere.16 However, this criterion would exclude from the same (auxiliary) paradigm verbs and/or constructions such as French falloir, Italian andare and bisogna, which are not invariably followed by an infinitive. Moreover, although Italian avere da + infinitive, essere da + infinitive, French avoir à and Spanish haber de do take an infinitival complement, they are still compositionally transparent and, consequently, probably only weakly grammaticalised. Romanian a trebui and a avea are never followed by an infinitive, but combine with other non-tensed verb forms such as a supine and in the case of a trebui (but not a avea) with a past participle. Another indication that may point to an ongoing process of grammaticalisation of a trebui, is the reduction in size of its own paradigms. Being used initially as an impersonal verb (thus having only the 3rd person singular (trebuie in the Indicative Present)), it gradually began to agree with the subject of the sentence, which resulted in the development of a parallel paradigm with suffixed forms (Indicative Present: trebuiesc, trebuieúti, trebuieúte, trebuim, trebuiĠi, trebuiesc). After 1940, this paradigm of ‘personal’ forms tended to disappear again (see Avram 1999: 220-222). These observations seem to confirm the widespread feeling that auxiliaries form a category that is difficult to define, at least when using exclusively morphological or syntactic criteria. This feeling explains why some grammarians prefer to refer to the modal verbs as semi- or quasiauxiliaries, suggesting that French verbs such as devoir, pouvoir, aller are not auxiliaries to the same extent as être or avoir. Auxiliation can be considered a fundamentally dynamic phenomenon, which is to be studied not from a synchronic, but from a panchronic perspective, where synchrony is seen as a reflection of language history. Auxiliaries can then be defined as “linguistic items covering some range of uses along the Verb-to-TAM chain” (Heine 1993: 70—the TAM chain being the grammaticalisation chain corresponding to the formation of Tense-Aspect-Modality markers). This might further point to the conclusion that auxiliaries are not to be defined in purely syntactic terms and that the category might be conceptually based (Kuteva 2001); thus Jones (1996) concludes his analysis of devoir and pouvoir by saying that these verbs are not syntactic auxiliaries, but that “this conclusion does not preclude the possibility that all of these verbs [devoir, pouvoir, aller, venir de] (and perhaps some others) form a coherent class of ‘semantic’ auxiliaries (e.g. verbs which express notions of time and modality rather than describing events or states) and that some syntactic processes are

126

Bert Cornillie et al.

sensitive to these semantic properties”. This conclusion is confirmed for Romanian: the Gramatica limbii române (1963: II, 98) does not consider modal verbs as a clearly delineated grammatical category, precisely because “the fundamental and almost only delimitation criterion is the semantic one” (cf. Avram 1999: 165-166). In sum, modal verbs may tend to form closed paradigms, but the degree to which this tendency toward paradigm formation is carried through is different from one language to another. The paradigms formed by modal verbs can be seen as part of a larger process by which languages turn lexical forms into auxiliaries. However, they are better seen as open and dynamic paradigms, their members being situated at various points on the grammaticalisation cline. In line with the desemanticisation processes, there is no sharp boundary between lexical and more grammaticalised modal verbs. 3.3. Variability/paradigmatic variability 3.3.1. Transparadigmatic variability Unlike temporal/aspectual auxiliaries and other tense/aspect markers that are grammatically required for certain temporal and aspectual oppositions in Romance languages, modal auxiliaries have no grammatical motivation, but rather depend on speaker’s intentions and choice. Their absence does not revert into a default modal interpretation of the main verb: absence/presence of a modal auxiliary signals absence/presence of modal reading of “obligation” or “necessity”. 3.3.2. Intraparadigmatic variability If one accepts that the paradigms are mainly founded on semantic criteria, it is possible to describe in detail the semantic elements that oppose the different verbs expressing obligation in a language to one another, as well as their (possible) syntactic consequences. Since Italian is the Romance language where the tendency to develop a paradigm is perhaps the most pronounced, we will illustrate this parameter for Italian in Table 1 (and see Pietrandrea 2005 for an account of epistemic modality in Italian).

127

Modals in Romance languages Table 1. Intraparadigmatic variability with Italian deontic modals Dovere

avere da

essere da

andare

bisogna conviene occorre

non-epistemic, participant-external, non-deontic

yes

yes

yes

no

yes

non-epistemic, participant-external, deontic

yes

yes

yes

yes

yes

non-epistemic, participant-internal

yes

yes

does not apply

no

no

One could hypothesise at this point that the apparent interchangeability of the forms is not a counter-argument to the grammaticalisation of these verbs, but could rather be seen as a way in which the individual verbs tend to loose their idiosyncratic properties (semantic, syntactic) in favour of a generalised distribution of a similar set of properties or interpretations within the (newly created) paradigm. This could perhaps be taken as an intermediate step that could lead towards further specifications for some verbs, and overlap for others. Table 1 illustrates considerable intraparadigmatic differences. Whereas all constructions can express non-epistemic participant-external deontic modality, the other modal readings are found with only some of the verbs. 3.4. Weight/structural scope “The (…) structural scope of a grammatical means is the structural size of the construction which it helps to form” (Lehmann 2002: 128); during a grammaticalisation process, an expression tends to reduce its grammatical scope. Thus, the Latin demonstratives were equivalent to NPs, but the definite articles that were derived from them became determiners inside the NP. It can easily be seen that when verbs turn into modals, they also reduce their structural scope. In fact, this follows in a way from the general tendency of modals to change their subcategorisation frames as they become more and more grammaticalised: whereas verbs such as dovere, devoir and deber were originally ditransitive verbs (and still are when they are used as lexical verbs)

128

Bert Cornillie et al.

and could take SN complements, they progressively become associated with the infinitive and lose their autonomy in their modal value. Clitic climbing may be an indication of reduced scope. Spanish haber de, haber que, tener que, Romanian a avea de, Italian avere da / andare + past participle and French avoir à clearly reduce their structural scope as modals when compared to their use with their possessive meaning. In present-day Spanish haber, unlike tener, does not have a transitive reading anymore, and instead reduces its structural scope to a prepositional complement: (32) a. *He tres libros de escribir. have-IND.PRS.1SG three book:M.PL to write-INF He de escribir tres libros. write-INF three book:M.PL have-IND.PRS.1SG to ‘I have three books to write / I have to write three books.’ b. Tengo have-IND.PRS.1SG Tengo have-IND.PRS.1SG

tres libros que three book:M.PL COMP que escribir tres COMP to write-INF three

escribir. write-INF libros. book:M.PL

However, the evolution has not always reached its endpoint, since avoir à can still be used in constructions such as in (33). (33) a. J’ ai trois enfants à I have-IND.PRS.1SG three child:M.PL to ‘I have three children to feed.’

nourrir. feed-INF

b. Il a deux maisons à entretenir. he have-IND.PRS.3SG two house:F.PL to maintain-INF ‘He has two houses to maintain.’ This also holds for falloir, tener que, avere da and a avea de which can be used in structures where they are still followed by an NP. When a verb allows both reduced and wide scope, the general tendency is that the scope is more reduced in the most clearly deontic meanings than in its more lexical, possessive interpretation. This can be illustrated by the following examples of the Italian construction avere da + infinitive.

Modals in Romance languages

129

(34) Ho da stirare camicie. have-IND.PRS.1SG to iron-INF shirt:F.PL ‘I have to iron shirts.’ (35) C’ho da fare. LOC-have-IND.PRS.1SG to do-INF ‘I have to do something.’ (36) Non ho da fare nulla oggi. not have-IND.PRS.1SG to do-INF nothing today ‘I don’t have to do anything today.’ Indeed, the clearer cases of deontic interpretation are those where avere loses its selectional capacity, and is used intransitively (while the infinitive takes – or does not take – a complement). In these cases, the possessive reading is absent; the scope of the negation is also more on the combined meaning of avere + da + infinitive: compare (36) to (37): in (36) above the sentence implies that there’s no obligation to do anything; in (37) below the sentence states that there isn’t anything that you could do. (37) Non ho nulla da fare oggi. not have-IND.PRS.1SG nothing to do-INF today ‘I don’t have anything to do today.’ The tendency is also confirmed by andare + past participle. The structural scope of the auxiliary is typically that of an auxiliary: it precedes an uninflected verb form (participle) creating a periphrastic construction with specific bondedness properties. 3.5. Cohesion/bondedness The idea here is that less grammaticalised forms are juxtaposed, whereas more grammaticalised forms cannot really be used independently and tend to become affixed. Thus, we mainly checked whether the modals analyzed here could be used alone. There is some variation in this respect, but the general tendency is that when a form has both a deontic and an epistemic interpretation, the form can only be used alone in its deontic reading. This comes as no surprise, of

130

Bert Cornillie et al.

course, if it is true that the form is more grammaticalised when it has an epistemic than when it has a deontic meaning. However, there is some variation: (i) tener que, a trebui, devoir, falloir, dovere and deber can be used alone, but only in their deontic reading, not in their epistemic use (if they have one); (ii) haber de, haber que, avoir à, a avea de cannot be used alone; (iii) in the same line, the modal interpretation of avere da and essere da typically hinges on a continuous verb + da + infinitive construction without interruption; (iv) andare + past participle, finally, is more tightly related than dovere; its past participle cannot be dislocated or omitted. 3.6. Variability/syntagmatic variability According to this parameter, more grammaticalised items can be less freely moved around in the sentence than less grammaticalised items. Moreover more grammaticalised forms tend to show up in fixed constructions – this of course again confirms the idea that constructions and not lexical items are grammaticalised. This parameter can easily be applied to the verbs that have evolved from debere. In Latin the auxiliary verb debere could either be preceded by the infinitival construction, or be followed by it. In French, Italian, Spanish and Romanian, infinitives have followed the general shift from left-branching to right-branching that took place in the evolution from Latin to modern Romance. Thus, modal dovere, devoir and deber are always used with an infinitive to their right. The same goes for avoir à, haber de, haber que, tener que, which always precede the infinitive, and for a avea de, which always precedes the supine. This has not always been the case, however: in Old, Middle and Classical Spanish, for instance, the modals are very often attested following the infinitive, as in (38) (38) Esto es saber hacer lo this be-IND.PRS.3SG know-INF do-INF it hacer debo. do-INF must-IND.PRS.1SG (Cadalso, José. Solaya o los circasianos. 1762) ‘This is knowing to do what I have to do.’

que COMP

Modals in Romance languages

131

4. Conclusions The modal verbs of obligation in the four Romance languages analyzed here confirm some general trends of grammaticalisation research: The modal forms are the result of a desemanticisation process in two ways. In accordance with the general definition of grammaticalisation, they have evolved from verbs with a more lexical meaning. The status of the deontic interpretation is not exactly the same for all items: whereas for debere the deontic meaning can be considered a separate meaning of the verb, it is rather a constructional meaning in the expressions derived from habere and probably at first a contextually induced reinterpretation for the verbs meaning ‘to need’. Some Romance deontic modals í but not all í also get a more grammaticalised, more subjective epistemic meaning. In order to explain why some modals get an epistemic meaning whereas others do not, it is probably necessary to start from the idea that grammaticalisation processes do not concern lexical items, but constructions or lexical items in particular constructions. The verbs in question change their subcategorisation frames: whereas most of them originally have NP-complements, these tend to be “replaced” by non-tensed infinitive complements and the modal verbs tend to turn increasingly into full fledged auxiliaries. It is less clear, however, whether these verbs also show signs of phonological erosion. In any case, none of these forms seems to be on its way to become affixed to the following verb form. These observations allow us to answer some of the more general questions discussed in this book: 1. Since most of the elements summed up above are exemplified to very different degrees by the verb forms under study, it is not really adequate to hold that Romance modals form paradigms; in Italian there seems to be a certain tendency to build a paradigm, mainly from a morphological point of view. There is, however, not really a morphologically or syntactically based paradigm of deontic modals. This might be due to the fact that modal auxiliaries have no grammatical motivation comparable to that of temporal/aspectual auxiliaries and other tense/aspect markers; as pointed out above, they largely depend on the speaker’s intention and choice. 2. It follows from all of this that, in our view, it is difficult if not downright impossible to define the class of (deontic) modals in terms of necessary

132

Bert Cornillie et al.

and sufficient conditions; the only definition that is possible would combine two elements: (i) a semantic characterisation of the class as certain uses of verbs or certain constructions that express the notion of obligation and, in their epistemic use, the notion of probability or certainty; (ii) a more syntactic categorisation of these verb uses as exemplifying one of the stages on the grammaticalisation cline. 3. Of course, this also implies that there are no strict borderlines enabling us to distinguish modals from the full lexical verbs they are derived from. The criteria described above can be used for a typology of the deontic modals, but it is to be expected that every verb will only display some of them. Notes 1. In this chapter we will use the classic terms of “deontic” and “epistemic modality”. For a detailed overview of the current terminology, cf. Nuyts (2005, 2006) and de Haan (2006). 2. Interestingly, Olbertz (1998: 246) suggests that the reason why epistemic examples in Old Spanish texts are less frequent than those in Latin lies in the type of texts these constructions appear in. Many Latin texts discuss philosophical problems, while the medieval Spanish texts are most often narratives. 3. The two senses are not to be attributed to two homonyms (an idea that has been defended for French, see, for instance, Sueur (1975, 1979)). Kronning (1996: 92-94) cites a whole series of unrelated languages that all have morphemes expressing both deontic and epistemic modality, an observation that in itself is sufficient to plead for an approach in terms of polysemy. (Kronning 1996:15, n. 2, signals that Sueur (1983) also adopts a polysemy approach for devoir and pouvoir). 4. Bybee, Perkins and Pagliuca (1994: 200-201) also point out that the obligation sense does not imply the certainty sense and that it is thus difficult to accept that the conventionalisation of implicatures would be the source of the epistemic sense, as proposed by Traugott (1989). Sweetser (1990) also analyzes the development of the epistemic sense as a metaphorical transfer from the deontic sense. For a criticism of her account, cf. e.g. Pelyvás (1996) and Traugott and Dasher (2002: 111, 119). 5. Invited inferences are inherently pragmatic extensions of the default reading of an expressions. They are implicatures “that arise in language use”, when “the speaker / writer (…) evokes implicatures and invites the addressee / reader (…) to infer them” (Traugott & Dasher 2002: 5). 6. Cf. Huot (1974), Sueur (1975, 1979), Kjærsgaard (2003).

Modals in Romance languages

133

7. Kronning (1996: 103), for instance, points out that French devoir also developed an alethic sense of which the epistemic one has been derived (for a discussion, also see Squartini, 2004, and Vetters, to appear). Moreover, there has always been a considerable degree of overlap between the different senses (e.g. Pelyvás’ (1996) reinterpretation of some of Traugott’s examples). 8. For a detailed argumentation in favour of Langacker’s model as an adequate representation of the polysemy of the French verb devoir, see Kronning (1996: 90-121). 9. Cadiot (1997: 69) notes indeed that à, in opposition to de presents the complement “projectivement à partir du verbe” and that the complement is seen (“saisi”) from the subject. 10. Grevisse (1993: §791, 1195) believes that the object, when it immediately follows avoir, is indeed its object; nevertheless, when the noun phrase does not immediately follow avoir, it is frequently analyzed as the object of the infinitive, as can also be seen from the fact that the past participle eu frequently stays invariable when the noun phrase precedes avoir, as in Quelque course que (...) il avait eu à faire ‘some shopping that he (…) had had to do’ (Gide, Faux-Monnayeurs, p.100, cited by Grevisse (2007: § 951 b) 5°, 1176). 11. The Corpus del español shows that haber de is very frequent from the 13th century until the second half of the 19th century, when tener que becomes increasingly more common. 12. This meaning later gives rise to the verb faillir (Herslund 2003). 13. The last two possible sources are not mentioned in the list of possible sources of modals of obligation of Bybee, Perkins and Pagliuca (1994: 182-183). 14. Kronning (1996: 90-121) thus rejects more “abstract” proposals (e.g. Kratzer (1978, 1981), and, for French, Meyer (1991)), that define an abstract, unitary meaning and consider all other readings as resulting from the combination of this abstract meaning with contextual elements. Also see Traugott and Dasher (2002: 15) for criticisms of this kind of approaches. 15. Corpus research (Italian newspaper La Repubblica from 1985-2000) gives some examples of longer clitic dependencies with avere da, as in ne ho da dire una che mi ha sconvolto ‘I have one (story) to tell that upset me'. 16. On the auxiliarity status of devoir, cf. also Huot (1974: 165-181). Cf. also Riegel, Pellat and Rioul (1994) and Grevisse (1993).

References Academia Română 2000 DicĠionarul Limbii Române (1913-2000) [Dictionary of the Romanian language]. Bucharest: Academia Română.

134

Bert Cornillie et al.

Academia Română 1963 Gramatica Limbii Române [A Grammar of the Romanian language]. Bucharest: Academia Română. Avram, Larisa 1999 Auxiliaries and the structure of language. Bucharest: Editura UniversităĠii din Bucureúti. Benincà, Paola, and Cecilia Poletto 1994 Bisogna and its companions: the verbs of necessity. In Paths towards Universal Grammar, Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, and Raffaella Zanuttini (eds.), 35-57. Washington: Georgetown University Press. Benincà, Paola, and Cecilia Poletto 1996 The diachronic development of a modal verb of necessity. Venice Working Papers in Linguistics 6/2: 1-30. Bolkestein, Machtelt 1980 Problems in the description of modal verbs. An investigation of Latin. Assen: Van Gorcum. Bybee, Joan, Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: University of Chicago Press. Cadiot, Pierre 1997 Les prépositions abstraites en français. Paris: Colin. Ciorănescu, Alexandru 2001 DicĠionarul etimologic al limbii române [Etymological dictionary of the Romanian language]. Bucharest: Saeculum. Cornillie, Bert 2007 Evidentiality and Epistemic modality in Spanish (semi-)auxiliaries. A cognitive-functional account. Berlin/New York: Mouton de Gruyter. de Haan, Ferdinand 2006 Typological approaches to modality. In The Expression of Modality, William Frawley (ed.), 27-69. Berlin/New York: Mouton de Gruyter. Desclés, Jean-Pierre 2003 Interactions entre les valeurs de pouvoir, vouloir, devoir. In Aspects de la modalité, Merete Birkelund, Gerhard Boysen, and Søren Kjærsgaard (eds.), 49-66. Tübingen: Niemeyer. Gougenheim, Georges 1929 Etude sur les périphrases verbales de la langue française. Paris: Nizet. Grevisse, Maurice and Goosse, André 2007 Le bon usage. Grammaire française [14th edition]. Paris / Louvainla-Neuve: DeBoeck Université.

Modals in Romance languages

135

Guéron, Jacqueline 2000 From need to necessity. A syntactic path to modality, Belgian Journal of Linguistics 14: 63-87. Heine, Bernd 1993 Auxiliaries, cognitive forces, and grammaticalization. New York: Oxford University Press. Herslund, Michael 2003 Faillir et falloir: la creation d’opérateurs modaux. In Aspects de la modalité, Merete Birkelund, Gerhard Boysen, and Søren Kjærsgaard (eds.), 67-73. Tübingen: Niemeyer. Hopper, Paul J. 1991 On some principles of grammaticization. In Approaches to Grammaticalization, Elizabeth Traugott, and Bernd Heine (eds.), 1735. Amsterdam/Philadelphia: John Benjamins. Huot, Hélène 1974 Le verbe devoir. Etude synchronique et diachronique. Paris: Klincksieck. Jones, Michael 1996 Foundations of French Syntax. Cambridge: Cambridge University Press. Kjellman, Hilding 1913 La construction de l’infinitif dépendant d’une locution impersonnelle en français des origines au XVe siècle. Uppsala: Almqvist & Wiksell. Kjærsgaard, Søren 2003 Les equivalents danois de la construction devoir/pouvoir + infinitif. In Aspects de la modalité, Merete Birkelund, Gerhard Boysen, and Søren Kjærsgaard (eds.), 89-111. Tübingen, Niemeyer. Kratzer, Angelika 1978 Semantik der Rede. Kontexttheorie – Modalwörter – Konditionalsätze. Königstein/Taunus: Scriptor. Kratzer, Angelika 1981 The notional category of modality. In Worlds, Words and Contexts. New Approaches in Word Semantics; H.-J. Eikmeyer, and H. Rieser (eds.), 38-74. Berlin/New York: Mouton de Gruyter. Kronning, Hans 1996 Modalité, cognition et polysémie: sémantique du verbe modal devoir. (Studia Romanica Upsaliensia 54) Uppsala: Acta Universitatis Upsaliensis. Kronning, Hans 2001 Pour une tripartition du modal devoir. Cahiers Chronos 8: 67-84.

136

Bert Cornillie et al.

Krug, Manfred G. 2000 Emerging English Modals: A Corpus-Based Study of Grammaticalization. Berlin/New York: Mouton de Gruyter. Kuteva, Tania 2001 Auxiliation: An enquiry into the nature of grammaticalization. Oxford: Oxford University Press. Lamiroy, Béatrice 1999 Auxiliaires, langues romanes et grammaticalisation. Langages 135: 63-75. Lehmann, Christian 2002 Thoughts on Grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Meyer, Wolfgang 1991 Modalverb und Semantische Funktion. Wiesbaden: Franz Steiner. Migliorini, Bruno 1960 Storia della lingua italiana. Firenze: Sansoni. Nuyts, Jan 2005 The modal confusion. On terminology and the concepts behind it. In Modality: Studies in form and function, Alex Klinge, and Henrik Høeg Müller (eds.), 5-38. London: Equinox. Nuyts, Jan 2006 Modality: Overview and linguistic issues. In The Expression of Modality, William Frawley (ed.), 1-26. Berlin/New York: Mouton de Gruyter. Olbertz, Hella 1998 Verbal Periphrases in a Functional Grammar of Spanish. Berlin/New York: Mouton de Gruyter. Pelyvás, Péter 1996 Subjectivity in English: Generative Grammar versus the Cognitive Theory of Epistemic Grounding. Frankfurt am Main: Lang Pietrandrea, Paola 2005 Epistemic modality. Functional properties and the Italian system. Amsterdam/Philadelphia: John Benjamins. Riegel, Martin, Jean-Christophe Pellat, and René Rioul 1994 Grammaire méthodique du français. Paris: PUF. Squartini, Mario 2004 Disentangling evidentiality and epistemic modality in romance. Lingua 114 (7): 873-889. Sueur, Jean-Philippe 1975 Etude sémantique et syntaxique des verbes “devoir” et “pouvoir”, Université Paris X. These de 3e cycle.

Modals in Romance languages

137

Sueur, Jean-Philippe 1979 Une analyse sémantique des verbes devoir et pouvoir. Français moderne 47: 97-120. Sueur, Jean-Philippe 1983 Les verbes modaux sont-ils ambigus? In La notion sémanticologique de modalité, Jean David, and Georges Kleiber (eds.), 164180. Paris: Klincksieck. Sweetser, Eve E. 1990 From etymology to pragmatics. Metaphoric and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 65: 31-55. Traugott, Elizabeth Closs, and Richard Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality's semantic map. Linguistic Typology 2: 79-124. van der Auwera, Johan, and Andreas Ammann 2005a Epistemic possibility. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gill, and Bernard Comrie (eds.), 306-309. Oxford: Oxford University Press. van der Auwera, Johan, and Andreas Ammann 2005b Overlap between situational and epistemic modal marking. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gill, and Bernard Comrie (eds.), 310-313. Oxford: Oxford University Press. van der Auwera, Johan, and Andreas Ammann 2005c Situational possibility. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gill, and Bernard Comrie (eds.), 302-305. Oxford: Oxford University Press. Vetters, Carl to appear Les verbes modaux pouvoir et devoir en français.

5. Modals in Greek* Anastasios Tsangalidis

1. Introduction The term ‘Greek’ is used in this paper to refer to the current stage of the language, also known as (Standard) Modern Greek. According to most researchers this is practically the sole descendant of the Ancient Greek language 1 . In terms of general typological features Greek is a highly inflecting pro-drop language in which word order is quite free. Thus, the main typological characteristics of the language that will be relevant to the discussion below involve the fact that (i) all finite verb forms in Greek need to inflect for tense, aspect, mood, voice, number and person 2 and (ii) the freedom of word order means both combinations of V + Complement and Complement + V are equally acceptable. The paper will focus on the presentation of the modal categories in the language along the lines of Lehmann’s parameters of grammaticalisation. Modal categories in Greek arguably include (i) a morphological mood distinction, (ii) periphrastic combinations with three modal particles and (iii) a set of two special verbs of necessity and possibility. Given the assumptions in van der Auwera and Plungian (1998) and van der Auwera, Ammann and Kindt (2005), perhaps the main focus of attention should be on modal verbs and less so on modal particles, as the former have been discussed in greater detail in terms of the grammaticalisation of modality. It will be shown, however, that Lehmann’s parameters are successful in describing the relative degree of grammaticalisation of each of the three Greek cases: the modals of necessity and possibility will always appear to be less grammaticalised than the modal particles just as these, in turn, are less grammaticalised than the morphological mood distinction, precisely as predicted by historical, typological and theoretical considerations. The paper is structured as follows: the data are briefly presented in section 2 in terms of morphological, syntactic and semantic properties of all three modal categories. Section 3 briefly mentions the main formal and functional characteristics of lexical verbs of modality to be contrasted with the two modals of possibility and necessity. In section 4 each of Lehmann’s

140

Anastasios Tsangalidis

parameters is examined in some detail in relation to the Greek facts. Problems and conclusions are summarised in section 5. 2. The category of modals in Greek The modal categories systematically expressed in Greek can be grouped into three subsystems: (i) an inflectional mood distinction, which is best described as [+/-imperative]; (ii) a periphrastic [+/-subjunctive] distinction, which involves the use of three modal particles; and (iii) the modal verbs of necessity and possibility – which have been described as ‘semi-auxiliary’ (Tsangalidis 2004a, 2004b). As often noted in the general literature (e.g., Palmer 2001: 4, 21, 104; Bybee, Perkins and Pagliuca 1994: 181; de Haan 2006: 37), it is worth distinguishing between (inflectional) mood on one hand and modal systems (e.g. of auxiliary verbs, particles or affixes) on the other, precisely because a language may employ both – and the Greek case appears to be highly interesting in this respect. 2.1. The inflectional moods All verb forms in Greek can be identified as either [+] or [-imperative] on the basis of their distinct morphological marking. As seen in (1), imperative endings are distinct for active and non-active voice and attach to both perfective and imperfective stems. Indeed the perfective is the “unmarked” form of the imperative (and the only one available in the non-active voice), whereas the imperfective will mostly be used for giving an order for a repeated or continuous action (cf. Holton, Mackridge and PhilippakiWarburton 1997: 206). They only appear in second person, singular and plural, and they do not inflect for tense. Unsurprisingly, they do not appear in subordinate clauses or interrogative sentences. Finally, imperatives do not negate, and various versions of the periphrastic ‘subjunctive’ can be used as prohibitions, as shown in (2). (1)

a. Ȗrafe / Ȗrafete write-2SG-IMP-IPFV-ACT / write-2PL-IMP-IPFV-ACT ‘Write/Keep writing!’

Modals in Greek

141

b. Ȗrapse / Ȗrapste write-2SG-IMP-PFV-ACT / write-2PL-IMP-PFV-ACT ‘Write!’ c. (Ȗrafu) / (Ȗrafeste) write-2SG-IMP-IPFV-NACT / write-2PL-IMP-IPFV-NACT ‘Get written/Keep on getting written!’ (rare) d. Ȗrapsu / Ȗraftite write-2SG-IMP-PFV-NACT / write-2PL-IMP-PFV-NACT ‘Get written!’ (2)

a. *mi Ȗrafe NEG write-2SG-IMP-IPFV-ACT b. na/as/ø mi Ȗrafis / Ȗrapsis SUBJ/ZERO NEG write-2SG-IPFV-ACT / write-2SG-PFV-ACT ‘Don’t write’

Verbal endings in Greek are “portmanteau realisations” of various inflectional categories. As in all inflectional languages, a word is not wellformed without the presence of the appropriate ending. Therefore, it is always possible to classify any verb form as either imperative or nonimperative. 2.2. The modal particles and their periphrases All non-imperative finite verb forms can be accompanied by one of a set of three modal particles, thus forming the periphrastic opposition that is often described as ‘indicative’ vs. ‘subjunctive’ (e.g. Holton, Mackridge and Philippaki-Warburton 1997: 203-205; cf. Tsangalidis 2004b and references therein). According to this view (most recently presented in PhilippakiWarburton and Spyropoulos 2004), free verbal forms and forms accompanied by the future marker șa constitute the indicative mood, whereas forms accompanied by the particles na or as are considered subjunctive. There are good reasons for distinguishing between the two ‘periphrastic moods’ and also for grouping the three particles together in a category of ‘modal particles’ (Tsangalidis 2002). For our purposes here, it

142

Anastasios Tsangalidis

should be noted that the historical development of all three particles has been well documented and widely discussed from a number of theoretical perspectives (e.g. Bybee, Perkins and Pagliuca 1994: 212-225; Roberts and Roussou 2003: 58-87; Joseph and Pappas 2002). As far as their uses are concerned, șa is a well-developed marker of futurity and the later uses of epistemic and speaker-oriented modality (Bybee, Perkins and Pagliuca 1994: 279; Tsangalidis 1999); na is a general subordinator and a marker of epistemic and speaker-oriented modality; finally, as is a marker of epistemic and speaker-oriented modality with main clause uses only. As Joseph and Philippaki-Warburton (1987: 181) put it, the “uniting factor” in all the uses of na and as is that “they express the attitude of the speaker towards the contents of the sentence which are presented as nonfactual.” In general, all three particles can combine with all verbal forms, marked for all tense, aspect, voice and person/number categories (i.e., excluding imperatives alone). Various generalisations concerning their interpretation have been made but will not be discussed here in any detail. Briefly, as with the modals of necessity and possibility, a number of grammatical factors may be seen to affect the precise modal interpretation in predictable ways. Thus, for example, modal combinations with past perfective verb forms are always epistemic, whereas non-past perfectives tend to be interpreted as non-epistemic, as shown in Tsangalidis (2004a, 2004b). The fact that the behaviour of modal particles is apparently the same as that of modal verbs constitutes evidence that we are dealing with members of the same broad category. 2.3. The modals of necessity and possibility Tsangalidis (2004a, 2004b) argues that the class of modals in Greek should include the two modals of necessity and possibility. Crucially, it is argued there that the term “auxiliary” is to be avoided on both formal and functional grounds. The two modals are in fact termed “semi-auxiliary”; this term is explicitly meant to reflect both their intermediate position on grammaticalisation clines as well as the fulfilment of about half the usual criteria for full vs auxiliary verb status. The class of modals is thus taken to comprise the two forms prepi (‘must’) and boro/bori (‘can / may’). Of these, only the former is clearly a reduced verbal form (in that it only appears in this ‘impersonal’ form – also available in the past), whereas the

Modals in Greek

143

invariable bori is most often described as a special use of the related full verb boro (which is fully inflecting). As discussed in some detail in Tsangalidis (2004a) the two verbs are not generally described as a special class – a situation reminiscent of that in the Slavonic languages (as described in Hansen 2004: 245). In fact, it is only Ben-Mayor (1980: 9-15) who uses the term “modal auxiliaries” and argues that the term is justified to the extent that the Greek verbs are translation equivalents of modal auxiliaries in other languages. This analysis is rather exceptional, because in most grammatical descriptions of Greek the modal verbs are simply regarded as a special case of impersonal verbs. Indeed, their main formal property that may distinguish them from lexical verbs is the lack of person / number agreement. However, the standard auxiliaries ime ‘be’ and exo ‘have’ do inflect for both person and number; therefore lack of agreement marking cannot reasonably be a criterion for auxiliary status within the grammatical system of Greek. Kallergi (2004) showed that the semantic map analysis proposed by van der Auwera and Plungian 1998 may be used for the description of the two Greek modals and Tsangalidis (2004a) capitalised on the systematic alternations between epistemic and non-epistemic interpretations of both prepi and boro/bori. Indeed, they both seem to cover the core of the relevant paths proposed in Bybee, Perkins and Pagliuca (1994: 240-241, for example) and the semantic map developed in van der Auwera and Plungian 1998. 3 Most of the examples below are the Greek counterparts of those offered in van der Auwera and Plungian (1998: 80-81). POSSIBILITY (3) participant-internal (personal boro only): Boro/*Bori na andekso me įio ores ipno kașe can-1SG/*can-3SG na get.by-1SG with two hours sleep every nixta. night ‘I can get by with sleeping two hours a night.’ (4)

participant-external (both personal boro and impersonal bori): a. Bori na andekso me įio ores ipno kașe nixta. can-3SG na get.by-1SG with two hours sleep every night ‘It is possible that I can get by with sleeping two hours a night.’

144

Anastasios Tsangalidis

b. Ja na pas sto stașmo boris/bori na For na go-2SG to.the station can-2SG/can-3SG na paris to leoforio 66. take-2SG the bus 66 ‘To get to the station, you can/may take bus 66.’ (5)

deontic (normally personal boro – impersonal bori clearly excluded when speaker-oriented): a. agent-oriented: Boris/?Bori na psifisis an can-2SG/can-3SG na vote-2SG if ‘You can vote if you are 18.’ b. speaker-oriented: Boris/*Bori na can-2SG/can-3SG na ‘You may go now.’

ise 18. be-2SG 18

pijenis tora. go-2SG now

c. Boris na aniksis tin porta? can-2SG na open-2SG the door ‘Can you open the door?’ d. Bori na aniksis tin porta? can-3SG na open-2SG the door ‘Is it likely that you will open the door?’ (6)

epistemic (impersonal bori only) Bori/*Borun na irșan. can-3SG/can-3PL na came-3PL ‘They may have arrived.’

NECESSITY (7) participant-internal: Prepi na kimiși įeka ores kașe nixta ja na hours every night for na must-3SG na sleep-3SG ten liturjisi kanonika. function-3SG properly ‘He needs to (lit.: must) sleep ten hours every night for him to function properly.’

Modals in Greek

(8)

(9)

participant-external: Ja na pas sto stașmo prepi na must-3SG na For na go-2SG to.the station leoforio 66. bus 66 ‘To get to the station, you must take bus 66.’ deontic: a. agent-oriented: Prepi na fiȖun must-3SG na leave-3PL ‘All minors must leave.’

oli all

145

paris to take-2SG the

i aniliki. the minors

b. speaker-oriented: Prepi na fijis tora. must-3SG na leave-2SG now ‘You must leave now.’ (10) epistemic: O Janis prepi na the John must-3SG na ‘John must have arrived.’

irșe. came-3SG

Thus, as has also been noted for a number of other languages, the exact value of any modalised utterance in Greek is often not clear and has to be calculated taking a number of factors into account. In other words, the interpretation cannot rely on any single meaning; thus, the Greek modals are ‘polyfunctional’ (in the sense of van der Auwera, Amman and Kindt 2005) and are comparable with their counterparts in many other European languages. 3. Other means of expression of possibility and necessity The three modal particles clearly constitute an alternative means of expressing possibility and necessity – in some of their uses. However, these are predictable uses of the relevant forms and are in principle accounted for by their grammaticalisation patterns; for example, epistemic possibility is an expected development in the case of a subordinator like na – and

146

Anastasios Tsangalidis

epistemic necessity (in the form of ‘inferred certainty’) is naturally related with the development of a future marker like șa (cf. Bybee, Perkins and Pagliuca 1994; Tsangalidis 2002, 2004b). Necessity and possibility are also expressed lexically by various words which belong to any one of the major word classes and which can be seen to constitute fully lexical means of expressing the relevant notions. One crucial difference between lexical expressions of modality and items that can be seen to even have started movement on a grammaticalisation path involves the availability of more than one meaning in the case of the latter; in other words, what van der Auwera, Ammann and Kindt 2005 call polyfunctionality can be (and has been) used as a criterion of grammatical(ised) status. Indeed, it has been argued that “lexical verbs of modality” can only “express either epistemic or non-epistemic modality, but, crucially, not both. What makes such verbs different from the two modals of necessity and possibility, prepi and bori/boro, is the unavailability of both epistemic and non-epistemic readings. A verb like nomizo (‘think’) can function as a lexical expression of ‘propositional modality’ only, and a verb like elpizo (‘hope’) as an expression of ‘event modality’ only.” (Tsangalidis 2004a: 237)

In other words, polyfunctionality may not be a prerequisite for modal status, but it is obviously one very clear indication of grammaticalisation. In this sense, polyfunctionality can be seen as (at least) a sufficient condition for grammatical status in the area of modality. 4. The grammaticalisation of the modals As already stated in the introduction, the aim of this paper is to show that Lehmann’s parameters correctly describe what is a straightforward descriptive fact: that the modal particles in Greek are more grammaticalised than the modal verbs of necessity and possibility. This analysis then constitutes clear evidence for the validity of these parameters as criteria for grammaticalisation. Thus, these same parameters may be used as arguments for the assumption that the modals of necessity and possibility are themselves grammaticalised elements compared to lexical verbs (although to a lesser extent than the particles) and are therefore best described as

Modals in Greek

147

instances of grammaticalisation processes that have been repeatedly attested cross-linguistically. 4.1. Parameter Integrity (bleaching, erosion of the form) According to Lehmann, the integrity of the sign can be discussed in terms of both content and form. More generally, it is expected that there must be some correlation between the two processes of semantic and formal reduction (captured by notions such as “the form-meaning co-variation hypothesis” in Bybee, Perkins and Pagliuca 1994). However, it is not always clear whether both occur at the same time, whether semantic change always comes first and whether both are required before a certain case can be regarded as an instance of grammaticalisation. Moreover, as will be seen at the end of this section, as Lehmann himself points out, there are phenomena (such as his “morphological degeneration”), which cannot be fully described as specifically formal or functional. In other words, such phenomena appear to involve both formal and semantic integrity, thus questioning the validity of the distinction. However, the distinction is generally worth making, at least for analytical purposes as there may be important differences between the erosion of the form, on one hand, and the extent of semantic bleaching, on the other (cf. the discussion of example (13) below). 4.1.1. Semantic integrity – Bleaching This part of the parameter involves the loss of some semantic content, described as ‘desemanticisation’ and also known as ‘bleaching’ in the literature (e.g. Lehmann 2002: 114). Briefly, grammaticalised elements are expected to lack fully lexical meanings; this is considered a gradual process as an element moves on grammaticalisation ‘paths’ or ‘clines’ through time. Such elements seem to lose lexical content first, reach grammatical status and then develop ‘later’ meanings – such as the ‘postmodal meanings’ discussed in Hansen (2004: 252-4); cf. Bybee, Perkins and Pagliuca (1994), van der Auwera and Plungian (1998). The contrast between modal particles and modal verbs is very obvious in terms of bleaching: the (de-verbal) particles (as and șa) can no longer express any meaning close to the original meanings of their source verbs. 4

148

Anastasios Tsangalidis

This means that they cannot theta-mark an agent precisely because they are no longer verbs. Thus, on the basis of this general observation, Tsangalidis (2004b) argued that the distinction introduced in Bybee (1985) between agent-oriented and speaker-oriented modality can capture the difference between modal verbs and modal particles: only verbs – and not particles – can theta-mark an agent. Therefore, any entity can be specified as that agent in the case of modal verbs, whereas, in the case of particles, the ‘deontic source’ (in the sense of Lyons 1977: 843) will be normally identified as the speaker – unless explicitly indicated otherwise. More interestingly perhaps, in the case of the development of na from a purpose conjunction to a subjunctive mood marker and then to a generalised subordinator, the original purpose meaning cannot be traced in most of its contemporary uses. Arguably na is highly polysemic (cf. Delveroudi, Tsamadou and Vassilaki 1994, among others) and, clearly, in the extreme case, when na is used as a general subordinator, it may be claimed to lack all traces of the source lexical meaning. In marked contrast, it can be argued that the modals of possibility and necessity always involve some lexical content (or at least some trace thereof) in their semantics (cf. Kallergi 2004 and section 2 above). In short, as expected, the most obviously grammaticalised, eroded forms, i.e. the particles appear to have lost (or, anyway, lack) full lexical meanings. Similarly, the impersonal and invariable bori has lost more lexical content than the fully inflecting boro: only the latter can signal agent-oriented notions such as participant-internal possibility (ability), whereas the former can only express the other, more grammaticalised notions. As for the development of late and postmodal meanings, the relevant predictions are also generally confirmed as can be seen, for example, in relation to the following use of bori: (11) Bori (*Borume / *Borusame / *Boruse / may-3SG (*may-1PL / *may-PST-1PL / *may-PST-3SG / *șa bori) na min peksame kala, ala įen *FUT may-3SG) na not played-1PL well, but not bori kanis na mas katiȖorisi ja aįjaforia. (HNC 5 ) can-3SG nobody na us accuse for indifference ‘We may not have played well, but nobody can accuse us of indifference.’

Modals in Greek

149

This concessive interpretation is considered a ‘late’ meaning for a modal of possibility (van der Auwera and Plungian 1998), but crucially, the use of bori as a concessive is also subject to further formal restrictions: it has to be the impersonal version only – so the fully inflecting boro can never be interpreted as a concessive. Moreover, the impersonal form has to be completely uninflected; in other words, it cannot be marked for tense, aspect or mood and still receive this interpretation. Finally, the structure appears to be subject to obligatory clause ordering: the concessive clause has to precede the main clause, otherwise the interpretation does not obtain, as shown in (12): (12) *įen bori kanis na mas katiȖorisi ja aįjaforia, not can-3SG nobody na us accuse for indifference, (ala) bori na min peksame kala. (but) may-3SG na not played-1PL well *‘Nobody can accuse us of indifference, (but) we may have not played well.’ (Ungrammatical in the relevant sense) More generally, it is clear that the three particles have developed late and / or postmodal meanings: the development of epistemic meaning in the case of șa is a late meaning for futures according to Bybee, Perkins and Pagliuca (1994: 279) and the development of conditional and concessive meanings in the case of na and as may be seen as postmodal (cf. Kallergi 2004, based on van der Auwera and Plungian 1998). Clearly, it is no coincidence that these late developments are more evident in the case of those elements which are more obviously reduced formally, too. 4.1.2. Formal integrity – Attrition Considerable formal reduction can only be seen in the case of șa and as (and perhaps na, too) but not prepi or bori/boro – thus matching Hansen’s (2004: 262) conclusion regarding the Slavonic modals that “substantial erosion happens only in the postmodal field. The transition from premodal meanings into modality does not seem to be accompanied by erosion”. In all cases it is quite clear that semantic change precedes formal change (as stated in van der Auwera and Plungian 1998: 115, “in the domain of modality semantic change seems to come first”). The following example

150

Anastasios Tsangalidis

from the mid-17th century Erotokritos, is discussed in Holton (1993: 124 – cf. Tsangalidis 1999: 152, here retranscribed): (13) Kimaste șeli alișina, ja kino įen sleep-INF want-3SG really, for that NEG ‘He will be sleeping really, and so he is not coming.’

proveni. walk-3SG

In this most interesting example, we encounter the full verb șelo (‘want’ – the lexical source of the future marker șa). Although it is not obvious whether it is a true 3rd singular or an ‘impersonal’ occurrence, it is certainly not reduced to șa (however, there are many occurrences of șa in the same text); moreover, the accompanying lexical verb is still in the (Classical Greek) infinitive form; and finally, the relative order of V and the modal is not the one that led to the reduction of șeli to șa as the modal appears post-verbally. In other words, the development of the epistemic use appears to have proceeded independently of the formal development of either the auxiliary itself or any part of the whole construction. To the extent that historical evidence is available, similar processes have been documented for all our cases: semantic change has always preceded formal change – in the development of futurity and most modal meanings. Finally, it may be noted that prepi and the impersonal bori display what Lehmann (2002: 118) calls “morphological degeneration”: “the loss of the ability to inflect … is an integral part of the reduction of the integrity of the sign”. Lehmann mentions the English modals as a “related example” and argues that “It would be wrong to explain the loss of the ability to inflect either by phonological attrition or by desemanticization alone. Both interact to constitute what might be called morphological degeneration. What is lost is not some arbitrary phonological or semantic feature, but an inflectional category. The loss of all inflectional categories is the symptom of a change in status.” (Lehmann 2002: 118)

Clearly, both prepi and bori have lost the inflectional categories of person and number and this is straightforward evidence for a change in categorial status.

Modals in Greek

151

4.2. Parameter Paradigmaticity (paradigmaticisation) Paradigmaticity is defined as follows: “What is meant here by paradigmatic cohesion or paradigmaticity is the formal and semantic integration both of a paradigm as a whole and of a single subcategory into the paradigm of its generic category. This requires that the members of the paradigm be linked to each other by clear-cut paradigmatic relations, especially opposition and complementarity.” (Lehmann 2002: 118)

As has been mentioned above, the modal verbs may be seen as special instances of the more general class of impersonal verbs in Greek – whereas the particles form a well-defined class on their own (cf. Tsangalidis 2002). Thus, the three particles may be seen to constitute a closed set whereas the class of modal verbs is not as well-defined or as homogeneous (on homogeneity of the paradigm, see Lehmann 2002: 120). For example, the fact that prepi is invariable whereas bori contrasts with the fully inflecting boro makes the two quite hard to analyse as instances of exactly the same formal category. The particles appear to be more grammaticalised on another count as well, namely the fact that they are indeed mutually exclusive. As Hansen (2004: 263) puts it, “The members of a paradigm usually exclude each other in one and the same position” and this is indeed the case with the particles, as in (14) though not with the modals of possibility and necessity, as in (15). (14) a. b. c. d. e. f. g. h. i. j. k. l.

*șa na Ȗrapsi 6 *na șa Ȗrapsi *as na Ȗrapsi *na as Ȗrapsi *șa as Ȗrapsi *as șa Ȗrapsi *as na șa Ȗrapsi *as șa na Ȗrapsi *na șa as Ȗrapsi *na as șa Ȗrapsi *șa as na Ȗrapsi *șa na as Ȗrapsi

152

Anastasios Tsangalidis

(15) a. b. c. d.

Bori na prepi/boris na Ȗrapsis Prepi na prepi/boris na Ȗrapsis Boris na prepi/boris na Ȗrapsis *Prepi/*Boris/*Bori na bori na eȖrapses

Thus, whereas the particles clearly exclude each other in the same sentence the modal verbs boro/bori and prepi can indeed coexist quite freely – the sole restriction being a semantic one, namely that it is not possible to embed an epistemic modal, as shown in (15d), which involves embedding the unambiguously epistemic combination of bori and a past perfective form.

4.3. Parameter Paradigmatic Variability (obligatorification) In the general case it may appear that neither particles nor verbs are obligatory; this can clearly support the idea that modality is less grammaticalised than tense, aspect, voice, number and person, which are all obligatorily marked in verbal morphology. However, there may be a case here for asserting a distinction between ‘mood’ and ‘modal system’ along the lines of Palmer 2001 (cf. Bybee, Perkins and Pagliuca 1994; similarly, de Haan 2006 distinguishes between ‘mood’ and ‘modal affixes’ 7 ). It can easily be argued that the [+/-imperative] opposition is truly a mood distinction in that all verbs, by virtue of their morphological makeup, can be classified as either [+imperative] or [-imperative]. The use of the particles (and, of course, the use of modal verbs) is always an extra option for all non-imperative verb forms, and, clearly, it is never obligatory in the same sense (thus conforming to de Haan’s 2006: 36 statement that the “difference between mood and modal affixes is that mood is an obligatory category”). This general freedom of combination with all non-imperative finite verb forms arguably puts both modal verbs and modal particles at the same level as the Slavonic modals discussed in Hansen (2004: 263); as Hansen suggests, the fact that they can be used with any verb can be taken as evidence that they “show a medium degree of obligatorification”. Clearly then, on the basis of this parameter, the analysis of the imperative as a more grammatical category can be supported. The question then arises as to whether this same parameter can distinguish in any way between modal verbs and modal particles. Indeed, there may be a case for obligatorification of the particles in the case of the main clause uses of the

Modals in Greek

153

‘dependent’ verb form – which combines the ‘contradictory’ features [+perfective] and [-past] and as such is very well suited for future-oriented subordinate clauses (of the type if/when/in case she writes…) but cannot occur by itself as a main clause verb, as shown in (16a). By virtue of its internal makeup, this form needs to be supported by one of the three particles in main clauses, as shown in (16b). However, the same form in earlier stages of the language did not depend on the presence of any other material (i.e., the Classical Greek counterpart of (16a) was a grammatical sentence). In other words, there is at least a very well defined morphosyntactic context, in which the use of the particles has been made obligatory in the course of time. 8 (16) a. *Ȗrapsi write-PFV-NONP-3SG b. șa/as/na Ȗrapsi șa/as/na write-PFV-NPST-3SG ‘S/he will/may write / Let him/her write’ In any event, both particles and verbs display the phenomenon of “expansion” discussed in Lehmann (2002: 126). To take just one quite clear example, the development of șa from a verb of volition to a marker of futurity necessarily involved expansion at some point or other from sentences with animate subjects (who might reasonably act as the agents of this volition) to sentences with inanimate subjects. Even in those cases where the historical evidence is not enough, we can assume that all these modal markers originally combined with non-past forms only and that their distribution was extended later on, to cover past verb forms (and also perfect forms which are themselves a relatively recent development – cf. Moser 1988). 4.4. Parameter Scope (condensation) Questions of scope are not as straightforward – not surprisingly, since the syntactic domain is by nature more complex than that of morphology. Besides, semantic relations do not always follow syntactic ones and this has been the cause of various misunderstandings (cf. the “fallacies” discussed by Lehmann 2002: 129-130). According to Lehmann this parameter refers

154

Anastasios Tsangalidis

primarily to syntactic structure and not to semantics. In this sense, both epistemic and non-epistemic modals are found in the same syntactic position. The fact that the latter may be seen to be valid for a whole clause whereas the former appear to affect the VP only is a matter of semantics and therefore irrelevant to the discussion. In general, it is quite clear that all Greek data comply with Hansen’s (2004: 264) observation that “[t]he syntactic development of modals is characterised by the reduction of argument positions. There is no instance attested where a modal expression has increased the number of argument positions”. Indeed, in this sense, as also argued by Hansen, it is easy to support the assumption of unidirectionality of the relevant grammaticalisation process (ignoring here all other possible objections, as in Joseph 2001, 2004, among others). Clearly, reduction of argument positions can be seen in the case of prepi which to this day contrasts with a very marginal use that survives from earlier times. This use reflects the original meaning of the source lexical verb ‘resemble, fit’ and, crucially, involves two (nominal) argument positions: (17) a. įen

tu eprepe tetia metaxirisi. him must-PST-3SG such treatment ‘He did not deserve such treatment.’

NEG

b. Tu prepun įokses ke times. him must-3PL glories and honours ‘He is worthy of great honours.’ However, there are also differences in argument structure which cannot be related to any earlier constructions, as in the following examples: (18) įen

ton boro pia. him can-1SG anymore ‘I can’t stand him any longer’

NEG

(19) Les na to boresi? (accepted by some speakers) say-2SG na it can-PFV-3SG ‘Do you think he will manage it?’ Examples such as (18) and (19) arguably involve an extra argument position compared with the modal uses we have seen so far; there are also

Modals in Greek

155

considerable meaning differences. However, it is not clear that they constitute later developments and that there is indeed an increase of argument positions. In any event, they cannot affect the general picture: as predicted, in most cases the number of argument positions has decreased over time. 4.5. Parameter Bondedness (coalescence) Bondedness is defined by Lehmann as follows: “The syntagmatic cohesion or bondedness of a sign is the intimacy with which it is connected with another sign to which it bears a syntagmatic relation. The degree of bondedness of a sign varies from juxtaposition to merger, in proportion to its degree of grammaticality.” (Lehmann 2002: 131)

It can be readily observed that the particles are more intimately related to the verb they accompany: a. modal particles (but not modal verbs) are phonologically dependent on the verb they accompany; they constitute one phonological word and cannot be separated from the verb in the general case; modal verbs, on the other hand, are autonomous entities both phonologically and syntactically: (20) a. șa tu to po. șa his it say-1SG ‘I will say it to him.’ b. *șa șa

ego tu I his

to po. it say-1SG

c. *șa ston patera mu to po. șa to.the father my it say-1SG ‘I will say it to my father.’ d. Bori / prepi ston patera mu eȖo na to po. may-3SG / must-3SG to.the father my I na it say-1SG ‘I may / must say it to my father.’

156

Anastasios Tsangalidis

b. modal particles (but not modal verbs) are normally strictly adjacent to the verb they accompany; the Particle+Verb sequence can only be interrupted by negation or pronominal clitics (as in example (20a) above) which can be argued to have a special affix-like status, whereas all kinds of constituents of practically any size may interrupt the sequence of prepi or bori and their complement, as in the following example: (21) Os pros to Kosovo, i Elaįa pistevi oti prepi se as to the Kosovo, the Greece believes that must-3SG in aftin ti fasi amesa na efarmostun ta kritiria.(HNC) this the phase immediately to be.applied-3PL the criteria ‘As far as Kosovo is concerned, Greece believes that at this stage the criteria must be applied immediately.’ c. the Particle+Verb sequence is inseparable under coordination reduction (whereas the modal verb can be separated in these contexts). Briefly: (22) a. *Na ke șa pandrefti. Na and șa marry-3SG ‘S/he should and will get married.’ b. Bori ke prepi na pandrefti. can-3SG and must-3SG marry-3SG ‘S/he can and must get married.’ (23) a. *șa pandrefti ke xorisi. șa marry-3SG and divorce-3SG ‘S/he will marry and divorce.’ b. Bori na pandrefti ke na can-3SG na marry-3SG and na ‘S/he can marry and divorce.’

xorisi. divorce-3SG

In more general terms, the particles are more clearly instances of what Lehmann (2002: 139) describes as “synsemantic” or “syncategorematic” signs, in that their semantics is always dependent on that of the items which they accompany – whereas the modal verbs can be seen to have some content on their own. This can be easily seen in the relevant dictionary entries, where although they are all presented as quite polysemous, the

Modals in Greek

157

particles are best described in relation to particular constructions, example sentences etc. The situation again seems to conform to the pattern described by Hansen in relation to the Slavonic modals: “In the Slavonic languages, modal meanings do not have a tendency towards tight bonding with the verb. Verb and modal are more or less independently juxtaposed. Cliticisation is attested only in the postmodal field of tense-aspect marking. Serbian-Croatian HTETI can cliticize with the verb if it has future meaning, but it never does so if it retains its original volitional meaning.” (Hansen 2004: 265)

Indeed, the whole development of șelo>șa might be summarised and contrasted with the cases of prepi and bori. At some historical stage, an impersonal șeli na V arguably participated in a structure very similar to prepi na V or bori na V in contemporary Greek. Having developed future and even epistemic meanings, that syntagm was gradually reduced to șa V – while no such reduction has occurred in the case of the more lexical modals of necessity and possibility (cf. Tsangalidis 1999; Joseph and Pappas 2002). 4.6. Parameter Syntagmatic Variability (fixation) This parameter concerns the extent to which an item occurs in a fixed position, as defined by Lehmann: “The syntagmatic variability of a sign is the ease with which it can be shifted around in its context. In the case of a grammaticalized sign, this concerns mainly its positional mutability with respect to those constituents with which it enters into construction. Syntagmatic variability decreases with increasing grammaticalization.” (Lehmann 2002: 140)

As can easily be seen in all examples presented so far, the case can be made quite clearly with the distinction between modal verbs and modal particles. The latter always appear in predictable positions: a modal particle is always the leftmost member of the verbal complex (unless the indicative negation marker is present). In the following examples, the elements in parentheses are optionally present – but they can only appear in this order:

158

Anastasios Tsangalidis

(24) a. (įen)

șa (tu) (to) pis. șa his it say-3SG ‘You will not say it to him.’

NEG

b. Na (min) (tu) (to) pis. Na NEG his it say-3SG ‘You should not say it to him.’ In contrast, modal verbs can surface either before or after the lexical verb. There may be a clear sense of a canonical position in which the modal precedes the verb, but still the reverse order is always available, as the following attested examples show: (25) a. įiloni oti bori na treksi ke pio ȖriȖora. (HNC) state-3SG that can-3SG na run-3SG and more fast ‘He states that he can run even faster.’ b. Etsi opos ton vlepo tora, ute na treksi this.way as him see-1SG now, not.even na run-3SG bori. (HNC) can-3SG ‘The way I see him now, he cannot even run.’ The restrictions on the use of the postmodal concessive bori that were presented in section 4.1 above are also of some relevance here: assuming that this is a more grammaticalised construction, it is predicted that its syntagmatic variability will be decreased. 5. Conclusion It is only very recently that modality and modal markers have started to play a part in the analysis of Greek. Up to the early 1990s there were hardly any available terms for even the most basic concepts in the area of modality. Most grammatical descriptions of the language do not have a section for modality and only some grammars include the particle facts in the discussion of mood. However, as far as the modals of necessity and possibility are concerned, no grammar has discussed them as such yet. At best prepi and

Modals in Greek

159

bori are mentioned as a special case of impersonal verbs – but this misses the semantic affinity between the uses of the impersonal bori and the fully inflecting boro. The exact content of each modal element and the way each contributes to a modal interpretation are still to be investigated in greater detail. However, what is clearly seen in the discussion of all modal elements – and perhaps more so in the case of elements like the modal verbs and particles – involves the unavailability of a single meaning that can account for all their uses. Indeed, the Greek modals can be shown to interact with particular elements in the linguistic context (such as particular tense and aspect features); the resulting interpretations, crucially and predictably, belong in the same broad area of modal meanings surrounding ‘possibility’ and ‘necessity’. In this sense, they can be argued to be the Greek instances of a cross-linguistically available category of markers of possibility and necessity, and, conversely, they can support the assumption of such a category. In this connection, the concept of ‘modal polyfunctionality’ can be of use. In view of the remarks in van der Auwera, Ammann and Kindt (2005: 262, n. 2) it may appear that this cannot be a prerequisite for modal status; however, it is a very clear indication of grammaticalisation (in the way mentioned in section 3 above). Modal polyfunctionality then, at least in a language like Greek, can be a valid criterion for modal status insofar as it is taken not as a necessary but as a sufficient condition. Notes *

I would like to thank the editors and the reviewers of this volume for their comments and suggestions. I have also benefited from comments provided by the audience at the workshop on the Modals in the Languages of Europe, held within the 38th Annual Meeting of the Societas Linguistica Europaea (Valencia 2005). All errors and inadequacies remain mine. 1. However, some of the dialects might be best described as distinct linguistic systems. General descriptions of the Standard variety include Joseph and Philippaki-Warburton (1987) and Holton, Mackridge and PhilippakiWarburton (1997). 2. Indeed, there are no infinitives in Greek of the type known from most European languages, including Classical Greek – cf. Joseph (1983). 3. The development from a fully inflecting ‘agent-oriented’ boro to an invariable epistemic bori is what may be expected in any approach to grammaticalisation;

160

4.

5. 6.

7.

8.

Anastasios Tsangalidis what appears to be quite extraordinary is the availability of ‘agent-oriented’ readings in the case of the impersonal prepi, assuming that it also developed from a fully inflecting verb. Apparently, semantic and formal changes do not always go hand in hand; the current status of prepi seems to be a case of “retention of earlier meanings” in the sense of Bybee, Perkins and Pagliuca (1994: 16). As discussed in greater detail elsewhere (Tsangalidis 2002, 2004 and references therein), the future particle șa is a textbook case of grammaticalisation of a verb of volition / desire (șelo ‘want’) which turned into a marker of futurity and then epistemic modality, eventually excluding all volition from its semantics. The particle na has been discussed in works such as Bybee, Perkins and Pagliuca (1994); having started as a purposive conjunction (hina) it has developed into a marker of general subordination as well as (main clause) marker of speaker-oriented modality. Finally, as is described as a hortative / concessive particle which historically derives from afes, the imperative form of the verb meaning ‘let/leave’. Examples marked ‘HNC’ are attested examples, included in the Hellenic National Corpus. The HNC was developed by the Institute for Language and Speech Processing and is available at . It should be noted that some of the more conservative dialects arguably include a combination of șa and na in various versions of earlier forms of șa which may appear to combine with na: Forms like șena, șana, șela, ișe na apparently involve a less grammaticalised șa which – by virtue of the fact that it is not completely grammaticalised – can be combined with na; whereas șa in the standard variety can no longer do it – precisely because it is fully grammatical. Moreover, it is a well documented historical fact that such forms have constituted intermediate stages in the development of the original desire construction into the modern Future marker. As de Haan (2006: 36-37) puts it: “It is not always easy to tell if we are dealing with a mood or with a modal affix. Many grammars do not make a distinction between the two categories and indeed many works on modality do not do so either. Nevertheless, the distinction is worth making because a language can have both moods and modal affixes.” In fact, a sentence like (16a) can also be saved by adding isos ‘perhaps’ alone; most analysts, however, assume that there is an underlying, dropped na or șa there. Alternatively, isos might be argued to belong to the same category of modal particles – but I will not pursue the matter here; cf. Tsangalidis (2001).

Modals in Greek

161

References Ben-Mayor, Isaac 1980 The Semantics of Some Modal Constructions in Modern Greek: The Modal Auxiliaries and the Particle THA. D.Phil. Dissertation. University of Oxford. Bybee, Joan L. 1985 Morphology: A Study of the Relation Between Meaning and Form. (Typological Studies in Language 9) Amsterdam/Philadelphia: Benjamins. Bybee, Joan, Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago and London: The University of Chicago Press. de Haan, Ferdinand 2006 Typological approaches to modality. In The expression of modality, William Frawley (ed.), 27-69. Berlin: Mouton de Gruyter. Delveroudi, Rhea, Irene Tsamadou, and Sophia Vassilaki 1994 Mood and modality in Modern Greek: the particle na. In Themes in Greek Linguistics (CILT 117), Irene Philippaki-Warburton, Katerina Nicolaidis, and Maria Sifianou (eds.), 185-192. Amsterdam and Philadelphia: John Benjamins. Hansen, Björn Modals and the boundaries of grammaticalization: The case of 2004 Russian, Polish and Serbian Croatian. In What makes Grammaticalization? A Look from its Fringes and its Components, Walter Bisang, Nikolaus P. Himmelmann, and Björn Wiemer (eds.), 245-270. Berlin: Mouton de Gruyter. Holton, David 1993 The Formation of the Future in Modern Greek Literary Texts Up to the 17th Century. Proceedings of Neograeca Medii Aevi, Vol. A, Venice, 118–128. Holton, David, Peter Mackridge, and Irene Philippaki-Warburton 1997 Greek: A Comprehensive Grammar of the Modern Language. Routledge Grammars. London and New York: Routledge. Joseph, Brian D. 1983 The Synchrony and Diachrony of the Balkan Infinitive: A Study in Areal, General and Historical Linguistics. (Cambridge Sudies in Linguistics Supplementary Volume) Cambridge University Press. Joseph, Brian D. 2001 Is There Such as Thing as ‘Grammaticalization’? Language Sciences 23, 163-186.

162

Anastasios Tsangalidis

Joseph, Brian D. 2004 Rescuing traditional (historical) linguistics from grammaticalization theory. In Up and down the Cline: The Nature of Grammaticalization, ȅlga Fischer, Muriel Norde, and Harry Perridon (eds), 45-71. (Typological Studies in Language, 59) ǹmsterdam and Philadelphia: John Benjamins. Joseph, Brian D., and Panayiotis Pappas 2002 On Some Recent Views Concerning the Development of the Greek Future System. Byzantine and Modern Greek Studies 26, 247-273. Joseph, Brian D., and Irene Philippaki-Warburton 1987 Modern Greek. London: Croom Helm. Kallergi, H. 2004 Greek Modality’s Semantic Map: particles, auxiliaries and adverbs. MA dissertation in General Linguistics. University of Amsterdam. Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt Nr. 9. Lyons, John 1977 Semantics. Cambridge University Press. Moser, Amalia 1988 The History of the Perfect Periphrases in Greek. Unpublished Ph.D. Dissertation. University of Cambridge. Palmer, Frank Robert 2001 Reprint. Mood and Modality. 2nd ed. (Cambridge Textbooks in Linguistics) Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Philippaki-Warburton, Irene 1994 ȉhe Subjunctive Mood and the Syntactic Status of the Particle na in Modern Greek. Folia Linguistica XXVIII/3-4: 297-328. Philippaki-Warburton, Irene, and Vassilios Spyropoulos 2004 A change of mood: the development of the Greek mood system. Linguistics 42–4, 791–817. Roberts, Ian, and Anna Roussou 2003 Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Tsangalidis, Anastasios 1999 Will and tha: a comparative study of the category future. Thessaloniki: University Studio Press. Tsangalidis, Anastasios 2001 Criteria of modalhood: The case of the Modern Greek modal particles [in Greek]. Studies in Greek Linguistics, 759-770.

Modals in Greek

163

Tsangalidis, Anastasios 2002 Homonymy, polysemy, category membership: The case of Greek modal particles. Belgian Journal of Linguistics 16, Particles (ed. by Ton van der Wouden, Ad Foolen, and Piet van de Craen), 135-150. Tsangalidis, Anastasios 2004a Disambiguating Modals in English and Greek. In English Modality in Perspective: Genre Analysis and Contrastive Studies, Roberta Facchinetti, and Frank Robert Palmer (eds), 231-268. (English Corpus Linguistics, 1) Frankfurt am Main: Peter Lang. Tsangalidis, Anastasios 2004b Unidirectionality in the grammaticalization of modality in Greek. In Up and down the Cline: The Nature of Grammaticalization, ȅlga Fischer, Muriel Norde, and Harry Perridon (eds.), 193-209. (Typological Studies in Language, 59) ǹmsterdam and Philadelphia: John Benjamins. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79-124. van der Auwera, Johan, Andreas Ammann, and Saskia Kindt 2005 Modal polyfunctionality and Standard Average European. In Modality. Studies in Form and Function, Alex Klinge, and Henrik Høeg Müller (eds.), 247-272. London: Equinox.

B. Modals in Indo-European languages (Eastern branch)

6. Modals in the Slavonic languages Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

1. Introduction The aim of this article is to give an outline of the semantics and syntax of modals in the Slavonic languages. We shall try to develop an understanding of modals as a fuzzy cross-linguistic category and propose a typology based on morpho-syntactic features. The chapter is organised as follows. First, we shall give some basic information about the Slavonic languages. Section 2 discusses the category of modals in the Slavonic languages in general. It demonstrates the essential semantic and syntactic properties of modals in contrast to lexical items with modal meanings, and pinpoints six types of modal constructions to be distinguished in the Slavonic languages. Section 3 focuses on the grammaticalisation of modals in the Slavonic languages, whereas Section 4 deals with the areal distribution patterns with regard to the six types of modal constructions. Section 5 summarises the results and presents some conclusions concerning the degree of grammaticalisation of modals in Slavonic. The Slavonic languages form a branch of the Indo-European language family spoken in Central and Eastern Europe, the Balkans, and parts of Northern Asia. With a total of approximately 275 million speakers, this family consists of 171 languages, which can be divided into four sub-groups (Mareš 1980), as presented in Table 1: Table 1. The Classification of the Slavonic languages: North vs South North-West Slavonic languages: Polish, Kashubian, Czech, Upper Sorbian, Lower Sorbian, Slovak

North-East Slavonic languages: Russian, Belorussian, Ukrainian, Ruthenian

South-West Slavonic languages: Slovenian, Bosnian, Croatian, Croatian of Burgenland (Austria), Serbian

South-East Slavonic languages: Macedonian, Bulgarian

168

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

This classification, based on both diachronic and synchronic phonological, morphological and, to a lesser extent, syntactic criteria can be reduced to a simpler one, viz. ‘North vs South’ and ‘East vs West Slavonic languages’ model. In the present study we will make use of the ‘North vs South’ classification. Although a considerable differentiation has taken place since the 6th century, there are some common typological features which distinguish the Slavonic from other Indo-European languages. Here are some features typical of the verbal system of the Slavonic languages:  They are typical inflectional languages,  Verbal aspect covers all finite and non-finite forms of the verb,  Slavonic aspect is found with past and non-past reference as opposed to the tripartite system found e.g. in Romance,  The converb is widespread,  The paradigm of the subjunctive is extremely simplified,  The logical subject (agent) often occurs in an oblique case,  The percentage of impersonal sentences is high. In the Slavonic languages modality can be expressed in a variety of ways. Apart from modals, which are in the focus of the present survey, the Slavonic languages possess a wide range of explicit lexical means for coding the modal notions of possibility and necessity, such as:     

lexical verbs like Bulgarian umeja ‚to be able1.SG’, nouns like Polish koniecznoĞü ‘necessity’, adjectives like Serbian dužan ‘obliged’, sentence adverbs: Russian navernoe ‘probably’, word formation suffixes deriving adjectives from verbs like Russian –þiv in nerazbor-þiv-yj ‘unreadable’.

There exist a considerable number of studies on the expression of modality in Slavonic. The reader is referred to the general works Bondarko (1990), Jachnow (1994), Hansen (2001, 2006, in press), and Hansen and Karlík (2005).

Modals in Slavonic languages

169

2. The category of modals in the Slavonic languages 2.1. Modals as a cross-linguistic category We understand modals as means of expression of modality, which have undergone a grammaticalisation process; they express the basic notions of ‘necessity’ and ‘possibility’ and show syntactic properties of auxiliaries. Modal is a gradient category; there are prototypical and peripheral instances. We can define modals in the following way: A modal is a polyfunctional expression of modality. It always occurs with main verbs in the predicate position and opens one and only one argument position, which is filled by a lexical verbal stem. A modal does not select its own nominal arguments but influences the encoding of the arguments of the verbal form. Modals are to be located at the “grammatical periphery” (see Plungjan 2003: 130) and tend to form a kind of fully analytical paradigm of the verb. Typical modals are polyfunctional in the sense that they express no less than two types of modality. One usually distinguishes dynamic, deontic and epistemic modality. Modals are polyfunctional, while socalled modal content words, i.e. words with modal meaning which are not subject to an auxiliarisation process, have only one modal meaning. Let us compare the fully-fledged modal auxiliary Slovak môct’ ‘can’ with the modal content word vládat’ ‘to be capable’. The former can express ‘capability’ (dynamic) (1), ‘objective possibility’ (dynamic) (2), ‘permission’ (deontic) (3) and ‘perhaps’ (epistemic) (4), while the latter is confined to ‘capability’: (1)

Slovak Nevidel som Ģa 3 dni, not-see-PST-SG.M AUX.1SG you.ACC.SG 3 day.PL nemôžem/nevládzem to vydržaĢ. not-can.PRS-1SG that stand-INF ‘I have not seen you for three days, I can’t stand that any longer.’

(2)

Naši our-NOM.PL stan tent.ACC.SG

priatelia nemohli/?nevládali postaviĢ friend-NOM.PL not-can-PST-3PL put.up-INF lebo nemali kladivo. because not-have-PST-3PL hammer-ACC.SG

170

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

‘Our friends couldn’t put up the tent because they didn’t have a hammer.’ (3)

Odsúdení na najvyšší trest convict-NOM.PL to SUPER-high-ACC.SG punishment môžu / *vládzu si v USA v štáte Utah can.PRS-3PL self.DAT in USA in state-PREP.SG Utah vybraĢ spôsob smrti. choose-INF way.ACC.SG death-GEN.SG ‘In the state of Utah the prisoners sentenced to death can choose the way they will die.’

(4)

V Tatrách môže / *vládze dnes pršaĢ. in Tatra-PREP.PL can.PRS-3SG today rain-INF ‘It might rain in the Tatra today.’

Semantic polyfunctionality is not restricted to the three types of modality. Some modals have developed functions beyond modality, i.e. post-modal grammatical meanings. This term coined by van der Auwera and Plungian (1998) denotes meanings, which according to the universal semantic map develop out of modal meanings. This has happened with Czech mít ‘should’ that has adopted the evidential meaning of ‘hearsay’. (5)

Czech Zítra má pršet. tomorrow have.PRS.3SG rain-INF ‘They say, it will rain tomorrow.’

From a morpho-syntactic point of view modals can be characterised as auxiliaries similar to markers of analytical verbal forms. A typical modal is part of the predicate and usually does not occur in other syntactic positions. In contrast to the Germanic languages, there is no dedicated morphological marking of modals in Slavonic languages. As modals are the result of grammaticalisation processes their morphology and syntax show traces of the part of speech they originally belonged to:  modals of verbal origin: e.g. Bulgarian moga ‘can1.SG’  modals of adjectival origin: e.g. Russian dolžen ‘should/must’;  modals of adverbial origin: e.g. Slovak možno ‘one can’.

Modals in Slavonic languages

171

Modals of verbal origin show verbal inflection; i.e. they are marked for person, number, mood and tense (6). In contrast to that, modals of adverbial origin are uninflected and need a tense auxiliary bearing the tense and finite features (7). (6)

Russian My možem rabotat’. we can.PRS-1PL work-INF ‘We can work.’

(7)

Možno bylo rabotat’. possible be-PST-SG.N work-INF ‘It was possible to work.’

Modals historically going back to adjectives exhibit agreement marking both on the modal and the tense auxiliary, whereas tense and mood is marked exclusively on the auxiliary; e.g. (8)

Russian Ivan dolžen byl rabotat’. Ivan must.SG.M be-PST work-INF ‘Ivan had to work.’

Modals have undergone a decategorialisation process and, therefore, they tend to show idiosyncratic morphological properties setting them apart from lexical verbs, nouns or adjectives. The Polish and Czech modals of verbal origin móc/moct ‘can’ and musieü/muset ‘must’, for example, can neither form imperatives, nor deverbal nouns. They also have no perfective aspect.2 2.2. A cross-linguistic typology of modal constructions In this paragraph we shall develop a typology of modal constructions which is based on the morpho-syntactic coding of the arguments of the verb modified by a modal (for previous versions see Hansen 2006 and 2007). We understand modals as operators modifying the predicational or the propositional layer of the clause. Modal constructions in Slavonic vary with respect to i) the syntactic encoding of the privileged syntactic argument, and

172

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

ii) the assignment of the agreement marking to the modal and/or the main verb. i) We distinguish three types of coding of the privileged syntactic argument: it can be coded either in the Nominative, the Dative or it can be omitted (on the behaviour and control properties of dative subjects in Russian see Testelec 2001). Compare: (9)

Russian My možem we.NOM can.PRS-1PL ‘We can work.’

rabotat’. work-INF

(10) Nam možno rabotat’. we.DAT possible work-INF ‘It is possible for us to work.’ (11) Možno rabotat’. possible work-INF ‘It is possible to work.’ ii) The agreement with the subject can be marked in three ways: only on the modal (12), only on the main verb (13) or on the modal and the main verb (14). (12) Serbian Ivan mora Ivan must.PRS.3SG ‘Ivan must work.’

raditi. work-INF

(13) Treba da rade. should.PRS.3SG COMP work.PRS-3PL ‘They should work.’ (14) Ivan mora da radi. Ivan.NOM.SG must.PRS-3SG COMP work.PRS-3SG ‘Ivan must work.’

Modals in Slavonic languages

173

Some modal constructions, however, lack subject agreement, as illustrated in (10). Not all logically possible combinations of the features are attested; e.g. as Dative case does not trigger verbal agreement, facultative Dative subjects do not co-occur with agreeing modals. Applying this typology to the Slavonic languages we get six types of constructions including modals listed in Table 2: Table 2. Types of modal constructions in Slavonic languages Main construction types

Subtypes (tense and mood marking)

Type 1: Nominative subject Modal = + agreement Verb = - agreement

a) marking on the modal b) marking on the auxiliary

Type 2: Nominative subject Modal = + agreement Verb = + agreement Type 3: Facultative ‘Dative subject’ Modal = - agreement Verb = - agreement

a) marking on the modal b) marking on the auxiliary

Type 4: Facultative ‘Dative subject’ Modal = - agreement Verb = + agreement Type 5: No subject Modal = - agreement Verb = - agreement

a) marking on the modal b) marking on the auxiliary

Type 6: Nominative subject Modal = - agreement Verb = + agreement

a) marking on the modal b) marking on the verb

174

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

2.2.1. Construction type 1 (Nominative subject; modal = + agreement; main verb = no agreement) The modal occurs in a construction with a subject in the Nominative case (pronouns can be omitted). The modal shows subject agreement with respect to person and number and sometimes to gender and combines with a verb in the infinitive whereas the main verb is not marked for agreement. We can distinguish between deverbal modals (like Russian moþ’ ‘can’) and deadjectival modals used with a tense auxiliary (like Russian dolžen ‘to be obliged / must’). (15) Russian My možem rabotat’. we.NOM can.PRS-1PL work-INF ‘We can work.’ (16) Ivan dolžen byl rabotat’. Ivan must.SG.M be-PST work-INF ‘Ivan had to work.’

2.2.2. Construction type 2 (Nominative subject; modal = + agreement; main verb = + agreement) In the South Slavonic languages which are affected by the loss of the infinitive we also find semi-finite forms marked for person and number. Here, the agreement with the subject is marked both on the modal and the main verb. (17) Serbian Ivan može da radi. Ivan can.PRS-3SG COMP work.PRS.3SG ‘Ivan can work.’ The verbal form with da carries the agreement features, but not tense or mood; these have to be assigned to the modal, as can be seen in examples (18) and (19). Therefore, the da-form can be called ‘semi-finite’ (for more details on the element da see sect. 3.4).

Modals in Slavonic languages

175

(18) Serbian Ivan je morao da radi Ivan AUX.3SG must-PST-M.SG COMP work.PRS-3SG ‘Ivan had to work.’ (19) Ivan bi morao da radi. Ivan COND.AUX.3SG can-PST-M.SG COMP work.PRS-3SG ‘Ivan should work.’ 2.2.3. Construction type 3 (‘Facultative Dative subject’; modal: no agreement; main verb: no agreement) In construction type 3 there is no Nominative subject and the first argument of the infinitival verb can be instantiated in the Dative case or can be omitted. As there is no subject agreement, the modal (complex) has the default ending third person singular neuter. We distinguish constructions with a deverbal modal, viz. 3a, and constructions with an uninflected deadverbial modal accompanied by a tense auxiliary marked for the third person neuter, viz. 3b, as illustrated in (20) and (21), respectively. (20) Serbian Onda valja raditi. then should.PRS.3SG work-INF ‘Then, one has to work.’ (21) Russian Ivanu nado bylo rabotat’ […]. Ivan-DAT must be-PST-N.SG work-INF ‘Ivan had to work […]’

2.2.4. Construction type 4 (‘Facultative Dative subject’; modal = no agreement; main verb = + agreement) In line with the opposition between type 1 and 2 the South Slavonic language show a Dative subject construction characterised by the agreement of the main verb:

176

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

(22) Serbian Ako nam treba da zamenimo jedan znak u sign in if we.DAT need.3SG COMP replace.PRS-1PL one izrazu koristimo symbol taþke. expression-LOC.SG use.PRS-1PL symbol full.stop-GEN.SG ‘If we have to replace an element of the expression, we use a full stop.’

2.2.5. Construction type 5 (no subject; modal: no agreement, main verb: no agreement) Some modals, especially in Polish, developed from type 3 to a construction, which in contrast to that does not allow the overt encoding of the privileged syntactic argument. As type 5 is the result of a still on-going language change there are still found some archaic examples of the use of the Dative subject. Type 5 is represented by 5a deverbal (23) and 5b deadverbial modals (24). (23) Polish a. NaleĪaáo pracowaü. must-PST-3SG.N work-INF ‘One had to work.’ b. *NaleĪaáo nam pracowaü. must-PST-3SG.N we.DAT work-INF (24) a. Trzeba byáo pracowaü. must be-PST-SG.N work-INF ‘One had to work.’ b. *Trzeba nam byáo pracowaü. must we.DAT be-PST-SG.N work-INF

Modals in Slavonic languages

177

2.2.6. Construction type 6 (Nominative subject; modal = no agreement, main verb = + agreement) This construction type is characterised by subject agreement being marked exclusively on the main verb. According to the predicative features of tense and mood we distinguish two subtypes. In Serbian we find a modal construction 6a where agreement is marked on the main verb and tense and aspect on the modal: (25) Serbian [Pa dobro, hajdemo.] I ja treba da se [Fine, let’s go.] and I should.PRS.3SG COMP self žurim. hurry.up.PRS-1SG ‘I should hurry up.’ (26) To bi ja trebalo vas da that COND.AUX.3SG I should-PST-SG.N you.ACC COMP pitam. ask.PRS-1SG ‘That’s what I should have asked you.’ The type 6b is represented by Slovene lahko which does not inflect at all. (27) Slovene Tu smo lahko sreþali […]. find-PST-PL here be.AUX.1PL can ‘Here we could find […].’ In the individual languages the core of the category of modals comprises a limited set of elements. Table 3 lists all elements which are characterised by the two crucial features polyfunctionality and auxiliary-like syntactic behaviour. The table does not include peripheral modals.

178

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

Table 3. Core modals in Slavonic (all Standard languages except Macedonian) Modals of possibility

Modals of necessity

Old Church Slavonic

mošti (1a)

Polish

móc (1a), moĪna (5b)

musieü (1a), mieü (1a), powinien (1b) wypada (3a), naleĪy (5a), trzeba (5b)

Czech

moct (1a)

muset (1a), mít (1a), tĜeba (5b)

Slovak

môct’ (1a), možno (3b)

musiet’ (1a), mat’ (1a), treba (3b)

Upper Sorbian

móc (1a), smČü (1a)

dyrbjeü (1a), mČü (1a), njetrjebaü (1a)

Lower Sorbian

móc (1a)

musaĞ (1a), mČĞ (1a), trjebaĞ (1a), dejaĞ (1a)

Russian

moþ’ (1a), možno (3b), nel`zja (3b)

dolžen (1b), sleduet (3a), nado (3b)

Ukrainian

mohty (1a), smity (type 1a), možna (3b)

musyty (1a), maty (1a), povynen (1b), naležyt’ (3a), treba (3b), slid (3b)

Belorussian

mjahþy (1a), l’ha (3b), nel’ha (3b), možna (3b)

music’ (1a), pavinen (1b)

Slovene Serb./Croat./Bosn. Bulgarian

moþi (1a), utegniti (1a), lahko (6b) moüi (1a, 2) moga (2, 5a)

morati (1a), treba (3b) morati(1a, 2), trebati (1a, 2, 3a, 4a, 6a), valjati (3a, 4a, 6a) trjabva (4a, 6a)

3. The grammaticalisation of modals In this section we are going to analyse the modal constructions in terms of the parameters of grammaticalisation developed by Ch. Lehmann (2002, first version 1982). Due to the lack of space we shall confine our analysis to

Modals in Slavonic languages

179

the modals of types 1, 2 and 3. Table 4 shows an overview of the parameters: Table 4. Parameters of grammaticalisation paradigmatic

syntagmatic

weight

integrity

structural scope

cohesion

paradigmaticity

bondedness

variability

paradigmatic variability

syntagmatic variability

3.1. Parameter Integrity (bleaching, erosion of the form) According to Ch. Lehmann (2002: 126) “[...] integrity of a sign is its possession of a certain substance which allows it to maintain its identity, its distinctness from other signs, and grants it a certain prominence in contrast to other signs in the syntagm.” It seems sensible to distinguish between semantic and formal integrity. The loss of semantic integrity is usually described as semantic bleaching or the loss of semantic components. As mentioned in ch. 2, modals are understood as polyfunctional expressions, covering usages on more than one level of modality (dynamic, deontic, epistemic). During the grammaticalisation process, the elements take over new functions, for example: dynamic elements adopt deontic functions and vice versa. According to the number of meanings or usages we can assume a lower or higher degree of semantic bleaching. The following figures contain frequently attested polysemy patterns and, therefore, can be interpreted as parts of the universal semantic map of modality.

Figure 1. Modality’s semantic maps

180

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

Generally speaking, modals of construction type 1 and 2 (with a Nominative subject) tend to be more polyfunctional than modals of type 3 (with a Dative subject); whereas the former can cover all meanings on the map,3 the latter do not express participant internal modality or epistemic meanings (probability), as can be seen in Figure 2.4

Figure 2. Semantics of type 3 modals

The process of semantic bleaching is reflected not only in the tendency towards polyfunctionality, but also in the loss of fully lexical pre-modal meanings. There are some cases where the pre-modal meaning of a modal either got lost or is currently getting out of use: e.g. Polish powinien ‘should’ lost its meaning ‘to owe’, attested until the 19th century or Russian možno ‘one can’ has lost its meaning ‘mighty’, found in Old Russian texts. We can spot several lexical sources feeding the semantic field of modality. The most frequent sources or pre-modal etymological meanings are:  ‘to be strong’ (cognates of Common Slavic *mogti): all Slavonic languages;  borrowing from German (German müssen and dürfen): Polish, Czech, Slovak, Upper Sorbian, Lower Sorbian, Ukrainian, Belorussian;  ‘to torment’ (cognates of Common Slavic *terb-): Polish, Czech, Slovak, Ukrainian, Slovene, Serbian/Croatian/Bosnian, Bulgarian;  ‘to have’ (cognates of Common Slavic *jɶmČti): Polish, Czech, Slovak, Upper Sorbian, Lower Sorbian, Ukrainian;  ‘to owe’: Russian dolžen, Polish powinien, Ukrainian povynen, Belorussian pavinen;  ‘lightly/easily’ (cognates of Common Slavic *lɶgɴkɴ): Russian, Belorussian, Slovene;

Modals in Slavonic languages

181

Some of the modals go ahead in their semantic development and adopt postmodal grammatical functions. In Russian, Polish and Serbian/Croatian/ Bosnian the following are attested: conditional, evidential, tense-aspect (future and impending action), and prohibitive. It is interesting to note that none of the Polish and Russian post-modal units developed into a fullyfledged grammatical subcategory. They are either stylistically restricted like ‘impending action’, soon fall into disuse again like the analytical imperative or they have the status of non-grammatical lexicalisations. In contrast to this, Serbian/Croatian/Bosnian modals developed further und some became grammatical subcategories like future and prohibitive (for more details cf. Hansen 2001 and 2004). 3.2. Parameter Paradigmaticity (paradigmaticisation) According to Lehmann, paradigmaticity is the integration of a paradigm as a whole and of a single subcategory into the paradigm of its generic category (Lehmann 2002: 118). Typical grammatical categories like case possess a closed set of items. The category of modals comprises between four and seven items and, thus, in respect of the sheer numerical size, shows a fairly high degree of paradigmaticity. However, the degree of paradigmaticity varies in the individual languages, if we take into account the heterogeneity of the modals. In Upper and Lower Sorbian, modals form a homogeneous morpho-syntactic class (only construction type 1a). In contrast to that, e.g. Polish has five different types. Apart from the heterogeneity, we have to take into account the fact that modals compete with a relatively large number of lexical items fulfilling similar functions. Modals can be replaced by a large set of non-grammaticalised lexical vocables with modal meaning; for example, instead of can the speaker can say to be able, to have the possibility, to be allowed etc. (cf. chapter 1). In this sense, Slavonic modals are much less grammatical than for example tense or case which form closed sets of sub-categories. Modals do, however, form a generic category with the subtypes possibility – necessity on the one hand, and participant internal vs participant external vs epistemic modality on the other. The members of a paradigm usually exclude each other in one and the same position; e.g. modals do not combine with each other; the only exception seems to be the rare combination of necessity and possibility (in that order).

182

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

(28) Polish Trzeba móc sfinansowaü tĊ produkcjĊ. should can-INF finance-INF that-SG.F.ACC production-SG.ACC ‘One should be able to finance the production.’ This would be evidence for a relatively low degree of paradigmaticity of the Slavonic modals. 3.3. Parameter Paradigmatic Variability (obligatorification) Paradigmatic variability is the freedom with which the language user chooses a sign. The speaker may have the freedom in activating the category by using one of its subcategories or leaving the category unspecified (Lehmann 2002: 123). As speakers are not compelled to activate the category of modals when they use a verb, modals are not obligatory. Whereas tense, aspect, mood and number have to be marked, modality can be left unspecified. Thus, the degree of obligatorification is very low. However, over the course of time modals expand their range of use and lose selection restriction. Old Polish powinien and Old Russian dolžen combined exclusively with nouns denoting humans and only later started to appear also with inanimate nouns as illustrated in (29): (29) Polish Zupa powinna byü gorąca. soup should-SG.F be-INF hot-SG.F ‘Soup should be hot.’ In the modern languages they have the same range of use as for example the future tense, i.e. they can be used with any verb. In respect of paradigmatic variability, construction types 1 and 2 (Nominative subject) differ from type 3 (Dative subject). Construction types 1 and 2 show no selection restrictions; i.e. they can take any kind of infinitival complement with any kind of subject (30a, 31a, and 32a-a’). In contrast to that, impersonal type 3 constructions cannot co-occur with avalent verbs and verbs with non-human/inanimate arguments in the subject position, as illustrated in (30b, 31b, and 32b):

Modals in Slavonic languages

183

(30) Russian a. Možet svetat’. can.PRS-3SG dawn-INF ‘The day may dawn.’ b. *Možno svetat’ can dawn-INF  verbs with non-human first argument (31) a. Xorošaja sobaka dolžna lajat’. good-NOM.SG.F dog-NOM.SG must-SG.F bark-INF ‘A good dog has to bark.’ b. *Nado lajat’. must bark-INF  passive verbal forms with a promoted non-human second argument (32) a. Kniga dolžna byt’ perevedena. book-NOM.SG must-SG.F be-INF translate-PTCP-SG.F ‘The book should be translated.’ Serbian a’. Knjiga mora da bude book-NOM.SG must.PRS.3SG COMP be.FUT.3SG prevedena. translate-PTCP-SG.F ‘The book has to be translated.’ Russian b. *Knige nado byt’ perevedenoj. book-DAT.SG must be-INF translate-PTCP-INS.SG.F ‘The book should be translated.’ We come to the conclusion that Slavonic modals show a medium degree of obligatorification. They can be combined with a wide range of verbs,

184

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

modals of type 1 and 2 show no restriction at all, type 3 modals do not combine with all verbs; but their use is not obligatory. Modality as a semantic category might be left unspecified.

3.4. Parameter Scope (condensation) The structural scope of a unit is the structural size of the construction, which it helps to form (Ch. Lehmann 2002: 128). In the history of the Slavonic modals we find many instances of a narrowing of scope, i.e. instances of condensation. The syntactic development of modals is characterised by the reduction of argument positions. There is no instance attested where a modal expression has increased the number of argument positions. Hence, as predicted by grammaticalisation theory we are dealing with a unidirectional process. In the history of many modals we find a tendency to restrict the propositional argument position p to infinitival verbs. Subordinated clauses as in Polish naleĪy, aby or trzeba, Īeby ‘it is necessary that’ have become very rare and have obtained an archaic feature. The loss of subordinated clauses is also a case of condensation in Lehmann’s sense. However, Serbian modals of type 2 seem to be at odds with this implication of grammaticalisation, as they can be complemented either by the infinitive, on the one hand, or finite clauses introduced by the complementiser da ‘that’ (henceforth: da-clauses) on the other, showing a strong preference for the latter. The loss of the infinitive and its replacement by a finite clause is one of the Balkan Sprachbund properties; in the case of Serbian, it has been regarded as the result of multiple factors, such as language-internal processes, and language contact (Grkoviü-Major 2004; Joseph 1983). Two types of da-clauses, viz. da1- and da2-clauses, have been distinguished thus far. Da1-clauses can take any tense occurring within a main clause; they function as reported statements. Da2-clauses, on the other hand, can contain the present tense; they appear after verbs of volition or function as a reported imperative, and are referred to as the infinitive substitute (Browne 1986; Browne and Alt 2004). However, after applying Noonan’s (1985 cited Cristofaro 2003) classification of complement-taking predicates on Serbian, and analyzing the morphosyntactic features of da-clauses complementing them, we come to the following conclusions: The da1 type complements predicates of utterance (kazati ‘say’, reüi ‘tell’, etc.), propositional attitude (misliti ‘think’, verovati

Modals in Slavonic languages

185

‘believe’, etc.), knowledge (znati ‘know’, razumeti ‘understand’, shvatiti ‘realise’, etc.), and ‘perception’ (videti ‘see’, þuti ‘hear’, etc.). The da2 type complements desideratives (želeti ‘want’, hteti ‘want’, etc.) and manipulatives (narediti ‘order’, naterati ‘make’, ubediti ‘persuade’, etc.), whereas a distinct type of da-clauses that we will refer to as da3 type complements modal, as well as tense and aspectual auxiliaries, i.e. Serbian TAM verbs.5 Apart from the fact that da3-clauses allow the present tense only (a characteristic they have in common with the da2 type), they display the following additional restrictions: pro-drop and co-reference between the subjects are obligatory. These restrictions (which do not apply to the verb of volition hoüu) are illustrated in (33) and (34), respectively. (33) Serbian a. Hoüu da ja idem sa tobom want.PRS-1SG COMP I go.PRS-1SG with you.INS.SG (a ne Ivan). and not Ivan ‘It’s me who wants to go with you, not Ivan.’ b. *Moram da ja idem sa tobom must.PRS-1SG COMP I go.PRS-1SG with you.INS.SG (34) a. Hoüu da ideš sa njima want.PRS.1SG COMP go.PRS-2SG with them.INS.PL ‘I want you to go with them’ b. *Moram da ideš sa njima must.PRS-1SG COMP go.PRS-2SG with them.INS.PL Hence, a da3-clause is not a finite clause in the proper sense of the word. Compared to da1 and da2, it displays reduced syntactic properties, very close to these of the infinitive, and it deserves, therefore, the ‘infinitive substitute’ tag even more than its da2 counterpart.

3.5. Parameter Bondedness (coalescence) Bondedness is the intimacy with which a sign is connected with another sign to which it bears a syntagmatic relation (Ch. Lehmann 2002: 131). A

186

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

high degree of bondedness is reached when the two elements coalesce. In our case we are dealing with the degree of bondedness with which the modal is connected to the infinitival verb. The Slavonic modals do not show any tendency towards cliticisation or agglutination to the verbal stem; they need not be adjacent to each other, other elements can be posited between them. (35) Russian On ne možet prosto nabljudat’. he NEG can.PRS-3SG simply look-INF ‘He cannot just look.’ (36) Polish Czy mogáa by mi pani poĪyczyü Q can-PST-SG.F COND I.DAT lady.SG.NOM lend-INF pieniądze na powrót do Europy? money-ACC.N for return to Europe-GEN.SG ‘Could you lend me the money for a return ticket to Europe?’ (37) Serbian/Croatian S pasošem sam bar mogao with passport-INST.SG AUX.1SG at.least can-PST-SG.M slobodno da se kreüem. COMP REFL move.PRS-1SG freely ‘With the passport I could at least move freely.’ In the Slavonic languages, modal elements do not have a tendency towards tight bonding with the verb. Verb and modal are more or less independently juxtaposed. Cliticisation is attested only in the postmodal field of tenseaspect marking. Serbian/Croatian/Bosnian hteti can cliticise with the verb if it has future meaning, but it never does so if it retains its original volitional meaning. Compare thus nastupiüe (nastupiti + üe o nastupiüe)6 ‘they will emerge’ in (38a), to hoüeš da te odvedem ‘do you want me to take you’ in (38b): (38) Serbian/Croatian a. Nastupiüe, ukratko, sve one neodreÿene, složene emerge-FUT.3PL in.a.word all those undetermined complex

Modals in Slavonic languages

187

i mnogobrojne odgovornosti [...] and numerous responsibilities ‘In the near future there will emerge those undetermined, complex, and numerous responsibilities [...].’ b. Ujak me je pitao: ‘Hoüeš da te uncle I.ACC AUX.3SG ask.PST want.PRS-2SG COMP you.SG.ACC odvedem kod kapetanice reþnog take.PRS-1SG to captain’s.wife-GEN.SG river-SG.GEN brodarstva?’ boating-SG.GEN ‘My uncle asked me: ‘Do you want me to bring you to the wife of a river boat captain?’’ 3.6. Parameter Syntagmatic Variability (fixation) Bondedness goes together with the last parameter of grammaticalisation, syntagmatic variability. The latter concerns the positional mutability with respect to those constituents with which the modal enters into construction. Russian and Polish modals can occupy the unmarked pre- and the marked postverbal position in the sentence: (39) Russian Nel’zja bylo razobrat’, plaþet li on [...] cry.PRS-3SG Q he.NOM impossible be-PST-3SG.N see-INF ‘One could not see if he was weeping […]’ (40) Tak postupat’ nel’zja. that.way act-INF impossible ‘One must not do something like that.’ (41) Polish Tak dalej w Polsce byü nie moĪe like.that further in Poland-PREP.SG be-INF NEG can.PRS-3SG ‘It can’t go on in Poland like that.’ In contrast to this, Serbian/Croatian/Bosnian and Bulgarian modals cannot appear after the main verb. It has to be before the verb irrespectively of the realisation of the main verb as an infinitive or the Serbian da-construction;

188

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

cf. moram iüi / moram da idem; *iüi moram / *da idem moram ‘I must go’. We come to the conclusion that syntagmatic variability has decreased in Serbian/Croatian/Bosnian, whereas Russian and Polish modals can be shifted around relatively freely. 4. Some areal distribution patterns of modal constructions The different types of modal constructions presented in chapter 2.2 are not evenly distributed among the Slavonic languages. If we combine the semantics of modals with the structural types 1-6 under the areal aspect, we get the following picture:7 Table 5. Areal distribution of types 1-6 (possibility) Dynamic POSS

Deontic POSS

Epistemic POSS

Type 1

Polish, Czech, Slovak, Upper Sorbian, Lower Sorbian, Russian, Ukrainian, Belorussian, Slovene, Bosnian, Croatian, Serbian

Polish, Czech, Slovak, Upper Sorbian, Russian, Ukrainian, Belorussian, Bosnian, Croatian, Serbian

Polish, Czech, Slovak, Russian, Ukrainian, Belorussian, Slovene, Bosnian, Croatian, Serbian

Type 2

Serbian, Bulgarian

Serbian

Serbian

Type 3

Russian

Russian, Ukrainian, Belorussian

not attested

Type 4

not attested

not attested

not attested

Type 5

Polish, Slovak, Ukrainian, Belorussian Slovene

Polish, Slovak

not attested

Slovene

Slovene, Bulgarian

Type 6

Modals in Slavonic languages

189

Table 6. Areal distribution of types 1-6 (necessity) Dynamic NEC

Deontic NEC

Epistemic NEC

Type 1

1a: Polish, Czech, Slovak, Upper Sorbian, Lower Sorbian, Ukrainian, Belorussian, Slovene, Bosnian, Croatian, Serbian 1b: Russian

1a: Polish, Czech, Slovak, Upper Sorbian, Lower Sorbian, Ukrainian, Belorussian, Slovene, Bosnian, Croatian, Serbian 1b: Polish, Russian, Ukrainian, Belorussian.

1a: Polish, Upper Sorbian, Lower Sorbian, Russian, Belorussian, Slovene, Bosnian, Croatian, Serbian 1b: Polish, Russian, Belorussian

Type 2

Serbian

Serbian

Serbian

Type 3

Czech, Slovak, Russian, Ukrainian, Slovene, Bosnian, Croatian, Serbian

Czech, Slovak, Polish, not attested Russian, Ukrainian, Slovene, Bosnian, Croatian, Serbian

Type 4

Serbian (±DAT), Bulgarian (-DAT)

Serbian (±DAT), Bulgarian (-DAT)

Serbian (-DAT) Bulgarian (-DAT)

Type 5

Polish, Czech, Ukrainian

Polish, Czech, Ukrainian

not attested

Type 6

Serbian, Bulgarian

Serbian, Bulgarian

Serbian, Bulgarian

As tables 5 and 6 show, there are certain correlations between construction type, semantics and area. 4.1. Correlations between construction type and semantics Type 1a) (deverbal modal) is widespread for all types of possibility, generally expressed by the verb *mogti. As an expression of necessity, type 1a is today also widespread, but these modals are in general younger (see ch. 4.2).

190

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

Type 1b) (deadjectival modal) exists only in North Slavonic languages (in a closely related group: Polish, Belorussian, Ukrainian, Russian) and is used for necessity, but is completely absent in the field of possibility. 1b) is a kind of intermediate step between typ 1a and 3, since an adjective is not as “active” as a verb, but the category of gender and number is marked. The etymology is “to owe”, either in a moral (vina) or in a more materialistic sense (dolg). The impersonal types 3 and 5 are closely connected with all types of necessity (except epistemic), but express possibility only in a limited group of languages (again Polish, Belorussian, Ukrainian, Russian, and additionally Slovak). Necessity is felt as something outside the person, hard to influence, therefore it was originally not expressed by a personal verb (type 1a) which would reflect a certain freedom of action of the individual. This is the general basic relation between semantics and type 1-6. It is not the division between deontic, dynamic and epistemic (which was probably introduced later), but rather the dichotomy possibility – necessity which influences deeply the structure of the Slavonic modals. Type 2 reflects type 1, adapted to the infinitiveless languages Bulgarian and (partially) Serbian; type 4 reflects type 3, adapted to the same languages. Type 2 and type 4 with an argument (+ DAT) express the acting person twice: type 2 in both personal endings, type 4 in the dative and in the ending of the finite verb. This is the case only in Serbian. The double marking of the person by the personal pronoun and the ending of the finite verb in type 6a (Serbian, Bulgarian) is a relatively young development. It shows that there is a tendency to develop a symmetric relation between a Nominative subject and a corresponding verbal form. This is the second factor which erases the traditional impersonal structure of modals with a Dative subject or without subject. The epistemic modality which occurred later than the two other meanings is sometimes developed from the corresponding dynamic, sometimes from the deontic modal. This is obvious in cases when deontic and dynamic modal do not coincide (e.g. Slovene dynamic moþi or utegniti, on the other hand Polish deontic powinien which all developed epistemic readings). It is conspicuous that type 3 and 5 cannot be used for epistemic modality in languages in which there are personal modals in combination with an infinitive (type 1). In these languages, epistemic modality requires a personal construction. Only languages of the da-type (Serbian, Bulgarian)

Modals in Slavonic languages

191

can have impersonal modals with epistemic meaning, because the person is marked by the finite verb. In some languages like Ukrainian, epistemic necessity is preferably expressed by modal adverbs, not by modals. Also speakers of other Slavonic languages often use modal adverbs in order to express epistemic meaning (cf. van der Auwera and Schalley and Nuyts 2005, 214). We must be aware that our investigation says nothing about frequency, in other words we present the language system, not the language use. 4.2. Correlation between construction type and area It is obvious that type 1, 3 and 5 exist in North and South Slavonic languages, whereas type 2, 4 and 6 are restricted to the South Slavonic languages. The only language which is represented by more than 3, namely 5 types, is Serbian, but, of course, with a strong “internal” regional differentiation (see below). Under the current political circumstances (promotion of language diversity) in former Yugoslavia, one can expect that some typical Serbian constructions will disappear in Croatian and perhaps also vice versa. So far Serbian shows an extraordinary structural wealth, especially in the domain of necessity. Polish, Ukrainian, Bulgarian and Slovene (thanks to the type 6b with lahko, a unique case in the whole Slavonic modal system [Roeder and Hansen 2006]) occur in 3 types, all other Slavonic languages are restricted to maximally 2 types. More interesting is the following observation: The picture described in 4.1 as the “typical” Slavonic relation between structure of the modals and its meaning is today obscured by internal parallel development (analogy) and especially by language contact phenomena. A “parallel development” is, for example, the verb morati ‘must’ in Slovene and Bosnian, Croatian, Serbian. This verb, in analogy to trebati, has probably been derived from the modal expressing possibility, from more (3SG of moþi “to can”). External influence has been discussed, but is not proved. Language contact is apparently much more important than analogy. Since German – like English – prefers personal verbs as modals, it has not only transmitted the old lexical forms of “müssen” and “dürfen” into NorthWest Slavonic languages (see Hansen 2000), but is probably also, together with Latin, responsible for the transfer or at least the reinforcement of “to

192

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

have” as a modal.8 Therefore, the North-West Slavonic languages show remarkably often type 1a, even in the area of necessity (for more details see Hansen 2005). The borrowings have also affected Ukrainian and Belorussian, but not Russian: type 3 is here more relevant. The South Slavonic modality area is quite heterogeneous because of the affiliation of Bulgarian, Macedonian and, to a lesser extent, Serbian with the “Balkan Sprachbund”: the loss of the infinitive makes it impossible to use type 1, 3 or 5 in Bulgarian and Macedonian and restricts their application in Serbian where we observe a regional, stylistic and apparently individual differentiation: In the South-East (near Bulgaria) and in oral communication we find more constructions with da + finite clause (see Section 5.4), in the other regions and in written texts more infinitive. The clear West – East movement of the modals in the North Slavonic languages (from German to Sorbian/Czech/Polish and then to Ukrainian/Belorussian) brings up the question: What was first, the grammaticalisation of the modals in the source language or the borrowing process? According to our observations, at least in this case the grammaticalisation process was finished before borrowing took place. The only exception might be the Upper Sorbian (without status as a modal also in Czech, Slovak, Lower Sorbian) modal dyrbjeü, where the Slavonic languages conserved an older meaning (necessity), in contrast to German “dürfen” (deontic possibility). Since the western and southern border areas of Slavonic have undergone deep, long lasting and far reaching non-Slavonic influences, it is difficult to find in one of the contemporary languages the most “authentic” Slavonic modal system. Russian seems to have preserved the original modal structure best of all, although we cannot exclude the fact that the tendency to impersonality (type 3) has been influenced by Uralic substrate. 5. Conclusion This chapter is devoted to a description of the category of modals in the Slavonic languages. We understand modals as more or less grammaticalised expressions of necessity and possibility which form categories with fuzzy boundaries. Modals represent a special type of auxiliaries and are found in all Slavonic and in many other languages. We have tried to show that modals differ from fully lexical expressions of modality both in semantics and morpho-syntax. Modals are polyfunctional and show the syntactic

Modals in Slavonic languages

193

behaviour of auxiliaries. In contrast to Germanic, Slavonic modals are heterogeneous and form different types of constructions. We propose a cross-linguistic typology of modal constructions which is based on the morpho-syntactic coding of the arguments of the verb modified by a modal. Modal constructions in Slavonic vary in respect to the syntactic encoding of the privileged syntactic argument and the assignation of the agreement marking to the modal and/or the main verb. It is shown that these morphosyntactic construction types correlate with semantic features and different degrees of grammaticalisation. As a matter of fact, impersonal modal constructions where the privileged syntactic argument is realised either as a Dative subject or as syntactic zero are less polyfunctional and show more selection restrictions that personal modals with Nominative subjects. Notes 1. 2.

3.

4.

5. 6. 7.

One more language should be added to this family, namely Church Slavonic, which is nowadays actively used only as the liturgical language of the Orthodox Church. An interesting case is Russian smoþ’ ‘to be able’ which seems to be the perfective equivalent of the modal moþ’. However, both verbs have different semantics. Whereas moþ’ covers all types of possibility, smoþ’ expresses the actualisation of a participant-internal possibility (see van der Auwera and Plungian 1998: 106). These polysemy patterns are found in the following personal modals; POSS: all cognates of Common Slavic *mogti, Slovene lahko; NEC: Polish musieü, Czech musiet’, Upper Sorbian dyrbjeü, Lower Sorbian musaĞ, Russian dolžen, Ukrainian musyty, Belorussian music’, Slovene/Serbian/Croatian morati. This holds true for the following impersonal modals of possibility: Polish moĪna Ukrainian/Belorussian možna, Russian/Slovak možno, Belorussian l’ha; modals of necessity: Polish trzeba, naleĪy, wypada, Russian nado, sleduet, Slovene/Slovak/Ukrainian treba, Czech tĜeba and Ukrainian slid, naležyt’. Note that instead of TAM verbs, Noonan’s (1985) classification deals with phasals and modals. We group phasals under aspectuals, and add tense auxiliaries given that in Serbian they take da-clauses too. üe is the eroded form of hoüe. In the following tables the Serbian language occurs sometimes together with Bosnian and Croatian, sometimes as an isolated language. In the former case these three languages share a common structure with an infinitive, in the latter we have in mind the da-construction (finite clause) which is much more typical for Serbian than for Bosnian and Croatian (see section 3.4).

194 8.

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen ‘To have’ as a verb expressing necessity existed already in Old Church Slavonic (9-10th century). Therefore one could admit that it is inherited from Common Slavic. But this opinion is not uncontested.

References Anderson, John M. 2000 Auxiliary. In Morphologie. Ein internationales Handbuch zur Flexion und Wortbildung, (Handbücher zur Sprach- und Kommunikationswissenschaft 17/1), Geert Booij, Christian Lehmann, and Joachim Mugdan (eds.), 808-820. Berlin: Mouton de Gruyter. Benešová, Eva 1971 Syntax slovesné modality. Klasifikace sémantickych jednotek slovesné modality [The syntax of verbal modality. A classification of the semantic elements of verbal modality]. In ýas a modalita v þeštinČ [Tense and modality in Czech], Jarmila Panevová, Eva Benešová, and Petr Sgall (eds.), 97-143. Praha: Univ. Karlova. Besters-Dilger, Juliane 1996 Modalität und Negation (am Beispiel des Russischen). Wiener Slavistisches Jahrbuch 42, 7-31. Besters-Dilger, Juliane 1997 Modal'nost' v pol'skom i russkom jazykach. Istoriþeskoe razvitie vyraženija neobchodimosti i vozmožnosti kak rezul'tat vne- i mežslavjanskogo vlijanija [Modality in Polish and Russian. The historical development of expressions of necessity and possibility as the result of language contact between Slavonic and Non-Slavonic languages]. Wiener Slavistisches Jahrbuch 43, 17-31. Bondarko, Aleksandr V. (ed.) 1990 Teorija funkcional’noj grammatiki. Temporal’nost’. Modal’nost’ [A theory of Functional Grammar. Temporality. Modality]. Leningrad: Nauka Browne, Wayles 1986. Relative clauses in Serbo-Croatian. PhD dissertation. University of Zagreb. Browne, Wayles, and Theresa Alt 2004 A handbook of Croatian, Serbian and Bosnian. Durham, NC: Duke University, SEELRC. Available from: http://www.seelrc.org:8080/ grammar/mainframe.jsp?nLanguageID=1 [accessed 16.01.2007] Bybee, Joan, Perkins, Revere and William Pagliuca 1994 The evolution of grammar. Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press.

Modals in Slavonic languages

195

Cristofaro, Sonia 2004 Subordination. Oxford: Oxford University Press. Drobnjakoviü, Ana 2006. Expression of possibility and necessity: Serbian vs English and French. In The CamLing Proceedings, 381-388. Cambridge, UK: Cambridge Institute of Language Research. Grkoviü-Major, Jasmina 2004 Razvoj hipotaktiþkog da u starosrpskom jeziku [The development of hypotactic ‚da’ in Old Serbian]. Zbornik Matice Srpske za filologiju i lingvistiku 47 (1-2): 185-203. Hansen, Björn 2000 The German modal verb ‘müssen’ and the Slavonic Languages – The reconstruction of a success story. Scando Slavica 46, 77-93 Hansen, Björn 2001 Das Modalauxiliar im Slavischen. Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen. (Slavolinguistica 2) München: Otto Sagner. Hansen, Björn 2004 The boundaries of grammaticalization. The case of modals in Russian, Polish and Serbian/Croatian. In What makes grammaticalization? A look from its fringes and its components, Bisang Walter, Nikolaus P. Himmelmann, and Björn Wiemer (eds.), 245-271. (Current trends in linguistics Studies and Monographs 158) Berlin: Mouton de Gruyter. Hansen, Björn 2005 How to measure areal convergence: a case study of contact-induced grammaticalization in the German-Hungarian-Slavonic contact area. In Modality in Slavonic Languages. New perspectives (Slavolinguistica 6), Björn Hansen and Petr Karlík (eds.), 239-257, München: Otto Sagner. Hansen, Björn 2006 Na polputi ot slovarja k grammatike: modal’nye vspomogatel’nye slova v slavjanskich jazykach [Halfway from lexicon to grammar: modals in the Slavonic languages]. Voprosy jazykoznanija 2006/2, 6884 Hansen, Björn 2007 A morpho-syntactic typology of constructions with modals in Serbian. In Sintaksiþka istraživanja (dijahrono-sinhroni plan) (Lingvistiþke sveske 6) [Syntactic studies – diachronic-synchronic level. Papers in Linguistics 6], 31-44. Novi Sad: Filozofski Fakultet Hansen, Björn in press Modals. In Slavic Languages. An International Handbook of their History, their Structure and their Investigation. (Handbücher zur

196

Juliane Besters-Dilger, Ana Drobnjakoviü and Björn Hansen

Sprach- und Kommunikationswissenschaft). Peter Kosta, Tilman Berger, Karl Gutschmidt, and Sebastian Kempgen (eds.): Berlin: Mouton de Gruyter Hansen, Björn, and Petr Karlík (eds.) 2005 Modality in Slavonic Languages. New perspectives. (Slavolinguistica 6) München: Otto Sagner Heine, Bernd 1993 Auxiliaries. Cognitive forces and grammaticalization. Oxford: Oxford University Press. Jachnow, Helmut (ed.) 1994 Modalität und Modus. Allgemeine Fragen und Realisierung im Slavischen. Wiesbaden: Harrasowitz. Joseph, Brian Daniel 1983 The synchrony and diachrony of the Balkan Infinitive: a study in areal, general, and historical linguistics. Cambridge: Cambridge University Press. Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Mareš, František Václav 1980 Die Tetrachotomie und doppelte Dichotomie der slavischen Sprachen. Wiener Slavistisches Jahrbuch 26, 33-45. Noonan, Michael 1985 Complementation. In Language typology and syntactic description, Volume 2: Complex constructions, Timothy Shopen (ed.), 42-140. Cambridge: Cambridge University Press. Palmer, Frank Robert 2001 Reprint. Mood and modality. 2d ed. Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Plungjan, Vladimir A. 2003 Obšþaja morfologija. Vvedenie v problematiku [General morphology. An introduction]. Moskva: Editorial Ramat, Paolo, and Davide Ricca 1998 Sentence adverbs in the languages of Europe. In Adverbial constructions in the Languages of Europe, Johan van der Auwera (ed.), 187-277. Berlin: Mouton de Gruyter. Roeder, Carolin F., and Björn Hansen 2006 Modals in Contemporary Slovene. Wiener Slavistisches Jahrbuch 52, 153-170.

Modals in Slavonic languages

197

Testelec, Jakov G. 2001 Vvedenie v obšþij sintaksis [Introduction to general syntax]. Moskva: Rossijsk. gos. gumanit. un-t van der Auwera, Johan 1999 On the Semantic and Pragmatic Polyfunctionality of Modal Verbs. In The semantics/pragmatics interface from different points of view, Ken Turner (ed.), 50-64. Oxford: Elsevier. van der Auwera, Johan, and Andreas Ammann, with Saskia Kindt 2005 Modal polyfunctionality and Standard Average European. In Modality: Studies in Form and Function, Alex Klinge and Henrik Høeg Müller (eds.), 247–272. London/Oakville CT: Equinox. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality's semantic map. Linguistic Typology 1/2, 79-124. van der Auwera, Johan, Ewa Schalley, and Jan Nuyts 2005 Epistemic possibility in a Slavonic parallel corpus – a pilot study. In Modality in Slavonic Languages. New perspectives (Slavolinguistica 6), Björn Hansen, and Petr Karlík (eds), 201-217. München: Otto Sagner.

7. Modals in Baltic Axel Holvoet

1. Introductory remarks. Central notions Among the several classifications that have been proposed for the different types of modality, we have chosen to use the notions of epistemic, deontic and dynamic modality proposed in Palmer (1986) and basically retained in Palmer (2001, 8–10). Palmer (2001) subsumes deontic and dynamic modality under the notion of event modality, opposing both together to propositional modality which is viewed as comprising epistemic and evidential modality. While subscribing to the higher-level distinction between propositional vs. event modality, we differ from Palmer in setting evidentiality apart from modality. This will hardly affect our overview of the Baltic modals as evidentiality is not expressed by modal auxiliaries in Baltic. 1 Subdivisions within the domain of event modality are viewed as necessary by most scholars but solutions differ. We have opted for a classification in which deontic modality, a well-known and broadly used notion that also represents a link from modality to mood (cf. the connection of the imperative with deontic modality), figures as one of the main types rather than as a subtype of ‘participant-external modality’ (as in van der Auwera and Plungian 1998). This entails (as noted by de Haan 2006, 29) the necessity of operating with the notion of dynamic modality subsuming the non-volition-based types of event modality. 2. The inventory The Lithuanian and Latvian inventories of modal auxiliaries and other means to express the modal meanings of (dynamic, deontic and epistemic) possibility and necessity show little common elements going back to ProtoBaltic. Neither for possibility nor for necessity is there any nuclear modal common to both languages; only marginal items (such as Lith. derơti, Latv. derƝt ‘be fitting’) are shared. This points to a separation between the languages prior to the rise of more or less grammaticalised modals.

200

Axel Holvoet

For possibility Lithuanian has galơti, cognate with galia ‘power, might’, whereas Latvian has varƝt, cognate to vara ‘power, might’. For deontic possibility (‘be allowed’) Latvian has introduced drƯkstƝt, originally ‘dare’ (still the only meaning of its Lithuanian cognate drƳsti). Furthermore, both languages have semi-verbal expressions for deontic possibility: Lithuanian has valia (originally a noun meaning ‘will, free will, freedom’) plus the auxiliary ‘be’, while Latvian has brƯv (a truncated form of brƯvi, an adverb based on brƯvs ‘free’), which is also combined with the verb ‘be’. For necessity, the differences are even more striking. They are partly connected with the differences between the possessive constructions, which are a well-known source of necessitive modals. Lithuanian being a ‘have’language, the possessive verb turơti ‘have’ (original meaning ‘hold’, a meaning retained in Latvian; for the transition cf. Spanish tener ‘hold’ o‘have’) has also become the basic verb for all varieties of necessity. Lithuanian also has the less frequent privalơti ‘be under an obligation’, now only deontic in meaning but originally probably an expression of dynamic necessity (the noun prievolơ denotes compulsory works or service). Latvian has developed no verb ‘to have’ and retains the probably inherited construction with ‘be’ and the dative (cf. Latin mihi est lit. ‘to me is’), which is vestigially attested in Lithuanian and may be posited as Common Baltic (the meaning shift of turơti from ‘hold’ to ‘have’ being a separate Lithuanian development). The construction with ‘be’ and the dative (Latvian man ir, lit. ‘to me is’) has served as a basis for several modal constructions in Baltic; one of them, ‘be’ with the infinitive (also referred to as the predicative infinitive or the dative-infinitive construction) is common to both Baltic languages and also characteristic of Slavonic (cf. Fortuin 2000 for a recent in-depth investigation of the Russian constructions of this type); some other, more recent, expressions are found only in Latvian. The most important among them involved a relative clause embedded in an existential construction which developed into an inflectional form of the Latvian verb, the debitive: proto-Latvian *man (1SG.DAT) nav (be.PRS3.NEG) jƗ (REL.GEN.SG.N) Ɲst (eat.INF) ‘I have nothing (which) to eat’ > Modern Latvian man (1SG.DAT) nav (be.PRS3.NEG) jƗƝd (DEB.eat) ‘I need not eat’, man (1SG.DAT) jƗƝd (DEB.eat) ‘I must eat’. 2 Both languages have impersonal verbs expressing necessity: Lithuanian reikơti and Latvian vajadzƝt. Both are relatively recent as verbs: in Old Lithuanian and Old Latvian they were semi-verbal expressions comprising the auxiliary ‘be’ and a non-verbal, non-inflecting predicative form; in

Modals in Baltic

201

Lithuanian this was reikia, probably from reika ‘necessity’ (a noun), whereas Latvian vajag(a) is a borrowing from Fennic (Livonian vajag, Estonian vaja 3 ). Both were made into verbs by the addition of verbal endings; hence modern Lithuanian reikơjo (be-necessary.PST3) instead of Old Lithuanian reikia (necessary) buvo (be.PST.3), Latv. vajadzƝja (benecessary.PST.3) instead of vajaga bija (same structure). Impersonal expressions are available for possibility as well. For instance, Lithuanian has galima ‘it is possible’ (originally the neuter form of the present passive participle of galơti ‘be able’), combined with the auxiliary ‘be’ alongside the personal galơti. On the whole, the large share of impersonals in the modal system is a characteristic feature of both Baltic and Slavonic. 4 This overview would not be complete without a brief mention of further non-lexical means of expressing modality. Participles are used in both languages to express both possibility and necessity. Lithuanian has the socalled participle of necessity in -tinas, as in sakytinas ‘that has to be said’; judging by certain lexicalised instances such as priimtinas ‘acceptable’ it could originally also have expressed possibility. In Latvian (which once used to have the participle of necessity in -tins as well but lost it except for certain dialects) the original present passive participle in -ams/-Ɨms has virtually lost its original participial function and now expresses possibility or necessity: pƺaujama labƯba (originally ‘the corn that is being reaped’) now means ‘the corn that can or should be reaped’. Both the possibility and necessity meanings are retained if the participle is used predicatively in a copular construction (labƯba ir pƺaujama ‘the corn can or should be reaped’), but when the construction (in the lack of a nominativ subject) ceases to be copular and turns into a possessive construction with a dative possessor/modal subject, only the necessity reading is left (the participle is then in the default masculine singular form): man (1SG.DAT) ejams (go.PTCP.PRS.PASS.NOM.SG.M) ‘I must go’. 3. The Latvian debitive among the modals The Latvian system of modals displays one conspicuous feature already hinted at: part of the semantic domain usually covered by necessitive modal auxiliaries like must, should etc. is covered by an inflectional form of the verb, the so-called debitive. This form is traditionally described as a mood; why it is here considered in conjunction with the modal verbs will be made clear further on. The debitive consists of an invariable form obtained by

202

Axel Holvoet

adding the prefix jƗ- to the 3rd person present indicative form or (in the case of ‘be’), the infinitive, and some form of the auxiliary ‘be’. The syntactic properties of the debitive are peculiar: the modal subject, i. e. the noun denoting the person on whom an obligation is laid or the inanimate entity similarly subject to a necessity, is in the dative, whereas the object of a verb in the debitive is usually in the nominative in the standard language. 5 The nominative noun phrase can hardly be regarded as a true subject, however. 6 In the case of 1st and 2nd person pronouns as well as of the reflexive pronoun the accusative is used: 7 (1)

Tev mani jƗpavada. you.DAT.SG me.ACC DEB.accompany ‘You must accompany me.’

The debitive can thus be characterised as an impersonal form in the sense that it never opens a position for a nominative subject and shows no variation according to person. The debitive is a compound verb form, as (in modern Latvian) the verb form containing the prefix jƗ- cannot be used independently, outside combinations with bnjt. The traditional description of the debitive as a mood is rather inadequate in view of the fact that the debitive can itself inflect for mood, so that we can, for instance, oppose a debitive irrealis (conditional) man (I.DAT) bnjtu (be.IRR) jƗiet (DEB.go) ‘I would have to go’ to a debitive realis man (1SG.DAT) ir (be.PRS.3) jƗiet (DEB.go) ‘I must go’. A better alternative would therefore seem to be the notion of ‘modal affix’ (de Haan 2005, 36–37): modal affixes are defined as non-obligatory markers of modality, thereby opposed to mood, which would account for the superposition of mood on debitive marking. But the notion of ‘modal affix’ is not quite satisfactory either, because jƗ- cannot be described as an affix introducing a modal predicate, all other things remaining equal. The addition of jƗ- entails the necessary addition of the auxiliary ‘be’, and it radically changes the syntactic properties of the whole verb form, making it impersonal. Elsewhere (Holvoet 2001, 32–43) I have attempted to demonstrate that the debitive behaves, in many ways, as a combination of a full verb with a modal auxiliary, so that it could actually be described as an agglutinative form incorporating a modal auxiliary, much in the same way as the so-called potential mood of the Hungarian verb incorporates a modal -hat- ‘can’ (the resulting form being itself inflected for mood as well). We could provide an even closer analogue if we imagined an impersonal necessitive expression like Russian nado (combined with the auxiliary ‘be’, which has zero realisation in the present) to be

Modals in Baltic

203

combined with a full verb in such a way that the lexical component (corresponding to Russian nado) would be incorporated in the main verb whereas the auxiliary would remain separate. We could then establish a structural parallel between the Russian and Latvian structures in (2) and (3): (2)

Mne nado bylo I.DAT be-necessary be.PST.N ‘I had to leave.’

(3)

Man bija I.DAT be.PST.3 ‘I had to leave.’

uechat’. leave.INF

jƗ-aizbrauc. DEB-leave

The difference between these two constructions (apart from word order) is that jƗ-, the counterpart of Russian nado, is incorporated. Evidence in favour of regarding jƗ- as a kind of incorporated modal verb is the fact that the debitive construction shows tense inflection in two different places, as constructions with modal verbs do. In a construction like John must go we can alter the tense form of the modal (John must/had to go) and we can mark relative tense on the infinitive (must go/must have gone). We can do the same in the Latvian debitive construction, cf. (4)

(5)

Tev ir bijis you.DAT be.PRS.3 be.PTCP.PST.ACT.NOM.SG.M šƯ grƗmata. this book.NOM ‘You have been obliged to read this book.’

jƗizlasa DEB.read

RƯt tev jƗbnjt izlasƯjušam tomorrow you.DAT DEB.be read.PTCP.PST.ACT.DAT.SG.M šo grƗmatu. this book.ACC ‘By tomorrow you must have read this book.’

The only respect in which the debitive construction differs from a combination with the impersonal necessitive verb vajadzƝt is that the incorporation of jƗ- triggers the replacement of other infinitives than bnjt ‘be’ with the present 3rd person form, and the accusative object is replaced, in part of the cases, with the nominative. However we interpret the

204

Axel Holvoet

structure of the debitive, it cannot be denied that it covers a range of meanings usually covered by modal auxiliaries rather than by mood, and that a description of the Latvian modals would not be complete if necessitive verbs were not considered in their relationship to the debitive. 4. Modal particles It is a well-know fact that modal particles 8 often compete with modal verbs, especially in the domain of epistemic modality. The Baltic languages are no exception here, cf. the following Lithuanian example (in (7) we have the modal adverb galbnjt from gali bnjti ‘may be’): (6)

Sveþiai gali atvažiuoti apie aštuntą. guests.NOM can.PRS.3 arrive.INF about eight ‘The guests could arrive about eight o’clock.’

(7)

Sveþiai galbnjt atvažiuos apie aštuntą. guests.NOM can.PRS.3 arrive.INF about eight ‘The guests will perhaps arrive about eight o’clock.’

Occasionally we also find modal particles competing with modal verbs in the deontic and dynamic domains. A well-known example is Slovenian with its modal particle lahko ‘easily’ covering all types of possibility, also deontic and dynamic. (8)

Kako lahko ti pomagam? how easily you.DAT help.PRS.1SG ‘How can I help you?’

Latvian has an interesting modal particle covering certain deontic and, secondarily, epistemic meanings. It is the particle lai, which is originally a hortative particle 9 but has considerably widened its scope. Holvoet (2005) described lai as a marker of “interpretively” used deontic modality. Based on the work of Sperber and Wilson, the notion of interpretive use refers to utterances used to describe (and evaluate) other utterances rather than states of affairs. “Interpretive deontic modality” refers, then, to the use of deontic expressions to describe not so much an actual obligation as a negatively evaluated, unreasonable demand by some other person. The use of lai

Modals in Baltic

205

always presupposes a deontic statement either made by some other person or imputed to him by the speaker. Some languages have modal auxiliaries (wholly or partly) specialised in this meaning; we could refer here to Daniel Weiss’ description of the function of Polish mieü and German sollen (unpublished MS; the main ideas are outlined in Hansen 2001, 124 ff.). The following example is from Polish: (9)

Inni poszli do domu, a ja others.NOM go.PST.3PL home but I mam tu siedzieü. have.to.PRS.1SG here sit.INF ‘The others have all gone home, whereas I am expected to sit here.’

Instead of using a specialised modal verb containing reference to a directive speech act, Latvian uses a kind of echo construction in which the function of the imperative or hortative construction that would be used in the utterance referred to or imputed to another person is rendered with the aid of a construction with the hortative lai. A characteristic use of lai is in deontic requests, i. e., in questions whose aim is to elicit a directive rather than information (see Palmer 1986, 106–108): (10) Ko lai es what.ACC PTL I.NOM saku? say.PRS.1SG ‘What should I tell them?’

viƼiem they.DAT

In deontic requests, a deontic utterance is anticipated (on the interpretive nature of questions in general cf. Sperber and Wilson 1986, 231). The other interpretive uses of lai are echoic, i. e. they refer to an unreasonable demand formulated by another person or imputed to him. (11) Es lai palƯdzu zaglim! I.NOM PTL help.PRS.1SG thief.DAT ‘Me helping a thief!’ ‘How can you expect me to help a thief?’ Interpretive modal constructions with lai are usualy deontic, but an epistemic use has also developed, probably secondarily. One of the means

206

Axel Holvoet

of expressing disbelief in other people’s statements is to represent these persons as perversely willing things to be otherwise than the speaker is convinced they are, and this accounts for deontic expressions used to convey irony or scepticism: (12) Es lai bnjtu palƯdzƝjis zaglim? I.NOM PTL be.IRR help.PTCP.PST.ACT thief.DAT ‘How can you suppose I could have helped a thief!’ Inasmuch as the domain of necessitive modality is usually covered by modal verbs, 10 irrespectively whether it is used interpretively or not, these Latvian uses of lai could be grouped with Slovenian lahko among the rather exceptional cases of modal particles expressing non-epistemic modality. 5. Domains of modality As observed in other languages, the Baltic modals show instances of both polyfunctionality and specialisation. The restrictions we will discuss here are: (i) lack of epistemic function, (ii) restriction to deontic function. Those verbs that come closest to the average European modals by virtue of their personal character, the Lithuanian and Latvian modals for possibility galơti and varƝt, as well as Lithuanian turơti ‘have to’, cover the whole domain of modality from dynamic to epistemic. Restrictions appear when we look at the impersonal modals. In Lithuanian neither of the impersonal modals (galima, reikia) can be used epistemically, as opposed to Latvian, which has epistemic readings for necessitative impersonal modals. We will discuss this further on. (13) Latv. Tur there

vajag bnjt be-necessary.PRS be.INF

apraktai naudai. buried.DAT money.DAT

(14) Lith. *Ten reikia bnjti užkastiems there be-necessary.PRS be.INF buried.DAT ‘The money must be buried here.’

pinigams. money.DAT

Modals in Baltic

207

Restrictedness to deontic modality can have several causes. One possible cause is a non-modal source leading at once to a deontic interpretation; this is the case with Latvian drƯkstƝt ‘be allowed’, originally ‘dare’ (the original meaning is still retained alongside the new one): (15) Latv. Es istabƗ nedrƯkstu I.NOM room.LOC NEG.dare.PRS.1SG ‘I dare not / may not enter the room.’

iet iekšƗ. go.INF inside

We may assume that in its modal meaning, Latvian drƯkstƝt was originally a negative polarity item; 11 it has, however, lost its polarity and is now also used in affirmative constructions: (16) Latv. Tu drƯksti izsvƯtrot to kandidƗtu you may.PRS.2SG strike-out.INF those candidates.GEN vƗrdus un uzvƗrdus, kurus tu names and surnames.ACC REL.ACC.PL you neatbalsti [...]. NEG.support.PRS.2SG ‘You are allowed to strike out the names and surnames of the candidates you don’t support.’ (Internet) Another possible cause of restrictedness to deontic modality is a dynamic origin with subsequent extension to a deontic meaning and a loss of the dynamic source construction. This is observed for Lithuanian privalơti ‘be obliged’, which is nowadays used only deontically. (17) Lith. Privalai padơti tơvams. be-obliged.PRS.2SG help.INF parents.DAT ‘You ought to help your parents.’ The Lithuanian dictionaries give ample evidence of former dynamic use, as in (18):

208

Axel Holvoet

(18) Lith. Aš tave kromu pernešiu, you.ACC pick-a-back carry-across.FUT.1SG I.NOM tai tu neprivalysi per tiltą eit. then you.NOM NEG.have-to.FUT.2SG over bridge go.INF ‘I’ll carry you across pick-a-back, so that you won’t have to cross the bridge on foot.’ (Lithuanian Academy Dictionary, s. v.) This meaning is probably the original one; this is of importance for this verb’s behaviour with negation, cf. below. A third instance of restrictedness to deontic modality, occurring together with syntactic properties, should be noted for Lithuanian galima. In itself, this modal can express both types of event modality (dynamic and deontic), but when a dative is added for the modal subject, only the deontic reading obtains. Compare (19) (dynamic or deontic) and (20) (exclusively deontic): (19) Lith. Negalima laikyti šuns šiame kambaryje. NEG.possible keep.INF dog.GEN in this room ‘One cannot/may not keep a dog in this room.’ (20) Lith. Tau negalima laikyti šuns šiame kambaryje. you.DAT NEG.possible keep.INF dog.GEN in this room ‘You are not allowed to keep a dog in this room.’ The same is noted for Russian možno by Hansen (2001, 169). It is not clear whether conclusions may be drawn from this with regard to the argument structure of deontic and dynamic modals, as this restriction does not hold for the impersonal necessitative reikia. (21) calls for a dynamic reading in spite of the presence of a dative subject: (21) Lith. Tau reikơs laikyti automobilƳ, gatvơje, you.DAT be-necessary.FUT keep.INF car.ACC street.LOC nes garaže nơra vietos. for garage.LOC be.PRS.3.NEG place.GEN ‘You’ll have to park your car in the street, because there’s no place in the garage.’

Modals in Baltic

209

Still, the occurrence of similar restrictions in the case of Lithuanian galima and Russian možno cannot be a mere coincidence. We could perhaps invoke, in this connection, the distinction between ‘participant-internal’ and ‘participant-internal’ modality proposed by van der Auwera and Plungian (1998). One type of event modality that one would a priori expect to show no predisposition towards impersonal expression is ‘participant-internal’ modality, which, in the classification used here, is a subtype of dynamic modality. If an impersonal modal is provided with a dative subject (cancelling what could be a basic functional motivation for the use of impersonal modals: the possibility of providing the modal with an implicit generic subject), then the difference obtaining between the personal and the impersonal expression seems to be reduced to a difference between more or less ‘personal’ notions of possibility or necessity. The impersonal expression could, for instance, be associated with a possibility or necessity imposed on the subject from without. This would typically apply to every instance of deontic possibility or necessity, but not to every instance of dynamic possibility or necessity. The line of division between the domains in which expressions like man (1SG.DAT) negalima (NEG.possible) may or may not be used does not exactly coincide with that between ‘participantinternal’ and ‘participant-external’ modality, but that could be because within the domain of dynamic possibility the transition between ‘participant-internal’ and ‘participant-external’ is smooth in the case of animate subjects (as the predisposition of a situation towards the occurrence of an event may comprise both participant-external and participant-internal factors). In the case of dynamic necessity, the borderline is obviously sharper: participant-internal necessity is actually just a kind of strong volition, clearly opposed to external necessity which rests on external coercion. This could perhaps account for the different treatment of galima ‘one can/may’ and reikia ‘one must’ in Lithuanian. Correlations between restrictions with respect to type of modality and argument structure and argument reducing operations will be discussed below. Other instances of specialisation in modals have the character of general tendencies rather than of rigid restrictions. Latvian tends to use vajadzƝt deontically and the debitive dynamically, but there are exceptions to this general rule. The general tendency is reflected in the slightly different behaviour of vajadzƝt and the debitive when they are combined with negation (cf. below).

210

Axel Holvoet

6. Tense marking Tense is sometimes marked differently on epistemic and event modals. Epistemic modals are inherently tenseless, but their embedded predicate can be marked for tense; event modals are tensed and the time reference of the embedded predicate is connected to that of the modal by a relation of near-simultaneity or posteriority (the moment of the rise or realisation of a possibility or necessity preceding that of the occurrence of the event). The tense marking of epistemic modals shows some diversity because the time reference actually applies to the embedded predicate but may be marked on the modal verb. Romance and Germanic language use compound (anterior) infinitives in combination with present tense forms of the modal verb. Slavonic languages, which have no compound infinitives, mark past tense references on the modal verb. Compare Polish example (22) with its English counterpart (23): (22) Musiaáem widzieü tego must.PST.1SG.M see.INF that (23) I must have seen that man before.

czáowieka wczeĞniej. man.ACC before

Baltic languages have, in principle, both possibilities, and predilection for either of these models seems to be sensitive to foreign influence. Lithuanian would normally prefer marking anteriority on the modal verb: (24) Lith. Turơjai girdơti apie šƳ rašytoją. must.PST.2SG hear.INF about this writer ‘You must have heard about this writer.’ In Latvian the choice is not so obvious. One occasionally finds constructions with compound infinitives in the Western European (German) fashion: (25) Lith. Noteikti daudziem jƗbnjt dzirdƝjušiem certainly many.DAT.PL DEB.be hear.PTCP.PST.ACT.DAT.PL.M par šo kinofilmu. about this picture ‘Certainly many people must have heard about this movie picture.’

Modals in Baltic

211

This construction is, however, rather heavy and in the modern language there seems to be a tendency to use constructions of the Slavonic type instead – with past time reference marked on the modal verb: (26) FotouzƼƝmumos vajadzƝja bnjt Plutonam – taþu photographs.LOC be-necessary.PST be.INF Pluto.DAT but tas tika norakstƯts kƗ fotodefekts. that be.PST.3 put-down.PTCP.PST.PASS as photographic.defect ‘[The planet] Pluto must have been [visible] on those photographs, but this was put down to a photographic defect.’ (Internet) The strategy of marking past time reference on the infinitive by using a compound infinitive is not old in Latvian. The period of introduction of compound infinitives in modal constructions (not only epistemic, but also deontic) was probably the late 19th century. In 1864, Bielenstein states explicitly that constructions like morgen werde ich den Brief geschrieben haben müssen cannot be rendered in Latvian (Bielenstein 1864, 211). In modern Latvian this is routinely done by using a debitive containing the compound infinitive: (27) Latv. LƯdz rƯtdienai man jƗbnjt uzrakstƯjušam by tomorrow I.DAT DEB.be write.PTCP.PST.ACT.DAT.SG.M. vƝstuli. letter.ACC ‘I must have written the letter by tomorrow.’ 7. Irrealis marking and factivity Several Western European languages regularly mark counterfactivity in expressions with modal verbs. This can be illustrated by French aurait dû venir ‘should have come’ as against a dû venir ‘had to come’. Baltic has relatively little marking of counterfactivity in modal constructions, either grammatical or lexical. Both Latvian and Lithuanian have optional marking of counterfactivity on the modal verbs for possibility by means of the compound (past tense) irrealis form:

212

Axel Holvoet

(28) Latv. Fidels [...] bnjtu varƝjis bez bƝdƗm Fidel.NOM be.IRR can.PTCP.PST.ACT without worries baudƯt dzƯvi, taþu izvƝlƝjƗs dumpinieka enjoy.INF life.ACC but choose.PST.3 mutineer.GEN ceƺus. ways.ACC ‘Fidel [...] could have enjoyed life without any worries, but he chose the path of a mutineer.’ (Internet) The optional character of this marking is conspicuous especially in those cases where the preterite of the modal verb in the apodosis of a conditional period contrasts with an irrealis in the protasis (on such cases cf. Palmer 2001, 211–214): (29) Latv. Fidels [...] varƝja bez bƝdƗm baudƯt Fidel.NOM can.PST3 without worries enjoy.INF dzƯvi, ja viƼš bnjtu gribƝjis. life.ACC if he.NOM be.IRR want.PTCP.PST.ACT ‘Fidel could have enjoyed live without worries, if only he had wanted to.’ In Lithuanian, marking of counterfactivity on the modal verb is even rarer than in Latvian, where it has probably been stimulated in the past by the influence of German. 8. Passivisation of modals Unamenability to passivisation has often been mentioned as a characteristic feature of modal verbs. In the case of epistemic modals this feature is confirmed by the Baltic evidence as well, though there are certain problematic cases to be discussed below. Modal verbs expressing event modality, on the other hand, occasionally provide unambiguous instances of passivisation.

Modals in Baltic

213

(30) Latv. Ko gan visu nebnjtu varƝts what.ACC PTL all.ACC NEG-be.IRR can.PTCP.PASS.NOM.SG.M paƼemt lƯdzi, ja vien laikus zinƗtu. take-along.INF if only in time know.IRR ‘What couldn’t one have taken along if only one had known in advance [sc. what would happen].’ (A. EglƯtis) Epistemic modals do not passivise in Latvian. In contrast to such constructions with event modals as illustrated in (30), it is the embedded verb that passivises in epistemic constructions: (31) Latv. [Skaistas formas keramikas trauki vedina beautiful shape.GEN ceramic vessels.NOM induce.PRS.3 domƗt, think.INF] ka no tiem varƝja tikt dzerts that from them can.PST.3 be.INF drink.PTCP.PST.PASS arƯ medalus. also meadbeer.NOM ‘The beautifully shaped ceramic vessels suggest that meadbeer could also have been drunk from them.’ (Internet) The corresponding construction with a passivised modal auxiliary would be exclusively dynamic, never epistemic. In generative grammar, the unamenability of epistemic modals to passivisation would be seen as evidence for their status as raising verbs. Epistemic modals are one-place predicates assigning no external theta-role, which precludes passivisation. Explanations along these lines seem convincing enough, and the fact of passivisation being ruled out only for epistemic modals in Latvian seems to confirm them. In this connection it is striking that Lithuanian provides instances of what seem to be passive constructions based on epistemic modals as well. An instance would be (32): (32) Lith. [Tai, kad advokato ir jo snjnaus knjne the fact that in the body of the lawyer and his son

214

Axel Holvoet

buvo padaryta itin daug durtiniǐ žaizdǐ, rodo, were done many cut wounds shows] jog žudiko galơta bnjti that murderer.GEN can.PTCP.PST.PASS.N be.INF psichiškai nesveiko. mentally ill.GEN ‘The fact that a large number of cut wounds were inflicted to the lawyer’s body and that of his son suggests that the killer could have been mentally ill.’ (Internet) However, cases like this call for a special kind of explanation as this construction is evidential, i. e, evidential marking is here added on the marking of modality. Constructions with passive (as well as active) participles instead of finite verb forms can convey various evidential meanings, notably the inferential meaning of (32) (‘may evidently have been mentally ill’). The evidential constructions with passive participles are traditionally described as a special variety of the passive, but there are several objections to this, e.g., there are difficulties with deriving certain passive-like evidential constructions by applying the usual rules of passivisation. The necessity of setting Lithuanian passive-like evidential constructions apart from passives proper cannot be argued in detail here (this is done in Holvoet 2007, 97-104) but if we leave passive-like evidentials out of consideration, the impossibility of passivising epistemic modals holds for Lithuanian as well. 9. Negation A negation can, in principle, be added to both the modal verb and its infinitival complement, and it can also be added to both. This possibility is, however, fully exploited only in the case of verbs expressing possibility. In the case of necessitive verbs, the possibility of negating only the full verb, not the modal verb, is also exploited. This is observed, for instance, with Lith. deontic reikơti, where this construction produces an unambiguous deontic reading sometimes preferred to nereikơti, which is ambiguous between deontic and dynamic readings (cf. below):

Modals in Baltic

215

(33) Lith. Reikơjo neƳsileisti tǐ triukšmadariǐ. be-necessary.PST NEG.let-in.INF those troublemakers.GEN ‘You shouldn’t have let those troublemakers in.’ Negation of both necessitive and full verb is probably hardly used, however. There are formal restrictions in the case of the debitive: an internal negation is impossible (there are no such forms as *jƗneiet). In the domain of necessity, several irregularities in the functioning of negation may be observed. In the case of deontic necessity a negation occurring with the modal verb is usually interpreted as affecting the embedded predicate and the expression refers to the deontic necessity of refraining from a certain type of action. This is known from many languages and has been explained by invoking implicatures (cf. van der Auwera 2001). If there is no specialisation among necessity verbs in terms of deontic and dynamic modality, the negative variety will be ambiguous between lack of dynamic and negative deontic necessity. Lithuanian nereikia can mean both ‘one need not’ and ‘one should not’: (34) Lith. Tau nereikia taip garsiai rơkti. you.DAT NEG.be-necessary.PRS so loud shout.INF ‘You needn’t / shouldn’t shout so loud.’ If there is a distinction, as is the case in Latvian, the negation used with expression for dynamic necessity (in this case, the debitive) will usually mean lack of necessity (the reading Ƒ ¬p is quite rare); whether this necessity is dynamic or deontic cannot be specified. (35) Lith. Tev šodien nav you.DAT today be.PRS.3.NEG ‘You need not work today.’

jƗstrƗdƗ. DEB.work

With vajag, which has a stronger preference for the deontic reading than the debitive, the construction is ambiguous between lack of necessity (again, whether lack of dynamic or deontic necessity cannot be determined), and negative deontic necessity, i.e., the necessity of refraining from a certain course of action (with ‘raised’ negation, i.e., a negation

216

Axel Holvoet

actually affecting the embedded predicate though formally occurring with the modal, according to the formula: Ƒ ¬p instead of ¬ Ƒ p). (36) Tev šodien nevajag you.DAT today NEG.be-necessary.PRS ‘You need / should not work today.’

strƗdƗt. work.INF

Perhaps this is also a general principle: if an action is not required, it is usually irrelevant whether lack of dynamic necessity or lack of deontic necessity is involved. If negated deontic modals are usually interpreted as having a ‘raised negation’ (i.e., they denote the deontic necessity of refraining from a certain action), this may be partly due to the fact of true negated deontic modals (expressing the lack of deontic necessity) being relatively unimportant functionally: the negated dynamic modal extends to negated deontic modality as well. This does not mean that negated deontic modals do not exist, however; Lithuanian privalơti, which is always deontic, retains this meaning when negated: (37) Lith. Tu neprivalai taip you.NOM NEG.be-obliged.2SG so ‘You are not obliged to work so hard.’

sunkiai dirbti. heavily work.INF

Privalơti, however, has lost its original dynamic meaning (illustrated in (17) above), and perhaps the negated neprivalơti has concomitantly shifted from lack of dynamic necessity to lack of deontic necessity, thereby escaping the shift from ¬ Ƒ p to Ƒ ¬ p characteristic of deontic modals. 10. Personal and impersonal modals Both Baltic languages have several impersonal verbs and constructions encoding modal meanings. For necessity Latvian has (in addition to the debitive) the impersonal verb vajadzƝt; Lithuanian has reikơti. Such impersonal modals, characteristic of Baltic and Slavonic, are peculiar against the background of the modal verbs of average European languages. In English, must go has roughly the same syntactic distribution as go, whereas the syntactic properties of the Latvian impersonal construction vajag iet ‘one should go, it is necessary to go’ are completely different from those of iet ‘go’. Impersonality is an issue deserving more attention in

Modals in Baltic

217

the literature on the grammaticalisation of modals; I will return to this point below. The impersonal modals of Baltic pose interesting problems against the background of the controversy whether modal verbs are raising verbs or control verbs. This problem has been discussed in connection with personal modals such as must, can. It makes sense to argue that at least epistemic modals are raising verbs, as an epistemic predicate is a one-place predicate: it has one propositional argument, whereas dynamic and deontic modals may be said to have two arguments: a modal subject and a kind of object denoting an event. Could the raising hypothesis be extended to impersonal modals as well? This problem does not arise for the impersonal verbs of Russian or Lithuanian, which do not allow of an epistemic interpretation. Latvian impersonal modals do, however, as shown by example (38) (= (13)) with vajadzƝt. (38) Latv. Tur vajag bnjt apraktai naudai. there be-necessary.PRS be.INF buried.DAT money.DAT ‘The money must be buried there.’ Should we assume raising of an embedded clause subject to the position of dative complement of vajadzƝt here (in analogy to the subject-to-subject raising with personal epistemic modals)? There is probably no compelling reason to do so, as we have the possibility of regarding nauda as an embedded clause argument. In Latvian (and also, less frequently, in Lithuanian), infinitival embedded clauses may have overt dative subjects. Does the dative belong to the matrix clause or to the embedded clause? Questions of this kind can often be answered by using both semantic and syntactic (word order) criteria. In a sentence like It is rare for foreigners to visit this region we will interpret foreigners as embedded clause subject because a PP for foreigners is not a plausible complement with rare. In similar circumstances, datives will be described as infinitival clause subjects in Baltic. In the Latvian construction with the modal vajadzƝt, there is hardly any positive evidence for the interpretation of the dative as belonging to the embedded clause. Word order is free in Latvian, so that little evidence can be derived from it. Semantically, both possible interpretations (the dative as a main clause complement or as an embedded clause subject) amount to the same as, in the infinitival construction, a PRO subject in the embedded clause would have to be coreferential with the

218

Axel Holvoet

main clause complement. There is also a construction in which vajag can be used both with a dative complement and a finite complement clause. There is no coreferentiality between the dative complement and the complement clause subject in this case (39): (39) Latv. Man vajag, lai you.DAT be-necessary.PRS PTL ‘I need you to help me.’

tu man you I.DAT

palƯdzƝtu. help.IRR

There is also a construction in which vajag governs only a finite embedded clause but no dative complement (40): (40) Latv. Vajag, lai tu palƯdzƝtu vacƗkiem. be-necessary.PRS PTL you.NOM help.IRR parents.IRR ‘It is necessary that you should help your parents.’ Only in the latter case is the meaning of vajag purely modal (‘it is necessary’), whereas in (39) we have the premodal meaning of ‘experience a need’. Only the construction in (40) could therefore be viewed as synonymous with the infinitival construction in (41), in which we could view vajadzƝt as a modal verb proper: (41) Latv. Tev vajadzƝtu palƯdzƝt you.DAT be-necessary.IRR help.INF ‘You should help your parents.’

vecƗkiem. parents.DAT

This suggests that we could view the dative tev in (41) as an embedded clause subject as well, even though in this case the interpretation is deontic rather than epistemic. But, as noted above, we not not really have any means of establishing whether the dative in constructions like these belongs to the main clause or to the embedded clause, and in this respect we cannot really observe any difference between epistemic and, say, deontic uses. The only reason we could invoke for associating different syntactic interpretations with the epistemic and modality reading would therefore be the a priori consideration that epistemic modals have a different argument structure. Observation of the actual syntactic behaviour and use of modal

Modals in Baltic

219

verbs does not seem to confirm this. Modals may gradually extend their scope from dynamic and deontic to epistemic use, and this does not seem to involve a radical change in syntactic structure. Moreover, the Latvian constructions just discussed show that dynamic and deontic constructions can also be interpreted as having an embedded-clause-internal subject. So perhaps we could say that modal verbs are an area of indeterminacy between “raising” and “control”. Still, there are demonstrable differences between epistemic and event modality revealed precisely by impersonal modals. Latvian is the only language within Balto-Slavonic which uses impersonal modals in epistemic meaning. Generally, the stage of epistemic meaning does not seem to be readily accessible to impersonal modal verbs and constructions. The epistemic interpretation is lacking for Polish trzeba (Hansen 2001, 334– 336), naleĪy (Hansen 2001, 330), Russian nado (Hansen 2001, 187), nel’zja (Hansen 2001, 176), nadležit (Hansen 2001, 193), Lithuanian galima. An obvious conclusion is that impersonal modals are generally slower to achieve the final step in the grammaticalisation of modals, viz. the epistemic use, though the Latvian example shows that this epistemic use is not impossible. The general unamenability of impersonal modals to epistemic use seems to be reflected in the fact that Lithuanian, Russian and Polish use only their personal modals (Lith. turơti, Russ. dolžen, Pol. musieü to cite only necessitive modals) in this function while restricting their impersonal modals (Lith. reikia, Russ. nado, Pol. trzeba) to deontic and dynamic use. In the case of Latvian there was obviously no alternative but to extend the impersonal modal vajag and the equally impersonal debitive to the epistemic sphere, as no personal modal verb was available to express epistemic necessitive meaning. Why are impersonal modals less amenable to an epistemic interpretation? Probably for the same reason for which epistemic modals are not passivised. A feature shared by impersonals and the crosslinguistically most typical passive is that they are agent-suppressing. Event modal constructions are often agent-suppressing, e. g. they express the dynamic or deontic possibility or necessity of performing a certain action for a generic subject. Modal sentences (sentences with modal predicates) are more often generic than non-modal sentences, e. g. it is more common to state that a certain course of action is admissible for any subject than to state that it is followed by every possible subject, which explains the existence of subject-suppressing impersonal modals. Epistemic modals are

220

Axel Holvoet

never subject-suppressing because they superimpose no argument structure of their own on the argument structure of the predicate they modify. The similarity of impersonal modals like Lithuanian galima, reikia to impersonal passives also extends to the suggestion of animacy and humanness adhering to their modal subjects. As Frajzyngier (1982) has noted, the implicit subject of an impersonal passive is normally human; so is the dative modal subject with Lithuanian galima and reikia: even if ‘it may rain tomorrow’ is interpreted deontically (i. e., as meaning ‘I don’t mind if it rains tomorrow’), the use of the impersonal Lith. galima is blocked because an ambient subject like ‘rain’ cannot be used as the modal subject with this modal: (42) Lith. Dơl manĊs for my part

rytoj gali tomorrow may.PRS.3

lyti rain.INF

lietus. rain.NOM

(43) *Dơl manĊs rytoj galima lyti for my part tomorrow be-allowed.PRS.3 rain.INF ‘As far as I’m concerned, it may rain tomorrow.’

lietui. rain.DAT

11. Grammaticalisation An answer to the question whether the Baltic modals can be described as modal auxiliaries must depend on the criteria for auxiliarisation. Languagespecific criteria are often used (as in the literature on English modals, cf. Huddleston 1980), but generalisations based on a broader empirical basis are also available, cf. Heine’s list of auxiliarisation features mentioned in the literature (Heine 1993, 22–24). Some of these criteria would appear to be mutually contradictory. Heine’s list includes both decategorialisation, i. e., partial loss of grammatical features and possibile defectiveness of the paradigm (illustrated by the English modals), and the property of ‘carrying all morphological information relating to a predicate’ (Heine 1993, 23); this makes one wonder whether the epistemic modals, which mark fewer categories than deontic and dynamic modals (they tend not to mark tense, are not passivised etc.), are more or less grammaticalised than other types of modal verbs. Auxiliarisation obviously consists of two verbs (auxiliary and main verb) dividing between them a predicative content and the grammatical information relating to it, but the way in which the relevant

Modals in Baltic

221

features are spread over auxiliary and main verb may be language-specific and dependent on historical (also external) factors, as can be seen in the case of tense and mood marking in Baltic, cf. above. The evidence of word order is ambiguous as well: auxiliaries ‘tend to occur in a fixed order and in a fixed position in the clause’ (Heine 1993, 24), but their position is often considered in relation to that of the verb (as when the positioning of auxiliaries in VSO and SOV languages is examined). Should a typical auxiliary tend to accrete to the verb or should it rather have a fixed place in the clause? Obviously both ways of restricting the freedom of movement of the auxiliary reflect a kind of grammaticalisation, but such oscillations do not make it any easier to formulate a set of clear-cut criteria for auxiliarisation. A feature absent from Heine’s list of auxiliarisation features, probably because it is negative in nature (i.e., it could be viewed as a precondition rather than as a positive feature of auxiliarisation), is valency preservation. Can a modal be described as a modal auxiliary if it (i) induces a change of morphosyntactic marking of noun phrases and/or (ii) induces changes in syntactic valency? Discussing the notion of ‘specifier’, Zwicky (1993) mentions the fact that in (it) may rain it is the main verb (rain) that is responsible for the external representation of the whole phrase, e. g., it determines subject selection etc. This feature, which reflects a shift of head properties from the governing head to its complement, is certainly a valid criterion for auxiliarisation. If may is an auxiliary, then we expect the distributional properties of may rain to be similar to those of rain rather than of may. In characterising these distributional properties, one could, in principle, leave the morphosyntactic marking of noun phrases out of consideration and concentrate on syntactic valency and argument structure. It is not easy, however, to keep these two things apart, because the possibility of suppressing modal subjects is sensitive to morphosyntactic marking. The dative expressing the modal subject with Lithuanian galima in (20) can be dropped, which triggers a generic interpretation of this subject but is not subject to syntactic restrictions; by contrast, a nominative subject cannot simply be dropped in Lithuanian. Besides, as shown by (43), galima has specific requirements with regard to the animacy of the modal subject, which is also a matter of argument structure rather than of mere morphosyntactic marking. This makes Lithuanian galima into a typically subject-suppressing modal. In this sense, units like Lithuanian galima, Russian možno etc. clearly differ from the (morpho)syntactically ‘selfeffacing’ English modal: they determine the morphosyntactic and syntactic

222

Axel Holvoet

requirements of the phrase ‘modal + main verb’. Impersonality is thus clearly an obstacle on the way of a modal expression towards grammaticalisation, and languages where impersonal modals figure prominently can be characterised as languages with a low degree of grammaticalisation of modals. If it is not always clear what kind of behaviour one should expect, specifically, from auxiliaries as opposed to full verbs, there is still the possibility of testing the Baltic auxiliaries against the general criteria for grammaticalisation found in the literature. The most obvious choice is Lehmann’s set of criteria (Lehmann 2002). It is useful but applicable principally to processes involving advanced degrees of grammaticalisation, ending up in the rise of new inflectional forms. Modal verbs are always on the periphery of grammar, and such symptoms of grammaticalisation as they may show are rarely spectacular. Lehmann’s notion of the loss of integrity comprises loss of phonological substance and desemanticisation. The first (phonetic attrition) is not observed in Baltic: nothing even remotely reminiscent of English ’ll < shall/will can be found. Desemanticisation comprises such typical paths as the shift from physical strength or power to possibility, or from possession to necessity. The target meanings are more abstract, but arguably still lexical; claiming them to be inherently grammatical would be begging the question. For shifts within the domain of modality cf. below. In terms of Lehmann’s paradigmaticity the Baltic modals do not constitute a clear-cut and closed class. First, there is no set of morphological and syntactic properties setting apart modal verbs from other verbs. Criteria like those formulated for the English modals are not helpful in Baltic. There is nothing similar to the opposition between bare infinitive complements and to-infinitives (constructions with modals conform to the general pattern for higher predicate constructions; as mentioned above, no strong case can be made for setting the Baltic modals apart from predicates taking clausal complements in terms of the raising vs. control distinction); Baltic modals have both finite and non-finite forms, even though their infinitives are rare; and the criteria concerning negation, inversion etc. known from the literature on the English modals (can’t rather than *don’t can; can you rather than *do you can etc.) simply do not apply to Baltic as they are connected with specific (morpho)syntactic properties of English. Secondly, the means of expression for modal predicates are highly heterogeneous in Baltic, especially in Latvian; in addition to modal verbs Baltic has special means of several types: inflectional forms (the Latvian

Modals in Baltic

223

debitive); finite forms with a modal particle (Latvian lai for interpretivelyused deontic modality); constructions with modally marked participles (Latvian present passive participles, the Lithuanian participle of necessity). This contrasts with the situation in many other languages, where (at least outside the epistemic domain, where modal particles are frequent) the basic means to express modal predicates is to use modal verbs. Paradigmatic variability is not highly relevant to a discussion of grammaticality in modals as these are on the periphery of grammar anyway, and cannot be classified with inflection unless they develop, say, into tense or mood markers. Neither of this can be observed in Baltic: the future tense, for instance, is synthetic in both Lithuanian and Latvian; and both languages have synthetic irrealis and imperative forms, not involving modal verbs. Questions of obligatoriness and expansion are therefore hardly relevant. Reduction of functional scope is worth discussing especially in connection with the non-verbal modal means. The use of the Latvian modal adverb lai in a function usually performed in other languages by modal verbs is reflected in a reduction in scope: as a hortative adverb, lai is used sentence-initially, which reflects its status as a sentential adverb; in the interpretive uses, the tendency is to position lai immediately before the verb form, which seems to indicate reduction of scope to the verb phrase, as in the case of (at least deontic and dynamic) modal verbs. Cf. (44) and (45): (44) Lai visi nƗk man lƯdzi. PTL all.NOM.PL come.PRS.3 you.DAT along ‘Let everybody come along with me.’ (45) Es tev lai palƯdzu! I.NOM you.DAT PTL help.PRS.1SG ‘How can you expect me to help you?’ An extreme case of this reduction is, of course, the development of jƗ- from relative pronoun to verbal prefix. Syntagmatic variability is a criterion which is hard to apply to the Baltic languages because of their relatively free word order. The Baltic modals need not be adjacent to their complement verbs. Again, it is the non-verbal modals that provide evidence of reduction of syntagmatic variability: in addition to the tendency of lai to occur before the verb, we could mention

224

Axel Holvoet

the presumably fixed positioning of the relative pronoun in the protodebitive construction prior to its incorporation. In terms of bondedness, one striking feature in Baltic is the Latvian debitive marker jƗ-, which could be viewed as an instance of incorporation, as argued above. As mentioned above, the debitive allows a certain amount of internal inflection, but not, for example, insertion of a negation. With regard to semantics, a grammaticalisation chain has been discerned in the historically well-attested meaning shift from dynamic through deontic towards epistemic. To account for this, Traugott (1989) invokes a process of subjectification. Loss of argument structure, described as an important symptom of the grammaticalisation of modals by Traugott (2005; citing Roberts and Roussou 2003) seems to be a concomitant of this. If we set these semantic parameters alongside the morphosyntactic ones discussed above, there is, in principle, a possibility of conflicting criteria. A Latvian modal like vajag (or, mutatis mutandis, the Latvian debitive) is highly grammaticalised semantically in that it has reached the stage of epistemic modality; on the other hand, its valency-changing morphosyntactic features are not those of a typical highly grammaticalised modal. However, the very rarity of such instances of conflicting values suggests that they are exceptions confirming the rule of parallel semantic and morphosyntactic grammaticalisation. 12. Final remarks The characteristic features of the Baltic systems of modals are not all peculiar to Baltic: the impersonal character of many modals is shared with neighbouring Slavonic. This feature can be connected generally with a low degree of grammaticalisation of the modals. Baltic provides many instances of alternative means of expression of modality where the average European language would have modal verbs. Many of them can be found in Latvian, and seem to be connected with this language’s failure to develop a personal verb for ‘to have’ on which necessitive constructions could be based or modelled. This accounts for an atypical formation like the Latvian debitive, or the use of the modal adverb or particle lai for interpretive use of necessitive predicates. The lack of a personal necessitive modal in Latvian has led, paradoxically, to the corresponding impersonal expressions achieving a degree of grammati-

Modals in Baltic

225

calisation not generally accessible to impersonal modals (including epistemic use). Generally speaking, the Baltic modals are not advanced on the way towards auxiliarisation. Notes 1. For evidential meanings, both Baltic languages use constructions with active and passive participles (without auxiliaries) instead of finite verb forms. For further details cf. Holvoet (2007). 2. Historically, the prefix of the debitive derives from the Indo-European relative pronoun *ye/o-; the form jƗ- is a genitive, probably a genitive of negation, which shows that the oldest variety of the debitive was negative and expressed lack of necessity. A structural parallel is Russian neþego þitat’ ‘there is no point in reading’, originally ‘there is nothing to read’. Whereas in Russian the relative pronoun þego accreted to the negated existential verb net ‘there is not’, in Latvian it accreted to the verb thereby giving rise to a new inflectional form. For further details of this process cf. Holvoet (2001, 9-27). 3. These Fennic words are modal predicators whose original categorial status is hard to reconstruct (Bernhard Wälchli, p. c.). 4. Though some Slavonic languages, such as Sorbian, lack impersonal modals, as Björn Hansen kindly pointed out to me. 5. The actual state of affairs is different. In non-standard Latvian, the accusative is used with increasing frequency, which seems to be a result, on the one hand, of a natural tendency to replace the historically motivated but synchronically deviant nominative marking with accusative marking, and, on the other hand, of the fact that part of the Latvian dialects traditionally use the accusative rather than the nominative for the object of the debitive. 6. Another fact which suggests that the nominative is actually a nominative object is that it does not control verb agreement: in the compound perfect we have, for instance: Man ir bijis jƗizlasa šƯ grƗmata ‘I have been obliged to read this book’, with the participle bijis in the default masc. sg. form rather than in the fem. sg. form to agree with grƗmata. 7. Though the full debitive consists of the form containing the prefx jƗ- and the auxiliary ‘be’, the tag DEB refers, for practical purposes, to the prefix jƗ-, and the auxiliary is tagged separately. 8. It would be more accurate to use the term ‘modal adverb’, as it is impossible to draw a sharp line of division between adverbs and particles; here the term ‘particle’ is used because it is probably more familiar from the literature on modality. 9. Lai is a truncated form of laid, the imperative of laist ‘let’.

226

Axel Holvoet

10. In all the uses listed here, from deontic requests to epistemic use of interpretive deontic expressions as markers of disbelief, German and Polish use their specialised modal verbs sollen and mieü respectively. Some Slavonic and Baltic languages can use imperatives interpretively to refer to an unreasonable demand, cf. Lithuanian: Visi sơdi, o man dirbk. all.NOM.PL.M sit.PRS3 but I.DAT work.IMP ‘All are sitting about, but I’m expected to work.’ Similar constructions are known from Russian, where, however, the interpretive imperative is combined with a nominative subject. 11. This assumption is based on the fact that verbs with the meaning ‘dare’ are ‘achievement’ or ‘implicative’ predicates (Noonan 1985, 129); their affirmative variety is factive (a nonfactive use of How dare you … would be infelicitous), and this is, as Hansen (2001, 219) observes, one of the considerations preventing investigators from classifying ‘dare’ with the modal verbs. The modal meaning could therefore arise only in the negative variety, where factive implications are absent.

References Bielenstein, August 1863-64 Die lettische Sprache nach ihren Lauten und Formen erklärend und vergleichend dargestellt. I–II. Berlin: Ferdinand Dümmler. de Haan, Ferdinand 2006 Typological approaches to modality. In The expression of modality, Frawley (ed.), 27–69. Berlin–New York: Mouton de Gruyter. Fortuin, Egbert Lambertus Johan 2000 Polysemy or monosemy: Interpretation of the imperative and the dative-infinitive construction in Russian. Amsterdam: Institute for Logic, Language and Computation. Frajzyngier, Zygmunt 1982. Indefinite agent, passive and impersonal passive. A functional study. Lingua 58, 267–290. Frawley, William (ed.) 2005 The expression of Modality. Berlin–New York: Mouton de Gruyter. Hansen, Björn 2001 Das slavische Modalauxiliar. Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen. (Slavolinguistica 2) München: Otto Sagner.

Modals in Baltic

227

Hansen, Björn, and Petr Karlík (eds.) 2005 Modality in Slavonic Languages. New Perspectives. (Slavolinguistica 6) München: Otto Sagner. Heine, Bernd 1993 Auxiliaries. Cognitive Forces and Grammaticalization. New York/London: Oxford University Press. Holvoet, Axel 2001 Studies in the Latvian Verb. Cracow: Wydawnictwo Uniwersytetu JagielloĔskiego. Holvoet, Axel 2005 Evidentialität, Modalität und interpretative Verwendung. In Modality in Slavonic Languages. New Perspectives, Björn Hansen, and Petr Karlík (eds.), 95–105. München: Otto Sagner. Holvoet, Axel 2007 Mood and Modality in Baltic. Cracow: Wydawnictwo Uniwersytetu JagielloĔskiego. Huddleston, Rodney D. 1980 Criteria for auxiliaries and modals. In Studies in English Linguistics for Randolph Quirk, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (eds.), 65–78. London: Longman. Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Noonan, Michael 1985 Complementation. In Language typology and syntactic description. II. Complex constructions, Timothy Shopen (ed.), 42–140. Cambridge: University Press. Palmer, Frank Robert 2001 Reprint. Mood and Modality. 2nd Edition. Cambridge: University Press. Original edition, Cambridge: University Press, 1986. Roberts, Ian, and Anna Roussou 2003 Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: University Press. Sperber, Dan and Wilson, Deirdre 1986 Relevance: Communication and Cognition. Cambridge MA: Harvard University Press. Traugott, Elisabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55.

228

Axel Holvoet

Traugott, Elisabeth Closs 2005 Historical aspects of modality. In The expression of Modality, William Frawley (ed.), 105–139. Berlin–New York: Mouton de Gruyter. van der Auwera, Johan 2001 On the typology of negative modals. In Perspectives on Negation and Polarity Items, Jack Hoeksema, Victor Sánchez-Valencia, and Ton van der Wouden (eds.), 23-48. Amsterdam-Philadelphia: John Benjamins. van der Auwera, Johan, and Vladimir Plungian 1998 Modality’s semantic map. Linguistic Typology 2, 79–124. Zwicky, Arnold M. 1993 Heads, bases and functors. In Heads in Grammatical Theory, Greville G. Corbett, Norman M. Fraser, and Scott McGlashan (eds.), 292–315. Cambridge: University Press.

8. Modals in Albanian Walter Breu

1. Introduction From a genetic point of view, Albanian is an Indo-European language of the Eastern Group. Given the scarce documentation of Illyrian, it is still not clear if this language was the parent language of Albanian, as Albanian linguists have been claiming for a long time (Çabej 1987b), or if this role has to be attributed to Thracian or to both. The third possibility would be that Albanian forms a branch of its own, perhaps not even to be classified as belonging to the Eastern (Satem) group, due to the fact that it shows remnants of all three original dorsal consonant series, palatals, velars, and labiovelars. 1 From an areal point of view, Albanian is a Balkan language and as such, it is a member of the Balkan sprachbund, which to a high degree determines its typological properties (Demiraj 2004: 83–170). In many respects, it can even be claimed that Albanian is the most typical Balkan language as it lacks – in its standard variety – an infinitive, has a completely analytical comparative, forms the future with will (want), shows object doubling in the dative and accusative, a postponed definite article, a morphological genitive-dative syncretism and a very complex verb system which includes, among other things, a great number of tense grammemes, at least two aspectual oppositions, an optative and an admirative mood. On the phonological level, Albanian has a schwa (written ë) like Rumanian and Bulgarian. It shows lexical correspondences with Rumanian which go back to ancient times of neighbourhood in the central Balkans. With Greek it shares the synthetic forms for the mediopassive. A typical feature of the Slavic Balkan languages Bulgarian and Macedonian, the lack of a nominal declension, is absent in Albanian, however, which has four morphological cases including an ablative in the standard variety and still more in some dialects (locative, vocative). Even the morphological genitive-dative syncretism is disambiguated on the syntactic level by using a linking particle in the case of the genitive. Contrary to other Balkan languages, Albanian has a rounded high front vowel y in the standard language and in many dialects.

230 Walter Breu It is generally claimed that there is only one Albanian language with a certain number of varieties and dialects. But it has to be emphasized that the differences between these varieties and dialects are substantial, not only with respect to phonology and morphology but also concerning syntactic, structural and typological criteria. For example the Northern Geg (Gheg) dialect – or dialect group as it is to be subdivided into considerably different regional and local forms – has nasal and long vowels and forms the future with kam ‘have’, while the Southern Tosk dialect (or dialect group) has neither nasals nor an opposition of vowel quantity and forms the future with dua ‘want, will’. The problem becomes even greater if we look at the Albanian minority languages, e.g. in Greece and Italy, which have changed considerably toward the respective donor languages in their centurieslasting contact situations. The grammatical differences between the modern varieties of Albanian are also relevant to the field of modality, including to the category of modals. In order to give a consistent picture of the systems, I will first describe the class of modals in the standard language and then add the most important deviant properties of the systems of some varieties. Today’s standard language is the so-called gjuha letrare e njësuar ‘unified literary language’ and has a Tosk basis. During the younger history of Albanian, the Geg dialect was the basis of a standard language too, in principle until up to the end of World War II, when it was suppressed by the communists. It had survived for two more decades in Kosova (former Yugoslavia), where it also was replaced in 1968 by the ‘unified’ language, despite the Geg dialect base of that province. Today, some writers and also linguists in Northern Albania try to revive the use of the Geg based standard language (Breu 1997: 248–257). Albanian as a written language is relatively young (Breu 1997: 237– 248). The oldest book 2 we know is Gjon Buzuku’s Meshari (a translation of the Catholic missal) from 1555, followed by Pjetër Budi’s Dottrina Christiana from 1618, a translation of a catechism, but with several originally Albanian religious poems added. The first originally Albanian prose work of some length is Pjetër Bogdani’s Cuneus Prophetarum from 1685. Frang Bardhi wrote the Dictionarium latino-epiroticum, the first Albanian dictionary, which was published in 1635. All three authors used Geg varieties of Albanian. The first book based on the Tosk dialect, E mbsuame e krështerë, once again a translation of a catechism, was written by an Arbëresh (Italo-Albanian) of Sicily, Lekë Matrënga. 3 Early texts on a Tosk basis are very scarce in the Albanian mainland, where the true roots of to-

Modals in Albanian

231

day’s standard language can be found from only as late as the second half of the 19th century. 2. The category of modals in Albanian In Albanian, modality is to a high degree expressed by modals derived from fully conjugated modal verbs that show various degrees of reduction of their inflection. In Standard Albanian, among these more or less particlelike functional words, mund for POSSIBILITY and duhet for NECESSITY can be understood as the “centre” of the functional category of modality, together with the normally fully inflected – dua, expressing VOLITION. 4 In the following paragraphs the semantic, morphological, and syntactic properties of these central modals will be described. Given their wide range of polyfunctionality a certain degree of vagueness as to specific levels of modality, e.g. between dynamic and deontic NECESSITY, is typical for the modals of Albanian (and other languages). We try to give examples that are unambiguous in this respect on the basis of the larger context, but not always so in the isolated sentences. 5 2.1. The expression of [+POSSIBILITY] Positive POSSIBILITY is most typically expressed by the modal particle mund followed by the subjunctive. 6 The subjunctive in Albanian is formed morphosyntactially by the subjunctive particle (PTL) të and a verb form which is identical with the indicative except for the 2nd and 3rd person singular present active, where the subjunctive has morphological forms of its own. 7 There is no subjunctive in the aorist, so uninflected mund cannot be used in combination with this tense-aspect. The modal mund has no influence on the government of the full verb. The sentences in (1) are examples for the expression of dynamic POSSIBILITY in the present (1a), the active imperfect (1b), and the mediopassive imperfect (1c). 8 Example (1d) shows that mund can also be used in the future which is constructed normally with the verbal particle do ‘will’ + 9 SBJV, but again mund remains unchanged. (1)

a. Në shtëpinë e in house:ACC.SG.DEF PTL:POSS.ACC.SG.DEF

tij mund his can

232 Walter Breu të

gjesh gjithfarë tabelash. find:SBJV.PRS.2SG all kinds chart:ABL.PL.INDF ‘In his house you can find all kinds of charts.’ PTL:SBJV

b. Librin e botës book:ACC.SG.DET PTL:POSS.ACC.SG.DEF world:GEN.SG.DEF mund ta shkruante can PTL:SBJV-it:ACC.SG write:IMPRF.3SG vetëm ai. 10 only he:NOM.SG ‘Only he could write the book of the world.’ c. Kështu mund të rregulloheshin punët në thus can PTL order:IMPRF.PASS.3PL work:PL in këtë planet. this planet ‘In this way things could be set in order on this planet.’ d. Prisnin atë pjesë ku do të wait:IMPRF.3PL that part where FUT PTL:SBJV mund të shqyheshin gazit. can PTL:SBJV tear:IMPRF.PASS.3PL laughter:DAT.SG.DEF ‘They waited for that part where they would be able to burst out laughing.’ The examples in (2) show the modal mund with the present (2a) and the imperfect (2b) of the main verb expressing permission (deontic): (2)

a. Në rregull, mund të in order can PTL ‘Okay, you can (may) leave.’

largohesh. leave:PRS.PASS.2SG

b. U tha se mund të shkonin them say-AOR.3SG that can PTL go:IMPRF.3SG të flinin. PTL sleep:IMPRF.3SG ‘He told them that they could (might) go to sleep.’

Modals in Albanian

233

In the case of (weak) assumptions, in addition to the present (3a) und the imperfect, we often find the perfect – again in the subjunctive – as in (3b): (3)

a. Sot mbledhja mund të ketë Today meeting-DET can PTL have:PRS.SBJV.3SG lidhje me Kinën. link with China-DET ‘Today, the meeting could be connected with China.’ b. N. Bonaparti mund të ketë N. Bonaparte can PTL have:PRS.SBJV.3SG vdekur nga arseniku. die:PST.PTCP from arsenic ‘Napoleon Bonaparte could have died from arsenic.’ [Internet]

2.2. The expression of [–POSSIBILITY] Negative POSSIBILITY (impossibility) of all modality levels can be expressed by negating the modal mund. In the case of wide scope negations it is combined with the proclitic negative particles nuk or s’: 11 (4)

a. Ajo nuk mund ta kuptonte PTL-him understand: IMPRF.3SG She not can në këtë rast. (dynamic) in this case ‘She could not understand him in this case.’ b. Në ushtrinë tonë askush nuk mund të In army-DET our nobody not can PTL jetë robot. (deontic) be:PRS.SBJV.3SG robot ‘In our army, nobody may [not] be a robot.’ c. Ai nuk mund të ketë qenë një He not can PTL have:PRS.SBJV.3SG be:PST.PTCP an ëngjëll. (epistemic) angel ‘He can’t have been an angel.’ [internet]

234 Walter Breu When narrow scope is at issue, the full verb in the subjunctive is normally negated by the negative particle mos placed after the modal, between the subjunctive particle and the main verb: (5)

a. Mund të mos takosh të can PTL not meet:PRS.SBJV.2SG ART:PL njohur në çdo hap. (dynamic) known in every step ‘It is possible that you don’t meet acquaintances at every turn.’ b. Ti mund të mos shkosh. (deontic) you can PTL not go:PRS.SBJV.2SG ‘You are allowed not to go.’ c. Megjithatë mund të mos vejë puna nevertheless can PTL not go:PRS.SBJV.3SG work-DET gjer atje. (epistemic) until there ‘Nevertheless, it could be that the matter won’t go up to that point.’

Double negation is also possible, with the result of NECESSITY in the case of a logic doubling as in (6a), 12 but with no special effect in the case of obligatory syntactic doubling of negative pronouns as in (6b): (6)

a. S’mund të mos e ketë parë. not-can PTL not it have:PRS.SBJV.3SG see:PST.PTCP ‘He must have seen him. [Buchholz and Fiedler 1987: 68] b. As ai nuk mund ta zgjidhë këtë problem. nor he not can PTL-it solve:PRS.SBJV.3SG this problem ‘Not even he would [not] be able to solve this problem.’ [internet]

2.3. The expression of [+NECESSITY] In Standard Albanian, the typical modal to express NECESSITIY is duhet with reduced inflection. 13 Just like mund it is not inflected for person and number and it governs the subjunctive. Contrary to it, it is inflected for

Modals in Albanian

235

tense (present duhet, imperfect duhej); see (7) for examples of objective necessity, obligation and strong assumption: (7)

a. Ekonomia duhej të gërshetohej me tjetër economy must:IMPRF PTL twist:IMPRF.PASS.3PL with other gjë. thing ‘The economy had to be interlaced with another thing.’ b. Ne duhet të dalim jashtë për të we must:PRS PTL go:PRS.1PL out for PTL mos dëgjuar. not hear: PST.PTCP ‘We have to leave the room in order not to hear.’ c. Tani darka duhet të ketë now supper must-PRS PTL have:PRS.SBJV.3SG mbaruar, mendoi. end:PST.PTCP think:AOR.1SG ‘Now, supper should have finished, he thought.’

There is, however, a certain tendency to use duhet as a completely uninflected modal: 14 (8)

Ata duhet të zgjoheshin më në they must:PRS PTL wake up:IMPRF.3PL in the nga gjumi. from sleep:NOM.SG.DEF ‘They had to wake up finally from their sleep.’

fund end

Moreover, duhet can optionally transform the subject of the main verb into a dative. The sentences (9a) with the subject in the nominative and (9b) with a corresponding dative do not show any clear difference in modality. (9)

a. Unë duhet të nisem në aeroport në I:NOM must:PRS PTL start:PRS.PASS.1SG in airport in orën katër. clock four ‘I have to leave for the airport at 4 o’clock.’

236 Walter Breu b. Më duhet të jem në tre I:DAT must:PRS PTL be:PRS.SBJV.3SG in three e gjysmë në shtëpin e tij. and half in house-DET PTL his ‘I have to be at half past three at his home.’ According to the general rules of object doubling in Albanian, nominal subjects transformed into a dative as in (10) have to be doubled by a clitic pronoun: (10) Drejtorit i duhej director:DAT.SG.DET him must:IMPRF zgjuar. awake ‘The director had to stay awake.’

të PTL

rrinte stay:IMPRF.3SG

2.4. The expression of [–NECESSITY] Negative necessity is expressed by the modal duhet + SBJV combined with a negative particle. It can be claimed that duhet is most frequently used only for narrow scope, giving it the sense of ‘must not’ (necessity for not doing = prohibition). This is independent of the position of the negative particles, though there is a general preference to put them, in this case nuk or s’, before the modal as in (11a). When the negative particle is put “iconically” after the modal, the prohibition or the necessity for not realizing the state of affairs in question is very strict. In this case the negative particle for the subjunctive mos is used, see (11b). (11) a. Nuk duhet ta vështrosh kurrë njeriun not must:PRS PTL-him look:PRS.SBJV.2SG never man-DET që fle. REL sleep:3SG.PRS ‘You must never look at a person that is sleeping.’ b. Ai duhej he must:IMPRF

të PTL

mos dallohej për not distinguish:IMPRF.PASS.3SG for

Modals in Albanian

237

asgjë origjinale. nothing original ‘He had not to distinguish himself for anything original.’ Example (12a) shows the semantically redundant type of a syntactic double negation (negative pronoun and negative particle). Logical double negation with the positive result of an existing necessity appears in example (12b). In this case both negation slots – before and after the modal – are filled. (12) a. Askush nuk duhet të jetë mbi ligjin. no-who not must:PRS PTL be:3SG.SBJV above law-DET ‘Nobody may be above the law.’ [internet] b. Por njëkohësisht nuk duhet të mos e but simultaneously not must:PRS PTL not it shohim të vërtetën see:2PL.PRS PTL truth ‘But at the same time, it must not be that we do not see the truth.’ = ‘But at the same time, we must see the truth.’ More rarely, duhet can also be found with wide scope, i.e. with an outer negation in the sense of an action that is not necessary (‘need not’ be realized), as in (13) [from the bible, Jeremiah 2:24]: (13) Të

gjithë ata që e kërkojnë nuk duhet all they REL her seek:PRS.3PL not must:PRS të lodhen. PTL tire:PRS.PASS.3PL ‘All those who are looking for her need not make themselves tired.’ ART:PL

Contrary to the positive case, negated duhet only very rarely has an epistemic meaning. I did not find any example with the negation after the modal, and only relatively few with the negative particle before it, i.e. in the position of vagueness with respect to scope. In the following two examples [from the internet], the wider context is in favour of a narrow scope interpretation. But in principle wide scope (may not, needn’t) is not excluded:

238 Walter Breu (14) a. Nuk duhet të ketë qenë thjesht not must:PRS PTL have:PRS.SBJV.3SG be:PST.PTCP simple një surprizë. a surprise ‘It simply can’t have been a surprise.’ b. Ky qytet nuk duhet të ketë this town not must:PRS PTL have:PRS.SBJV.3SG pasur vikarë. have:PST.PTCP vicars ‘This town can’t have had apostolic vicars.’ To summarize, in both negated modality types, [–POSSIBILITY] and [–NEthe same lexical item is used for both scopes. This means that Albanian could be called an NPS (negation placement strategy) language with biscopal modals in the classification of de Haan (1997: 86–87). 15 In the case of [–NECESSITY] the position before the modal is vague, and only the position after it refers unambiguously to narrow scope, i.e. we have NT (negative transportation) in the sense of de Haan (1997: 87). 16 CESSITY],

2.5. The expression of [+VOLITION] Volition is normally expressed by the modal verb dua ‘want’ + SBJV of the main verb, which contrary to mund and duhet is fully inflected: (15) a. Dua të jetoj si njeri këto want:PRS.1SG PTL live:PRS.1SG like human these pak ditë. few days ‘I want to live like a human these few days.’ b. Gjithë ditën plakat duan të presin all day old women want:PRS.3PL PTL wait:PRS.3PL shoqet. friends ‘All day long the old ladies want to wait for their friends.’ As a main verb in the sense of ‘want (to have), love’ dua governs a direct object. The mediopassive counterpart to active dua is fully inflected duhem

Modals in Albanian

239

‘love each other, need, be needed’, from which the modal of necessity duhet is derived. Contrary to the cases of mund and duhet, the subjects of the modal verb dua and of the main verb it governs can be different. As the inflected form of the 2nd and 3rd person SG.PRS do, expressing VOLITION, is homonymous with the auxiliary particle expressing the volitional future (see below), an ambiguity arises in these persons which sometimes has to be avoided. One of the means to make clear that VOLITION is at issue is to insert the conjunction që ‘that’ as in (16a). This is a transfer from sentences like (16b), where the subject of dua and that of the main verb are different: 17 (16) a. Ti do që të vish? – you want:PRS.2SG that PTL come:PRS.SBJV.2.SG e pyeti ajo. him AOR:3SG she ‘Do you want to come? – she asked him.’ (=that you come) b. Duan që të tjerët të dëgjojnë. want:PRS.3PL that ART:PL others PTL listen:PRES.3PL ‘They want the others to listen.’ (=that the others listen) Epistemic modality seems to be excluded from the functions of the fully inflected modal verb of VOLITION. It can, however, be expressed in the same way as the future, including the future perfect as in (17), which is especially frequent in epistemic use, i.e. by means of uninflected do + SBJV. Again, this cannot be concluded from the 3SG (as for example in (17a)) but it is obvious in those persons where the inflected verb form does not coincide with the particle as in the 3PL (duan ‘they want’) in (17b). (17) a. Të

tillë boshllëk do të ketë such emptiness FUT PTL have:PRS.SBJV.3SG ndier edhe ai. feel:PST.PTCP also he ‘Such an emptiness he will (might) have felt, too.’ ART

b. Do të

kenë qenë në ndonjë mbledhje. FUT PTL have:PRS.SBJV.3PL be: PST.PTCP in some meeting ‘They will (might) have been at some meeting.’

240 Walter Breu In other words it is indeed possible to express epistemic modality by means of dua, but only as a secondary property of its post-modal future function where we find the uninflected form derived from it and not in the realm of VOLITION. It has to be kept in mind that in (17a-b) the do-construction has lost its temporal function of referring to the future (prediction) but simply expresses the speaker’s assumption about what could have happened in the past. 2.6. The expression of [–VOLITION] Negative VOLITION (i.e., rejection) is also expressed by fully inflected dua + SBJV. In principle narrow and wide scope negation are possible, but contrary to POSSIBILITY and NECESSITY the meanings of both are very close to each other, at least when the subjects of the modal and of the main verb are different. Again, a negative particle before the modal as in (18a) does not exclude narrow scope, but “iconic” positioning after the modal as in (18b) expresses the rejection of the action expressed by the main verb more rigidly. When the two subjects are identical, iconically expressed narrow scope is rather rare, but not excluded, see (18c): (18) a. Nuk donte të dëgjonte shkaqet not want:IMPRF.3SG PTL hear:IMPRF.3SG reasons:DET e refuzimit. PTL refusal-DET ‘He did not want to hear the reasons of the refusal.’ b. Unë dua të mos më keqkuptoni. I want:PRS.1SG PTL not me misunderstand:PRS.2PL ‘I want you not to misunderstand me.’ [internet] c. Donte të mos e bënte want:IMPRF.3SG PTL not it do:IMPRF.3SG të tillë. ART such ‘She wanted not to do such a thing.’

një a

gjë thing

Modals in Albanian

241

3. Other means of expression of POSSIBILITY and NECESSITY Apart from the modals discussed so far, Standard Albanian has three more means to express possibility: other uninflected forms, fully inflected modal verbs or verbal complexes, and grammemes of the category mood. Only some of them will be presented in the following sections. 3.1. Additional uninflected modals The modal feature [-POSSIBILITY] can be expressed by means of a negative sentence with the modal particle dot instead of a negated mund. This particle can never express [+POSSIBILITY], it does not govern the full verb (no subjunctive), and its typical position is after the verb. 18 It is restricted to the expression of objective (dynamic) impossibility/inability: (19) Gjergji nuk e gjente dot fjalën e George not it find:IMPRF.3SG can word:DET PTL përshtatshme. appropriate ‘George could not find the appropriate word.’ Contrary to the modal mund, the particle dot can also be combined with the aorist: 19 (20) As nga zyrat e aeroportit s'mësoi neither from offices:DET PTL airport:DET not- learn:AOR.3SG dot më shumë. can MORE much ‘From the offices of the airport he could not find out more either.’ It can also be combined with mund, which makes the impossibility more rigid, in the sense of ‘by no means, not at all’: (21) a. Gjysma e Evropës s'mund ta nxinte dot. half PTL Europe not-can PTL -it contain:IMPRF.3SG can ‘Half of Europe could not contain it (at all).’

242 Walter Breu b. Nuk mund të të harroj dot. not can PTL you forget:PRS.1SG can ‘I cannot forget you (at all).’ [+NECESSITY] can be expressed by lipset/lypset + SBJV or passive participle, but this construction is rare in the literature and is nowadays considered dialectal (Tosk, see Buchholz and Fiedler 1987: 189). For [+VOLITION] there is no uninflected modal in mainland Albanian, but see below for the situation in the diaspora (linguistic enclaves). 3.2. Fully inflected modal verbs and verb complexes 3.2.1. POSSIBILITY Although rarer than as an uninflected modal, mund can be used as a fully inflected modal verb. 20 This is true for all three modality levels of [+POSSIBILITY] and without any obvious semantic difference. Instead of the active forms, the mediopassive ones (mundem and so on) can be used. 21 As the uninflected modal mund cannot be combined with an aorist of the main verb (because of the general lack of an aorist subjunctive), the only way to express POSSIBILITY in the aorist is to use the fully inflected modal verb mund and put it into the aorist itself as in (22a). 22 As the aorist of the modal already expresses past tense, the temporally unmarked subjunctive present (instead of the imperfect) can also be combined with it (Demiraj 1986: 870). When the modal is used without a main verb, as in (22b), it normally has to be inflected, too. 23 Contrary to these cases, the inflected form is optional in the subjunctive construction of the “future in the past” as in (22c). 24 (22) a. Pastaj mundi ta kapte. then can:AOR.3SG PTL -it catch:IMPRF.3SG ‘Then he could catch it.’ (= ‘he succeeded in catching it’) b. As Marksi vetë s’do të mundte. neither Marx self not-FUT PTL can:IMPRF.3SG ‘Not even Marx himself would [not] be able to.’

Modals in Albanian

243

c. Po e dinte se s’do të PTL it know:IMPRF.3SG that not-FUT PTL mundte të rrinte. can:IMPRF.3SG PTL stay:IMPRF.3SG ‘But he knew that he wouldn’t be able to stay.’ Restricted to the sense of ability (capacity), the full verb di + SBJV can be used. As a full verb it means ‘know’, e.g. dinte të harronte ‘was able to forget’, nuk dinë të përqafohen ‘they are not able to embrace’. 3.2.2. NECESSITY Instead of governing the subjunctive, duhet can also be combined with the past participle. In this case it is most frequently used impersonally as in the positive example (23a) and the negative one (23b): (23) a. Duhet nisur pa humbur kohë. must:PRS start:PST.PTCP without lose:PST.PTCP time ‘We have to start without loosing time.’ (it must be started) b. Sidoqoftë fëmijëve s'u duhet treguar. anyway children:DAT.PL not-them must:PRS tell:PST.PTCP ‘Anyway, the children should not be told.’ Examples with an overt subject like (24a) and (24b) show, however, that in the participle construction duhet has to be classified as a form of the mediopassive fully inflected modal verb duhem, whose inflection is determined by the subject (mostly in the 3SG and 3PL). (24) a. U duhej mbyllur goja Them must:IMPRF.3SG close:PST.PTCP mouth:NOM.SG.DEF me çdo mënyrë. in any way. ‘Their mouths had to be closed in any way.’ b. Shpresat s'duheshin hope:NOM.PL.DEF not-must:IMPRF.3PL

humbur lose:PST.PTCP

244 Walter Breu kaq shpejt. so quickly ‘Hopes should not be lost so quickly.’ As shown in (23b) and (24b) negation normally has a narrow scope in the construction of duhet + past participle, just like in the case of duhet + SBJV. But contrary to the subjunctive case “iconic” negation after the modal is excluded. 25 An example for wide scope negation is (25): (25) Në të vërtetë s'duhet çuditur. in ART truth not-must:IMPRF.3PL surprise:PST.PTCP ‘Indeed, there is no need to be surprised.’ Inflected duhet (3SG.PRS) is of course the etymological source of uninflected (for person and number) duhet + SBJV, and duhej (3SG.IMPRF) is the source for its counterpart in the past. The expression of [+NECESSITY] is, however, not restricted to such mediopassive forms, but can also be expressed by the active do + PST.PTCP as in (26), i.e., by the same form with which the future is expressed when followed by the subjunctive: 26 (26) Do pasur parasysh cilat kanë want:PRS.3SG have:PST.PTCP before-eye which have:PRS.3SG qenë kriterët e emërtimit të popujve. be:PST.PTCP criteria-DET PTL nomination-DET PTL people. ‘It has to be considered which were the criteria in naming people.’ In Standard Albanian the construction do + PST.PTCP is normally restricted to the 3SG.PRS and could therefore also be classified as uninflected. But as the nevertheless existing plural examples show (duan ‘3PL.PRS’), 27 do has to be classified as an inflected form of the verb dua as well. As in the other cases mentioned, we are confronted here with the general problem of the historical derivation of auxiliary and modal particles from the 3SG.PRS in the Balkan languages and their still widely existing homonymy when referring to this grammatical person. With respect to their modal functions, both participle constructions with the mediopassive and the active modal verb duhem/dua seem to be more or less equivalent in expressing [+NECESSITY] in an impersonal or passive way, with the formal difference of diathesis being neutralized on the se-

Modals in Albanian

245

mantic level. 28 Contrary to the type duhet + SBJV both seem to be restricted to non-epistemic (active) modality. In colloquial speech, mediopassive fully inflected duhem is also used with the subjunctive, replacing the constructions of duhet (uninflected for person and number) + SBJV and fully inflected duhem/(dua) + PST.PTCP, preferred in the literary standard [examples from the internet]: 29 (27) a. Dokumentet duhen të paraqiten në document:DEF must:PRS.3PL PTL submit:PRS.PASS.3PL in gjuhën shqipe. language Albanian ‘The documents have to be submitted in the Albanian language.’ b. Këtu duhemi të jepim emrin tonë personal. here must:PRS.1PL PTL give:PRS.1PL name:DEF our personal ‘Here we have to give our personal name.’ c. Duhen të kenë marrë edhe must:PRS.3PL PTL have:PRS.SBJV.3PL take:PST.PTCP also lojën e shahut. game:DET PTL chess-DET ‘They must have taken over also the game of chess.’ (epistemic) Rarer than the modals of the dua-type, the fully inflected verb kam ‘have’ is also used to express necessity in Standard Albanian. In this function it is normally combined with the final complement construction, the so-called “compensatory” or “Tosk infinitive” of the type për të + PST.PTCP (Buchholz and Fiedler 1987: 85-86, 151). 3.3. Other lexical means to express modality [+POSSIBILITY] can also be expressed with the help of several lexical items that have the same root as the modal mund, e.g., in the constructions (nuk) ka mundësi + SBJV ‘there is a (no) possibility that’, (nuk) është i mundur që ‘it is (im)possible that’, është i pamundur që ‘it is impossible that’. They can replace mund in deontic and epistemic statements. Dynamic possibility can also be expressed by these constructions except for ‘ability’. There are also more specific lexical items, for example deontic lejohet + SBJV ‘it is

246 Walter Breu allowed’ uninflected for person and number, identical to the 3SG.PASS of lejoj ‘allow’. The modality levels of [+necessity] can be expressed by specialized lexical items like the dynamic kam nevojë + SBJV ‘I have to’, nuk kam nevojë + SBJV ‘have no need to’, which often replaces duhet in the case of wide scope negation, or deontic (nuk) detyrohem + SBJV ‘(not) to be obliged to’. None of them expresses more than one modality level and they have no epistemic functions. 3.4. Grammemes of the category mood As for grammemes of the grammatical category mood, the optative (as a wish), the free subjunctive (expressing that something should be done) and especially the jussive are used to express modality functions in Albanian. The jussive, which is formed with the uninflected auxiliary le + SBJV, 30 can refer to all three deontic modalities. It expresses most directly the position of the speaker in a sense of “resignation” (and therefore permission) or “preference” (and therefore weak volition or obligation). Even the imperative can sometimes have a modal character. But it seems that in all these cases, modality in the narrow sense – as it is understood in this volume – is expressed only as a secondary nuance of the primary functions of these grammemes. 31 4. Modals in non-standard varieties of Albanian The non-standard varieties of the Albanian language or language family have modality systems that are to a large extent very different from the Standard. In this paper I can only refer to some essential points. 4.1. The category of modals in modern Geg A very important feature of the Geg verb system is the existence of an infinitive. It is formed by the particle me + PST.PTCP and is used in many cases where Tosk-based Standard Albanian has to make use of the subjunctive. It cannot be used with the uninflected modal mund to express [+POSSIBILITY], however. Here and in the case of uninflected (for person and number) duhet to express [+NECESSITY], just like in the standard, the

Modals in Albanian

247

construction with the subjunctive is used. But contrary to the standard, the inflected modal verbs mund (active) and mundem (mediopassive) ‘can’, duhem ‘must’, and due ‘want’ can govern the infinitive, e.g. mundet me thanë ‘she can say’, duhet me qenë ‘he must be’, due me ditë ‘I want to know’. The construction of duhet + PST.PCTP is missing in Geg (Buchholz and Fiedler 1987: 189). In addition to the do-future, Geg forms the future with kam + Infinitive. This construction is also used to express NECESSITY, including epistemic modality (Buchholz and Fiedler 1987: 152–153). 4.2. The category of modals in Arbrisht (Italy) The Tosk-based Arbrisht (Italo-Albanian) dialects are characterized by a greater number of uninflected verb forms than Standard Albanian. This is true for auxiliaries that express temporal, aspectual, causative and modal meanings alike (Breu 1994). As far as the sphere of modality is concerned, mund (with phonetic variants) continues to express all kinds of POSSIBILITY in the Italian diaspora and has almost completely ousted (or never had) the fully inflected alternatives of Standard Albanian. It always governs the morphological subjunctive, but the subjunctive particle t(ë) is used only rarely. On the other hand, duhet ‘must’ is missing completely. Instead, the original full verb of ‘possession’ kam ‘have’, apart from continuing to function as an auxiliary for compound tenses, has extended its semantics to the full range of NECESSITY. This is a result of language contact, as in Southern Italian dialects the verb for ‘have’ also means ‘must’. 32 From this starting point, a post-modal future of the have-must-type has developed, formally still identical with the verbal complex expressing necessity, e.g. kat shkruaj ‘will/must write’, where kat < ka (have/ must:3SG.PRS) + të (PTL) governs the subjunctive and shows a tendency to be used also for other grammatical persons (reduction of inflection). The corresponding form of the imperfect is kish(t) < kish të, which has completely lost its inflection for person and number. 33 Moreover, many ItaloAlbanian dialects have developed a modal presumptive perfect, formed with kam ‘have/must’ + PST.PTCP. In spite of its formal identity with the Standard Albanian perfect, it has lost (or never had) the indicative meaning of a perfect, which can only be expressed by the aorist. 34 The construction of inflected dua ‘want’ + PST.PTCP continues to express NECESSITY and is used more frequently than in Standard Albanian,

248 Walter Breu while dua + SBJV continues to express VOLITION in verbal complexes, but only in the present. In other tenses uninflected forms are used, e.g. in the imperfect dot < do (3SG.PRS) + të (PTL) or dojt < doj (3SG.IMPRF) + të (PTL), where – as in the case of the uninflected modals of Standard Albanian – the inflectional categories are only expressed by means of the inflected subjunctive forms of the main verb; see Breu (1994: 373-374). While Standard Albanian allows a modal construction with different subjects for the modal dua and the governed main verb to be expanded into a construction of dua + object clause (initiated by the complementizer që ‘that’), this is excluded in many (all?) Italo-Albanian dialects. 35 4.3. Other varieties As was shown by Sasse (1991: 390–405), the Tosk-based Arvanitika (Greek-Albanian) variety in North-eastern Attica continues to use uninflected mund(Ω) tΩ + SBJV ‘can’ to express all modality levels of POSSIBILITY and inflected di tΩ + SBJV to express capability. It also uses mostly uninflected Greek loans for single levels, e.g. mborés ‘be able’, jé ikanó ‘be in a position to’, epitrépsetΩ ‘it is allowed to’, apaȖorépsetΩ ‘it is forbidden to’, 36 all governing the subjunctive (including the particle tΩ). In the same way, NECESSITY can be expressed by the fully inflected indigenous verb kám ‘have’, but also with the impersonal Greek loans préps ‘it is necessary’ and núkë xriasetë ‘need not’, as well as the fully inflected loan verb anangásem ‘be forced to’, all of them with the subjunctive of the main verb (Sasse 1991: 390, 393, and personal communication). In addition, a construction with active do ‘want’ is used, but – contrary to its Standard Albanian and Arbrisht counterpart – not with the past participle, but with a neutral verbal noun, which Sasse (1991: 406) calls an “infinitive”. Just like in Italo-Albanian, the negation is always put before the modal, irrespective of its scope (Sasse 1991: 375). The modality of VOLITION is expressed in the Standard Albanian way by means of fully inflected dúa + tΩ + SBJV. The post-modal future, which also expresses epistemic modality, is formed in the Standard Albanian way, too, at least as far as the particle do is concerned. It is often attached immediately to the subjunctive without the particle tΩ (Sasse 1991: 416), however. This means that the morphologization of the analytical future has gone even further than in Standard Albanian, where it is already morphologized to a higher degree than the modality constructions. The negative particle mos

Modals in Albanian

249

can, however, be put between do(t) and the main verb (Sasse 1991: 323, 373), thus preventing univerbalization. It is worth mentioning that in the Geg-based variety of the Arvanas in Croatia, uninflected dot + subjunctive is reported to be one of the means to express NECESSITY, 37 while VOLITION is expressed by the fully inflected auxiliary dua ‘want’ with the infinitive. The same verb serves also as a future auxiliary, in which case it governs the indicative (!). 38 In the Bulgarian diaspora, the Tosk-based Albanian variety of Mandrica has a wide variation for the expression of NECESSITY, including the Standard Albanian means of do ‘want’ (active base) + past participle and duhet ‘must’ (mediopassive base) + subjunctive, but in addition with fully inflected kam ‘have’ + subjunctive. The latter construction is also used to express the future. When negated kam loses its inflection, e.g. s’ka të zalláhim “it must we don’t speak”, i.e. ‘we must/will (PRS.3SG) not speak (PRS.SBJV.1PL)’. This has been induced by the corresponding Bulgarian construction with njama da. Another possibility of forming the future corresponds, in principle, to the Standard-Albanian way of using uninflected do, originally the 3SG of dua ‘want’, with the subjunctive. However, several types of phonological erosion have occurred in this construction. On the one hand, the subjunctive particle is always reduced, giving univerbalized dot, or it is even omitted. On the other hand, in both cases the initial consonant d can be omitted, yielding the proclitic future particles ot and even o, yielding the most highly grammaticalized future form of Albanian as a whole (Sokolova 1983: 137–138, 165). 5. The grammaticalization of the modals As stated in the introduction, Albanian is rather young as a written language. Our best historical basis for a comparison with modern Albanian is Gjon Buzuku’s missal from 1555, which fortunately is not a word-by-word translation but differs in many respects from its source. We can therefore presume that his system of modals is mainly the original Albanian one. Nevertheless, it has to be kept in mind that we have to do with the system of a Geg variety, which can be compared with today’s literary language only to a limited extent. This holds true especially for the use of the infinitive. 39 Nevertheless Buzuku’s work can give important information with respect to the historical development of the modern constructions. Besides such diachronic information, we can also draw some conclusions about the

250 Walter Breu grammaticalization path of Albanian modals from the synchronic system and its variation itself. In the following sections the Albanian modals are discussed with respect to the criteria given by Lehmann (2002: 108–143) and applied to Slavic modals by Hansen (2001: 402–425). 5.1. Parameter Integrity Apart from the tendency of the modals to develop into particles, which is strongest in the case of IMPOSSIBILITY expressed by dot, followed by mund for [+POSSIBILITY] and less so in the case of NECESSITY, expressed by duhet, and almost non-existent in the case of VOLITION, expressed by dua, phonological erosion is rather rare in the standard language. A case of erosion can be noted at least in Geg dialects where mund and its negative counterpart s’mund are strongly reduced. 40 As mentioned above, in ItaloAlbanian a tendency towards reduction is found for kam expressing NECESSITY (contrary to the epistemic presumptive), but also for dua (VOLITION). On the other hand, mund shows a certain variation of its vowel with schwa (mënd), 41 and a “loss” of its final consonant. This could, however, also be the original form, if we follow the widely accepted etymology of mund < mundë < mun + të (PTL). 42 It is precisely in this original amalgamation and the subsequent loss of the final schwa that we find the only clear case of a phonological erosion in the standard language. This seems to be repeated in modern trends of the colloquial language to omit the subjunctive particle after mund, and it is therefore possible to interpret these cases as mund < mund të. But here again we could also assume the preservation of an older system before the renewal of të after its original amalgamation with mun. 43 As for semantics, first we have to establish the pre-modal meanings of the words in question. When the fully inflected verb mund or the mediopassive mundem have a nominal complement, they mean ‘defeat, overpower, surpass, beat’. The same was true for Buzuku’s active full verbs mund and mundinj (Mansaku 1979: 140 and Fiedler 2004: 536, fn. 25). This cluster of meanings, which has “exert force (successfully)” as its common denominator, can thus also be assumed to be the source meaning of the uninflected modal mund and its inflected counterpart in modal complexes, the more so as all inflected verb forms with a -d- are secondary derivations from the impersonal modal mund. The etymology of the modal particle dot, which is restricted to objective IMPOSSIBILITY, remains unclear. Çabej (1987a: 296-297) discusses many

Modals in Albanian

251

possibilities from a Latin loan (in toto, totum) over an Indo-European *dhƝtim (as in Greek IJȓșȘȝȚ ‘put’) or another one cognate to Russian da, Greek įİ ‘yes’, but finally declares himself in favour of M. Lambertz’ etymology based on do të, which in modern Standard Albanian is used to form the future. 44 As for VOLITION, the original meaning of dua ‘want’ with its irregular aorist desha and past participle dashur 45 could have been ‘to love/like’ or ‘want’ or both of them, 46 as its cognates are Avestan zaoša ‘desire’, Old Persian dauštar ‘friend’, Old Indian juúáte ‘is content’, Greek ȖİȪȠȝĮȚ, Latin gustare ‘enjoy’. 47 The semantic change from ‘want’ or ‘want to have’ into the sphere of necessity in the sense of ‘need’ and then to ‘must have’, ‘must, have to’, to be found in the constructions of the type do + PST.PTCP, is a well known transition, for example in the Romance languages. 48 As mentioned above, the most typical modal for the expression of NECESSITY, duhet with its different degrees of reduced inflection, goes back to the mediopassive of dua ‘want’. Here again we have the transition from ‘want’, or rather ‘be wanted’ to ‘must’. 49 Used with a nominal complement the inflected mediopassive forms mean ‘be needed’. This could in principle be the linking point for the use of the (passive) past participle “is needed (to be) done”, from where the impersonal subjunctive construction with duhet të “it is needed to” could have been derived. The optional use of the “subject” in the dative with impersonal duhet is the original state of affairs, while the use of the nominative is a step towards the loss of its syntactic independence and thus its morphologization. Another one is, of course, its tendency to lose its inflection for tense and number. The opposite development of using fully inflected duhem + SBJV to express NECESSITY, found in substandard varieties, could indicate a new strategy of Albanian, away from the old Balkan way of grammaticalization of verbal complexes by the loss of inflection in the auxiliary. This can be considered a case of degrammaticalization. 5.2. Parameter Paradigmaticity Albanian modals form a small group with three members in the centre. This could be taken as a tendency to form a paradigm of modals. It has to be considered, however, that their morphosyntactic behaviour is rather different. While mund (POSSIBILITY) is normally uninflected and always governs the subjunctive, duhet (NECESSITY) is normally either inflected only for

252 Walter Breu tense, when it governs the subjunctive, or fully inflected, when it governs the past participle. The third modal, dua (VOLITION), is normally inflected and behaves morphologically like a full verb, when it means ‘want (to have)’, but can also govern a past participle, when it means ‘must’. Active dua and mediopassive duhet are etymologically forms of the same verb. On the other hand, mund has a fully inflected synonym with active and mediopassive inflection to be used obligatorily in the aorist and in isolated position with the governed main verb omitted, while full mediopassive inflection of duhem + SBJV is apparently restricted to (regional) substandard. In any case, dot seems to be outside the paradigm, as it can be combined with mund, in order to reinforce its meaning, and also because it is fixed in precisely one meaning, i.e. that of dynamic impossibility, while mund and duhet express the whole range of modality levels, from dynamic over deontic up to epistemic. In the short time span in which Albanian has been documented, little change has occurred in this “paradigm”, at least as far as the mainland varieties are concerned. In Italo-Albanian, the verb kam ‘have’ has spread all over the sphere of NECESSITY, perhaps replacing an older duhet, while the active paradigm of dua has kept or even strengthened this function when used with the past participle. For Italo-Albanian it is especially important that there is a tendency to turn kam ‘have’ and dua ‘want’ into particles, which started in the imperfect, from where it has spread to other tenseaspects. In this way, a real paradigm seems to be developing with similar morphological characteristics for all modal verbs, although this tendency is not restricted to modals alone, but to all types of auxiliaries, including even those forming the causative or expressing aktionsarten. 5.3. Parameter Paradigmatic Variability All modals are in principle compatible with all verbs, but they are never obligatory, i.e., there is no need or overt means to express the absence of all three modalities by a morpheme or a specific modal.

Modals in Albanian

253

5.4. Parameter Scope Though it seems obvious for the modal verbs dua, duhet and also kam (Italo-Albanian) that they have started from constructions with a nominal complement which could later have been substituted by a verb, we have no historical evidence for such a development. In all these cases both possibilities continue to exist. Only in the case of mund and, to the extent in which they have transformed into uninflected particles, also in the case of duhet, dua and kam, a nominal complement is not possible. 5.5. Parameter Bondedness Standard Albanian modals have not developed into clitics, nor are they agglutinated to the main verb. In dialects there is, however, a tendency for mund to amalgamate with the negative particle s’. In Italo-Albanian the position immediately before the main verb (in the subjunctive) is so characteristic for all kinds of uninflected auxiliaries including the modals that one could think of proclitics or even prefixes, the more so as they have incorporated the subjunctive particle that separated them originally from the main verb. The fixed position of the negative particle before the modal also contributes to such a tendency. 5.6. Parameter Syntagmatic Variability In the standard language relatively few elements can be put between the more or less uninflected modal and the main verb, be it in the subjunctive or the past participle. Nevertheless neither subject, nor objects or adverbs are excluded (Buchholz and Fiedler 1987: 68). For example in the case of VOLITION, the modal verb dua, which in the 2/3SG is identical with uninflected do expressing future, can be separated from the subjunctive particle të, with the reflexive vetë between do and the subjunctive complex. When expressing the future, however, do cannot be separated form të and could be classified as proclitic to the verb: (28) Ti më duket se nuk you me seem:PRS.PASS.3SG that not

do vetë want:PRS.2SG self

254 Walter Breu të

telefonosh. phone:PRS.SBJV.2SG ‘It seems to me that you yourself don’t want to phone.’

PTL

The negative particles are placed before the modal in the case of wider scope, but inside or outside the modal-verb complex in the case of narrow scope. As was mentioned above, in some varieties there is a certain tendency for the uninflected modals to lose all syntactic links to the main verb by giving up their characteristics of governing the subjunctive and even by losing their typical position before the main verb. For dot expressing impossibility this seems to always have been the case, due to its origin from the reduplicating predicate tag. The Geg infinitive is excluded from being governed by the uninflected modal mund ‘can’, a situation which can be found already in Buzuku and the other ancient texts. The same is true for uninflected duhet and do. The infinitive is used, however, from the very beginning with the inflected modal verbs mund, mundem (POSSIBILITY), 50 kam (NECESSITY), and dua (VOLITION). The reason for this difference can be found in the requirement of having at least one element in the verbal complex expressing the finite categories, either the fully inflected modal verb or the fully inflected subjunctive of the main verb. For Albanian as a whole, this is part of a tendency to inflect modal complexes in exactly one position, which means in the case of the Tosk substitution of the infinitive by the subjunctive (typical for Balkan languages) that it is the modals’ turn to lose their inflection. A confirmation for this assumption can be seen in the tendency of duhet ‘must’ to lose its remaining inflection for tense and ultimately also in a tendency of the colloquial language to use the present of the subjunctive with mund ‘can’ in the aorist, in order to prevent the past-tense feature from being expressed twice. 6. Conclusions I started this paper with a general overview of the history and the varieties of Albanian. The subsequent description of the modal systems was mainly based on the situation of modern Standard Albanian. I pointed out that the three modalities of (positive and negative) POSSIBILITY, NECESSITY and VOLITION are expressed firstly by a set of more or less uninflected modals, gov-

Modals in Albanian

255

erning the subjunctive of the main verb and being able to express all modality levels: dynamic, deontic, and epistemic. Out of these three modals, mund ‘can’ remains completely uninflected, while mediopassive duhet ‘must’ is inflected only for tense, and its active counterpart from the same root, dua ‘want’, is fully inflected when it expresses modality. It becomes, however, an uninflected auxiliary when it is used to form the post-modal future. This situation is seriously complicated by the fact that the fully inflected verbs from which mund and duhet are derived continue to express modality, too. Under certain conditions the use of inflected mund (active) and mundem (mediopassive), also governing the subjunctive, is even obligatory. Fully inflected mediopassive duhem governs the past participle, and so does its active counterpart dua when it expresses NECESSITY instead of VOLITION. The picture becomes still more intricate if we consider the particle dot, the most grammaticalized item from a morphosyntactic point of view (fixed position after the verb, no inflection, no government), but restricted to the particular case of dynamic IMPOSSIBILITY and freely combinable with mund ‘can’, whose meaning of “inability” it makes more rigid. In summary it may be concluded by and large that the degree of inflection of the modal depends on the modality type and that within the individual types there is a tendency to reduce the degree of inflection of the main verb to the extent the modal verb is inflected. Table 1 below tries to give a rough sketch of the main features of the modal system of Standard Albanian, including the post-modal future (tense, not modal particle do) with its secondary epistemic function, but ignoring minor important secondary means of expression like purely lexical ones or certain grammatical categories. The colloquial tendencies towards a complete loss of inflection in the case of duhet + SBJV and towards a reduced expression of tense in the subjunctive in the case of fully inflected mund(em), as well as the substandard use of inflected duhem + SBJV, are not included either. With respect to negation, Albanian is a “negation placement strategy” language with biscopal modals and “negative transportation” to the position before the modal in the case of narrow scope. In the non-standard varieties of Albanian, the modal systems show substantial differences compared with the Standard. The most evident case is probably the use of kam ‘have’ as a modal expressing all levels of NECESSITY and partially or fully ousting duhet (or preventing its development). In the Geg varieties the infinitive is generally used instead of the participle, but also instead of the subjunctive which resulted in a more restricted use of the uninflected modals. In Italo-Albanian varieties an oppo-

256 Walter Breu site development has taken place, with a strong tendency to increase the number of uninflected particles in the whole sphere of modality and beyond, thus strengthening the position of the subjunctive. Many ItaloAlbanian dialects have developed a deductive evidential (presumptive), formed with fully inflected kam, which is clearly restricted to inference and could thus be included in the realm of epistemic modality. In all Albanian linguistic enclaves language contact has played an important role in the realm of the modals. Table 1. The Standard Albanian modal system

increasing inflection (modal)

increasing inflection (modal)

main modal

alternative 1

alternative 2 dot only [–POSS.] modal particle

[+POSSIBILITY]

mund + SBJV mund(em) + SBJV uninflected modal inflected modal

[+NECESSITY]

duhet + SBJV duhem + PST.PTCP dua + PST.PTCP inflected for tense inflected modal inflected modal

[+VOLITION]

FUTURE

dua + SBJV inflected modal do + SBJV uninflected AUX

Notes 1. See Jokl (1963: 123-127) who confirmed Holger Pedersens theory with the supposed developments: kj > th [T@ k > q [c], and kw > s [s] before a front vowel. 2. There are a few smaller testimonies of ancient Albanian writing, the most important being the baptism formula from 1462, and there are rumours that one or more unknown large texts of medieval Albanian are kept in the Vatican Library. 3. See Fiedler (2004) for a description of the verbal system of Buzuku’s work and Matzinger (2006) for a linguistic introduction to Matrënga’s catechism.

Modals in Albanian

257

4. There are many definitions for modality in the linguistic literature with different terminologies and different extensions as to the categories concerned. This paper is based on a fairly restricted concept which corresponds more or less to the narrow understanding of modality claimed by Hansen (2001: 54–81) and – with the addition of VOLITION – to that of van der Auwera and Plungian (1998) as well as – with some restrictions – to that of Palmer (2001: 24-106). In the following, I will use the term “POSSIBILITY” as a cover term for all its levels, e.g. – in the positive case – objective (alethic/dynamic) possibility and ability/capacity, permission (deontic) and subjective (epistemic) possibility, i.e. weak assumptions. Vice versa, the same is true for negative POSSIBILITY (inability, prohibition etc.). “NECESSITY” covers in the same way objective (alethic/dynamic) necessity, obligation (deontic) and strong assumptions (subjective/epistemic necessity) and its negative counterparts. “VOLITION” refers to the (deontic) will and to assumptions expressed by the same formal means. These onomasiological (from function to form) terms have to be kept apart from the semasiological terms “possibilitative” and “necessitative”, used for morphosyntactic categories with similar means of expression in Albanian (Buchholz and Fiedler 1987: 133–153) but referring in the first place to (postmodal) future grammemes. We will use the terms “necessitative” and “volitional future” instead. On the other hand, Buchholz and Fiedler’s (1987: 160163) “volitive” refers clearly to the functions of [+POSSIBILITY] expressed by the formal marker mund. 5. It would be an interesting task to draw the borderlines between vagueness and polysemy (and also homonymy) in the semantics of Albanian modals following the criteria given by van der Auwera (1999: 58–63), but this is beyond the scope of this paper. 6. Modals like mund ‘can’ and also duhet ‘must’ are normally called “semi-auxiliaries” by Albanian grammarians, as they stand between fully inflected auxiliaries as jam ‘be’ and kam ‘have’, used to form complex tense grammemes, and particles like po, used to form a progressive aspect, in having a strongly reduced inflection but a firm syntactic connection to the governed main verb in the subjunctive. As Albanian lacks an infinitive, we use the 1st person singular present indicative active as the citation form of full verbs. But impersonal verbs have to be cited in what is historically the 3rd person singular. 7. The irregular verbs jam ‘be’ and kam ‘have’ also have separate subjunctive forms for the 1st person singular and for the 1st and 3rd person plural. 8. If not stated differently, Standard Albanian examples come from ELSNET’s European Corpus Initiative Multilingual Corpus I (ECI/MCI) version of the romance of Ismail Kadare, Koncert në fund të dimrit (Tiranë 1978) and are cited, as a rule, in abridged form or with omissions. 9. As a matter of fact, this is an example of the “future in the past” (homonymous with the conditional), where fully inflected verbs show the subjunctive of the imperfect, morphologically identical with the imperfect indicative, but with the

258 Walter Breu

10.

11.

12.

13. 14.

15.

16.

particle të as its morphosyntactic marker. Together with the fact that mund governs the subjunctive, its use in such complex forms as the future of the past restricts its status of a “particle” to a strictly morphological point of view. Despite its lack of inflection it remains a verb form. We will see in section 3.2.1 that inflected verb forms expressing POSSIBILITY can also be used in such complex constructions. In this sentence, we find a genitive expressed by the genitive particle e (here in the ACC because librin is in this case) and the GEN/DAT-form botës. The subjunctive particle të appears here as part of the fused ta < të + e (crasis), where e is the weak pronoun for ACC.SG, which doubles the object librin e botës. Given their very complicated agreement, which is irrelevant for the discussion of modality, I will refrain from fully tagging Albanian genitive and possessive particles (linkers) in the examples below. The subjunctive particle will also be glossed simply by PTL, and glossing of non-verb forms will be restricted to a minimum. Albanian regularly uses two different kinds of negative particles, nuk and s’ for the indicative and mos for the subjunctive, imperative, optative etc. A similar differentiation is found in Greek with the negative particles įİȞ and ȝȘȞ respectively. As de Haan (1997: 136) pointed out, it is not always the case that two negations cancel each other out in modal logic. Thus “not (not POSSIBILITY)” does not lead to “POSSIBILITY”, but to NECESSITY, when one negation is used to negate the modal element and the other is used to negate the main verb. There is also a fully inflected modal verb duhem, which will be discussed later. See Demiraj (1986: 869-870). Buchholz and Fiedler (1987: 194, fn. 25) call this a feature of the colloquial language. Their example is a negated one: Po edhe në vend nuk duhet (PRS) të qendronim (IMPRF.1PL) ‘But we were not even allowed to stop’. The other basic strategy in de Haan’s (1997: 58–85) classification is MSS (modal suppletion strategy), to be found for example in English or German, where uniscopal modals are used: need not / nicht brauchen for external negation (wide scope) and must not / nicht dürfen for internal negation (narrow scope). For an alternative analysis based on a cluster of languages and especially dealing with the parameter of lexicalization of negative modals see van der Auwera (2001). Apart from being more or less uninflected, the most typical Albanian modals of POSSIBILITY and NECESSITY, mund and duhet, behave in a very similar way like the Italian modals potere and dovere in de Haan’s (1997: 92-93) sample. The only obvious difference would be that Albanian mund does not allow for NT in the case of narrow scope (neither in my corpus nor in the opinion of Albanian informants), unlike Italian potere, if de Haan’s claim based on the judgement of “some speakers” – but not confirmed by my Italian informants – is true.

Modals in Albanian

259

17. In (16b) the future with the auxiliary particle do would, of course, have only one subject, i.e. të tjerët do të dëgjojnë ‘the others will listen’. As claimed by Buchholz and Fiedler (1987: 143, 145), uninflected do + SBJV can sometimes express a strong volitional component which is, however, always secondary to the meaning of the future, e.g. Ku do të veni? ‘Where will you go?’ = ‘Where do you want to go?. 18. Only in highly affective sentences and in folk poetry dot is also used in a sentence-initial position (Çabej 1987a: 297). 19. With the aorist, in this case mësoi ‘he came to know, learnt’, the supposed impossibility expressed by dot is transformed into certainty, i.e. ‘he did not succeed in getting the information’. This change of the modal meaning in the narrow sense is a result of the semantic interaction with the aspectual meaning of the aorist, see below. 20. While the singular forms of the present are identical with the uninflected form, others are different, e.g. munden 3PL.PRS, mundej 3SG.IMPRF or munda 1SG.AOR. 21. For some details of the difficult relationship between these three means of expressing possibility in Standard Albanian see Buchholz and Fiedler (1987: 162-163). From an etymological point of view the modern inflected forms are clearly secondary to the uninflected mund as they show the d-extension which stems from the subjunctive particle të, originally amalgamated with the older form mun, which still exists in dialects; see section 5.1 below. In addition, the older form of the aorist, found in Buzuku and in Geg dialects is mujta without such a d. The d-extension can also be found in other derivational forms, especially also in sëmundje ‘illness’, literally ‘unableness’; see Schumacher (2005: 164–167). 22. It should be noted that it is typical for the aorist to express completed (totalized) states of affair, which means in the case of mund that the original POSSIBILITY becomes reality (Mansaku 1979: 142). Therefore modality remains only as a subordinated connotation, dominated by the aspectual characteristics of the verb form. This is an interesting parallel to the perfective verb smoþ’ in Russian, which unlike imperfective moþ’ not only expresses POSSIBILITY but also its realization (Hansen 2001: 174-175, van der Auwera and Plungian 1998: 106). The same interaction of the aorist (limitative) meaning with modal verbs is found in Romance languages; see Breu (2005: 65–67) for Italian and also for the Slavic minority language in Southern Italy. A similar interaction can be noticed with the modals of NECESSITY and VOLITION, i.e. ‘must’ gets the interaction meaning “did, because s/he had to” and ‘want’ becomes “did as s/he wanted to” (vice versa for negation). The perfect (present and past) shows the same interaction, as it expresses a state which has come about by an action realized in the past, e.g. Shiqipëria kishte mundur [PST.PRF.3SG] prapë t'i ikte Azisë në fillim të shekullit ‘Albania had been able to escape (had succeeded in escaping) again from Asia in the beginning of the century’ or ...ndonjë afishe e

260 Walter Breu

23.

24. 25.

26. 27.

28. 29.

30. 31. 32.

vjetër që shiu, era apo pastruesit e pandërgjegjshëm të rrugëve s'kanë mundur [PRF.3PL] ta shqitin akoma. ‘...some old poster which the rain, the wind, or the irresponsible street cleaners have not yet been able (achieved) to tear off. In any case, it should be noted that the semantic (modal) feature of POSSIBILITY is not simply absent in aorist or perfect forms and that its logical subordination in the surface meaning of the verb forms is the regular result of its interaction with the aspect meanings in question. When the imperfect of modals is used, similar to imperfective verbs in Slavic, the question of a realization of the state of affairs expressed by the main verb is left open. There are, however, examples that show that in certain cases of ellipsis mund can be uninflected even when the governed main verb is not present, e.g. Linda mendonte se as ai dhe as ajo s'mund të flisnin ashtu. Dhe jo vetëm s’mund, por s'duhej. ‘Linda thought that neither he nor she could [not] speak in this way. And not only couldn’t, but mustn’t.’ The inflected forms would be s’mundte (active) and s’mundej (mediopassive). See section 2.5 above for the use of uninflected mund in this construction. As a consequence, many examples are ambiguous (or vague) as to narrow or wide scope negation of necessity, for example: Kam një parandjenjë se gjithçka do të sqarohet. S'duhet dëshpëruar para kohe ‘I have a presentiment that something will be cleared up. There is no need to (~ one must/should not) despair too early’. Buchholz and Fiedler (1987: 189) describe this construction as mainly colloquial. It can, however, be found even in scholarly literature. The example given here is from Çabej (1987b: 198). For example: Përpara çdo krahasimi duan marrë elementet e veçanta të sistemit të shqipes, dhe duan vendosur sipas moshës së tyre ‘Before any comparison, the important elements of the system of Albanian have to be picked and have to be arranged according to their age’ (Çabej 1987b: 204). See Buchholz and Fiedler (1987: 188–189) for a more detailed description of these constructions under the joint heading “nezessitative Passivdiathese”. Informants attributed such NECESSITY constructions with two inflected verbs as “substandard” or “illiterate”. They are not mentioned in grammars and dictionaries and seem to be restricted to Northern areas including Kosova, where in the dialectal base duhem normally governs the infinitive. The standard forms in the first example would be duhet të paraqiten and duhen paraqitur respectively. In Standard Albanian the uninflected auxiliary le is identical only with the 2nd person singular of the aorist of lë ‘let’, but just like in the cases of mund, duhet and do, its historical base is certainly an older form of the 3SG.PRS. For more details see Buchholz and Fiedler (1987: 133–149). An alternative solution is presented by Altimari (2005), who links the ItaloAlbanian functions of kam with an old Balkan feature. It has to be admitted that kam can sometimes express NECESSITY in mainland Albanian, too, but

Modals in Albanian

261

only in combination with the infinitive (Geg), including some rare examples in Buzuku’s Old Albanian text (Fiedler 2004: 598-599), or with a final clause (Tosk). But as even “kam + SBJV” with this function can be found in other diaspora dialects (Greece, Bulgaria, see below), separated from mainland Albania a long time ago, it cannot be excluded that first steps in this direction go back to older stages of common Albanian. Nevertheless, the almost exclusive and full-range expression of NECESSITY by the construction of kam + SBJV in Italo-Albanian has to be attributed to language contact. 33. It is worth noting that in Arbrisht the negative particle must always be put before the modal, i.e. an iconic narrow scope negation with mos between the modal and the main verb is not possible. Together with the reduced inflection of the modals, this indicates a higher degree of grammaticalization of NECESSITY than we find in Standard Albanian. The situation described here refers to the Calabro-Albanian dialect of Frascineto in the province of Cosenza. There are other dialects in which the modal expressing NECESSITY has already completely lost its inflection including that for tense, e.g. ket in the dialect of San Nicola dell’Alto, province of Crotone. The formal restriction for (narrow scope) negation to the position before the modal is valid there, too (Turano 1995: 136-139). 34. The Arbrisht presumptive perfect is an inferential (deductive) evidential, whose meaning is very close to epistemic necessity, e.g. Nëng ësht; ka partirtur ‘He is not here; he must have left’, literally ‘...he has left’. Palmer (2001: 35-52) ranks all types of evidentiality with modality, while de Haan (2001) excludes them completely from this sphere. I would like to follow the golden mean suggested by van der Auwera and Plungian (1998: 85–86) and include the inferential type of evidentiality like the one found in Arbrisht under modality, as an assumption made by the speaker is at issue here. 35. See example (16b) above and also Turano’s (1995: 142) example Dua që burri të hajë bukën (standard) ‘I want the man to eat the bread’, where (in the dialect of San Nicola) the subject of the main verb burri ‘the man’ could only be put at the end of the sentence with the modal complex “dua të + main verb” left intact. 36. This corresponds to an equal variety of expressions in Greek itself. 37. In this case dot has nothing to do with dot expressing IMPOSSIBILITY in Standard Albanian, nor with the uninflected future auxiliary of other varieties, but is a contracted variant of düet that corresponds to duhet ‘must’. A more frequent future is formed with fully inflected kam ‘have’ + infinitive. For information about this nowadays probably completely disappearing variety see Tagliavini (1937: 36, 100, 107). 38. Language contact with Croatian could have played a role in the restricted use of the subjunctive and the strong position of the infinitive governed by the modals.

262 Walter Breu 39. The Geg infinitive of the type me + PST.PTCP is either a secondary formation restricted to Geg, substituting perhaps an older common Albanian synthetic infinitive (Mansaku 1979: 154–155), or it existed already in Protoalbanian and was lost afterwards in the Tosk varieties, as Gabinskij (1970: 14–37) thinks, who also gives a good overview of the arguments of the restricted Geg position (1970: 7–14). Anyway, the Geg me-infinitive should not be mixed up with the aforementioned analytic final constructions with the Tosk “infinitive” për të ‘in order to’ + PST.PTCP in Standard Albanian. 40. Demiraj (1986: 871) cites the forms s’und, s’un, sun ‘cannot’ from Geg dialects and sund, sunt from older Italo-Albanian poetry. We can add the (inflected) aorist s’undi te njoh ‘was able to detect’ from Sokolova’s (1983: 91) description of the Mandrica Dialect in Bulgaria. Mansaku (1979: 138-139) notices that in southern Geg dialects sun – and sometimes also its positive counterpart mun – has even lost its verbal character by not governing the subjunctive any more, by being combined with all forms of the main verb (i.e. even the aorist), and also by leaving its place before the main verb. In these dialects the marker of [+POSSIBILITY] seems to have developed into a more or less grammaticalized particle like dot ‘cannot’. 41. In some dialects additional changes can be seen, which for example in the Molise-Albanian dialect of Montecilfone have resulted in a form bΩt, which is homonymous with the causative particle. 42. Buzuku has different graphic forms, e.g. mNJde, munde, mundeh (Fiedler 2004: 537) and so on, which all can be interpreted as mundë. A clear confirmation of the amalgamation of the subjunctive particle with the modal follows from the fact that older writers do not use të after the modal, even in the case of a morphologically marked subjunctive form of the main verb (which normally always requires të). In addition the modal shows the same effects of crasis (vowel fusion) in combination with weak pronouns as të, e.g. mundë + e (pers. pronoun 3SG.ACC.) > munda like të + e > ta (Mansaku 1979: 146, Fiedler 2004: 537–538). 43. For a formal reconstruction of a present 2./3. Sg. mun based on the aorist nasal stem mûj- that could ultimately go back to an Indo-European root mewH- ‘to be reproductively powerful’ with an n-Extension, see Schumacher (2005: 159). A common root of the present and the aorist was already assumed by Çabej (1964: 44–45) and Mansaku (1979: 140). 44. Çabej’s argument is based on the fact that dot, being absent in Geg and ItaloAlbanian, seems to be restricted to those areas where the will-future is used. As a starting point he sees affective reduplications at the end of the sentence of the type nuk do t vij dot ‘I will not come, I will’. But why this form should have spread from the future to the present and so on, and how it obtained the meaning ‘cannot’ remains unclear in this theory.

Modals in Albanian

263

45. The ancient root seems to be dash-. The other forms are secondary, but of the same root. An older assumption, claiming the present dua to be a Latin loanword (from debeo ‘must’), is nowadays rejected (Çabej 1987a: 173). 46. Given the South Slavic development of the root volj- from ‘want’ to ‘like’ (Old Church Slavonic) to ‘love’ (Serbo-Croatian, besides ljubiti), see Skok (1973: 614), one could think of a typical Balkan development to be assumed also for Albanian. 47. Another possible etymology that puts dua together with Greek ΈΉϧ ‘must’, įȑȠȝĮȚ ‘request, invoke, need’ is rejected by Çabej (1987a: 173). 48. Compare e.g. Italian ci vuole ‘is needed’, vuol (essere) fatto ‘has to be made’, where the verb volere ‘want’ is used. The same is true for non-Romance languages as well, e.g. German die Pflanze will Wasser ‘the plant wants (= needs / must get) water’. This transition can be found for both nominal and verbal government alike (Çabej 1987a: 173-175). A well known example for the inverse development from ‘lack, be needed’ to the sphere of VOLITION is English to want. 49. Although the mediopassive form is mostly used with the modal meaning of NECESSITY, it nevertheless is still productive in the function of a fully inflected passive or reflexive/reciprocal of ‘love’ or ‘want’, e.g. duhem ‘I am wanted/loved’, duhen ‘they love themselves / each other’. In addition, it can also have the meaning ‘fall in love’. 50. In Buzuku’s work the construction of inflected mund + infinitive clearly prevails over uninflected mund + subjunctive. As Mansaku (1979: 136–138) points out, the conservatism of the modal ‘can’ with respect to the use of an infinitive can also be found in other Balkan languages. It can be explained by the fact that in the case of ‘can’ there can never be a second subject in the same clause, while this is easily possible in the case of ‘want’. This can give rise to misunderstandings when the infinitive is used. In the case of NECESSITY there is also only one subject in the clause. Thus it is no surprise that we find in Buzuku the same prevalence of an infinitival government over the subjunctive for the modals duhet and kam like in the case of mund (Mansaku 1979: 149).

References Altimari, Francesco 2005 Il “futuro necessitativo” dell’albanese d’Italia: Influenza italo-romanza o arcaismo balcanico? In L’influsso dell’italiano sulla grammatica delle lingue minoritarie, Walter Breu (ed.), 1–12. Rende: Centro Editoriale e Librario dell’Università della Calabria. Breu, Walter 1991 Das italoalbanische ‘Perfekt’ in sprachvergleichender Sicht. In Atti del Congresso Internazionale di Studi sulla Lingua, la Storia e la

264 Walter Breu Cultura degli Albanesi d’Italia, Francesco Altimari, Gabriele Birken-Silverman, Martin Camaj, and Rupprecht Rohr (eds.), 51–66. Rende: Centro Editoriale e Librario dell’Università della Calabria. Breu, Walter 1994 Forme verbali perifrastiche arbërisht. In I dialetti italo-albanesi. Studi linguistici e storico-culturali sulle comunità arbëreshe, Francesco Altimari and Leonardo M. Savoia (eds.), 365–385. Roma: Bulzoni. Breu, Walter 1997 Variantenstreit und Normierung im Albanischen. In Standardisierung und Destandardisierung europäischer Nationalsprachen, Klaus J. Mattheier, and Edgar Radtke (eds.), 237–257. Frankfurt am Main etc.: Peter Lang. Breu, Walter 2005 Verbalaspekt und Sprachkontakt. Ein Vergleich der Systeme zweier slavischer Minderheitensprachen (SWR/MSL). In Slavistische Linguistik 2003, Sebastian Kempgen (ed.), 37–95. München: Otto Sagner. Buchholz, Oda, and Wilfried Fiedler 1987 Albanische Grammatik. Leipzig: Enzyklopädie. Çabej, Eqrem 1964 Studime rreth etimologjisë së gjuhës shqipe XIV [Studies on the Etymology of the Albanian Language]. Studime Filologjike 3: 44-45. Çabej, Eqrem 1987a Studime etimologjike në fushë të shqipes [Etymological studies in the field of Albanian]. Vol. 3. Tiranë: Shkenca. Çabej, Eqrem 1987b Ilirishtja dhe shqipja [Illyrian and Albanian]. In Studime gjuhësore IV: Nga historia e gjuhës shqipe [Linguistic Studies IV. From the History of the Albanian Language], 197–207. Prishtinë: Rilindja. de Haan, Ferdinand 1997 The Interaction of Modality and Negation: A Typological Study. New York/London: Garland Publishing, Inc. de Haan, Ferdinand 2001 The Relation between Modality and Evidentiality. In Modalität und Modalverben im Deutschen, Reimar Müller, and Marga Reis (eds.), 201–216. Hamburg: Buske. Demiraj, Shaban 1986 Gramatikë historike e gjuhës shqipe [Historical Grammar of the Albanian Language]. Tiranë: 8 Nëntori. Demiraj, Shaban 2004 Gjuhësi Ballkanike [Balkan Linguistics]. Tiranë: Shkenca.

Modals in Albanian

265

Fiedler, Wilfried 2004 Das albanische Verbalsystem in der Sprache des Gjon Buzuku (1555). Prishtinë: Akademia e Shkencave dhe e Arteve të Kosovës. Gabinskij, Mark Aleksandroviþ 1970 Pojavlenie i utrata perviþnogo albanskogo infinitiva. [The Rise and Loss of the Primary Albanian Infinitive]. Leningrad: Nauka. Hansen, Björn 2001 Das slavische Modalauxiliar. Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen. (Slavolinguistica 2). München: Otto Sagner. Jokl, Norbert 1963 Die Verwandtschaftsverhältnisse des Albanischen zu den übrigen indogermanischen Sprachen. Die Sprache 9: 113-156. Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Mansaku, Seit 1979 Ndërtimet me foljet modale mund dhe duhet te Gjon Buzuku [Constructions with the modal verbs mund and duhet in Gjon Buzuku]. Studime Filologjike 23, 3: 135–157. Matzinger, Joachim 2006 Der Altalbanische Text Mbsuame e Krështerë (Dottrina cristiana) des Lekë Matrënga von 1592. Dettelbach: J. H. Röll. Palmer, Frank Robert 2001 Reprint. Mood and Modality. 2nd ed. (Cambridge Textbooks in Linguistics) Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Sasse, Hans-Jürgen 1991 Arvanitika: Die albanischen Sprachreste in Griechenland. Teil 1. Wiesbaden: Otto Harrassowitz. Schumacher, Stefan 2005 Altalbanisch munde /mundë/ ‘können’ und semuhem, -ete /sëmuhem, -etë/ ‘krank werden’. In Albanologische und balkanologische Studien: Festschrift für Wilfried Fiedler, Monica Genesin, and Joachim Matzinger (eds.), 151–168. Hamburg: Dr. Kovaþ. Skok, Petar 1973 Etimologijski rjeþnik hrvatskoga ili srpskoga jezika [Etymological dictionary of the Croatian or Serbian Language]. Vol 3. Zagreb: Jugoslavenska akademija znanosti i umjetnosti. Sokolova, Bojka 1983 Die albanische Mundart von Mandrica. Berlin: Harrassowitz.

266 Walter Breu Tagliavini, Carlo 1937 L’albanese di Dalmazia: contributo alla conoscenza del dialetto ghego di Borgo Erizzo presso Zara. Firenze: Leo S. Olschki. Turano, Giuseppina 1995 Dipendenze sintattiche in albanese. Padova: UNIPRESS. van der Auwera, Johan 1999 On the semantic and pragmatic polyfunctionality of modal verbs. In The Semantics/Pragmatics Interface from Different Points of View, Ken Turner (ed.), 49–64. Oxford etc.: Elsevier. van der Auwera, Johan 2001 On the typology of negative modals. In Perspectives on Negation and Polarity Items, Jack Hoeksema, Hotze Rullmann, Víctor Sánchez-Valencia, and Ton van der Wouden (eds.), 23-48. Amsterdam: Benjamins. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79–124.

9. Modality in Romani Viktor Elšík and Yaron Matras

1. Introduction Romani 1 (Indo-Aryan) is an inflectional language with a ‘thin’ layer of agglutinating morphology. Grammatical material is generally suffixed. Lexical roots that entered the language in the Byzantine period or later typically draw to some extent at least on Greek-derived morphology, which has remained productive since Byzantine contact for all subsequent loans from European languages. In its syntactic typology, Romani closely resembles typical Balkan formations, having undergone intense convergence with Greek, and later with other languages of the area. Romani is the only New Indo-Aryan language that relies exclusively on prepositions as analytic markers of semantic case roles, and the only IndoAryan language that possesses a definite article (pre-posed, as in Greek). Word order is flexible, generally alternating between verb-middle and verbinitial. Attributes are pre-posed. Subordinations are based on conjunctions, usually deriving from interrogatives. In complement clauses a distinction is made between factual and non-factual complements. Relative clauses are post-posed, usually introduced by an uninflected relativiser, and they usually require resumptive pronouns in the subordinate clause. The impact of diverse contact languages is a major feature differentiating the individual Romani dialects, and it is therefore difficult to generalise when discussing the structures of Romani. In the present contribution we try to exploit this structural diversity within Romani in order to draw some generalisations about the language based on a sample of dialects. 2 Our subsequent discussion is therefore devoted to general tendencies in Romani. Alongside internal pathways of grammaticalisation, we shall also make use of the opportunity offered by the Romani sample and pay special attention to grammatical borrowing and language convergence and their role in shaping the inventory of modality-related structures in the language. Our data examples draw on both published sources wherever indicated; where a source is not identified, data derive from our own ongoing survey of European Romani dialects, as part of the

268

Viktor Elšík, and Yaron Matras

RMS (Romani Morpho-Syntax) Database project (see footnote 2, and see Manchester Romani Project). Much of these data were collected by questionnaire elicitation, which in some cases puts certain limitations on the discussion and the scope for interpretation. Our present work may be understood as a kind of follow-up to Boretzky’s (1996b) earlier work devoted to modals in Romani. In the present contribution, a vast amount of data is taken into consideration on dialects that had not been described and were in fact unknown in the research context at the time of Boretzky’s publication. Our discussion is also anchored in a wider morphosyntactic context, one which takes into account not just the etymology of individual expressions of modality, but attempts a more typologically informed evaluation of modal constructions, inspired by models in grammaticalization theory. 2. Modals in Romani Early Romani, the ancestor of all Romani dialects, can be reconstructed to have possessed three or four dedicated modals: a volition modal and two or three possibility modals. In all likelihood, there were no dedicated necessity modals, although periphrastic constructions that could express necessity must have been available. Today we find a bewildering variety of modals and modal constructions within Romani, especially in the domain of necessity and, to a lesser extent, possibility. Volition is clearly the most stable form of modality. There are several sources of dialect divergence with regard to modals. The majority of necessity modals, numerous possibility modals and a couple of volition modals are dialect-specific loanwords from Romani’s European L2s. In addition to lexical borrowing, there are also several instances of contact-induced or ‘replica’ grammaticalization (cf. Heine and Kuteva 2005) of modals based (mostly) on indigenous lexical material. Finally, several modals may result from autonomous, contact-independent, grammaticalization within Romani. In order to abbreviate the subsequent discussion we first provide a morphosyntactic classification of Romani modal constructions (Section 2.1), which will then be referred to in the actual data subsections on volition, possibility and necessity modals (Sections 2.2–4). The last part of Section 2 is devoted to the interaction between modality and negation (Section 2.5).

Modals in Romani

269

2.1. Modal constructions Romani modal constructions consist of two major parts: the modal and the main verb of the clause that encodes the modalized proposition. Since the clause need not be a complement of the modal in syntactic terms (though it frequently is) and since the main verb in this clause need not be lexical, we use the term ‘modalized verb’. Romani modal constructions vary on several parameters, both across and within dialects. Most importantly, there are differences with regard to the location of subject marking. The modal subject (the subject of the modalized proposition) is expressed through an NP, which is mostly optional if the subject is pronominal. In addition, person, number and, rarely, gender of the modal subject are, in most constructions, also cross-referenced on the modal and/or on the modalized verb. We distinguish between personal vs impersonal constructions, according to whether the modal inflects for the subject categories or not; and between finite vs nonfinite constructions, according to whether the modalized verb inflects for the subject categories or not. The criterion of subject cross-referencing thus renders four types of modal constructions. The following examples from Hameln Sinti illustrate three of these types: the impersonal–finite construction in (1a) may express different kinds of possibility, while the personal–finite (1b) and the personal– nonfinite (1c) constructions are specialized for participant-internal possibility. (1)

Hameln Sinti (Northwestern, Germany; Holzinger 1993) a. Me givau našte. I.NOM sing.PFUT.1SG can ‘I can sing.’ (p. 94) b. Me hajevau te givap. I.NOM understand.PFUT.1SG COMP sing.SBJV.1SG c. Me hajevau I.NOM understand.PFUT.1SG ‘I am able to sing.’ (p. 94)

te COMP

givel. sing.SBJV.3SG[=INF]

The structural difference between the two synonymous personal constructions is noteworthy. In (1b) both the modal and the (subjunctive) modalized verb cross-reference the first-person singular subject, whereas in

270

Viktor Elšík, and Yaron Matras

(1c) only the modal does so. The modalized verb instead assumes a default, third-person singular, subjunctive form. While the finite construction has been inherited from Early Romani, the nonfinite subjunctive, or the ‘new infinitive’ (Boretzky 1996b, cf. also Matras 2002: 161–162), is an imperfect copy of the infinitive used in corresponding constructions in German, the current L2 of Hameln Sinti. Numerous further Romani dialects outside of the Balkans have developed the ‘new infinitive’ in same-subject non-factual complement clauses (and some other tightly-integrated subordinate clauses), due to pattern borrowing from L2s that, like German, possess an infinitive verb form. 3 Nonfinite modal constructions are restricted to those Romani dialects that have developed the ‘new infinitive’. Personal–finite modal constructions, on the other hand, are restricted to those dialects that retain finite non-factual complements. Impersonal–finite modal constructions can occur in dialects of both types. The one type of modal construction that has not been illustrated so far is the impersonal–nonfinite construction. In this type there is no crossreferencing of the modal subject, and so it must be overtly encoded through an NP even if it is pronominal. No subject marking at all may only occur if generic modal subject is intended. In the example from modern Finnish Romani (2) the impersonal modal takes a nonfinite (default third-person singular subjunctive) modalized verb and the first-person singular modal subject is marked solely through an accusative pronoun. In the example from Selice Romani (3) the impersonal modal takes nonfinite (default thirdperson plural subjunctive) modalized verbs and the generic subject remains unexpressed. (2)

(3)

Finnish Romani (Northwestern, Finland) MƗn mote lel tauva I.ACC must take.SBJV.3SG[=INF] this ‘I have to take this medicine.’ Selice Rumungro (South Central, Slovakia) Te te dživen kampe, also COMP live.SBJV.3PL[=INF] need te dǀgozinen. COMP work.SBJV.3PL[=INF] ‘One also needs to live, not only to work.’

tram. medicine.NOM

na NEG

þak only

Modals in Romani

271

Romani modal constructions also differ with regard to the location of TAM marking. While personal modals always do, impersonal modals may but need not inflect for TAM categories. In most Romani dialects modalized verbs in personal constructions and in TAM-inflected impersonal constructions assume a (finite or nonfinite) subjunctive form that does not encode tense or aspect. If however impersonal modals do not inflect for TAM, then TAM categories must be marked on the modalized verb. In other words, while the location of TAM marking is usually predictable from the location of subject marking in most constructions, the second criterion does differentiate between two types of impersonal–finite constructions: those where TAM categories are marked on the modal and those where they are marked on the modalized verb. In the dialect of Sliven Muzikanta, for instance, past reference can be marked on the impersonal modal trjabvizela, in which case the modalized verb assumes the tenseless subjunctive (4a), or it can be marked on the modalized verb, in which case trjabvizela assumes the default present form (4b). For the functional difference between the two examples see Section 2.4. (4)

Sliven Muzikanta (Balkan, Bulgaria) a. Trjabvizelas te užaras. must.IMPRF.3SG COMP wait.SBJV.1PL ‘We had to wait.’ b. Trjabvizela te diþhan must.PRS.3SG COMP see.PST.2SG ‘You must have seen him.’

les. he.ACC

All of the above patterns show TAM inflection either on the modal, or on the modalized verb. Nevertheless, instances of double TAM marking are also attested, though restricted to a few dialects. In the dialect of Parakalamos, as elsewhere in Romani, modalized verbs are in the subjunctive when the modal is in the present form (5a–b). However, unlike most dialects, Parakalamos Romani shows imperfect (rather than tenseless subjunctive) marking on the modalized verb when the modal is in the imperfect form. This kind of tense ‘agreement’ occurs irrespective of whether the modal is personal (5c) or impersonal (5d). (5)

Parakalamos Romani (Balkan, Greece; Matras 2004) a. Kamama te avav demosiȖráfos.

272

Viktor Elšík, and Yaron Matras

want.PRS.1SG COMP come.SBJV.1SG journalist ‘I want to become a journalist.’ (p. 89) b. Prepi te džas othe. must.PRS COMP go.SBJV.1PL there ‘We have to go to town.’ (p. 72) c. Kamamas te džavas ti poli. want.IMPRF.1SG COMP go.IMPRF.1SG to town ‘I wanted to go to town.’ (p. 75) d. Eprepe te džakerasas. must.IMPRF COMP wait.IMPRF.1PL ‘We had to wait.’ (p. 88) The NP that expresses the modal subject may be in the nominative case or in an oblique case (the dative, the locative, or the markerless oblique, whose main function is to mark animate direct objects and which is usually labelled the ‘accusative’). The two types of case marking of the modal subject are close to complementary. Nominative (canonical) marking always occurs in personal constructions (cf. 1b–c) and in those impersonal– finite constructions where TAM categories are marked on the modalized verb (cf. 1a). Oblique (noncanonical) marking, on the other hand, is obligatory in impersonal–nonfinite constructions (cf. 2). Nevertheless, we find variation between nominative and oblique marking in those impersonal–finite constructions where the modalized verb does not inflect for TAM categories. Typically the differentiation is lexically determined, though instances of (apparently) free variation are also attested. For example, in the dialect of Servy Ghympeny there are several impersonal modals, all of which take finite but tenseless complements: while našty always has a nominative subject (6a) and javel-pe always has a dative subject (6b), the subjects of treb’i alternate between nominative (6c) and dative marking (6d). (6)

Servy Ghympeny (Northeastern, Ukraine) a. Me našty isys te otþhak’iraw udara. I.NOM cannot COP.3.PST COMP uncover.1SG.SBJV door.PL ‘I could not open the door.’

Modals in Romani

b. Mange javja I.DAT come.PST.3SG ‘I had to wait.’

-pe

te

REFL

COMP

273

užakiraw. wait.SBJV.1SG

c. Tu treb’i vark’edys’ te javes thou.NOM need sometimes COMP come.SBJV.2SG ke me. at/to I.NOM ‘You must come to me sometimes.’ d. Tuke treb’i dor’ik thou.DAT need there ‘You must go there.’

te COMP

džas. go.SBJV.2SG

The role of finiteness in the case marking of the modal subject can be illustrated from a dialect that shows a different type of variation. In Kubanskie Servy the subject of impersonal modals nasþi and trebun’i is nominative when the modalized verb is finite (7a–b) but dative when the verb is nonfinite (7c–d). (7)

Kubanskie Servy (Ukrainian, Ukraine and Russia) a. Me nasþi st’irav e ural’i. I.NOM cannot repair.SBJV.1SG DEF car ‘I can’t repair the car.’ b. Tu trebun’i var’ikala mande thou.NOM must sometimes I.LOC avex. come.SBJV.2SG ‘You should visit me sometimes.’ c. Mand’i nasþi te makhe I.DAT cannot COMP smear.INF ‘I can’t paint the house.’

o DEF

d. Tut’i trebun’i korde te dža. thou.DAT must there COMP go.INF ‘You have to go there.’

te COMP

cer. house

274

Viktor Elšík, and Yaron Matras

Next, there are differences with regard to the internal constituency and structural complexity of modal expressions. While modals that inflect neither for TAM nor for subject categories are the least complex, one can further distinguish between synthetic vs analytic inflected modals. In synthetic modals TAM and/or subject categories are marked within the same word as modality. In analytic modals, on the other hand, the actual modality word is uninflected and inflectional categories are marked separately, on an inflectional auxiliary, which is structurally identical to the copula (and the verb of existence). The following examples from Podhradie Romani illustrate the three types of modals: an uninflected modal (8a), a synthetic inflected modal (8b) and an analytic inflected modal (8c). In most instances of analytic modals the inflectional auxiliary assumes the default, third-person singular, subject category, thus only inflecting for TAM categories, and it is omitted in the present. The personal and obligatory auxiliary in (8c) is thus rather untypical in both respects. Nominal expressions of modality tend to require the copula, and so they resemble the analytic inflected modals. Adjectival modals, which are rather infrequent in Romani, show distributed marking of subject categories: the adjectival cross-references the number and (sometimes) gender of the modal subject, while the copula encodes the TAM categories and mostly cross-references the subject’s person and number, as in Crimean Romani (9). (8)

Podhradie Romani (North Central, Slovakia) a. Šaj džav khƝre. can go.PRS.1SG home ‘I can go home.’ b. Kamav khƝre want.PRS.1SG home ‘I want to go home.’

te

c. Musaj som must COP.PRS.1SG džan. go.SBJV.3PL[=INF] ‘I have to go home.’

khƝre home

COMP

džan. go.SBJV.3PL[=INF] te COMP

Modals in Romani

(9)

275

Crimean Romani (Balkan; Ukraine and Russia) Tume sanusas dolžn’a te raskeld’ijen les. COP.IMPRF.2PL obliged.PL COMP meet.SBJV.2PL he.ACC you ‘You guys must have met him.’

Although modalized verbs are typically introduced by the non-factual complementizer te, 4 there are several kinds of exceptions. Some modals simply do not allow a complementizer. These include, but are not restricted to, expressions where an original complementizer has become an integral part of the modal. In ýáry Romani, for instance, some modals require the complementizer (10a), whereas others do not allow it (10c). Despite an etymological presence of *te in the modal moste, the construction (10b) is of the latter type: while the complementizer te is separable from the modal and inseparable from the modalized verb (cf. 10a), the opposite holds for *te in (10b). In many dialects some modals that usually take the complementizer may allow its occasional dropping. Importantly, the obligatory lack of a complementizer is restricted to uninflected modals, i.e. to constructions where TAM categories are marked solely on the modalized verb. There is no such limitation with the optional dropping of te. (10) ýáry Romani (North Central, Slovakia) a. Kamlom ole þhavenca want.PST.1SG that.OBL.PL boy.INS.PL vakerel. speak.SBJV.3SG[=INF] ‘I wanted to talk to those boys.’

te COMP

b. Moste prekal odƗ prelezinćom. must over that crowl.over.PST.1SG ‘I had to climb over it.’ c. Šaj tut vareso dƗs. can you.ACC something give.IMPRF.1SG ‘I could give you something.’ Even if licenced by the modal, the complementizer is mostly omitted when the modalized verb is borrowed and retains its L2 inflection. For example, the modal može in Kaspiþan Xoraxane requires te with indigenous modalized verbs (11a) but the complementizer must be omitted with

276

Viktor Elšík, and Yaron Matras

Turkish-inflected verbs (11b). Similarly, some Russian Romani modals require te with indigenous modalized verbs but there is no complementizer when the verb is borrowed from Russian, in which case it retains its Russian infinitive form. There is one dialect, Dolenjski Romani, where the complementizer te has been lost altogether, in modal constructions and elsewhere (see Matras 2002: 210 for details), having been replaced by the Slovene complementizer da (12a–b) or dropped altogether, especially in nonfinite constructions (12c). (11) Kaspiþan Xoraxane (Balkan, Bulgaria) a. Može te džas kaj gav. can COMP go.SBJV.1PL to village ‘We can go to town.’ b. Može inanasÕnÕs leske. can believe.SBJV.2PL he.DAT ‘You can believe him.’ (12) Dolenjski Romani (Slovene/Istrian, Slovenia; Cech and Heinschink 2001) a. Triba da leske dav love. need COMP he.DAT give.SBJV.1SG money.PL ‘I need to give him money.’ (p. 357) b. Hoþemo da lam [want.PRS.1PL]SLOVENE COMP take.SBJV.1PL ‘We want to take two apples.’ (p. 357)

duj two

phabaja. apple.PL

c. Morinave lake del love. must.IMPRF.1SG she.DAT give.SBJV.3SG[=INF] money.PL ‘I had to give her money.’ (p. 356) Several generalizations regarding the linear order within modal constructions can be made. The inflectional auxiliary, if there is one, immediately follows the modality word proper, thus conforming to the position of synthetic verb inflections. This order, however, is not required in nominal modal constructions with a copula (cf. 9). The complementizer, if any, precedes the modalized verb, usually immediately, though it may be separated from the verb by negators or pronominal clitics in some dialects.

277

Modals in Romani

In pragmatically neutral contexts modals precede modalized verbs; they may be separated by other constituents, including the subject NP. However, most dialects probably allow pragmatically motivated fronting of modalized verbs, which results in postposed modals, cf. focus fronting in (3). German Sinti stands out in allowing postposition of uninflected modals even in neutral contexts, cf. (1a). Table 1 summarizes the structure of five major types of modal constructions in Romani (COMP+ means that the complementizer is at least optionally present with Romani-inflected modalized verbs): 5 Table 1. Major types of modal constructions in Romani constr.

SUBJECT INFL.

PF PnF nPF1 nPF2 nPnF

modal + + – – –

verb + – + + –

TAM INFL.

modal + + + – +

verb ± – ± + –

NOM NP

AUX

COMP

+ + ± + –

± ± ± – (–)

+ + + ± +

Personal–finite (PF) constructions mark TAM categories on the modal, use the complementizer and have a nominative subject; they vary with respect to the marking of TAM categories on the modalized verb and with respect to the presence of an inflectional auxiliary. Personal–nonfinite (PnF) constructions mark TAM categories solely on the modal, use the complementizer, and have a nominative subject; there may but need not be an auxiliary. There are two major types of impersonal–finite constructions. Those that mark TAM categories on the modal (nPF1) always allow the complementizer but vary on a number of parameters: TAM marking on the modalized verb, case marking of the subject NP and the presence of an inflectional auxiliary. Those impersonal–finite constructions that do not mark TAM categories on the modal (nPF2) are rather uniform: they mark TAM on the modalized verb, have a nominative subject, and cannot have an inflectional auxiliary since the modal is uninflected; the complementizer may be obligatorily lacking. Finally, impersonal–nonfinite (nPnF) constructions mark TAM categories solely on the modal, have an oblique subject, and allow the complementizer. 6

278

Viktor Elšík, and Yaron Matras

2.2. Volition There are two widespread expressions of volition in Romani. The verb kam- is the exclusive expression of volition in most Romani dialects, and can hence be reconstructed as the Early Romani form. This verb can still be used with nominal objects, and in most dialects it retains its lexical meanings ‘to love, to like’ and ‘to desire’. The lexical kam- ‘to love’ is certainly pre-European, though it might be a loanword from West Iranian or Armenian (cf. Boretzky 1995b: 141) rather than continuing an Old IndoAryan etymon (as suggested by Sampson 1926, II: 131). In the Romani dialects of the Balkans kam- competes with a more recent volition expression, the verb mang-. The latter has its origin in the common Romani verb mang- ‘to ask for, to demand’, also used for ‘to beg’ (and from the latter meaning sometimes generalised for ‘to make a living by hawking, fortune-telling or begging’). The two volition verbs appear alongside each other in several Romani dialects of the Balkans (see also Section 2.5), whereas in other dialects of this area mang- has replaced kam-, taking over all of its functions, both lexical and modal. While the grammaticalization of kam- into a volition modal must pre-date the split of Romani dialects, the grammaticalisation of mang- appears to be a fairly recent phenomenon that emerged after the outwards migration from the Balkans of some of the groups, and one that is still spreading within the area. Its origin can be traced in all likelihood to replica grammaticalisation of the South Slavic construction, e.g. Bulgarian iskam ‘I want’ as well as ‘I demand’. Lexical borrowings of volition modals are infrequent: we only find the verb wånt-(a)s- from English in Welsh Romani (Sampson 1926, II: 401) and the verb hoþ- from Slovene in Dolenjski Romani of Slovenia, both of which alternate with indigenous volition verbs. All volition verbs are personal (Type PF or PnF) and mostly take the complementizer te (cf. 8b, 10a, 13a, 53), with the usual exceptions (see Section 2.1, ex. 12c). Occasional dropping of the complementizer is attested, for example, in Piedmontese Sinti (Franzese 1985: 126). In some Romani dialects of Slovakia, including Klenovec Rumungro, the personal construction (13a) alternates with a less agentive volitional construction, which is impersonal (Type nPnF). The impersonal volition verb contains middle marking: either synthetic (13b), or analytic, reflexive-like (cf. 37c), as in Slovak, the source of this construction. In several dialects of northeastern Europe and the Balkans the verb kam- preserves an archaic first-person singular non-perfective suffix -am, rather than the regular -(a)v.

Modals in Romani

279

In the past tense, the modal verb may, in some dialects, take the simple preterite (14), but the imperfect (a non-perfective remote tense) is often preferred (15), which is typical of the portrayal of mental states. The Dolenjski volition loanword from Slovene retains its L2 inflection (cf. 12b). (13) Klenovec Rumungro (South Central, Slovakia) a. Na kames te džan ano gaw. NEG want.PRS.2SG COMP go.SBJV.3PL[=INF] in.DEF town ‘You don’t want to go to town.’ b. Musaj odoj te džas must there COMP go.SBJV.2SG te iš tuke na kamisaĐol if too thou.DAT NEG want.MIDDLE.PRS.3SG te džan. COMP go.SBJV.3PL[=INF] ‘You have to go there, even if you don’t want to go.’ (14) Kohila Romani (Northeastern, Estonia) Me kamjom khere te džal. I.NOM want.PST.1SG home COMP go.SBJV.3SG[=INF] ‘I wanted to go home.’ (15) Kaspiþan Xoraxane (Balkan, Bulgaria) Mangavas te džav mange want.IMPRF.1SG COMP go.SBJV.1SG I.DAT ‘I wanted to go home.’

khere. home

2.3. Possibility Two or three possibility modals can be reconstructed for Early Romani: certainly the affirmative šaj and the negative našti, and probably also the affirmative ašti. Though šaj is clearly of pre-European origin in Romani, its precise etymology remains open: it may either continue some form of the Old Indo-Aryan possibility verb Ğakno- (Boretzky 1996a: 3), or it may be a borrowing of the Persian possibility modal šƗje into Proto-Romani (Matras 2002: 162–163, also mentioned in Boretzky and Igla 1994: 268). The origin

280

Viktor Elšík, and Yaron Matras

of ašti is likewise disputed. Matras (2002: 162–163, following Sampson 1926: 216) hints at the Old Indo-Aryan third-person singular present copula form asti as its source, 7 thus suggesting that Early Romani possessed both the original ašti and the pre-European borrowing šaj. Boretzky (1996a: 5– 6), on the other hand, considers ašti to result from analogical decomposition of the negative našti, whereby n- of našti was identified with the indicative negator na. He also claims that this decomposition took place in individual dialects rather than before the dialect split of Romani, thus implying that šaj was the only possibility modal in Early Romani. 8 Boretzky and Igla (1994: 196) derive našti from a construction consisting of the indicative negator, the affirmative modal šaj and the non-factual complementizer te, i.e. from *na šaj te. However, našti may also continue a regularly negated form of the possibility modal ašti (Matras 2002: 162– 163). The disputed modal ašti has two etymologically unclear variants: šašty or sašty in Lithuanian and Latvian Romani and dašti- in Kalderaš and related North Vlax varieties (also diffused into some Ursari varieties). Assuming that both šaj and ašti were present in Early Romani, the question arises of what the functional difference between them was. This issue will be taken up in Section 3.1. In most Romani dialects the inherited possibility modals are uninflected and occur in impersonal–finite constructions with TAM marking on the modalized verb (Type nPF2). They do not allow a complementizer in some dialects (those spoken in central-eastern Europe and occasionally in the Balkans and elsewhere), whereas in other dialects the complementizer is required or at least allowed. Several dialects spoken in, or originating from, Romania show synthetic personal inflection (Type PF) of the inherited possibility modals, at least variantly, e.g. Pikulešti–Kurturare šaj(-in)- and nasþi(-n)-, Kalderaš-type dašti(-sar)- and našti(-sar)-, and Ursari (d)ašti(-z)-. In fact, uninflected dašti is attested in a single dialect (Boretzky and Igla 2004: 183). Though indigenous, these subject-inflected modals display morphological adaptation typical of loanverbs. Impersonal inflection (Type nPF1) of the inherited possibility modals is rare: we find synthetic marking of TAM categories in Piedmontese Sinti, e.g. stik-o-l-a [-middle-3SG-PFUT], alternating with uninflected stik < *ašti; and analytical marking of TAM categories in some dialects of Ukraine (cf. 6a). The inflected possibility modals, as a rule, require the complementizer. In addition to the possibility modals inherited from Early Romani, many dialects employ possibility expressions that result from dialect-specific grammaticalizations of pre-modal constructions or lexical verbs. An

Modals in Romani

281

impersonal copula construction, which has been grammaticalized into a dedicated necessity modal in several dialects (see Section 2.4), is reported to have developed into a possibility expression in Crimean Romani (cf. Boretzky 1999a: 113). A negative impersonal copula construction probably also underlies the negative possibility modals nahi or naj < *na hi [NEG COP.PRS.3] in a few dialects of the Balkans and in Core Sinti (Boretzky and Igla 2004: 184). Note that if we are to assume that ašti developed from an Old Indo-Aryan copula, then the dialect-specific development of possibility modals from the Romani copula represents a ‘second round’ of the same type of grammaticalization. Quite commonly, participant-internal possibility (ability or capability) is expressed by personal verbs meaning ‘to know’, ‘to understand’ or ‘to manage’, which as a rule retain their pre-modal meanings as well. The ability function of the indigenous verb (a)xaljov- ‘to understand’ is only attested in some dialects of Ukraine, Germany, and Finland, e.g. in Kotka Romani (16). On the other hand, the (cap)ability function of the indigenous verb džan- ‘to know’ is widespread within Romani, to the extent that Boretzky (1999b: 178) considers it to be reconstructable for Early Romani. At least in some Romani dialects the extension of these lexical verbs to (cap)ability functions is likely to have resulted from contact-induced grammaticalization, as this kind of polyfunctionality is common Romani’s L2s. In several dialects spoken in and around Hungary, possibly due to pattern borrowing from Hungarian (cf. Boretzky and Igla 2004: 184), the (cap)ability verbs (17a) have been extended to express participant-external possibility functions as well (17b). (16) Kotka Romani (Nortwestern, Finland) Hajuveha -ko tu cƝrel understand.PFUT.2SG -Q thou do.SBJV.3SG[=INF] butti touveria? work ax.INS ‘Are you able to work with an ax?’ (17) Gabor Romani (North Vlax, Romania) a. ýokanosa žane te kere buþi? hammer.INS know.PRS.2SG COMP do.SBJV.2SG work ‘Are you able to work with a hammer?’

282

Viktor Elšík, and Yaron Matras

b. ýi

žanel te žal -tar know.PRS.3SG COMP go.SBJV.3SG away pînë na arakël e þie. until NEG find.SBJV.3SG DEF key ‘He can’t leave until he finds the key.’ NEG

In some Romani dialects of central Europe the Early Romani verb troma‘to dare’ of Greek origin has developed deontic possibility (permission) functions (cf. Boretzky and Igla 2004: 187). An identical semantic development is attested in Slavic languages (cf. Hansen 2003: 11) and German, and so contact-induced grammaticalization may be the source of this extension (Boretzky and Igla 1994: 183). While the permission modal retains its personal verb inflection in the Central dialects, it has fused with the original complementizer in Sinti, losing all inflection on the way, e.g. *troma- te > Hungarian Sinti trunti (18; Type nPF2). In some Rumungro dialects the verb troma- is now a dedicated permission modal (19a; Type PnF), while the lexical meaning ‘to dare’ is expressed by a construction involving a secondary nominal back-formation from the verb (19b). (18) Hungarian Sinti (Northwestern, Hungary; Mészáros 1980: 10) Trunti džal dren. be_allowed go.SBJV.3SG in ‘S/he is allowed to come in.’ (19) Klenovec Rumungro (South Central, Slovakia) a. Me tromaw odƗ te I.NOM be_allowed.PRS.1SG that COMP keren. do.SBJV.3PL[=INF] ‘I am allowed to do it.’ b. Man hi troma odƗ I.ACC COP.PRS.3 courage that keren. do.SBJV.3PL[=INF] ‘I dare [lit. I have courage] to do it.’

te COMP

Several possibility modals have quite interesting lexical sources. For example, the verb vydža- ‘to go out’ (consisting of a Slavic aktionsart

Modals in Romani

283

prefix and the indigenous verb dža- ‘to go’) is used impersonally, with locative or dative marking of the subject NP, as the basic possibility modal in the Xandžary dialect of Ukraine. The construction is probably based on a Ukrainian idiomatic expression (‘it goes out to someone’ meaning ‘it works out for someone’), though the extension to possibility (‘it is possible [can work out] for someone’) is an autonomous innovation of the Romani dialect. The modal is attested in a wide range of possibility uses: ability (20a; Type nPnF), participant-external possibility (20b; Type nPF1) and permission (20c; Type nPnF). Further curious examples of lexical resources for possibility include the personal verbs dol- ‘to get, become; to get, receive; to get [somewhere]’ (21a; Type PnF), which consists of a Slavic aktionsart prefix and the indigenous verb l- ‘take’, and pƝr- ‘to fall; to get [somewhere]’ (21b) in Kohila Romani. Both expressions share the meaning ‘to get [somewhere]’, which might be the immediate pre-modal meaning. 9 (20) Xandžary (Ukrainian, Ukraine) a. Mande na vydžal I.LOC NEG go_out.PRS.3SG ‘I am not able to repair the car.’

te COMP

t’ire do.INF

ural’i. car

b. Lest’i na vydžal te udžal he.DAT NEG go_out.PRS.3SG COMP leave.SBJV.3SG poka vov na rakhel t’ii. until he.NOM NEG find.PRS.3SG key ‘He can’t leave until he finds the key.’ c. Tute vydžala te dža mansa. thou.LOC go_out.FUT.3SG COMP go.INF I.INS ‘You may come with me.’ (21) Kohila Romani (Northeastern, Estonia) a. Me dolƗ tuke lǀve tašša I.NOM get.PFUT.1SG thou.DAT money.PL tomorrow te plajskyrel. COMP pay.SBJV.3SG[=INF] ‘I can pay you [back] tomorrow.’

284

Viktor Elšík, and Yaron Matras

b. Jov kindža peske neve idža he.NOM buy.PST.3SG REFL.DAT new.PL clothes.PL sob te pƝrel džal ande foros. so_that COMP fall.SBJV.3SG go.SBJV.3SG in town ‘He bought himself new clothes so that he could go into town.’ Lexical borrowing of possibility modals into Romani is well attested, though many dialects do without a possibility loanword. Loans of the following possibility modals are attested within Romani: Greek boro, Macedonian and Bulgarian može, Slovene lahko, Polish móc, East Slavic moþ’ and možno, German dürfen, Italian potere, and Finnish voida. Possibility loanwords tend to be borrowed from the current L2s of the relevant Romani dialects, although Slovene-derived lako and lax are also attested in some Italian Romani varieties, e.g. Venetian Sinti (cf. Boretzky and Igla 2004: 183). Most loanwords have a wide range of possibility functions, with the exception of the loan of German dürfen into German Sinti, which is, like the source form, specialized for deontic possibility (permission). The Greek, Polish and Finnish possibility modals plus East Slavic moþ’ are personal verbs and they are always borrowed as such into Romani. Most commonly the loans are morphologically integrated into Romani verb inflection in the same way as other borrowed verbs are, through adaptation of a frequent inflectional stem of the L2 modal by means of a dialectspecific loanverb suffix, e.g. bor-in- or bor-iz- (< Greek) in some varieties of Greek Romani; mog-in-, moȖ-in- or mož-in- (< Slavic) in numerous Northeastern dialects of Poland, Lithuania and Ukraine; and voj-uv- or vojp-uv- (< Finnish) in Finnish Romani. Nevertheless, in some dialects in current contact with Greek and East Slavic, the possibility loanverbs are borrowed together with their L2 inflection, e.g. Parakalamos Romani (22), Russian Romani, some Servy dialects of Ukraine, and Crimean Romani (23). In all of these latter dialects this borrowing strategy is also common with lexical verbs. (22) Parakalamos Romani (Balkan, Greece; Matras 2004) a. Boro te džav. can.PRS.1SG COMP go.SBJV.1SG ‘I can go.’ (p. 87)

Modals in Romani

b. Na

borusa te phiravavas i can.PST.1SG COMP open.IMPRF.1SG DEF ‘I could not open the door.’ (p. 94) NEG

285

vudar. door

(23) Crimean Romani (Balkan, Ukraine and Russia) Tu možeš te pak’as leske thou.NOM can.PRS.2SG COMP believe.SBJV.2SG he.DAT ‘You can believe him.’ Although the East South Slavic possibility verb može may also be subjectinflected, it is frequently used as an impersonal modal, assuming the default, third-person singular, subject category. While some Romani dialects that borrow this modal allow both constructions, others have generalized the impersonal construction. Personal loanwords of može are always morphologically adapted through dialect-specific loanverb suffixes, e.g. mož-in-, mož-ij-, mož-iz-, mož-is-, mož-i(sar)-, i.e. there is no retention of L2 subject inflection. Impersonal loanwords of the verb, on the other hand, may but need not be adapted. Most dialects borrow the third-person singular present L2 form može, although they may also allow its adaptation and indigenous third-person singular inflection, especially in more complex TAM categories. Obligatory adaptation of the impersonal verb even in the present is exceptional, attested for example in Rakitovo Yerli, cf. mož-ijΩl-a/as [-LOAN-3SG-PRS/REM] (24). (24) Rakitovo Yerli (Balkan, Bulgaria) a. MožijΩla li tΩ kΩrΩs buti þukosa? can.PRS.3SG Q COMP do.SBJV.2SG work hammer.INSTR ‘Can you work with a hammer?’ b. Of na daralas þe možijΩlas tΩ he.NOM NEG fear.IMPRF.3SG COMP can.IMPRF.3SG COMP pΩrΩl. fall.SBJV.3SG ‘He was not afraid that he could fall.’ The East modality auxiliary auxiliary

Slavic construction of možno is retained in Romani: though the word itself is uninflected, it takes an impersonal inflectional that must be omitted in the present indicative (Type nPF). The is either borrowed in its East Slavic form or rendered by the

286

Viktor Elšík, and Yaron Matras

Romani copula. The uninflected possibility modal lako in Dolenjski Romani is, like its Slovene source, used in impersonal–finite constructions with TAM marking on the modalized verb (Type nPF2) and without a complementizer (25). The German Sinti permission modal darfte occurs in the same type of construction, as the complementizer te has become an integral part of the modal; darf- is a petrified singular present stem of the German modal dürfen. (25) Dolenjski Romani (Slovene/Istrian; Slovenia) a. I brzo lende lako živinamo. and without they.LOC can live.PRS.1PL ‘And we can live without them.’ (p. 357) b. Pe

domislindža da ruv lako realize.PST.3SG COMP wolf can haja porana daja eat.PST.3SG old.ACC mother.ACC ‘He realized that the wolf may have eaten the grandmother.’ (p. 359) REFL.ACC

2.4. Necessity In all likelihood, there were no dedicated necessity modals in Early Romani. The necessity modals we find in present-day Romani are loanwords or result from recent, dialect-specific, grammaticalizations. Nevertheless, we also find in various Romani dialects several weakly grammaticalized modal constructions that indicate how necessity could have been expressed in Early Romani. These constructions do not contain a dedicated modal word (or affix) and quite often they are not even specialized for expressing a certain kind of modality. For example, in Sepeþides the independent subjunctive, introduced by the non-factual connector ti < *te, can be used to express not only necessity (26a), but also permission (26b), volition (26c) and other modal meanings (Cech and Heinschink 1999b: 119–120). And in Welsh Romani a personal construction involving the copula and a finite (subjunctive or present indicative) complement, also introduced by te, can be used to express not only necessity (27a), but also possibility (27b) or general engagement 10 in an event (27c). A similar range of modal functions is also attested for impersonal constructions involving the copula in some dialects of the

Modals in Romani

287

Balkans (cf. Boretzky 1999a: 113). Necessity is not explicitely encoded in examples such as (26a) and (27a), and so their necessity reading (and translation) must involve some sort of pragmatic inference. (26) Sepeþides (Balkan, Turkey; Cech and Heinschink 1999b: 120) a. Okulestar ti daras! that.ABL COMP fear.SBJV.1PL ‘We must be afraid of that!’ b. Ti

bešav akate? sit.SBJV.1SG here ‘May I sit here?’ COMP

c. Akana me tuke ti vakerav… now I.NOM thou.DAT COMP speak.SBJV.1SG ‘Now I want to tell you…’ (27) Welsh Romani (British, Wales; Sampson 1926, II: 119) a. Šomaš te lƗ phabƗ COP.PST.1SG COMP take.SBJV.1SG apple.PL kjathakja thaneste. such_and_such place.ABL ‘I had to fetch apples from such and such a place.’ b. Na šom mƝ te reperava NEG COP.PRS.1SG I.NOM COMP remember.PRS.1SG ‘I cannot remember.’ c. Šom mƝ COP.PRS.1SG I.NOM ‘I am living with him.’

te COMP

dživava live.PRS.1SG

kek. NEG

lesa. he.INS

Nevertheless, the above modal constructions may become specialized for necessity, especially deontic necessity (but also for possibility, see Section 2.3). In Hameln Sinti, for instance, the independent subjunctive without te always conveys obligation (28; Holzinger 1993: 92), as does the impersonal copula construction in Welsh Romani (29a–b). In the latter construction the modal subject is not only cross-referenced on the

288

Viktor Elšík, and Yaron Matras

subjunctive verb but also encoded through an obligatory accusative-marked NP. Necessity here is modelled on predicative possession, whereby the modal subject is construed as the possessor of the modalized event: in most Romani dialects, including Welsh Romani, the predicative possessive construction consists of a nominative-marked possessee NP, the copula agreeing with it in number (and person, if the possessee is pronominal, which is rare) and an oblique-marked possessor NP (29c). The possession model for necessity is exactly what we find in English, one of the L2’s of Welsh Romani. Boretzky and Igla (2004: 185) suggest that Romani possessive-like necessity constructions, which are well attested in several dialects, may result from contact-induced grammaticalization on the model of possessive-like necessity modals in various European languages. In some North Central dialects of Slovakia, including ýáry Romani, a personal (rather than impersonal–possessive) copula construction is specialized to express weak necessity, especially weak obligation (30). Even though Slovak, the current L2 of the dialect, expresses weak necessity by the verb of possession, here too we might be dealing with an instance of pattern borrowing, though a less straightforward one. (28) Hameln Sinti (Northwestern, Germany; Holzinger 1993: 325) Štrap i grajes, krap brush.SBJV.1SG DEF.OBL horse.ACC do.SBJV.1SG leske futera. he.DAT animal_food ‘I must brush the horse, I must prepare food for him.’ (29) Welsh Romani (British, Wales; Sampson 1926) a. ŠƯ man te dƗ les COP.PRS.3 I.ACC COMP give.SBJV.1SG he.ACC bnjt lǀvǀ. many money ‘I have to give him a lot of money.’ (II, p. 119) b. Sas

Ư

Džakes te del COP.PST.3 DEF.OBL Jack.ACC COMP give. les vavƝr stƗdƯ sunakaj. SBJV.3SG other hat gold ‘Jack had to give him another hatful of gold.’ (I, p. 213)

Modals in Romani

c. Sas

289

phurƯƗ trin gurunƯƗ. old.F.ACC three cow.PL ‘The old woman had three cows.’ (I, p. 213) COP.PST.3

Ư

DEF.OBL

(30) ýáry Romani (North Central, Slovakia) Sal tuke te COP.PRS.2SG thou.DAT COMP pametinel so tuke phenƗ. remember.SBJV.3SG[=INF] what thou.DAT say.FUT ‘You ought to remember what I’m going to tell you.’ In several dialects the weakly grammaticalized copula constructions have undergone considerable grammaticalization. In most South Central the impersonal copula construction developed into a strongly grammaticalized necessity modal site or iste, which continues the third-person present indicative copula form (i)si plus the complementizer te. And in the Core Sinti dialects the personal copula construction developed into the strongly grammaticalized necessity modal homte, hunte, unti etc., 11 which continues the generalized first-person singular 12 present indicative copula form hom < *(i)som plus the complementizer (cf. already Pott 1844). In either case the two etymological components have fused together and the resulting univerbal modal is separable from the modalized verb. In Selice Rumungro site may constitute a separate utterance (31b) and in Hameln Sinti hunte may be postposed to the verb (32b). Although both modals are historically based on a present indicative copula form, neither inflects for TAM categories: their marking has shifted from the copula to the modalized verb (31b, 32c; Type nPF2). The Rumungro modal takes a regular negator rather than involve the irregular negative third-person present copula form, i.e. na site rather than *nane te (31b). Both modal constructions now show nominative marking of the modal subject (31a, 32a–b), although in older Rumungro oblique marking was still available in the impersonal si te construction (cf. Müller 1869: 161). Unlike their less grammaticalized source constructions, site and homte serve as the basic necessity modals in the respective dialects and have a wide range of necessity functions. (31) Selice Rumungro (South Central, Slovakia) a. Amen site džas onća. we.NOM must go.SBJV.1PL there. ‘We have to go there.’

290

Viktor Elšík, and Yaron Matras

b. Na

site mange papaleg palikerćal. De, site! must I.DAT again thank.PST.2SG but must ‘[A:] You need not have thanked me again. [B:] I did have to!’

NEG

(32) Hameln Sinti (Northwestern, Germany; Holzinger 1993: 92–93) a. Tu hunte kres kova. thou.NOM must do.SBJV.2SG that b. Tu kres kova hunte. thou.NOM do.SBJV.2SG that must ‘You have to do that.’ c. Hunte džajom must go.PST.1SG ‘I had to go home.’

khere. home

Like the South Central dialects, the Lovari-type dialects spoken in and around Hungary, including Austrian Lovari, possess the necessity modal site or iste, which is used in constructions with TAM marking on the modalized verb (33a; Type nPF2). Unlike its South Central counterpart, however, the Lovari modal is not separable from the modalized verb. In addition, the dialect has also retained a less grammaticalized copula construction in the past, where tense is marked on the copula rather than on the modalized verb and where the complementizer clearly remains a separate syntactic unit (33b; Type nPF1). The latter construction is only attested as an expression of participant-external necessity, while site appears to have a wider range of necessity functions. (33) Austrian Lovari (North Vlax, Austria) a. Site ašasas khere. must stay.IMPRF.1PL home ‘We had to stay at home.’ (Cech et al. 1999: 28) b. Me sas te dikhav pel I.NOM COP.PST.3 COMP see.SBJV.1SG on.DEF.PL grast. Sas te pijavav, te horse COP.PST.3 COMP make_drink.SBJV.1SG COMP xaxavav le. make_eat.SBJV.1SG they.ACC

Modals in Romani

291

‘I had to take care of the horses. I had to give them to drink, to feed them.’ (Fennesz-Juhasz and Heinschink 1999: 66) There are several further sources of dialect-specific grammaticalization of necessity modals in Romani. A few Balkan and Appenine dialects of Romani have grammaticalized the verb ther- ‘to hold’ into a personal verb of possession (‘to have’) and in one of these dialects, Rumelian Romani, this verb may also express necessity (34; Type PF). Both developments are contact-induced (cf. Boretzky and Igla 2004: 185), the latter once again, as with the impersonal copula construction, exemplifying the possessive model for necessity. Another common source of necessity expressions in Romani are volition verbs (cf. Boretzky 1996a: 14–17). Personal volition verbs may occasionally be used to express necessity in some Romani dialects of Greece, Slovakia, Ukraine (35; Type PF) and perhaps elsewhere. Pattern borrowing from Greek can be invoked to explain these uses at least in Greek Romani (cf. Boretzky 1996a: 16). In quite a few dialects the polysemous verb kam- (see Section 2.2) may also mean ‘to owe’ and this meaning, rather than the verb’s volition meaning, might be the immediate source of its necessity uses (cf. English ought, the historical past of owe). (34) Rumelian Romani (Balkan, Turkey; Paspati 1973 [1870], Boretzky 1999a: 113) Therava te phenav tuke. have.PRS.1SG COMP say.SBJV.1SG thou.DAT ‘I have to tell you.’ (35) Servy Ghympeny (Northeastern, Ukraine) Me na kamam dala draba I.NOM NEG like.PRS.1SG this.PL drug.PL save me kamam te prelaw. which.PL I.NOM want.PRS.1SG COMP take.SBJV.1SG ‘I do not like these pills that I have to take.’ Impersonal de-volitional constructions show a greater degree of grammaticalization. In these constructions the volition verb takes the default, third-person singular form and is accompanied by a reflexive pronoun in the accusative case, which is a way of decreasing the construction’s valency; the modal subject NP is in an oblique case. In Prizren Arli the verb mang- encodes volition in a personal construction, but necessity in an impersonal–reflexive construction; this may have been

292

Viktor Elšík, and Yaron Matras

modelled on Albanian (Boretzky 1999a: 114). In Kumanovo Kovaþki and Crimean Romani, on the other hand, the two verbs are now distinct: the original volition verb kam- had acquired a wide range of necessity functions within the impersonal–reflexive construction, but its volition meaning was later taken over by the verb mang- (see Section 2.2). Though the reflexive necessity construction (Type nPF1) is still attested in Crimean Romani (36a), the dialect now prefers dropping of the reflexive pronoun (36b), apparently without any semantic effect. (36) Crimean Romani (Balkan, Ukraine and Russia) a. Tumenge kamela pes te dž’an you.DAT necessary.3SG.PRS REFL COMP go.2PL.SBJV othe. there ‘You have to go there.’ b. Amenge kamelas te udž’aras e we.DAT necessary.3SG.IMPRF COMP wait.1PL.SBJV DEF.OBL brešÕndestar. rain.ABL ‘We had to wait because of the rain.’ The Central dialects of Romani exhibit further developments of the impersonal–reflexive de-volitional construction (cf. Boretzky 1996a: 14– 15). The accusative form of the singular reflexive pronoun pe(s) has been agglutinated to the volition verb kam-. In the Vendic subgroup of the Central dialects the reflexive was prefixed and some of the forms then underwent further phonological erosion, e.g. present *pe+kam-el [REFL+want-3SG] > peka-l. In the Rumungro and North Central subgroups, on the other hand, the reflexive was suffixed to the volition verb and the forms underwent not only phonological erosion but also externalization of verb inflections, e.g. present *kam-el+pe(s) [want-3SG+REFL] > ka(m)pe(l). Typically, these impersonal necessity modals are used as expressions of participant-internal necessity, though in several Central dialects they have been extended to other necessity functions as well. Some of these latter dialects (viz. the outliers Gurvari, Cerhari and Plašþuny) retain the impersonal construction in all necessity functions, while others reserve the impersonal construction (Types nPnF or nPF1) for expressing participantinternal necessity (37a, 38a) but develop a functionally distinct personal

Modals in Romani

293

construction (Type PnF) for other types of necessity. For example, in numerous Central dialects of Slovakia, e.g. Luþivná Romani, subjectinflected kamp- expresses weak obligation (37b) and in the dialect of Dobšiná it expresses participant-external necessity in general (38b). 13 Note that the strongly grammaticalized de-volitional necessity modal (e.g. 37a) is functionally distinct from the less grammaticalized impersonal–reflexive construction involving the verb kam- (37c; Type nPnF), which still expresses volition (see Section 2.2). (37) Luþivná Romani (North Central, Slovakia) a. Kampel mange te džal be_needed.PRS.3SG I.DAT COMP go.SBJV.3SG[=INF] andre skĐepa. in.DEF shop ‘I need to go to the shop.’ b. Furt kampes avka always be_needed.PRS.2SG so kerel. do.SBJV.2SG ‘You should always act like this.’

te

c. KamĐas pes mange want.PST.3SG REFL.ACC I.DAT džal. go.SBJV.3SG[=INF] ‘I wanted to go there.’

odoj there

COMP

te COMP

(38) Dobšiná Romani (North Central, Slovakia) a. Kampel mange te zav andro be_needed.PRS.3SG I.DAT COMP go.SBJV.1SG in.DEF foro. town ‘I need to go to town.’ b. Kampav te zal andro foro. must.PRS.1SG COMP go.SBJV.3SG[=INF] in.DEF town ‘I have to go to town.’

294

Viktor Elšík, and Yaron Matras

The majority of Romani necessity modals are dialect-specific lexical borrowings. Only few Romani dialects do without a necessity loanword, namely some of those that possess strongly grammaticalized de-copular and/or de-volitional necessity modals. Loans of the following necessity expressions are attested within Romani: Turkish lâzÕm, Greek prepi, South Slavic trjabva treba(ti), mora(ti) and valjati, Romanian trebui and musai, Hungarian muszáj, Slovak musieĢ, maĢ and potrebovaĢ, Polish musieü, trzeba and powinien, East Slavic treba, dolžen, nado and prixoditsja, German müssen and brauchen, English ought, Swedish måste and må, and Finnish pitää. 14 The dialectal distribution of these loanwords is too complex to be discussed here (see Boretzky and Igla 2004: 184–187 for a partial overview). Some loanwords are restricted to dialects whose speakers are currently bilingual in the source language, while others have been retained even after the Romani speakers shifted to a different L2. For example, the loan of Romanian trebui is found not only in most Romani dialects of Romanian bilinguals but also in most non-Romanian Vlax dialects, some of which lost contact with Romanian centuries ago; it has also diffused from Vlax into some non-Vlax dialects of Macedonia and Bulgaria. Many, perhaps most, borrowed necessity modals retain the range of functions they have in their source language, especially when the Romani speakers are currently bilingual in it. For example, the loan of English ought in Welsh Romani expresses weak obligation or probability and the loan of Slovak potrebovaĢ in some varieties of Slovak Romani expresses participant-internal need. However, the function of Romani necessity loanwords may also differ from that of the source modal, especially if the source language is not the dialect’s current L2 and/or if the loan acquires a different morphosyntactic status than it has in the source language. For example, the loan of the Polish uninflected and generic modal trzeba ‘one must, one should’ (cf. Hansen, in press) need not have a generic subject in Kohila Romani (39a) and may even show personal inflection (39b). On the other hand, loanwords from an older L2 may also show functional specialization, often due to competition with loanwords from the current L2. For example, Ub Gurbet trubu- appears to be specialized for weak obligation (40a), although its source modal, Romanian trebui, is much more general; the current Serbian loan mora- has taken over most necessity functions, including strong obligation (40b).

Modals in Romani

295

(39) Kohila Romani (Northeastern, Estonia) a. Jöj þebi lǀve te janel. she must money.PL COMP bring.SBJV.3SG ‘She must bring the money.’ b. Tu þebisos te jƝs an gošti. thou must.PRS.2SG COMP come.SBJV.2SG in guest.PL ‘You must come for a visit.’ (40) Ub Gurbet (South Vlax, Serbia) a. Tu trubus te posetis thou ought.PRS.2SG COMP visit.SBJV.2SG ma jekar. I.ACC once ‘You should visit me once.’ b. Uvek moraš taes pažljivo. always must.PRS.2SG COMP come.SBJV.2SG careful ‘You must always be careful.’ Necessity loanwords inflecting for subject categories are mostly morphologically adapted through loanverb suffixes, e.g. Dolenjski Romani mor-in- < Slovene mor-, and show Romani verb inflection. Retention of L2 subject inflection is rare, being only attested in some of those Romani dialects that borrow the Serbian or Macedonian personal verb mora-. For example, in Kosovo Bugurdži (Boretzky 1993: 74) singular present forms of the modal retain their Serbian inflection, i.e. 1SG mora-m, 2SG mora-š, and 3SG mora. However, the third-person singular form mora is also used with plural subjects of any person, like in colloquial Serbian. In further dialects of Serbian and Macedonian Romani the generalization of the thirdperson singular form has been completed and mora is now used irrespective of person or number of the modal subject (Type nPF1). The impersonal mora mostly encodes TAM categories, either analytically, through the inflectional auxiliary, e.g. Serbian Kalderaš past mora sas [must COP.PST.3], or synthetically, through indigenous TAM morphology, e.g. Šutka Arli imperfect mora-ine [must-REM]. On the other hand, impersonal modals may also acquire subject inflection in Romani. For example, the loan of trebui, which only inflects for TAM categories in Romanian, has developed subject inflection in several Vlax dialects. 15 In

296

Viktor Elšík, and Yaron Matras

some of them the borrowed modal shows defective subject inflection and personal constructions alternate with impersonal ones. For example, the Austrian Lovari loan trubu- only has third-person and second-person forms (Cech and Heinschink 1999a: 63), which are used personally (41a–b; Type PF). First-person subjects, on the other hand, require an impersonal construction (41c–d; Type nPF1). Though there are distinct second-person forms, the impersonal construction may also be employed with secondperson subjects (41e). (41) Austrian Lovari (North Vlax, Austria) a. Trubun te den e love. need.PRS.3PL COMP give.SBJV.3PL DEF.PL money.PL ‘They need to give money.’ (Cech and Heinschink 1999a: 73) b. Trubundanas sama te les tu. need.PLPF.2SG attention COMP take.SBJV.2SG thou ‘You should have paid attention.’ (Cech et al. 1999: 64) c. Me trubujas lašo šavo tavav. I.NOM need.IMPRF.3SG good son COMP come.SBJV.1SG ‘I should be a good son.’ (Cech et al. 1999: 48) d. Trobundas i hera opre te las. need.PST.3SG DEF clover up COMP take.SBJV.1PL ‘We needed to lift the clover.’ (Cech et al. 1999: 82) e. Trubujas tu taves o need.IMPRF.3SG thou COMP come.SBJV.2SG DEF kraj ande gado them. king in this country ‘You should become the king of this country.’ (Cech et al. 1999: 50) Those loans of impersonal necessity verbs that remain impersonal in Romani are often morphologically integrated as Romani-inflected verb forms of the default, third-person singular, subject category, e.g. Kumanovo Arli present treb-el [-3SG], treb-el-aine [-3SG-REM], preterite treb-in-g’-a [-LOAN-PFV-3SG.PFV] < Macedonian treba. Only rarely do we find internally derived TAM forms that do not contain the default subject morphology. For example, the Sliven Muzikanta imperfect trjabv-as is

Modals in Romani

297

derived directly from the Bulgarian present form trjabva by means of a Romani tense/mood suffix. Quite commonly, on the other hand, impersonal necessity verbs are borrowed without any morphological integration and retain their L2 inflection for TAM categories, e.g. present prepi and imperfect eprepe from Greek, present trjabva and past trjabvaše from Bulgarian, present trebuje < trebuie and imperfect trebuja < trebuea from Romanian, or present pittää and conditional pitäs < pittäis from (colloquial) Finnish. Several Romani modals inflect for TAM categories even though their source forms do not. For example, the uninflected Swedish modal måste has been borrowed as an impersonal verb (Type nPnF) into Finnish Romani, showing synthetic TAM marking, e.g. present– future most-u-l-a [-LOAN-3SG-PFUT] and preterite most-u-dƯl-o [-LOAN-PFV3SG.PFV.M]. A common development is a complete loss of inflection of borrowed modal verbs. For example, in Srem Gurbet the loan of the Serbian personal verb mora- lost not only subject inflection but also TAM inflection: the original third-person singular present form mora is now used in all TAM categories, which are thus marked solely on the modalized verb (Type nPF2). Similarly, Welsh Romani and some varieties of Slovak and Lithuanian Romani have generalized forms such as mus muš mos musi, which go back to inflectional stems or the thirdperson singular present forms of the German personal modal müssen and its loans into Western Slavic, e.g. musi < Slovak 3SG mus-í. These generalized forms may then fuse with the complementizer te, resulting in uninflected univerbal modals such as mos-te in ýáry Romani (see Section 2.1, 10b), which parallel the strongly grammaticalized indigenous modals hun-te and si-te in their morphosyntax (Type nPF2). A similar development also gave rise to the uninflected modal braux-te in German Sinti < German brauchen. Sometimes we find, within a single Romani variety, competition or complementary distribution between different constructions involving necessity loanwords. Several dialects reveal a TAM asymmetry here: the present indicative is more susceptible to borrowing, more likely to remain unintegrated, and less likely to show subject inflection than other TAM categories. In Vidin Kalajdži, for instance, the integrated impersonal loanverb trubu-z- (from Romanian, an old L2) is used in all TAM categories but in the present indicative it is currently being replaced by the unintegrated trebva (from Bulgarian, the current L2). In Slavošovce Romani the Hungarian-origin modal mušaj must show analytical subject

298

Viktor Elšík, and Yaron Matras

and TAM inflection (Type PnF) in most TAM categories (42a), while in the present the personal construction (42b) is rare and the modal is usually uninflected (42c; Type nPF2). Similar patterns are also attested in Finnish Romani, where the uninflected mote (probably from Swedish må plus the complementizer te) is restricted to the present, while the impersonal most-u- (from Swedish måste) is used in all TAM categories; in some varieties of Macedonian Arli, where the loan of Macedonian mora- inflects for subject categories in the past, e.g. mora-n-dil-jum [must-LOAN-PFV1SG.PFV] ‘I had to’, even though the generalized impersonal mora is used in the present; and elsewhere. (42) Slavošovce Romani (North Central, Slovakia) a. Mušaj šþamaš te užarel. must COP.PST.1PL COMP wait.SBJV.3SG[=INF] ‘We had to wait.’ b. Mušaj šþom must COP.PRS.1SG ‘I have to go there.’

odoj there

te COMP

dzal. go.SBJV.3SG[=INF]

c. Adádžive mušaj megin te dzav odoj. today must again COMP go.SBJV.1SG there ‘Today I have to go there again.’ Several dialects of Bulgarian and Greek Romani, including the dialect of Vălþi Dol, show competition in the past between impersonal constructions with TAM marking on the modal ((43b); Type nPF1) and impersonal constructions with TAM marking on the modalized verb ((43c); Type nPF2). 16 In some of these dialects, e.g. Cerovo Cocomanja, Sliven Muzikanta (cf. Section 2.1, ex. (4)) or Kalamata Romani, the former construction expresses situational necessity (44a), while the latter construction appears to be specialized for epistemic necessity (44b). (43) Vălþi Dol Romani (South Vlax, Bulgaria) a. LazÕmi te džas ote. necessary COMP go.SBJV.2SG there ‘You have to go there.’

Modals in Romani

b. LazÕmi sasdÕ necessary COP.PST.3 ‘I had to go around her.’

te COMP

299

nakav latar. walk.SBJV.1SG she.ABL

c. LazÕmi te aþasas sostar necessary COMP stay.IMPRF.1PL what.ABL delas brÕšind. give.IMPRF.3SG rain ‘We had to stay because it was raining.’ (44) Kalamata Romani (Balkan, Greece) a. Eprepe te cikaras jati must.IMPRF COMP wait.SBJV.1PL because delas brisim. give.IMPRF.3SG rain ‘We had to wait because it was raining.’ b. SiȖura prepi te bidzardan surely must.PRS COMP get_to_know.PST.2SG -es ta apo prin. OBJ.3SG.M also of before ‘Surely you must have got to know him before.’ In addition to borrowed necessity modals we also find their loan translations. While Russian and Ukrainian dialects of Romani usually borrow the East Slavic impersonal verb prixoditsja (lit. reflexive ‘come’) in its L2 form, also retaining its L2 inflection, some varieties prefer to use its ‘semicalqued’ or ‘calqued’ equivalent, e.g. present pri-dža-l-pe [AKTIONSART-go-3SG-REFL] or jav-el-pe [come-3SG-REFL]. Similarly, the East Slavic adjectival modal dolžen ‘obliged’, which still also retains its lexical meaning ‘indebted’ in the source languages, is mostly borrowed (and, in some varieties, morphologically integrated in the way borrowed adjectives are) but can also be translated by the adjective bango ‘crooked; indebted’. Interestingly, since the Russian copula does not encode person in the past (45a), the default, third-person, form of the Romani copula is selected in the ‘calqued’ construction (45b; Type nPF1) in Russian Romani. However, personal copula occurs in Crimean Romani (cf. 9).

300

Viktor Elšík, and Yaron Matras

(45) Russian Romani (Northeastern, Russia) a. Tu dolžen byl thou [obliged.SG.M COP.PTC.SG.M]Russian dykxes les. see.SBJV.2SG he.ACC b. Tu bango sys thou obliged.SG.M COP.PST.3 ‘You must have seen him.’

te COMP

te COMP

dykxes les. see.SBJV.2SG he.ACC

2.5. Modality and negation Negation in modal constructions may have scope over the modalized proposition or over the modality itself (cf. Palmer 2001: 90–98). In Romani the differences in the scope of negation are reflected especially in the linear order of the negator, as exemplified for the Rumungro (South Central) variety of Selice, Slovakia. In (46b) and (47b) the negator immediately precedes the main verb, having scope over the modalized proposition: the constructions express, respectively, possibility and necessity of a negative proposition. In (46c) and (47c), on the other hand, the negator immediately precedes the modal, being cumulated with it in (46c), and has scope over the modality: the constructions express, respectively, negative possibility and negative necessity of an affirmative proposition. Finally, in (46d) and (47d) both negators are used: the constructions express, respectively, negative possibility and negative necessity of a negative proposition. 17 (46) Possibility a. šaj le keras can he.ACC do.SBJV.1PL ‘we can/may do it’

(47) Necessity a. site le keras must he.ACC do.SBJV.1PL ‘we must do it’

b. šaj le na keras can he.ACC NEG do.SBJV.1PL ‘it is possible that we don’t do it’ i.e. ‘we need not do it’ c. naštig le keras cannot he.ACC do.SBJV.1PL

b. site le na keras must he.ACC NEG do.SBJV.1PL ‘it is necessary that we don’t do it’ i.e. ‘we must not do it’ c. na site le keras NEG must he.ACC do.SBJV.1PL

Modals in Romani ‘it isn’t possible that we do it’ i.e. ‘we cannot/may not do it’ d. naštig le na keras cannot he.ACC NEG do.1PL.SBJV ‘it isn’t possible that we don’t do it’ i.e. ‘we must do it’

301

‘it isn’t necessary that we do it’ i.e. ‘we need not do it’ d. na site le NEG

must he.ACC

na

keras

NEG

do.SBJV.1PL

‘it isn’t necessary that we don’t do it’ i.e. ‘we can/may do it’

In most Romani dialects, as in Selice Rumungro, negators preceding (or cumulated with) the modal have scope over modality. A single exception is attested: in Welsh Romani the negator preceding the necessity modal mus (48a) has scope over the proposition, so that na mus (48b) encodes necessity of a negative proposition (‘must not’) rather than negative necessity (‘need not’). This is clearly due to pattern borrowing from English, although mus itself is more likely to be of German origin (Sampson 1926: I, 216). In the remainder of this section we will be concerned solely with negative modality, focusing on the formal relationship between affirmative and negative modality. (48) Welsh Romani (British, Wales; Sampson 1926, II: 236) a. Mus te ves mansa. must COMP come.SBJV.2SG I.INS ‘You must come with me.’ b. Na

mus te þas odoj must COMP stay.SBJV.1PL there ‘We must not stop there long.’ NEG

þΩĚlƗ. long

Although negative volition is expressed by regular negation of the affirmative volition modal in most Romani dialects, a few dialects of Bulgaria employ a different modal for negative volition than for affirmative volition. For example, in Pazardžik Malokonare the original volition modal kam- is retained in negative contexts (49b), while the newly grammaticalized volition modal mang- (see Section 2.2) is used in affirmative contexts (49a). The Rešitare dialect of Velingrad represents a further stage in the replacement of kam- by mang-: while the former is restricted to negative contexts, the latter may now be used in both polarities. Another rare type of irregularity concerns the shape of the negator: in Cerovo Cocomanja most verbs take the indicative negator na,

302

Viktor Elšík, and Yaron Matras

while the volition verb kam- takes nan, which probably derives from the third-person present negative copula form nanaj. (49) Pazardžik Malokonare (Balkan, Bulgaria) a. Mangava te žav andi zi. want/demand.PRS.1SG COMP go.SBJV.1SG in.DEF town ‘I want to go to town.’ b. Na

kamam te žav andi zi. want/love.PRS.1SG COMP go.SBJV.1SG in.DEF town ‘I don’t want to go to town.’ NEG

Negative possibility is frequently irregular in that it is often not expressed by regular negation of the affirmative possibility modal. Although the Early Romani negative possibility modal našti may continue a regularly negated form of the affirmative possibility modal ašti or derive from a construction involving the negator and the affirmative possibility modal šaj (see Section 2.3), in neither case is našti completely regular in synchronic terms. Indeed, pairs such as šaj vs našti can be even considered to be suppletive. Nevertheless, many Romani dialects have discontinued this Early Romani heritage as a result of hyperanalysis (cf. Croft 2000: 121–126), analogical regularization, grammaticalization and borrowing. The first source of regular negation of possibility is the semantic shift of negative possibility modals into modals of affirmative possibility. This development, which has occurred in most modern Sinti dialects, consists in an addition of a more transparent negator (the German-origin nit/nix or gar) to a former negative possibility modal (našti or naj) and a subsequent hyperanalysis of the inherent negative value of the modal as a property of the negator alone (cf. Boretzky and Igla 2004: 183–184, Elšík and Matras 2006: 159). Examples from Austrian Sinti, a dialect in a transitional stage of the development, are illustrative: the modal naj may still express the negative possibility function by itself (50a), though it is more likely to be negated by a borrowed negator in this function (50b); the prevailing function of plain naj is affirmative possibility (50c). Those Sinti varieties that have undergone the complete shift now possess regularly negated possibility modals, e.g. Auvergne Manuš našti ‘can’ vs našti gar ‘cannot’. Although the addition of a more transparent negator to the negative possibility modal has also occurred in several Balkan dialects of Romani, giving rise to forms such as na-našti, there is no hyperanalysis of negation

Modals in Romani

303

and so plain našti, if there is one, continues to function as a negative possibility modal in these dialects. 18 (50) Austrian Sinti (Northwestern, Austria) a. Naj bešƝ imΩr kheri. cannot sit.PRS.2SG always home ‘You can’t sit at home all the time.’ b. Naj fƗrena butΩr nit kheri. can[not] drive.PRS.3PL more NEG home ‘They can’t drive home anymore.’ c. Kola nicΩrens man vri kaj naj that.PL use.IMPRF.3PL I.ACC out where can nicΩrens lΩ man vri. use.IMPRF.3PL they I.NOM out ‘They used me as much as they were able to use me.’ Next, the negative possibility modal našti may be re-analyzed as containing the regular verb negator and a novel affirmative possibility modal may be created by analogy. Boretzky’s (1996a: 5–6) claim that this kind of analogical decomposition gave rise to the affirmative ašti in general (see Section 2.3) is somewhat problematic in that it does not account well for the form’s initial vowel. Nevertheless, analogical decomposition is likely to be the source of šti ĸ na-šti in Moravian and Abbruzian Romani and of sti(k) ĸ na-sti(k) in Lombardian and Piedmontese Sinti (these dialects retain the Early Romani negator na). 19 Regular negation of inherited possibility modals may also result from an analogy in the opposite direction. Thus several Romani dialects of, or originating, in Romania have supplemented or even replaced the original negative possibility modal našti by a construction consisting of a regular verb negator and affirmative possibility modal, e.g. Kurturare þi šaj-, Šanxajcy þi dasþi(-sar)-, Laješa– Kišinjevcy–Rakarenge ni dašti(-sar)-, Ursari ni (d)ašti(-z)-, and probably also Kԥrԥmidarja na aštik. Finally, regular negation is also found with those possibility modals that have been grammaticalized from lexical verbs (see Section 2.3), e.g. džan‘to be able’ vs. na džan- ‘not to be able’, and in those dialects that borrow a possibility modal and negate it by a regular negator. The negator may be indigenous, in which case Romani ‘semicalques’ the regular negative

304

Viktor Elšík, and Yaron Matras

possibility expression of the source language of the possibility modal, e.g. Parakalamos Romani na bor-, Finnish Romani na vojuv-, Lithuanian Romani na možyn-. Or, less commonly, the possibility modal may be borrowed together with its L2 negator, as in Russian Romani and in some Ukrainian dialects, e.g. in Kiev Servy (51). (51) Kiev Servy (Ukrainian, Ukraine) Ame n’it’ev n’e maȖl’i we.NOM nowhere [NEG can.PST.PL]Russian kada te arakhe. this COMP find.INF ‘We couldn’t find this anywhere.’ However, semicalquing and matter borrowing of negative possibility modals only represents a minor pattern within Romani: most dialects that borrow an affirmative possibility modal retain the indigenous negative našti, and so they re-iterate or introduce suppletion between their affirmative and negative possibility modals. Also, the use of a borrowed possibility modal in negative possibility does not automatically result in a regularly negated negative possibility expression. Both points may be illustrated by examples from Karditsa Romani. This dialect borrows its possibility modal from Greek (52a) but retains the indigenous negative possibility modal (52b). There is nevertheless another way to express negative possibility: the borrowed affirmative possibility modal may combine with the indigenous negative possibility modal, which in effect functions as an irregular negator in the domain of possibility (52c). (52) Karditsa Romani (Balkan, Greece) a. Borinea te keres buti can.PRS.2SG COMP do.SBJV.2SG work ‘Can you work with a hammer?’ b. Ame nasti džaa we.NOM cannot go.PRS.1PL ‘We can’t go back.’

me sfiri? with hammer

pala. back

c. Nasti borinava ti dav boja o kher. cannot can.PRS.1SG COMP give.SBJV.1SG dye DEF house ‘I can’t paint the house.’

Modals in Romani

305

Negative necessity is mostly expressed by regular negation of affirmative necessity modals. Exceptions to this general pattern are contact induced. For example, in Hameln Sinti the negative counterpart of the necessity particle hunte is a regular negation of the modal brauxte, which cannot occur without a negator (Holzinger 1993: 92). 20 This restriction derives from an identical restriction on the source element, the German verb brauchen ‘to need’, when used with a clausal complement. A similar, though somewhat more complex, pattern applies to the necessity verb peka(m)- (see Section 2.4) in Burgenland Roman. When used impersonally (Type nPnF) the verb expresses participant-internal necessity (53a) and can be regularly negated (53b). However, when the verb is inflected for subject categories within a personal construction (Type PnF) it must be accompanied by a negator and the construction expresses negative participant-external necessity (53c). Regular negation of the participanexternal necessity modal iste appears to be extremely rare (53d). 21 Halwachs (1998: 158) explains this asymmetry as a result of pattern borrowing from German. (53) Burgenland Roman (South Central, Austria; Halwachs et al. 1999) a. Le faþunge eklik te DEF.OBL child.PL.DAT a_little COMP hal pekalahi. eat.SBJV.3SG[=INF] be_needed.IMPRF.3SG ‘The children would need to eat a bit.’ (p. 48) b. Ni te hal, ništa lake not_even COMP eat.SBJV.3SG[=INF] nothing she.DAT na pekal. NEG be_needed.PRS.3SG ‘She does not even need to eat or anything else.’ (p. 66) c. Na

pekajs mandar te daral. must.PRS.2SG I.ABL COMP fear.SBJV.3SG[=INF] ‘You need not be afraid of me.’ (p. 96) NEG

d. Na

iste uso sikajipe džan. must to.DEF teaching go.PRS.3PL ‘They need not go to the course.’ (Romani Patrin 1999: 8) NEG

306

Viktor Elšík, and Yaron Matras

3. Grammaticalization and borrowing asymmetries 3.1. Functional aspects of grammaticalization Most of those Romani modal expressions that are not material borrowings derive from Romani lexical verbs or other pre-modal constructions. The sources for modals were discussed in the data sections (2.2–4) and are summarized in (54–56). Most of the developments are dialect-specific, with the exception of the pre-split extension of the verb ‘to love’ to volition functions (see Section 2.1). Modal uses of copula constructions are also likely to have existed in Early Romani, though grammaticalization of these constructions into dedicated modals is certainly dialect-specific (see Section 2.4). The majority of the developments in (54–56) involved contact-induced grammaticalization on the model of Romani’s L2s and some grammaticalized modals even result from straightforward calquing or semicalquing (e.g. those based on Russian prixoditsja, see Section 2.3). (54)

volition less agentive volition

< ‘love, like’, ‘ask for, demand’ < ‘want’ reflexive/middle

(55)

possibility capability permission

< ‘be/have’, ‘work out, go out’, ‘get [somewhere]’ < ‘know’, ‘understand’, ‘manage’ < ‘dare’

(56)

necessity

< ‘be/have’, ‘have, hold’, ‘want, owe’, ‘come’ refl, ‘indebted’ < ‘want’ reflexive

need, weak obligation

Some modals are less grammaticalized than others in that they still retain their lexical meanings as well, thus possessing greater semantic weight or integrity (cf. Lehmann 2002: 112). This holds especially for the volition modals, including the verb kam- (e.g. Selice Romani ‘to love, to like; to wish, to desire; to owe; to want’), whose modal function must have existed since Early Romani and which thus shows remarkable semantic stability. Lexical meanings are also retained in several more recently grammaticalized modals, such as those expressing capability (e.g. Crimean Romani hal’ov- ‘to understand; to feel; to guess; to manage; to be able’) and those that calque polysemous source expressions (e.g. Russian Romani bango ‘crooked; indebted; obliged’, see Section 2.4). Modals expressing participant-internal necessity mostly also possess the

Modals in Romani

307

lexical meaning ‘to need [sth.]’, though several dialects differentiate between expressions of event-oriented/modal need and objectoriented/lexical need, e.g. de-volitional kamp- vs. borrowed potrebin- in some dialects of Slovakia. Semantic developments between different kinds of modality include the extension of modals specialized for participant-internal possibility and necessity (capability and need) into other possibility and necessity functions as well (cf. the discussion of džan- and kamp-, respectively, in Sections 2.3 and 2.4). This is in line with van der Auwera and Plungian’s (1998) modality’s semantic map. We have suggested in Section 2.3 that Early Romani possessed two possibility modals, šaj and ašti. The evidence from Kalderaš, a dialect cluster that preserves both šaj and dašti- (most likely a variant of ašti), shows that the latter tends to be less grammaticalized semantically (according to van der Auwera and Plungian’s map). 22 In Piteúti Kalderaš šaj is only attested as an expression of epistemic possibility, the most grammaticalized semantic function, whereas daštidominates all possibility functions. In Taikon Kalderaš of Sweden, on the other hand, the modal šaj expresses participant-external, deontic or epistemic possibility (57a–c; Type nPF2), while dašti- is specialised for participant-internal possibility (cf. the glossary item ‘be able, be capable, can, manage’ in Gjerdman and Ljungberg 1963: 222; Type PF). This distinction is parallelled by a similar, if not identical, functional distinction in the negative domain, probably due to morphological analogy: uninflected našti can express most impossibility functions (57d–e; Type nPF2), but inflected našti- is specialized for participant-internal impossibility (57f; Type PF). The Kalderaš data suggest that the Early Romani ašti was specialized for participant-internal possibility (perhaps due to a relatively late grammaticalization from a copula construction) and that its semantic extension to participant-external possibility is a recent, dialect-specific, development. 23 (57) Taikon Kalderaš (North Vlax, Sweden; Gjerdman and Ljungberg 1963: 127) a. Šaj dav tu deš kronurjä. can give.PRS.1SG thou.ACC ten crown.PL ‘I can give you ten crowns.’

308

Viktor Elšík, and Yaron Matras

b. Šaj keres sar kames. can do.PRS.2SG how want.PRS.2SG ‘You may do as you like.’ c. Šaj avel voj tehara. can come.PRS.3SG she tomorrow ‘She may/might come tomorrow.’ d. Našti ankΩrdine lako trajo. cannot hold.PST.3PL she.GEN life ‘They couldn’t save her life.’ e. Našti vazdel o gono korkoro. cannot lift.PRS.3SG DEF bag alone.M ‘He can’t lift the bag himself.’ f. ýi

naštin te keren amenge khanþi. unable.PRS.3PL COMP do.SBJV.3PL we.DAT nothing ‘They aren’t able to cause any harm to us.’ NEG

Of all modalities, volition is most prone to further grammaticalisation processes in Romani. The development of volition modals into necessity modals has been discussed in Section 2.4. In addition, in some North Vlax dialects we find the epistemic possibility adverb kam ‘perhaps’, which might derive from the verb kam-, although a Romanian etymology is also possible (Boretzky 2003: 67, Boretzky and Igla 2004: 173). Volition modals can also grammaticalize into post-modal meanings. Some varieties of Slovak Romani have developed the de-volitional free-choice postfix -kam in de-interrogative indefinites, e.g. so-kam ‘anything whatsoever’ (< *‘what one wants’). More importantly, in all Romani dialects of the Balkans the volition verbs have given rise to the future proclitics kam kan ka < *kam- or ma mΩ < *mang-, which are modelled on the future markers deriving from volitional verbs in other Balkan languages (Greek șa and so on; Boretzky 1999b: 170–172). Future meanings may also be encoded by the impersonal copula construction, which othewise tends to grammaticalize into possibility and, more often, necessity modals (see Sections 2.3–4). In several Romani dialects of Macedonia and Bulgaria this development has occurred especially in negative polarity due to pattern borrowing from East South

Modals in Romani

309

Slavic (Boretzky 1999b: 172–173). For example, in Montana Kalajdži affirmative future is marked by the de-volitional proclitic ka(m) (58a), while the negative future is encoded through a construction containing the negative third-person present copula form and the non-factual complementizer (58b), calquing Bulgarian šte and njama da, respectively. In several dialects of Romania and Kosovo impersonal copula constructions encode future in both polarities, due to an imperfect replication of the personal possessive-like future construction in Romanian and Albanian. For example, we find future constructions with present copula forms in Ursari (cf. Boretzky 1999a: 113) and constructions with future copula forms in Kԥrԥmidarja (59a–b). Although the development of future from necessity is well attested cross-linguistically, it is not obvious that this is what happened in the above Romani dialects. It is quite possible that, at least in some of the dialects, the semantically vague copula construction acquired the future function directly, due to pattern borrowing from an L2, without ever passing through the necessity stage. In any case, post-modal developments of strongly grammaticalized possibility and necessity modals are unattested in Romani. (58) Montana Kalajdži (Balkan, Bulgaria) a. Tase kam- ovav amende. tomorrow FUT become.SBJV.1SG we.LOC ‘Tomorrow I will be at home [at our place].’ b. Tase nanaj te ove tomorrow COP.NEG.PRS.3 COMP become.SBJV.2SG tumende. you.LOC ‘Tomorrow you won’t be at home [at your place].’ (59) Kԥrԥmidarja (North Central, Romania) 24 a. Tese ala te avaw tomorrow come.FUT.3SG COMP come.SBJV.1SG ‘Tomorrow I will be at home.’

khere. home

b. Tese na ala te aves khere. tomorrow NEG come.FUT.3SG COMP come.SBJV.2SG home ‘Tomorrow you won’t be at home.’

310

Viktor Elšík, and Yaron Matras

3.2. Formal aspects of grammaticalization The least grammatical Romani modals are fully fledged morphological verbs that, like most lexical verbs, inflect for TAM and subject categories. Syntactically, these modal verbs have scope over the whole clause that expresses the modalized proposition: the modalized verb occurs in a subordinate form and is, disregarding some exceptions that are not specific to modal constructions (see Section 2.1), introduced by the non-factual complementizer. Also, the nominative marking of the modal subject NP, which is the only option with personal modals, can be interpreted as marking of the grammatical subject of the modal verb. There are three major diachronic classes of personal modal verbs in Romani. First, many are lexical borrowings of L2 personal modal verbs, thus simply retaining the morphosyntactic properties of their source forms. Second, several modal verbs such as volition kam- and mang-, (cap)ability džan- and axaljov-, permission troma- or necessity ther- derive from lexical verbs. Significantly, this class of modals overlaps almost perfectly with the class of modals that still retain their pre-modal meanings as well, and so the low degree of morphosyntactic grammaticalization of these modal verbs parallels their low degree of semantic grammaticalization. Finally, and perhaps most interestingly, dedicated modals, too, may be personal verbs. This is the case of the possibility modals šaj(-in)-, (d)ašti(-z/sar)- and našti(-n/sar)- in a few (formerly contiguous) dialects (see Section 2.2). These possibility verbs do not represent an earlier, less grammaticalized, stage of the uninflected šaj, ašti and našti, which are found in the rest of Romani. Rather, they must have acquired their personal inflection secondarily, through analogy, one kind of evidence being their morphology otherwise typical of loanverbs. Similarly, the personal inflection of kamp- and pekam- in some dialects must be an analogical innovation based on the impersonal kamp- and pekam- (see Section 2.3), since the forms contain a reflex of the reflexive pronoun, which is only used in the third person. Despite their analogical rather than grammaticalization origin, the above possibility verbs tend to have less grammatical functions than their uninflected counterparts, provided they co-exist within a variety (cf. 57). On the other hand, if both personal and impersonal kamp- or pekam- co-exist within a variety, the personal (participant-external) modal is more grammatical semantically than the impersonal (participant-internal) modal, which appears to contradict the degree of their morphosyntactic grammaticality. The fact that the

Modals in Romani

311

impersonal (participant-internal) necessity modals kamp- or pekam- also function as lexical verbs of need seems to be relevant here. Impersonal modal verbs, i.e. modals inflecting for TAM but not subject categories, are obviously less differentiated morphologically than personal modal verbs. The lack of subject inflection may result from obligatorification of an impersonal modal construction (as in the case of the loans of Bulgarian može in some dialects, see Section 2.2) or from fossilization of a frequent person–number form of a personal modal (as in the case of the borrowed mora in some dialects, see Section 2.3), and so, given that obligatorification and fossilization are recognized paradigmatic aspects of grammaticalization (cf. Croft 2000: 157), impersonal modals can be considered to be more grammatical than personal modals. If there is alternation between personal and impersonal modal forms, the latter tend to occur in the more frequent TAM categories. Like personal modal verbs, impersonal modal verbs have scope over the whole clause that expresses the modalized proposition: the modalized verb occurs in a subordinate form and is introduced by the complementizer. However, some impersonal modal verbs with finite complements (Type nPF1) have nominative marking of the modal subject NP, which is to be interpreted as the grammatical subject of the modalized verb rather than of the modal. This indicates a certain degree of syntactic degradation of the modal. Impersonal verb inflection is common especially with possibility and necessity modals, often due to the impersonality of the source or model expressions in Romani’s L2s. Clearly the most grammatical are those modals that do not inflect either for TAM or subject categories (Type nPF2), thus functioning as invariant markers of modality that modify a fully inflected verb. Syntactically, such modals do not influence the argument structure of the modalized verb, and so the modal subject is invariably marked as the grammatical subject of the modalized verb (i.e. nominative-marked with most verbs). Also, they are the only type of modals that allow an obligatory absence of the non-factual complementizer, which is the last remnant of the embedded status of the modalized verb. The development of uninflected modals thus involves both paradigmatic and syntagmatic morphosyntactic processes of grammaticalization: fossilization and/or morphological loss of verbal inflection on the one hand and structural condensation, scope decrease and loss of independent syntactic status on the other (cf. Croft 2000: 157). In addition to modals that were inherited as uninflected from Early Romani (šaj, ašti and našti) and borrowed modals that are uninflected in the source language

312

Viktor Elšík, and Yaron Matras

as well, there are quite a few instances of recently grammaticalized uninflected modals in the domain of possibility (naj, trunti, darfte) and especially necessity (hunte, site, mus, moste, brauxte, mote and more). A complete loss of inflection and independent syntactic status in expressions of volition, on the other hand, is associated with post-modal functions. If there is alternation between constructions with inflected vs uninflected modals, the latter tend to occur in the more frequent TAM categories. More importantly, they also tend to occur in the more grammaticalized functions within a certain modality, as the distinction between the the tensed modals expressing situational necessity and the default-tense (and so in a sense uninflected) modals expressing epistemic necessity in (4) and (44) illustrates. All Romani modals are free forms, i.e. none is bound to the modalized verb as a clitic or even as an affix. This also holds for those Romani dialects that have been in contact with languages that possess bound modals (cf. possibility suffixes in Turkish and Hungarian): unlike free modals, bound modals are never borrowed in their material form; and if there is pattern replication it does not concern the degree of coalescence. Coalescence with the modalized verb has only affected the post-volition future markers (see Section 3.1), which are proclitics. Instances of coalescence of a modal with the non-factual complementizer te are restricted to uninflected modals, and so to possibility and necessity. Coalescence with the reflexive pronoun is typical of de-volitonal necessity. 3.3. Borrowing asymmetries Modality is a domain that is conspicuously susceptible to structural borrowing in Romani. However, the borrowing behaviour is asymmetrical: Some modality categories are more likely to be borrowed than others. The overall likelihood modality categories to be affected by structural borrowing is expressed by the following hierarchy: (60) necessity > possibility > volition Necessity appears at the top of the borrowability scale, meaning that it is both the most frequently borrowed modality category in the cross-dialect sample, and that lower positions on the hierarchy – volition and possibility – are not likely to be borrowed unless necessity is borrowed too. The

Modals in Romani

313

hierarchy seems compatible to some degree with the hierarchy of grammaticalization: Both possibility and necessity are targets in the grammaticalization process, rather than points of departure; in other words, they occupy higher positions on the grammaticalization hierarchy. This appears to suggest that the absence of a highly grammaticalized, semantically dedicated expression for modality attracts borrowing as an enrichment to the system. Such an argument follows a traditional “gaphypothesis” approach to borrowing: items that are “missing” in the recipient system are more likely to be borrowed from a donor system. It does not, however, explain the relative low susceptibility of volition to borrowing – which in Romani is the least grammaticalized modality category. It appears, rather, that the trigger for borrowing is to be found in the semantic-pragmatic functions involved. Prone to borrowing are those modality categories that are participant-external, rather than participantinternal. Arguably, necessity involves greater intensity of external pressure than possibility, allowing us to postulate the following hierarchy: (61) external pressure (necessity) > external facilitation (possibility) > internal (volition) Borrowing, in effect the speaker accommodating ‘intrusions’ from a (dominant) external language, is thus in a sense a metaphorical reflection of the actor succumbing to external pressure. In relation to language processing in communicative interaction, it is yet another expression of the susceptibility to borrowing (in an oral, minority language) of structures expressing contrast, discontinuity, uncertainty, or exemption. As discussed already in Elšík and Matras (2006: 385-386) and in Matras (1998), we may conclude that borrowing targets firstly those grammatical devices that are employed to process instances of potential tension between the message conveyed by the speaker, and the hearer’s expectations. It is, therefore, the ‘stress-effect’ in processing discourse that leads the bilingual speaker to lose control over the ‘correct’ selection of language form, a production error which, unaffected by normative attitudes may lead to convergence or ‘fusion’ of the structural representation of the category in both languages – i.e. to structural ‘borrowing’. A rather different picture emerges from the distribution of borrowed inflection accompanying borrowed modality structures. While such borrowed inflection is common with modals expressing possibility, it is rather rare with necessity – in a way reversing the hierarchy observed in

314

Viktor Elšík, and Yaron Matras

(60). (Borrowed inflection with volition is attested in only one dialects, Dolenjski). The explanation is likely to be related to the availability of impersonal modality markers in the contact languages, as well as the overall semantic-pragmatic tendency to express necessity as an external, impersonal force (cf. even the language-internal grammaticalisation from volition in kam- to an impersonal, reflexive necessity expression kampel). 4. Conclusions Figure 1 presents an integrated map of the grammaticalization paths of Romani modals. Two principal dimensions appear as relevant to the evaluation of modality and its grammaticalization paths. Modality categories can firstly be arranged on a cline reflecting the degree of intensity or determination – a conscious force driving the modal evaluation of a target action. This cline interacts with a split between the anchoring of modality through internal and external force. On the internal-force side, determination is expressed as agentivity: It is strongest in volition, weaker in capability, and ambiguous or irrelevant in possibility. On the externalfocre side, determination is expressed as intensive outside pressure: This is strong with necessity, weaker with permission, and ambiguous or irrelevant in possibility. Internal-force modality is of course present in volition and capability, but ambiguous in possibility, while external-force modality is present in necessity and permission, but also ambiguous in possibility. Possibility thus takes up a rather neutral position on the chart, free of explicit determination, and undefined for either internal-agentive or external-causative involvement.

Modals in Romani

Figure 1. Romani modality: An integrated map of grammaticalization paths

315

316

Viktor Elšík, and Yaron Matras

The grammaticalization paths of Romani modals generally lead from predications depicting a mental-emotional state, a material state, or, marginally, movement. There is a strong tendency for internal-force predications to feed into internal-force modality, and for predications that are situated at least on the borderline between internal and external semantic conceptualizations to feed into external-force modality. Thus, existence is arguably externally-driven, as it is taken at face value without reference to the speaker’s own emotions or intentions. Possession is somewhat ambiguous, as it entails a claim for possession on the part of the participant, but arguably also recognition of the possession relationship from an external source. In terms of its own grammaticalization path, Romani possession tends to draw on existence, supporting this ambiguity, at the very least, and quite possibly further biasing its external orientation. While the (marginal) grammaticalization of ‘to dare’ for permission is an exception to this tendency, the split between internal and grammaticalization paths receives support from the fact that possibility, occupying an ambiguous position, shows mixed source deriving both from expressions of internal and external states-of-affairs. The two principal dimensions depicted in Figure 1 also serve as a useful map to interpret the role of structural borrowing in the Romani modality system. Borrowing is most frequently encountered when there is close correlation of external-force modality with intensity (in this case, intensive outside pressure), as discussed in 3.3. On the entire chart, volition clearly stands out as the category with the greatest potential for further grammaticalization. Its grammatical derivations may cross the agentivity line to serve as impersonal, externalforce modals, which in turn may serve as sources for the grammaticalization of necessity expressions. Volition expressions may also end up in functions beyond strict modality, such as the epistemic future or the marking of indefiniteness. Notes 1. Information on the history of Romani and dialect differentiation within the language, accompanied by maps, examples, and sound samples, can be found on the website of the Manchester Romani Project: http://www.llc.manchester.ac.uk/Research/Projects/romani/. See also Matras (2002), Elšík and Matras (2006).

Modals in Romani

317

2. See also Elšík and Matras (2006), Matras (2002), and the Romani MorphoSyntax (RMS) database on the Manchester Romani Project website. The sample contains over 350 recordings and questionnaire elicitation of different Romani dialects from across Europe, collected between 2000-2007. We acknowledge support for our work during this period from the Arts and Humanities Research Council, the Economic and Social Research Council, and the Open Society Institute. 3. Dialects differ in what the default form of the nonfinite subjunctive is: it may be, like in Hameln Sinti, the third-person singular, but also the third-person plural or the second-person singular. Some varieties of Ukrainian Romani have reduced the second-person singular nonfinite subjunctive into a form that is now distinct from any subjunctive form, e.g. 2SG subjunctive t’ir-ex ‘[that] you do’ vs infinitive t’ir-e ‘to do’. 4. Apart from introducing non-factual complements, this polysemous connector may also introduce optative predications, conditional clauses, purpose clauses and further types of adverbial subordinations. 5. As it is not restricted to modal constructions, the obligatory lack of a complementizer in Dolenjski Romani nonfinite constructions is disregarded in Table 1. 6. Inflectional auxiliaries are unattested in Romani impersonal–nonfinite modal constructions. This does not appear to be due to any structural reason, however. 7. As Boretzky (1995a: 5) points out correctly, the OIA copula form asti is usually considered to be the source of the Romani third person present copula form isi. Nonetheless, a split development might also be considered, accounting for parallel forms. 8. The current dialect distribution of ašti suggests inheritance from Early Romani rather than independent dialect-specific innovations. While šaj has been retained in a geographically contiguous area (stretching from Czechia to northwestern Bulgaria plus recent out-migrant dialects), reflexes of ašti are attested within most dialect groups of Romani. 9. The grammaticalization of motion verbs meaning ‘to arrive at, to reach’ into ability modals is well attested (cf. Heine and Kuteva 2002: 45–46). 10. As Sampson’s (1926, II: 119) lexicographic definition goes: ‘to be in the act of, to be occupied or engaged in, to be the fact that’. 11. The modal humte has also diffused from Sinti into Bohemian (North Central) Romani. It appears to have been present also in some Para-Romani varieties (cf. Boretzky and Igla 2004: 184–185). 12. Boretzky and Igla (2004: 185) find the generalization of this person–number category suprising. It should be noted, however, that the first person singular is, cross-linguistically, the second most frequent person–number category in verbal paradigms (Bybee 1985).

318

Viktor Elšík, and Yaron Matras

13. See Section 2.5, ex. (53) for a more complex example from Burgenland Roman. 14. Several of these necessity modals are themselves borrowings into the languages that serve as the immediate source for Romani: Turkish lâzÕm from Arabic, Romanian trebui from Slavic, West Slavic musieĢ musieü, Hungarian muszáj and Romanian musai from German (the latter two cases from muss sein ‘must be’). 15. However, trebui has personal inflections in some non-standard varieties of Romanian, and so some instances of the loanword’s personal inflection in Romani may in fact simply retain this property. 16. Note that the Turkish modal lâzÕm must be morphologically adapted by the nominal suffix -i in those dialects that allow its analytic inflection through the inflectional auxiliary, including in Vălþi Dol Romani. The loan remains unadapted in those dialects that do not allow its analytic inflection, i.e. in those where TAM categories are marked solely on the modalized verb (Type nPF2). 17. The example (47d) has been elicited but might not occur in natural discourse. 18. In Taikon Kalderaš of Sweden inflected našti- ‘not to be capable’ requires an additional analytical negator, while uninflected našti ‘cannot’ does not allow one (cf. Section 2.3). Again, there is no hyperanalysis of negation in the former case. 19. Diffusion from another Italian Romani dialect is likely to be the source of šti in Italian Xoraxane, unless one wants to assume that the analogical decomposition of našti took place before the development of the innovative (South Vlax) indicative negator ni. 20. The source (Holzinger 1993) is not explicit about whether hunte can be regularly negated, or not. 21. It is unattested in a large collection of authentic Burgenland Roman narratives (Halwachs et al. 1999) and not discussed in the grammar of the dialect (Halwachs 1998). 22. In some Kalderaš varieties both possibility modals appear to have identical functions (e.g. in a Šanxajcy variety of Ukraine and in the Bunkuleš and Markuleš varieties of Banat and northern Serbia, cf. Boretzky 1994: 137). Unfortunately, the sources on other dialects that possess reflexes of both modals, Moravian Romani and Italian Xoraxane with šaj and šti, contain no detailed information on modal functions. 23. The evidence from a variety of Šanxajcy Kitajake is puzzling: in this Kalderaštype dialect dašti- only has the lexical meaning ‘to dare’, while all possibility functions are expressed by šaj. This either assumes de-grammaticalization of dašti- or questions the etymological connection between ašti and dašti-. 24. Note that the Kԥrԥmidarja future construction is not an instance of (directly) de-allative future (cf. Heine and Kuteva 2005: 103). Although ala is a form of the verb ‘to come’, this verb generally provides some TAM (viz. non-

Modals in Romani

319

indicative and future) forms of the copula and the verb of existence (‘to be, to become’) in this dialect.

References Boretzky, Norbert 1993 Bugurdži. Deskriptiver und historischer Abriss eines RomaniDialekts. (Balkanologische Veröffentlichungen, 21.) Wiesbaden: Harrasowitz. Boretzky, Norbert 1994 Romani: Grammatik des Kalderaš-Dialekts mit Texten und Glossar. Berlin: Harrassowitz Verlag. Boretzky, Norbert 1995a Die Entwicklung der Copula im Romani. Grazer Linguistische Studien 43, 1–50. Boretzky, Norbert 1995b Armenisches im Zigeunerischen (Romani und Lomavren). Indogermanische Forschungen 100, 137–155. Boretzky, Norbert 1996a Zu den Modalia in den Romani-Dialekten. Zeitschrift für Balkanologie 32, 1, 1–27. Boretzky, Norbert 1996b The “new” infinitive in Romani. Journal of the Gypsy Lore Society, V, 6, 1–51. Boretzky, Norbert 1999a Die Verwandschaftbeziehungen zwischen den Südbalkanischen Romani-Dialekten. Mit einem Kartenanhang. Frankfurt am Main: Peter Lang. Boretzky, Norbert 1999b Grammatical interference in Romani: Loan formations for foreign categories. Acta Linguistica Hungarica 46, 3–4, 169–200. Boretzky, Norbert, and Birgit Igla 1994 Wörterbuch Romani-Deutsch-Englisch für den südosteuropäischen Raum: mit einer Grammatik der Dialektvarianten. Wiesbaden: Harrassowitz. Boretzky, Norbert, and Birgit Igla 2004 Kommentierter Dialektatlas des Romani. Wiesbaden: Harrassowitz. Bybee, Joan L 1985 Morphology: A study of the relation between meaning and form. Amsterdam: John Benjamins.

320

Viktor Elšík, and Yaron Matras

Cech, Petra, Christiane Fennesz-Juhasz, and Mozes F. Heinschink (eds.) 1999 Lovarenge paramiþi taj textura anda Österreich – Texte Österreichischer Lovara. (= Arbeitsbericht 2 des Projekts Kodifizierung der Romanes-Variante der Österreichischen Lovara). Wien: Verein Romano Centro. Second, revised edition. Cech, Petra, and Mozes F. Heinschink 1999a Basisgrammatik. (= Arbeitsbericht 1a des Projekts Kodifizierung der Romanes-Variante der Österreichischen Lovara). Wien: Verein Romano Centro. Cech, Petra, and Mozes F. Heinschink 1999b Sepeþides-Romani: Grammatik, Texte und Glossar eines türkischen Romani-Dialekts. (Balkanologische Veröffentlichungen, 34.) Wiesbaden: Harrassowitz. Cech, Petra, and Mozes F. Heinschink 2001 Sinti Istriani, Liþki Šijaci, Gopti und Cigani Brajdiþi. Die Dolenjski Roma in Slovenien. In “Was ich noch sagen wollte …”: A multilingual Festschrift for Norbert Boretzky on occasion of his 65th birthday, Birgit Igla, and Thomas Stolz (eds.), 341–369. Berlin: Akademie Verlag. Croft, William 2000 Explaining language change: An evolutionary approach. Harlow: Longman. Elšík, Viktor, and Yaron Matras 2006 Markedness and language change: The Romani sample. Berlin: Mouton de Gruyter. Fennesz-Juhasz, Christiane, and Mozes F. Heinschink (eds.) 1999 Lovarenge paramiþi taj textura anda Österreich, kotor II – Texte Österreichischer Lovara II. (= Arbeitsbericht 4 des Projekts Kodifizierung der Romanes-Variante der Österreichischen Lovara). Wien: Verein Romano Centro. Franzese, Sergio 1985 Il dialetto dei Sinti Piemontesi. Note grammaticali. Glossario. Torino: centro studi Zingari di Torino. Gjerdman, Olof, and Erik Lundberg 1963 The language of the Swedish Coppersmith Gipsy Johan Dimitri Taikon: Grammar, texts, vocabulary and English word index. Uppsala: Lundequist. Halwachs, Dieter W 1998 Amaro vakeripe Roman hi — Unsere Sprache ist Roman. Texte, Glossar und Grammatik der burgenländischen Romani-variante. Klagenfurt: Drava. Halwachs, Dieter W. in co-operation with Gerd Ambrosch, Ursula Glaeser, Katharina Martens, and Michael Wogg

Modals in Romani 1998

321

Morphologie des Roman. (= Arbeitsbericht 3a des Projekts Kodifizierung und Didaktisierung des Roman). Oberwart: Verein Roma. Halwachs Dieter W., Gerd Ambrosch, Michael Wogg in co-operation with Ursula Glaeser, and Katharina Martens 1999 Märchen und Erzählungen der Burgenland-Roma. (= Arbeitsbericht 8a des Projekts Kodifizierung und Didaktisierung des Roman). Oberwart: Verein Roma. Hansen, Björn 2000 The German modal verb müssen and the Slavonic languages: The reconstruction of a success story. Scando-Slavica 46, 77–92. Hansen, Björn 2003 The Grammaticalization (and Degrammaticalization?) of Modals in Slavonic. In Funktionale Beschreibung slavischer Sprachen. Beiträge zum 13. Internationalen Slavistenkongress Ljubljana 2003, Tilman Berger, and Karl Gutschmidt (eds.), 97-115. München. Hansen, Björn In Press Modals. In Slavic languages: An international handbook of their history, their structure and their investigation, Peter Kosta, Tilman Berger, Karl Gutschmidt and Sebastian Kempgen (eds.). (Handbücher zur Sprachund Kommunikationswissenschaft) Berlin: Mouton de Gruyter. Heine, Bernd, and Tania Kuteva 2005 Language contact and grammatical change. Cambridge: Cambridge University Press. Holzinger, Daniel 1993 Das Romanes: Grammatik und Diskursanalyse der Sprache der Sinte. (Innsbrucker Beiträge zur Kulturwissenschaft, 85.) Innsbruck: Verlag des Instituts für Sprachwissenschaft der Universität Innsbruck. Lehmann, Christian 2002 Thoughts on grammaticaliszation. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Manchester Romani Project website: http://romani.humanities.manchester.ac.uk/ Matras, Yaron 1998 Utterance modifiers and universals of grammatical borrowing. Linguistics 36, 281-331. Matras, Yaron 2002 Romani: a linguistic introduction. Cambridge: Cambridge University Press.

322

Viktor Elšík, and Yaron Matras

Matras, Yaron 2004 Romacilikanes – The Romani dialect of Parakalamos. Romani Studies 5, 14:1, 59–109. Mészáros, György 1980 A Magyarországi szinto cigányok (történetük és nyelvük) [The Sinti in Hungary (their history and language)]. Budapest: Magyar nyelvtudományi társaság. Müller, Friedrich 1869 Beiträge zur Kenntniss der Rom-Sprache. Sitzungsberichte der philosophisch-historischen Classe der Kaiserlichen Akademie der Wissenschaften zu Wien 61, 1, 149–206. Palmer, Frank R 2001 Reprint. Mood and modality. Second edition. Cambridge: Cambridge University Press. Original edition, Cambridge: CUP, 1986. Paspati, Alexandre G. 1973 Reprint. Études sur les Tchinghianés ou Bohèmiens de l’empire ottoman. Osnabrück: Biblio Verlag. Original edition, Costantinople: Antoine Koroméla, 1870. Pott, August 1844–1845 Die Zigeuner in Europa und Asien. Ethnographisch-linguistische Untersuchung vornehmlich ihrer Herkunft und Sprache. Halle: Heynemann. RMS = Romani Morpho-Syntax Database. Viktor Elšík and Yaron Matras supported by Christopher White, Charlotte Jones, Christa Schubert, Hazel Gardner, Ruth Hill, Charlotte Jones, Daniele Viktor Leggio, Christa Schubert, Veronica Schulman, Anton Tenser, Eliška Vránová. On-line: http://romani.humanities.manchester.ac.uk/rms Romani Patrin December 1999, No. 4. Oberwart: Verein Roma. Sampson, John 1926 The dialect of Gypsies of Wales, being an older form of British Romani preserved in the speech of the clan of Abram Wood. Oxford: Clarendon Press. van der Auwera, Johann, and Plungian V. A. 1998 Modality’s semantic map. Linguistic Typology 1:2, 79–124.

C. Modals in Non-Indo-European languages

10. The grammaticalisation of modal auxiliaries in Maltese and Arabic vernaculars of the Mediterranean area Martine Vanhove, Catherine Miller and Dominique Caubet

1. Introduction The grammaticalisation processes of modal auxiliaries in Arabic vernaculars have rarely been a key issue of studies among specialists of these languages. Moreover, different theoretical frameworks and terminologies have been used, and the descriptions and analyses are far from being developed to the same extent. Such a situation does not ease the comparison between the different systems, but, as will be shown below, does not jeopardize it altogether. In order to fit in the general pattern of this book, this chapter will focus exclusively on the expression of epistemic and intersubjective (also called agent-oriented) modalities in a sample of four varieties of Arabic spoken along the Mediterranean coast, belonging to the two main branches, (i) Western (Maghribi) Arabic: Maltese, the official and national language of Malta, a Western Arabic vernacular in origin, and Moroccan Arabic; (ii) Eastern Arabic: Egyptian and Levantine (i.e. Jordanian, Syrian, Palestinian) Arabic. 1 For the first two languages we relied on our own fieldwork data (Vanhove 1993, Caubet 1993), and we made use of Mitchell and al-Hasan (1994), Lentin (1997), Woidich (2006), and Salame and Lentin (forthcoming), for the latter two. We will first describe the systems of the vernaculars (section 2, 3, and 4), 2 then compare the grammaticalisation patterns and grammaticalisation chains of the different modal values (section 5). For the description of Maltese and Moroccan Arabic, we used the framework of the enunciative theory, which can be translated, albeit somewhat roughly, into Bybee and Fleishman’s (1995) approach of modality. 3 Epistemic modality is defined from an enunciative viewpoint as “establishing a relation between the enunciator and the propositional content represented by the predicative relation”, 4 a definition in line with

326 Martine Vanhove, Catherine Miller and Dominique Caubet the traditional approach of logicians: epistemic elements are considered “clausal-scope indicators of a speaker’s commitment to the truth of a proposition” (Bybee and Fleishman 1995: 6). In terms of the framework of the enunciative theory, this means that the enunciator is not certain that the predicative relation is valid, that he does not choose between its validity (‘truth’ in the traditional approach) or non-validity, but simply gives a quantitative estimation of the chances of the predicative relation to be validated, hence these elements denote values of probability, and contingency. The second type of modality is called ‘intersubjective modality’, and concerns the relations between the subject and the predicate within the predicative relation. The truth of the propositional content is not at stake, but the validation of the predicative relation depends on the will of the subject of the utterance, or on the will, the pressure or the demand that the enunciator is exerting over the subject of the utterance. This includes the so-called deontic or ‘root’ modalities. In general, this corresponds to the ‘agent-oriented’ modality defined by Bybee and Fleischman (1995: 6) as follows: “all modal meanings that predicate conditions on an agent with regard to the completion of an action referred to by the main predicate, e.g. obligation, desire, ability, permission and root possibility.” Regarding Egyptian and Levantine Arabic, we transposed the existing descriptions into the above described framework in order to ease the comparisons between the various systems. It should be borne in mind that the enunciative theory distinguishes two other types of modality (assertive and appreciative), and explicitly claims that they may overlap. One must also be aware that the following study only represents part of the modal systems of the Arabic vernaculars under study, and that other devices are used (e.g., intonation, moods, tenses, verbs, adverbs, and adverbial constructions). Furthermore, the boundaries between aspectual and modal values, like in many other languages, are not clear-cut, and subtle modal and semantic values are also conveyed by tenses and the type of assertive modality. We do not deal with these questions in this paper. Following are the criterions we used to characterize the verbal auxiliary category. These criterions are valid for Arabic vernaculars, but also apply cross-linguistically: 5 – A verbal periphrastic construction made of an auxiliary verb and a main verb is a morphological, syntactic and semantic unit.

Modals in Maltese and Arabic vernaculars

327

– Within the utterance, the syntactic scope of the auxiliary is the predicative kernel, and not the utterance itself. – No coordinating or subordinating elements can occur between the two verbs. – Both verbs have the same subject. – The complement, if any, is that of the main verb. – Most often, a semantic abstractification (rather than bleaching) of the auxiliary occurs, in the sense that the meaning of the auxiliary verb is more “general” (its original meaning as a main verb does not need to be general) than that of the main verb, because it needs to combine with all (or at least most) semantic classes of verbs. – Paradigmatic substitution is only possible for the auxiliary. Verbal particles, whether of verbal origin or not, only differ from auxiliaries in their degree of grammaticalisation, in the sense that they are more grammaticalised than verbal auxiliaries. For readers unfamiliar with the verbal system of Arabic vernaculars, it is necessary to remind that, unlike what is the case in many European languages, the morphology of periphrastic constructions with an auxiliary is usually characterized by the juxtaposition of two finite verbal forms, i.e. both forms are fully inflected for TAM, gender, and number. The rare exceptions mostly concern the modal auxiliary forms themselves: they may be participial forms inflected only for gender and number, and if they are non-verbal in origin, they may have acquired a different type of inflection, based on suffix pronouns, or have no inflection at all. In the latter case they are more accurately analyzed as particles. 2. Maltese The verbal system of Maltese is particularly rich in auxiliaries and verbal particles which express almost as many aspectual, temporal and modal values as the forty-five forms that have been analyzed in Vanhove (1993). Most of them are the result of internal grammaticalisation processes, but in a few instances (see Vanhove 1994, 2000, and 2001) the Maltese auxiliary can clearly be traced back as a calque from the contact languages (Sicilian and Italian for a millennium, English for two centuries). 6 Six auxiliaries, which are also used as full verbs, are relevant for the study of modality as defined for the purpose of this book: seta’ ‘can’, jaf ‘know’, ried ‘want’, gƫandu (and its ‘irregular’ forms kellu and ikollu)

328 Martine Vanhove, Catherine Miller and Dominique Caubet ‘have’, mess ‘touch’, and ikun ‘be’. When used as modal auxiliaries, they may acquire new meanings as will be seen below. 2.1. Seta’ ‘can’ The Maltese modal verb seta’ ‘can’ is a cognate of the Arabic root ŸW¤ whose tenth derived form ista6CCbC also means ‘can’. 7 Like most modal auxiliaries and particles, seta’ is polysemous for both epistemic and intersubjective modal values. Still the latter are far more frequent than the former. The precise modal values depend on the TAM inflection of both verbal forms of the periphrasis, the semantics of the main verb, as well as on contextual factors. The degree of grammaticalisation of seta’ can be considered as minimal, as the auxiliary retains all its verbal and semantic characteristics, and can even be separated from the main verb by a subject, an object or an adverb. 2.1.1. Intersubjective modality The modalities of capacity (1), possibility (2), and permission (3) and their negative counterparts are the most frequent values of seta’ ‘can’: (1)

il-mara 8 tiegƫ=i ma tista=x ART-wife of-1SG NEG can.IPFV3F.SG-NEG gƫax marid=a because ill-F ‘My wife cannot cook because she is sick.’

(2)

kullƫadd jista’ jikkultiva l-gƫelieqi every_one can.IPFV3M.SG cultivate.IPFV3M.SG ART-fields tiegƫ=u of-3M.SG ‘Every one will be able to cultivate his own lands.’ il-ƫobĪ biex seta’ jinbieƫ… ART-bread PURP can.PFV3M.SG be.sold.IPFV3M.SG ‘The bread, so that it could be sold.’ [had to be stamped by an official stamp]

(3)

issajjar cook.IPFV3F.SG

Modals in Maltese and Arabic vernaculars

329

When both verbs of the periphrasis are in the perfective, the value is that of an unfulfilled possibility: (4)

dawn

kwaĪi setgƫu kienu almost can.PFV3PL be.PFV.3PL ‘These could almost have been sonnets.’ DEM.PL

sunetti sonnets

The construction can also take, in this case, a value of a posteriori advice, often with a connotation of reproach: (5)

stajt gdimt ilsien=ek can.IPFV2SG bite.IPFV2SG tongue-2SG ‘You could have hold your tongue!’

int ukoll 2.SG also

2.1.2. Epistemic modality Seta’ can also be used to express epistemic values, although rarely. It should be noted that this is only possible with the imperfective form for both the auxiliary and the main verb. This is regularly the case when the main verb is kien ‘be’, in the 3rd person masculine singular of the imperfective. The periphrasis means ‘it is probable that’, and is equivalent to Arabic forms such as yumkin/yªmkªn or mumkin ‘possible’ (this is lost in Maltese; see below section 3.5. and 4.5.): (6)

jista’ can.IPFV3M.SG ‘I might forget.’

jkun be.IPFV3M.SG

ma NEG

niftakar=x remember.IPFV1SG-NEG

With other verbs it may also be used to express an eventuality, a probability, the feasibility of an action: (7)

taƫt il-maskra tista’ tinheba r-rejaltà under ART-mask can.IPFV3F.SG be.hidden.IPFV3F.SG ART-reality ‘Under the mask, it happens that the reality is hidden.’

(8)

wara l-iskrivanja tiegƫ=i behind ART-desk of-1SG jidƫol jistrieƫ

wieƫed one sew

jista’ can.IPFV3M.SG mhux hekk?

330 Martine Vanhove, Catherine Miller and Dominique Caubet enter.IPFV3M.SG rest.IPFV3M.SG precisely NEG thus ‘Behind my desk, someone could come and rest, precisely, don’t you think?’ 2.2. Jaf ‘know’ Jaf is a defective verb in Maltese limited to the imperfective inflection. It is cognate with Arabic barafa ‘know’. 9 Aquilina (1987: 8), in his dictionary, mentions what he calls an ‘adverbial’ use of the verb jaf ‘know’ when used before another verb. The examples given are translated by ‘maybe, it is possible that’, showing that jaf loses its original meaning as a full verb in this particular construction. The examples provided by Aquilina all have epistemic values: (9)

jaf jagƫmel ix-xita llum know.IPFV3M.SG do.IPFV3M.SG ART-rain today ‘It is quite possible that it may rain today.’

As a matter of fact, the use of jaf as a modal auxiliary is very rare 10 in spontaneous speech and writing, and all occurred with the verb qal~igƫid ‘say’. Still, both epistemic (possibility) and intersubjective (capacity) values are possible: (10) Ma taf=x tgƫidi=l=na forsi xi NEG know.IPFV2SG-NEG say.IPFV2SG-to-1PL maybe some ƫaƥa… thing ‘Could you not tell us maybe something…’ (11) gƫandu Īewƥ skopijiet sa fejn naf have.PRS1SG two objectives till where know.IPFV1SG ngƫid say.IPFV1SG ‘He has two objectives as far as I can judge.’ The scarcity of the modal use of jaf combined with the native speakers’ judgement about Aquilina’s examples as ‘archaic’ is probably an indication that the auxiliary use of jaf is dying out.

Modals in Maltese and Arabic vernaculars

331

2.3. Ried ‘want’ The modality of volition is usually expressed by the verb ried ‘want’, a cognate of Arabic aaraada ‘want’. 11 It can either be used in a completive structure with the complementizer li ‘that’, introducing another predicative relation (the subjects are not co-referential), or in an asyndetic construction. In the latter, if the subjects of the two verbs are not co-referential, the subject of the second verb is the pronominal suffix object of the first one (see (13)), which rules out an interpretation of ried as an auxiliary in this case (see section 1 for the criterions of auxiliarihood): (12) iridu li int want.IPFV3PL that 2SG ‘They want you to eat.’

tiekol eat.IPFV2SG

(13) ma

rrid=ek=x tgƫid li jien dittur want.IPFV1SG-2SG-NEG say.IPFV2SG that 1SG vain ‘I don’t want you to say that I am vain.’ NEG

(14) ma

ried=x jinnegozja magƫ=hom want.PFV3M.SG-NEG bargain.IPFV3M.SG with-3PL ‘He didn’t want to bargain with them.’

NEG

More important for the purpose of this study, the verb ried ‘want’ is also used as a modal auxiliary in a periphrastic construction (formally similar to that in 14 above) for the expression of the intersubjective modalities of necessity, obligation, ineluctability, advice, and interdiction. If the subject is animate, only contextual factors can help decide if the original meaning of volition is retained or not (compare 14 and 18). But with inanimate subjects, no such a polysemy is possible, and only the intersubjective interpretations are possible. No epistemic use has been noted for the modal auxiliary ried. (15) il-garaxx ridna noborxu=h koll=u ART-garage want.PFV1PL brush.IPFV1PL-3M.SG all-3M.SG mil=l-qieƫ from-ART-foot ‘The garage, we had to brush it all from bottom up.’

332 Martine Vanhove, Catherine Miller and Dominique Caubet (16) biex

tikber trid grow_up.IPFV2SG want.IPFV2SG ‘In order to grow up, you must eat.’

PURP

tiekol eat.IPFV2SG

(17) trid tmur lejn ix-Xagƫra jew iĪ-ĩebbuƥ want.IPFV2SG go.IPFV2SG towards ART-Xagƫra or ART-ĩebbuƥ [there are things like that for sale at Xagƫra and ĩebbuƥ] ‘you should go to Xagƫra or ĩebbuƥ.’ (18) ma

rrid=x ninnegozja qalt=l=i NEG want.IPFV1SG-NEG bargain.IPFV1SG say.PFV3F.SG-to-1SG omm=i mother-1SG ‘I must not bargain, my mother told me.’

2.4. Gƫandu / kellu / ikollu ‘have’ The Maltese language has grammaticalised two different lexical sources and grammatical constructions into so-called pseudo-verbs meaning ‘have’. Gƫandu, as in many Arabic vernaculars, is the result of the agglutination of the preposition *band ‘at’, cognate with Classical Arabic binda, with the suffix object pronouns, which have become subject indices and the regular inflection of this pseudo-verb. It has the temporal value of present tense. Kellu and ikollu are both cognate with the verb kien ‘be’ in the perfective and imperfective forms respectively, to which the preposition lil ‘towards, to’, shortened to l, has been suffixed, and to which the suffix object pronouns are added as inflectional morphemes, also functioning as subject indices. They are past and future tense, respectively. 2.4.1. Intersubjective modality When used as a modal auxiliary, the three forms of this pseudo-verb mainly have intersubjective values of necessity, obligation, ineluctability, a posteriori advice, and interdiction. The difference with the auxiliary verb ried (see section 2.3. above) is mainly that of register, the use of gƫandu / kellu / ikollu being regarded as higher than that of ried. This feeling of Maltese native speakers might be linked to the long-standing contact with Italian

Modals in Maltese and Arabic vernaculars

333

and English: 12 the contact situation may be the reason behind the Maltese auxiliation construction, unknown or very marginal in other Arabic vernaculars (see section 4.8.). (19) tmur fejn gƫandek tmur go.IPFV2SG where have.PRS2SG go.IPFV2SG ‘You go where you have to go.’ (20) kellek taƫseb qabel have.PST2SG think.IPFV2SG before ‘You should have thought about it before.’ (21) biex

tƫit=ha ikollna nxarrbu=ha sew.IPFV2SG-3F.SG have.FUT1PL wet.IPFV1PL-3F.SG ‘In order to sew it, we’ll have to wet it.’ PURP

2.4.2. Epistemic modality These pseudo-verbs can have epistemic values, more frequently than seta’ ‘can’. These are values of logical probability (or inference), near-certainty, and calculation. They are more likely to occur with gƫandu than with the two other forms. (22) tgƫarraf=ni gƫandek tifhem x’ know.IPFV2SG-1SG have.PRS2SG understand.IPFV2SG what irried ngƫid bi=ha want.IPVF1SG say.IPFV1SG with-3F.SG ‘You know me, you should understand what I mean by that.’ (23) gƫandhom ikunu hawn gƫal s-sebgƫa have.PRS3PL be.IPFV3PL here for ART-seven ‘They should be here around seven’ (24) k ikollok ir-riƫ fuq sa erbgƫat ijiem ƫamsa if have.FUT2SG ART-wind on till four days five ikollha tinxef have.FUT3F.SG dry.IPFV3F.SG ‘If you have a North wind, it should dry within four or five days.’

334 Martine Vanhove, Catherine Miller and Dominique Caubet 2.5. Mess ‘touch’ The verb mess ‘touch’ (cognate with Arabic massa ‘touch’), a full verb regularly inflected for perfective and imperfective, can also be used as a modal auxiliary. When it functions as an auxiliary, the morphology is different from what it is as a full verb: to the 3rd person of either the perfective or the imperfective are added the object suffix pronouns which co-refer to the subject of the periphrasis. These bound pronouns constitute the inflection of the auxiliary. Although the verb itself is of Arabic origin, the auxiliary construction is obviously a calque of Italian. 13 The semantics and the grammaticalisation process can thus be considered as the result of language contact. The periphrastic construction with mess only has intersubjective interpretations. When both or either of the two verbs are in the perfective form, the modal value is that of an a posteriori advice or suggestion, always with a connotation of reproach or regret: (25) messek tkellimt qabel touch.PFV2SG talk.PFV2SG before ‘You should have talked before!’ When both verbs are in the imperfective, unachieved necessity is expressed, again with a connotation of reproach or regret: (26) imissek tistƫi touch.IPFV2SG be.ashamed.IPFV2SG ‘You should be ashamed of yourself!’ 2.6. Ikun ‘be’ When used as an auxiliary preceding a main verb in the imperfective, the imperfective form ikun of kien ‘be’ expresses modal values which are limited to the epistemic domain. The precise value is linked to the syntactic context. In an adverbial, completive or relative clause, ikun conveys a value of eventuality, probability:

Modals in Maltese and Arabic vernaculars

335

(27) taƫseb li l-iswed ikun think.IPFV2SG REL ART-black be.IPFV3M.SG jixraq=l=i suit.IPFV3M.SG-to-1SG ‘Do you think black would suit me?’ In independent or coordinated clauses, the epistemic value is that of a logical consequence, a logical probability (inference): (28) sib ƫmar ƛkejken u saqaj=k ikunu find.IMP.SG donkey small and feet-2SG be.IPFV3PL imissu ma’ l-art touch.IPFV3PL with ART-ground ‘Find a small donkey and your feet will touch the ground!’

2.7. Summary From the above study of Maltese modal auxiliaries, it can be concluded that intersubjective modalities are more commonly expressed with modal auxiliaries that epistemic ones. The different values of the periphrastic constructions and the forms of the auxiliaries are summarized in the following table (note that, apart from the forms of ‘have’, it does not provide information on the tenses of the two verbs of the periphrasis. The negative values are not listed either): Although polysemy is the rule for most modal auxiliaries, it should be noted that one of them is limited to epistemic values (ikun ‘be’). Conversely, two others are limited to intersubjective values (ried ‘want’ and mess ‘touch’). Several auxiliaries have modal values that are similar, but their semantic functions do not always fully overlap. They may be specialized in a restricted number of modal values within each type of modality. Within the intersubjective modality, gƫandu, kellu and ikollu are not used with an advice 14 value, only ried and mess can be. The latter, although sharing the value of advice and necessity with its four synonyms, is actually specialized for the expression of a posteriori advice, and for unachieved necessity. On the other hand, in the domain of necessity, only the pseudo-

336 Martine Vanhove, Catherine Miller and Dominique Caubet verbs meaning ‘have’, gƫandu, kellu and ikollu, can also occur with epistemic values Table 1. Values of Maltese modal auxiliaries

INTERSUBJECTIVE

EPISTEMIC

Modal value

Auxiliary

Possibility Capacity Permission Obligation Necessity Ineluctability Advice a posteriori advice unachieved necessity Eventuality Feasibility Probability logical probability (inference) Calculation

seta’ seta’, jaf (rare) seta’ ried, gƫandu, kellu, ikollu ried, gƫandu, kellu, ikollu ried, gƫandu, ikollu (rare) ried mess, kellu (rare) mess seta’, jaf (rare) seta’ seta’, ikun gƫandu, ikollu, ikun kellu

3. Moroccan Arabic Moroccan Arabic is one of the most innovative vernaculars, the present koinè being very much influenced by what is referred to as “prehilali dialects”, pertaining to the first centuries of arabicisation. One of the consequences of this historical situation is that innovation, and thus grammaticalisation, are pervasive in the morphosyntax of this language. In the verbal system itself, the use of a preverb has led to the development of two imperfective forms: ka-yªktªb and yªktªb (the oldest form). The preverb ka- is most probably a truncated form of the verb kan (‘be’). The bare form (i.e. without the preverb) is reduced to either syntactic dependency or modal values, whereas the new preverbal form marks the “real” imperfective. This includes the notions of habitual, repetition, general truths or progressive (see Caubet 1993, and 1994). ka-yªktªb means ‘he reads’ or ‘he is reading’, according to the context, whereas yªktªb can only mean ‘let him read, he’ll read, he’d read’ (hortative, vague future, eventual, etc.), or be used in concatenatives like:

Modals in Maltese and Arabic vernaculars

(29) baaȖi want.ACT.PTCP.M.SG ‘He wants to write.’

337

yªktªb write.IPFV3M.SG

As can be seen in the above example with baaȖi, the active participle has also developed into a pseudo-verbal conjugation, adopting the syntactic behaviour of a verb whereas it remains a nominal form, morphologically, with nominal marks of gender and number. With a negation, the construction of objects is similar to that of verbs, but the form has no personal indices and can only take feminine and plural markers like all adjectives and participles, as in: m5ªbbn-a-š ªl-ƫwaayªž NEG wash.ACT.PTCP-F.SG-NEG ART-clothes ‘I (fem.)/she hasn’t washed the clothes.’

(30) ma

m5ªbbn-iin-š ªl-ƫwaayªž wash.ACT.PTCP-PL-NEG ART-clothes ‘We/you/they haven’t washed the clothes.’

(31) ma

NEG

As for modals, innovation and grammaticalisation are also one of the characteristics of Moroccan Arabic. There are various degrees of grammaticalisation, from conjugated auxiliaries to frozen particles or adverbial forms. Four verbs will be analyzed here. They are currently still used as verbs with their full semantic values: qdª4 ‘can, may’, qªdd ‘can’, kaan ‘be’, Zª55 ‘must’. We’ll also consider the case of yªmkªn ‘may, maybe’, a verb in origin, but which is mostly used in its frozen form as an adverbial. 3.1. qdª4 ‘can, may’ The original meaning of the verb qdª4 is ‘to have the capacity, the strength, the power, the faculty, the possibility to’; the nominal qod4a (derived from the same root QDR) is linked semantically to divine power, predestination, fate or chance. 15 It is used in Moroccan Arabic to express both intersubjective and epistemic modalities. The differentiation between the two main types of modality is reflected in the morphology of the auxiliary verb.

338 Martine Vanhove, Catherine Miller and Dominique Caubet 3.1.1. Intersubjective modality When expressing intersubjective values, qdª4 is inflected as a regular verb in all its forms, perfective, imperfective or active participle. It either marks general capacity or permission: (32) ka-nªqdª4 nhªzz ª66ªbla TAM-can.IPFV1SG lift.IPFV1SG ART-table ‘I can lift the table.’ (33) - waš tªqdª4 tži? Q can.IPFV2SG come.IPFV2SG - nªqdª4! can.IPFV1SG ‘- Will you be able to come?’ ‘- Yes, I will!’ (34) tªqdª4 tži mbCC-ya? can.IPFV2SG come.IPFV2SG with-1SG ‘Can/could you come with me?’ (35) tªqdª4 tªmši daaba can.IPFV2SG go.IPFV2SG now ‘You can go now!’ 3.1.2. Epistemic modality As an epistemic auxiliary, qdª4 is invariable and frozen in the imperfective 3rd person singular form, yªqdª4. It expresses values of possibility, probability or plausibility: (36) yªqdª4 yži ƫmªd can.IPFV3M.SG come.IPFV3.MSG Ahmed ‘Ahmed may come.’

Modals in Maltese and Arabic vernaculars

339

(37) yªqdª4 ykuun PCCbªU can.IPFV3M.SG be.IPFV3M.SG sleep.ACT.PTCP.M.SG ‘He may be asleep.’ 3.2. qªdd ‘can’ The verb qªdd has very similar intersubjective values as qdª4. Etymologically, it means ‘to be enough or to suffice, to be of sufficient size, to be able (physically or morally)’: (38) ka-nqªdd nhªzz haad TAM-can.IPFV1SG lift.IPFV1SG this ‘I can lift this table.’

ª66ªDNC ART-table

(39) - PbCCYPªM? help.IPFV1SG-2SG - la, nqªdd nhªzz-ha NEG can.IPFV2SG lift.IPFV1SG-3F.SG ‘- Shall I help you?’ ‘- No, I can lift it!’ (I’ll manage) (40) qªddiiti tži Ȗªdda? tomorrow can.PFV2F.SG come.IPFV2F.SG ‘Can/could you come tomorrow?’ 3.3. ykuun ‘be’ As a modal auxiliary, kaan ‘to be’ is only used in the imperfective form ykuun, and marks logical probability, i.e. inference, an epistemic modality derived from the basic ‘vague future’ value the bare form of the imperfective conveys (see above, section 3): (41) ykuun mša l-ªl-barubiya be.IPFV3M.SG go.PFV3M.SG to-ART-country ‘He’ll be gone to the country!’

340 Martine Vanhove, Catherine Miller and Dominique Caubet (42) ykuunu baad ka-y6ªyybu ªl-Ȗda be.IPFV3M.PL still TAM-cook.IPFV3M.PL ART-lunch ‘They must still be preparing lunch!’ 3.4. xª55 ‘must’ The original meaning of the verb xª55 is ‘to be missing or lacking, to be insufficient, to be absent, nonexistent’. When used as a modal auxiliary, it can express both epistemic and intersubjective modalities. 3.4.1. Intersubjective modality The morphology of xª55 as a modal auxiliary is different from its full verb morphology. Two forms are used, and intersubjective values vary according to aspect. xª55 can be used as an impersonal frozen form in the 3rd person masculine singular of the ka- imperfective, and suffixed with personal pronouns co-referencial with the subject. It then conveys the idea of an obligation imposed from the outside, by the circumstances or by somebody, i.e. not self-imposed (cp. English ‘you have to’, as opposed to ‘you must’ 16 ): (43) ka-yxª55-ha taakul ªl-formaaž TAM-lack.IPFV3M.SG-3F.SG eat.IPFV3F.SG ART-cheese ‘She has to eat cheese!’ (it is good for her health) (44) ka-yxª55-ªk tšuufi daak-ªš-ši TAM-lack.IPFV3M.SG-PR2SG see.IPFV2F.SG that-ART-thing dyaal ªl-Mªžduub of El-Mejdoub ‘You have to see what El Mejdoub said!’ (you can’t escape it) xª55 can also be used in a frozen form of the perfective, the 3rd M.SG to which suffix pronouns are added as inflectional markers. It then means ‘to be under the obligation’, whether one likes it or not:

Modals in Maltese and Arabic vernaculars

341

(45) xª55-u yªmši go.IPFV1SG lack.PFV3M.SG-3M.SG ‘He has to go!’ (even if he does not feel like it) (46) kaan xª55-ni nªktªb-l-ha be.PFV3M.SG lack.PFV3M.SG-1SG go.IPFV1SG-to-3F.SG ‘I should have written to her.’ (but I didn’t) 3.4.2. Epistemic modality As an epistemic auxiliary, xª55 expresses near-certainty. Again, its morphology is different from what it is when a full verb. For the expression of the value of ‘to be under the obligation’ described above, it is frozen in the 3rd person masculine singular of the perfective, suffixed with personal pronouns, but in addition, it is followed with the imperfective form of ‘be’, ykuun. It thus forms a pseudo-conjugation, xª55-u ykuun ‘he/it must be’: (47) xª55-u lack.PFV3M.SG-3M.SG ‘He must be gone!’

ykuun be.IPFV3M.SG

mša go.PFV3M.SG

3.5. yªmkªn ‘may, maybe’ The original meaning of the verb mkªn is ‘to be possible or able’. Used as a modal, it can mean ‘to be possible, probable, likely’. It is nearly always used in a frozen form, that of the 3rd person masculine of the imperfective yªmkªn, and has even become an adverb meaning ‘maybe, could be!’. 3.5.1. Intersubjective modality For the expression of permission with second person subjects, the invariable verbal particle yªmkªn is used in a periphrastic construction before the main verb, which is in a relation of syntactic dependency (concatenation) with the particle, and thus occurs in the bare form of the imperfective (see Caubet 1994):

342 Martine Vanhove, Catherine Miller and Dominique Caubet (48) yªmkªn tži can.IPFV3.M.SG come.IPFV2SG ‘You may come.’ But to mark capacity, the set form of the frozen verb yªmkªn is constructed with suffix pronouns referring to the subject, and introduced by a preposition l ‘to’: yªmkªn-l-ek, yªmkªn-l-u, etc. (litt. ‘it is possible to you/ him’, ‘you can, he can’...): (49) yªmkªn-l-ek can.IPFV3.M.SG-to-2SG ‘Can you come ?’

tži? come.IPFV2SG

3.5.2. Epistemic modality The syntactic status of the form yªmkªn as an epistemic modal is hybrid: still verbal, but mostly adverbial. Within the utterance, its syntactic position can be that of an adverb, i.e. after the main verb or at the end of the utterance (see (50) and (51)), or that of an auxiliary verb, i.e. before the main verb (see (52) and (53)), in which case it can be analyzed as a verbal particle, a degree higher on the scale of grammaticalisation than auxiliaries. When placed in this position, it always requires syntactic dependency: (50) Ȗaadi yži, yªmkªn FUT come.IPFV3M.SG maybe ‘He’ll come, maybe.’ or: (51) yži yªmkªn come.IPFV3M.SG maybe ‘He’ll come, maybe.’ (52) yªmkªn nži can.IPFV3M.SG come.IPFV1SG ‘I may come.’

Modals in Maltese and Arabic vernaculars

343

(53) yªmkªn ykuun naabªU can.IPFV3M.SG be.IPFV3M.SG sleep.ACT.PTCP.M.SG ‘He may be asleep.’ 3.6. Summary In Moroccan Arabic, even though there are more adverbs implicated in the expression of modality, some modal auxiliaries are also used. The use of the verb xª55 ‘to be missing or lacking’ to express obligation is very characteristic of this vernacular. Most auxiliaries, be they verbal or pseudoverbal, and particles, are used for both types of modality, epistemic and intersubjective. The sole exception is ykuun ‘be’ which can only express epistemic modality. There are some morphosyntactic constraints for most of them (they may be used with only one type of inflection, or in an invariable form). One case of grammaticalisation to an adverb has been recorded, that of yªmkªn ‘maybe’. The different values of the periphrastic constructions and the forms of the auxiliaries are summed up in the following table: Table 2. Values of Moroccan modal auxiliaries

INTERSUBJECTIVE

EPISTEMIC

Modal value

Auxiliary

Possibility Capacity Permission Obligation Necessity Ineluctability Advice a posteriori advice unachieved necessity Eventuality Feasibility Probability logical probability Calculation

qdª4, qªdd, yªmkªn-l-ek qdª4, qªdd, yªmkªn-l-ek qdª4 xª55 xª55 xª55 xª55 xª55 yªmkªn qdª4 qdª4 xª55-u, ykuun xª55-u, ykuun

344 Martine Vanhove, Catherine Miller and Dominique Caubet 4. Egyptian and Levantine Arabic Very similar to Moroccan Arabic, the verbal systems of both Egyptian Arabic (EA) and Levantine Arabic (LA, i.e. Syrian, Jordanian and Palestinian) have split the imperfective aspect into two forms. One is the continuation of the old form, the so-called bare-form (yiktib), and the other is an innovative one, namely a b- preverb (b-yiktib), with values similar to those mentioned for Moroccan. The verb class contains numerous verbs expressing epistemic and intersubjective modal values, a few functioning also as auxiliaries. It may be noted that the most frequent ones are either pseudo-verb forms (i.e. based on a nominal construction but functioning as verbs), participles, and frozen forms in the 3rd SG.M imperfective. The auxiliary class cuts across word boundaries, and auxiliaries occur in the form of participles, nouns, adjectives, finite verbs and prepositional phrases. EA and LA share a number of modal auxiliaries but not all, as will be seen, and differ with regards to the TAM markers of the main verb. The selected forms for this paper are aidir ‘can’, birif ‘know’, bªdd- (bidd-) ‘wish’, bawiz (bayiz) ‘want’, laazim ‘must’, yimkin/mumkin ‘possible’, and ykuun ‘be’. It ends with a mention of the grammaticalisation of a noun (*aaa ‘truth’) and a prepositional phrase (bale ‘on’) as modal auxiliaries. 4.1. aidir-yiadar (~yiadªr), aaadir ‘can’ EA and LA modal verb aidir is a cognate of the Arabic root QDR ‘be able to’. aidir carries intersubjective modal values. The precise modal values depend on the TAM inflection of both verbal forms of the periphrasis, as well as on the semantics of the main verb. aidir is always inflected for gender and number. Although it can be used in the perfective, imperfective, and active participle forms when expressing capacity and possibility, it is only used in the imperfective for permission. 4.1.1. Intersubjective aidir expresses the deontic notions of (i) capacity, ability (ii) possibility, potentiality, and (iii) permission, and their negative counterparts. It is frequently associated with physical and intellectual ability:

Modals in Maltese and Arabic vernaculars

345

(54) miš aadr-a tauum mi-s-siriir NEG can.ACT.PTCP-F stand.IPFV3F.SG from-ART-bed ‘She cannot stand up from the bed.’ (EA, WOI:319) 17 (55) ma-*addi-š aidir yiaul-l-u *kaaya NEG-person-NEG can.PFV3M.SG say.IPFV3M.SG-to-3M.SG story ‘Nobody could tell him a story.’ (EA, WOI:319) (56) niadĮr nitaaabil ba-l-Ȗada bukrĮ can.IPFV1PL meet.IPFV1PL over-ART-lunch tomorrow ‘Could we meet over lunch tomorrow?’ (M&H:48) (57) iza ma get-š is-saaba sabba tKaFC4 if NEG come-PFV1SG-NEG ART-time seven can.IPFV2M.SG  ti4CYYC* go.IPFV2M.SG ‘If I don’t arrive at 7 o’clock you can leave.’ (EA, WOI:317) Already in Levantine Arabic texts of the 16th-18th century (Middle LA), qadar 18 was the most frequent ‘can’ auxiliary and was used in both the perfective and imperfective (bare and b- imperfective) forms to express capacity and ability. Note however that in (58) and (59) the imperfective form of qadar (i.e. naqdir) appears in combination with either another modal (yumkin in (59)) or (as in (58)) with the perfective form of qadar. This could be an indication that its modal value still needed to be emphasised through repetition: (58) maa qadarnaa naqdir narqud tilka al-layla NEG can.PFV1PL can.IPFV1PL sleep.IPFV1PL that ART-night ‘We could not sleep that night.’ (LA, LEN:629) (59) wa

yumkin naqdir be.possible.IPFV3M.SG can.IPFV1SG na5if-hu bal *aqiiqat-uh describe.IPFV1SG-3M.SG over fact-3M.SG ‘I can’t describe (in details) all these events.’ (LA, LEN:629) COORD



laa

NEG

346 Martine Vanhove, Catherine Miller and Dominique Caubet (60) maa b-aqdar b-atkallam NEG TAM-can.IPFV1SG TAM-speak.IPFV1SG ‘I can’t speak (Arabic).’ (LA, LEN:629) 4.1.2. Epistemic yiadar alone does not express epistemic values. It needs to be combined with either the future particle *a- or the adverb mumkin ‘maybe’, thus expressing eventuality: (61) *a-adár ašuuf-ak taani márrĮ FUT-can.IPFV1SG see.IPFV1SG-2MSG second time ‘Shall I be able to see you again?’ (M&H:49) 4.2. birif, yib4af, baarif ‘know, can’ The verb birif can function as a full-fledged verb keeping its original meaning of ‘know’. It can also function as an auxiliary and convey the intersubjective values of ability and capacity. It overlaps with aidir to express physical and intellectual ability. Like aidir, it can be used in the perfective, bare imperfective and b- imperfective forms as well as in the active participial form: brifti-š tit5arraf know.PFV2F.SG-NEG behave.IPFV2F.SG ‘You were not able to behave (properly).’ (EA, WOI:320)

(62) ma

NEG

(63) b-ab4af aa4a w aktib TAM-know.IPFV1SG read.IPFV1SG COORD write.IPFV1SG ‘I can read and write.’ (EA, WOI:320) The same modal value is recorded in Middle LA, mainly in negative sentences:

Modals in Maltese and Arabic vernaculars

347

(64) maktuub maa barafuu NEG know.PFV3M.PL write.PASS.PTCT.M.SG idabbiruu-h prepare.IPFV3M.PL-3M.SG ‘They were unable to write an answer.’ (LA, LEN:631) 4.3. bªdd- (bidd-) ‘want’ bªdd- (bidd-) is a frequent Levantine form, but it is rarer in EA. It is based on the prepositional phrase b-wudd, lit. ‘with-love’ and is attested in Middle LA since the 18th c. The suffix object pronouns added to this nominal base co-refer with the subject of the periphrasis and have become the regular inflection of this pseudo-verb. Syntactically, bΩdd- can function as a full-fledged verb: bªdd-u min-hu X aalf kiis COORD want-3M.SG from.3M.SG X thousand bag ‘X wants from him 500,000 piasters.’ (LA, LEN:624)

(65) wa

(66) bidd-hum is-sayyaara VKV5CNNC* repair.IPFV3F.SG want-3PL ART-car ‘They want the car to be repaired.’ (LA, M&H:39) 4.3.1. Intersubjective Preceding a bare imperfective form, bªdd- functions as a modal intersubjective auxiliary expressing (i) volition, (ii) ineluctability, necessity, and obligation imposed from the outside: 19 (67) bªdd-i aª*ki mab-ak kªlªmt-een want-1SG say.IPFV1SG with-2SG word-DU ‘I want to tell you a few words.’ (LA, LEN:624)

348 Martine Vanhove, Catherine Miller and Dominique Caubet (68) maa kunt bªdd-i aªnzel NEG be.PFV1SG want-1SG go.IPFV1SG ‘I did not have the intention to go there.’ (LA, S&L) (69) al-yoom bªdd-ak tsaafer ART-day want-2M.SG travel.IPFV.2M.SG ‘You have to travel today.’ [because the Pacha is coming tomorrow] (LA, LEN:626) (70) bidd-u(h) yikuun yib4af ingliizi want-3M.SG be.IPFV3M.SG know.IPFV3M.SG English ‘He has to know English.’ (i.e. as a condition of his employment) (M&L:40) Note that in the above example, the value of necessity is reinforced by the combination of two auxiliaries (bidd- and yikuun). bidd often functions with an inanimate agent or impersonal subject: (71) ban6aloon-i bidd-u(h) kawy  trouser-1SG want-3M.SG pressing ‘My trouser needs pressing.’ (LA, M&H:40) 4.3.2. Epistemic The periphrastic construction bªdd- + yikun + b-imperfective form expresses a hypothesis given as a logical probability or inference (cp. the intersubjective value in (71) above): (72) ha-l-*ajar bªdd-o ykuun zmªrrod DEM-ART-stone want-3M.SG be.IPFV3M.SG emerald ‘This precious stone might be an emerald.’ (LA, S&L) (73) bidd-uh ykuun b-i5ªlli want-3M.SG be.IPFV3M.SG TAM-pray.IPFV3M.SG ‘He must be praying.’ (inferential) (LA, M&H:40)

Modals in Maltese and Arabic vernaculars

349

The above sentence is to be distinguished from bidd-uh y5ªlli ‘he wants to pray’ or ‘he has to pray’ which has an intersubjective value of volition or necessity, according to the context. (74) bidd-uh ykuun D[Kb4ĮH ingliizi want-3M.SG. be.IPFV3M.SG TAM-know.IPFV3M.SG English ‘He must know English.’ (since he lived in England; as opposed to 70 above) (LA, M&H:40) 4.4. baawiz~baayiz ‘want’ The EA participial form baawiz~baayiz of the verb baaz~yibuuz ‘want, need’ functions as a full lexical verb as well as an auxiliary expressing the intersubjective values of (i) volition, intention, and (ii) need, obligation: (75) fi wa*da sitti bayz-a taaabil siyadt-ak there.is one woman want-F.SG meet.IPFV3F.SG Sir-2SG ‘There is a woman who wants to meet you.’ (EA, WOI:317) (76) il-barabiyya bawz-a titȖisil ART-car want-F.SG wash.IPFV3F.SG ‘The car needs to be washed.’ (EA, WOI:316) This use is similar to that of bidd- in section 4.3. Note that in both cases, the subject is an inanimate agent, and the predicate is either a verbal noun like kawy ‘pressing’ (71), or a verbal form with a passive meaning (cp. also Maltese ried, section 2.3.). 20 4.5. laazim ‘must’ laazim is originally the participial form of the Arabic root LZM, whose bare verbal form means ‘be necessary’. In both EA and LA it has become invariable, and can be considered as being grammaticalised as a modal particle, and, when occurring at the very beginning of an utterance, as an adverb.

350 Martine Vanhove, Catherine Miller and Dominique Caubet 4.5.1 Intersubjective laazim is the most common deontic modal of necessity and obligation in both EA and LA: 21 (77) laazim timši dilwaati must go.IPFV2SG now ‘You have to go now.’ (M&H: 51) 4.5.2 Epistemic laazim can convey an epistemic value of (i) logical probability (inference), or (ii) near-certainty, when combined with a perfective verb or an imperfective verb prefixed with the TAM markers *a- (in EA), b- (in LA), or with the auxiliary ykuun ‘be’: (78) inta laazim širibta-ha 2M.SG must drink.PFV2F.SG-3F.SG ‘You have certainly drunk it.’ (M&H: 43) (79) laazim *a-(/bi-)yiigi dilwaati TAM-(/TAM-)come.IPFV3M.SG now must ‘He must come soon.’ (I feel sure, I judge that) (M&H:43) The above utterance can be contrasted to laazim yiigi dilwaati, expressing the intersubjective value of ‘he must (he has to) come soon’ (i.e. ‘is under an obligation to’). (80) laazim yikuun wi5il maktab-uh must be.IPFV3M.SG arrived.PFV3SG office-PR3SG ‘He must (probably) have reached his office.’ (M&H:51) It may be noted that in LA bi-ykuun regularly occurs together with laazim (bi-ykuun wi5il) in epistemic inferential sentences to reinforce this value:

Modals in Maltese and Arabic vernaculars

351

(81) laazim *a-(/bi-)ykuun bi-yu6bux must TAM-(/TAM-)be.IPFV3M.SG TAM-cook.IPFV3M.SG ‘He will certainly be cooking.’ (M&H:51) 4.6. yimkin/mumkin ‘be possible’ yimkin is originally the 3rd M.SG imperfective form of the base verbal form of the Arabic root MKN ‘be possible’. mumkin is the passive participle form. Similar to the Moroccan form yªmkªn, both forms remain invariable. When used in a periphrastic construction they function as modal particles, but do not seem in these vernaculars to have become adverbs, as their position within the utterance does not vary. 22 Both are polysemous and express intersubjective values of possibility, capability, permission and an epistemic value of probability, the latter being more frequent that the former, the distinction between the two being a matter of context (see (84) and (85)): (82) amma ykuun is-sawwaaa miš mawguud fa if be.IPFV3M.SG ART-driver NEG find.PASS.PTCP and aana mumkin aruu* bidaal-u l-mugammab 1SG possible go.IPFV1SG instead-3M.SG ART-Mugamma‘ ‘If the driver is not present, then I can go instead of him in the Mugamma.’ (ability, potentiality) (EA, WOI:319) (83) mumkin bali izuur-ni bukra babd i&&uhr possible Ali visit.IPFV3M.SG-1SG tomorrow after ART-noon ‘Ali may well visit me tomorrow afternoon’ (probability) or ‘Ali is allowed to visit me tomorrow afternoon.’ (permission) (M&H:48) According to Mitchell and al-Hassan (1994:47), mumkin ‘very likely’ rates higher than yimkin ‘might be possible’ on the likehood-scale: (84) mumkin tišuuf-hum hinaak possible see.IPFV2M.SG-3PL there ‘You can see them there.’

352 Martine Vanhove, Catherine Miller and Dominique Caubet (85) yimkin tišuuf-hum hinaak possible see.IPFV2M.SG-3PL there ‘You may (perhaps) see them there.’ 4.7. yikuun ‘be’ yikuun is the imperfective form of the TAM auxiliary kaan~ykuun ‘be’. It has already been mentioned (section 4.3.2.) that yikuun added to another modal marker reinforces the epistemic value of probability, eventuality, doubt, etc. It can also precede a perfective, an imperfective (bare or b-form) and a participial form with a value of logical probability (inference): (86) walla akun-ši Ȗli6ti fi š-šaarib or be.IPFV1SG-Q mistake.PFV1SG in ART-street ‘Or did I mistake the street?’ (EA, WOI:306) (87) in ša a..aah il-aa*waal tikuun mašy-a ART-state be.IPFV3F.SG go.ACT.PTCP-F if wish God kwayyis maba-ak good with-2M.SG ‘I hope that everything will be OK for you.’ (EA, WOI:308) Numerous examples with subtle semantic modal values can be found in Woidich (2006). 4.8. Grammaticalisation of nominal and prepositional constructions In order to express the intersubjective values of necessity and obligation, in addition to laazim (described in section 4.5.) EA has grammaticalised certain nominal and prepositional constructions.

Modals in Maltese and Arabic vernaculars

353

4.8.1. *Caa ‘truth’ One of them is based on the noun *Caa ‘truth’, to which suffix pronouns are added. In the domain of modality, when preceding a finite verb, it functions as a modal auxiliary with an intersubjective value of advice: (88) *Caaik tiruu*i taȖayya4i hduum-ik truth-2F.SG go.IPFV2F.SG change.IPFV2F.SG clothe-2F.SG ‘You should go and change your clothes.’ (EA, WOI:316) 4.8.2. bale ‘on; have’ The preposition bale ‘on, over’ when inflected with suffix pronouns functions as a pseudo-verb ‘have’, albeit marginally in EA. 23 (89) ma-bale-huu-š zamb NEG-on-3M.SG-NEG responsible ‘He has no responsibility / he is not responsible.’ (EA, WOI:139) Similar to what has occurred in Maltese (see section 2.4.), but apparently to a much lesser extent, this construction can also be used as an intersubjective auxiliary when preceding a verb in the imperfective. Its modal value is that of advice: (90) bale-ek tiruu* on-2M.SG go.IPFV2M.SG ‘You should go.’ (M&H:52) 4.9. Summary From this brief survey, it can be concluded that, like in Maltese, intersubjective modalities are more commonly expressed by modal auxiliaries than epistemic ones. A number of modals are polysemous and express both intersubjective and epistemic values (laazim, mumkin). It should be noted that epistemic modalities are more often expressed by adverbial or nominal constructions, or by TAM particles like b- and *a- than by auxiliary verbs.

354 Martine Vanhove, Catherine Miller and Dominique Caubet In fact the only productive epistemic auxiliary is ykuun ‘be’. Apart from aidir ‘be able’, it is noteworthy that most modals are either pseudo-verbs of nominal origin such as bªdd-, *aaa-, or participial forms, or verbal forms frozen in the 3rd M.SG imperfective. The grammaticalisation of a nominal base + a suffix pronoun co-referential with the subject of the periphrasis is particularly developed in EA, but not all such constructions have been grammaticalised as auxiliaries. Table 3. Values of Egyptian and Levantine modal auxiliaries

INTERSUBJECTIVE

EPISTEMIC

Modal value

Auxiliary

possibility capacity permission obligation necessity ineluctability advice a posteriori advice unachieved necessity eventuality feasibility probability

aidir, mumkin aidir, birif, mumkin aidir, mumkin laazim laazim, bªdd-, baawiz bªdd-, laazim *aaa-, bale

logical probability calculation

bi-(*a-)yiadar mumkin, /yimkin yimkin, bªdd- + yikuun, laazim + *a-/bibªdd- + yikuun, laazim + yikuun

5. Conclusion: Grammaticalisation chains of modal auxiliaries Morphologically, four different source types for the grammaticalisation into modal auxiliaries have been identified for Maltese and the Arabic vernaculars in this survey: finite verbs, active participles, nouns, and prepositional phrases. Semantically, they belong to the following categories: modal verbs, be they finite or non finite (capacity, possibility, necessity, volition), verbs of cognition (‘know’), perception verbs (‘touch’), quantitative verbs (‘lack, be

Modals in Maltese and Arabic vernaculars

355

insufficient’, ‘suffice’), existentials (‘be’ > ‘have’), locative prepositions (‘at’, ‘on’ > ‘have’), and nouns (‘truth’, ‘love’). The languages discussed here have all grammaticalised two semantic domains as modal auxiliaries (or particles): a modal verb meaning ‘can’, and the existential verb meaning ‘be’. Three categories are shared by two or three languages, but not all four: ‘know’, ‘want’, and locative prepositions. This does not mean that the lexical items involved are all cognate forms. The other sources are language specific. One of them, the Maltese perception verb, can be directly attributed to language contact. It should be noted that the grammaticalisation of prepositions into modal auxiliaries involves an intermediary stage, that of a predicative function as the verb of possession ‘have’. Several morphological and functional changes have also been noted: (i) reduction of the verbal or participial paradigms up to invariability; (ii) whatever the morphological category of the source item, the use of suffix pronouns which are both possessive and object pronouns and have become, in the course of the grammaticalisation process, co-referential with the subject of the main verb; (iii) agglutination of prepositions to nouns or verbs. Below is a table summing up the sources and goals involved in the grammaticalisation of modal auxiliaries and particles in Maltese and the Arabic vernaculars. If we look at the whole system of intersubjective and epistemic modalities, whatever their grammatical category in the Arabic vernaculars and Maltese, it seems that there might exist some kind of correlation between the grammaticalisation strategies and the extension of the system of modal auxiliaries as a whole. The languages that tend to grammaticalise the TAM morphemes in general as verbal auxiliaries, as is the case for Maltese, have developed this strategy extensively for modals as well. Those which have already fully morphologized part of their TAM system (with preverbs which are either reduced forms of former verbs, participles, or prepositions), as are found in Moroccan, Egyptian and Levantine Arabic, tend to use other devices than auxiliaries more frequently, and/or to grammaticalise them further towards adverbs, even though modal auxiliaries do exist.

356 Martine Vanhove, Catherine Miller and Dominique Caubet Table 4. Grammaticalisation chains of modal auxiliaries and particles Source

Intermediate stage (if known)

Modal value as auxiliary or particle

Language

Forms

capacity verb

I.S. capacity possibility permission E. eventual probable feasible

Maltese Moroccan EA, LA Maltese Moroccan EA, LA

seta’ qdª4 (inflected) aidir seta’ qdª4 (frozen) TAM+ aidir

possibility verb

I.S. capacity permission

Moroccan

E. possible probable

Moroccan EA, LA

yªmkªn (frozen)/ yªmkªn +PREP+PR yimkin, mumkin (frozen), yªmkªn yimkin, mumkin

I.S. capacity

Maltese EA, LA Maltese

jaf birif jaf

Maltese

ried

EA, LA

baawiz

I.S. necessity obligation E. inference near-certain

EA, LA

laazim (frozen)

EA, LA

laazim (frozen)

I.S. capacity possibility permission

Moroccan

qªdd

EA, LA

cognitive verb (‘know’)

volition - verb - ACT.PART necessity (ACT.PART)

verb ‘suffice’

E. potential Possible volition auxiliary

I.S. necessity obligation

Modals in Maltese and Arabic vernaculars

verb ‘lack’ verb ‘touch’

calque

verb ‘be’

‘be’+prep+PR ‘have’

locative prep. ‘have’ ‘at’ ‘have’ ‘on’ nouns ‘truth’ prep+‘love’

volition verb + aux.

I.S. obligation Moroccan E.near-certain Moroccan

Zª55-PR Zª55 -PR

I.S. advice

Maltese

mess-PR

E. inference eventual probable inference

Maltese

ikun

Moroccan EA, LA

ykuun ikuun

I.S. necessity obligation E. inference

Maltese

kellu, ikollu

Maltese

kellu, ikollu

I.S. necessity obligation E. inference near-certain I.S. advice

Maltese

gƫand-PR

Maltese

gƫand-PR

EA, LA

bale-PR

I.S. advice I.S. necessity obligation E. inference

EA EA, LA

*CaaPR bªdd-PR

357

I.S. = Intersubjective modality; E. = Epistemic modality

It should also be mentioned that the imbalance between intersubjective and epistemic values in favour of the former at a quantitative and frequency level, could be the result of the fact that epistemic values are frequently taken in charge by the imperfective forms (bare or preverbed forms depending on the language) in all the languages concerned.

358 Martine Vanhove, Catherine Miller and Dominique Caubet Acknowledgments Special thanks are due to Jérôme Lentin for his help with Levantine Arabic and for having given us access to part of the manuscript of his co-authored dictionary with Claude Salame. Of course any error would be ours. Notes 1. Within these national boundaries, there also exists a great variety of sub-divisions, but for the purpose of this study we use the geographical terms as convenient cover terms. 2. We aimed at being exhaustive for each variety, which means that not all varieties described have the same modal auxiliaries. They vary in their semantic origin and in number. 3. Or the refined version of van der Auwera and Plungian (1998). 4. Our translation of Bouscaren and Chuquet’s definition (1987: 37) of epistemic modality: “la modalité épistémique établit une relation entre l’énonciateur et le contenu propositionnel représenté par la relation prédicative”. 5. The criteria are taken from David Cohen’s approach (still unpublished, but see Simeone-Senelle and Vanhove (1997, and 2003) for a sketch of the criteria). It differs from Heine’s approach (1993), but not fundamentally. 6. Today, tri-lingualism (Maltese, English, Italian) is the rule for most Maltese speakers, as well as code-switching with English (see Camilleri 1995). A millennium of linguistic contact with Sicilian and Italian led to a radical change in the language, as compared with Arabic vernaculars, at all linguistic levels: phonetic, phonological, morphological, syntactic, and lexical (see Brincat 2004). 7. This root has been reinterpreted in Maltese as *ST¤, and the verb seta’ is morphologically a base form. 8. The Maltese examples are given in the official orthography, which does not note vowel length, nor, apart from the article, the morphological boundaries, materialized by a dash in the glosses. In order to facilitate the reading of examples and glosses an = sign has been added to the Maltese orthography to materialize suffix and clitic boundaries. 9. The root consonants b and R have been lost. They surface again at the derived forms. 10. In the 50 hour-recording, the dozen novels, and the newspapers surveyed in Vanhove (1993), only three occurrences of jaf as a modal auxiliary were found. 11. There is also another volition verb: xtieq ‘desire, want’. 12. See footnote 5.

Modals in Maltese and Arabic vernaculars

359

13. Colloquial Italian (Venetian, Lombardian, Sicilian, Corsican…) as well as literary Italian (see e.g. Boch’s Italian-French, French-Italian dictionary under toccare) have grammaticalised toccare ‘touch’ and its cognates as modal auxiliaries expressing necessity, obligation. No such modal use of a verb meaning ‘touch’ is recorded in Arabic vernaculars. 14. In table 1, kellu is listed as rare in the line concerning a posteriori advice. As a matter of fact, there are only two written examples in the data, both taken from the same novel. They were considered to belong to a very high register by native speakers and were said not to occur in spoken discourse. 15. See Iraqi-Sinaceur (1993: 1539). 16. For a detailed analysis see Bouscaren and Chuquet (1987: 48 and 68). 17. WOI = Woidich, M&H = Mitchell & El-Hassan, LEN = Lentin, S&L= Salame and Lentin. WOI are all Egyptian Arabic, LEN and S&L are all Levantine Arabic, M&H do not always specify whether the examples are Egyptian or Levantine. 18. aKFKTof the contemporary vernaculars is the reflex of Middle Levantine Arabic qadar. 19. It may be noted, however, that the most frequent value of the auxiliary bªdd/bidd- is that of imminent future (see Lentin 1997, Salame and Lentin forthcoming, as well as Mitchell and El-Hassan 1994). 20. It should be noted that the grammaticalisation of nominal phrases into modal auxiliaries meaning ‘want’ or ‘intend’ is also recorded in other constructions in EA. All are agglutinations of a nominal base + a suffix object pronoun: aC5F‘intention’, ȖCTC& ‘desire’ and nifs- ‘self, mind’ (see Woidich 2006: 318). 21. Many other constructions express the deontic values of necessity and obligation such as the nominals &CTWWTK ‘necessary’, il-OCHTWW& ‘the duty’, or the negative compound form laa-budd ‘no way out’, etc. None of them has been grammaticalised as an auxiliary. 22. The subject of the periphrasis can be inserted between the particle and the main verb, but this is not a categorical criteria for adverbs, and the same can occur with auxiliaries. 23. More commonly used are the constructions with the prepositions mab ‘with’, band ‘at’, li ‘at’, wayya ‘with’, all inflected with suffix pronouns.

360 Martine Vanhove, Catherine Miller and Dominique Caubet References Aquilina, Joseph 1987 Maltese-English Dictionary. vol. 1. Malta: Midsea Books. Bouscaren, Janine, and Jean Chuquet 1987 Grammaire et Textes Anglais. Guide pour l’Analyse Linguistique. Gap, Paris: Ophrys. Boch, Raoul (con la collaborazione di Carla Salvioni) 2007 Dizionario Francese-Italiano, Italiano-Francese (quarta edizione). Bologna: Zanichelli. Brincat, Joseph M. 2004 Malta. Una storia linguistica. [Malta: a linguistic history] Genova: Le Mani. Bybee, Joan, and Suzanne Fleischman (eds.) 1995 Modality in Grammar and Discourse. Amsterdam: Benjamins. Camilleri, Antoinette 1995 Bilingualism in Education. The Maltese Experience. Heidelberg: Julius Groos Verlag. Caubet, Dominique 1993 L’Arabe Marocain. Tome II. Louvain : Peeters. Caubet, Dominique 1994 La conjugaison préfixale nue en arabe marocain: une forme en attente de repérage. Quaderni di Studi Arabi 11: 83-105. Heine, Bernd 1993 Auxiliaries. Cognitive Forces and Grammaticalisation. New York/ Oxford: Oxford University Press. Iraqi Sinaceur, Zakia (ed.) 1993 Le Dictionnaire COLIN d’Arabe Dialectal Marocain (8 volumes). Rabat: Editions Al Manahil. Lentin, Jérôme 1997 Recherche sur l’Histoire de la Langue Arabe au Proche Orient à l’Epoque Moderne. Lille: Service de publication des thèses. Mitchell, Terrence F., and Shahir El-Hasan 1994 Modality, Mood and Aspect in Spoken Arabic (with Special Reference to Egypt and the Levant). London/New York: Kegan Paul International. Salame, Claude, and Jérôme Lentin forth. Dictionnaire d’Arabe Dialectal Syrien (Parler de Damas). Damascus: Institut Français du Proche Orient (IFPO). Simeone-Senelle, Marie-Claude, and Martine Vanhove 1997 La formation et l’évolution d’auxiliaires et particules verbales dans des langues sémitiques: les langues sudarabiques modernes et le

Modals in Maltese and Arabic vernaculars

361

maltais. In Mémoires de la Société de Linguistique de Paris. Grammaticalisation et Reconstruction, 85-102. Paris: Klincksieck. Simeone-Senelle, Marie-Claude, and Martine Vanhove 2003 Le fonctionnement d’auxiliaires en afar. In Mélanges David Cohen. Etudes sur le Langage, les Langues, les Dialectes, les Littératures, Offertes par ses Élèves, ses Collègues, ses Amis. Présentés à l’Occasion de son Quatre-Vingtième Anniversaire, Jérome Lentin, and Antoine Lonnet (eds.), 615-34. Paris: Maisonneuve et Larose. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79-124. Vanhove, Martine 1993 La langue maltaise. Etudes syntaxiques d’un dialecte arabe “périphérique”. Wiesbaden, Otto Harrassowitz. Vanhove, Martine 1994 The Island of Malta, a Crossroad for Influences? The Case of Auxiliary Verbs. In Languages of the Mediterranean. Substrata. The Islands. Malta. Proceedings of the Conference held in Malta, 26-29 September 1991, Joseph M. Brincat (ed.), 286-292. Malta: University of Malta. Vanhove, Martine 2000 Le maltais et les interférences linguistiques. In Language in the Mediterranean Area. Typology and Convergence. Il Progetto MEDTYP: Studio dell'area linguistica mediterranea, Sonia Cristofaro, and Ignazio Putzu (eds.), 187-199. Milano: FrancoAngeli. Vanhove, Martine 2001 Contacts de langues et complexification des systèmes: le cas du maltais. Faits de Langues. Langues de diaspora. Langues en contact 18: 65-74. Woidich, Manfred 2006 Das Kairenisch-Arabische. Eine Grammatik. Wiesbaden: Otto Harrassowitz.

11. Modal verbs in Balto-Finnic Petar Kehayov and Reeli Torn-Leesik

1. Aims This article outlines an account of modal verbs in the Balto-Finnic languages, focussing on seven core modal verbs. The analysis of the modal verbs into different types of modality is based on that of van der Auwera and Plungian (1998). The central aim of the article is to describe the BaltoFinnic modal verbs in terms of the grammaticalisation parameters proposed in Lehmann (2002). The last section of the article looks at the effects of language contact on the modal verbs in Balto-Finnic languages. 2. Types of modality Definitions of modality vary across different approaches to the topic. In analysing Balto-Finnic modal verbs the present study employs the classification of modality types provided by van der Auwera and Plungian (1998), who define “modality” as applying to “semantic domains that involve possibility and necessity as paradigmatic variants, that is, as constituting a paradigm with two possible choices, possibility and necessity” (ibid.: 80). The semantic domains involve four types of modality: participant-internal modality, participant-external modality, deontic modality, and epistemic modality. Participant-internal modality refers to possibility or necessity that is “internal to a participant engaged in the state of affairs” (ibid.: 80). While internal possibility concerns a participant’s ability, internal necessity implies a participant’s internal need. Participant-external modality involves possibility and necessity external to the participant, that is, it refers to circumstances external to the participant. Deontic modality, a subtype of participant-external modality, “identifies the enabling or compelling circumstances external to the participants as some person(s), often the speaker, and/or as some social or ethical norm(s) permitting or obliging the participant to engage in the state of affairs” (ibid.: 81). Epistemic modality refers to the speaker’s judgment about whether something is certain or probable. Table 1, repeated here from van der Auwera and Plungian (1998: 82), summarises these types of modality.

364

Petar Keyahov and Reeli Torn-Leesik

Table 1. Modality types (van der Auwera and Plungian 1998: 82). Possibility Non-epistemic possibility Participantinternal possibility (Dynamic possibility, Ability, Capacity) Participantinternal necessity (Need)

Participant-external possibility (Non-deontic possibility)

Deontic possibility (Permission)

(Non-deontic necessity)

Deontic necessity (Obligation)

Participant-external necessity

Epistemic possibility (uncertainty)

Epistemic necessity (Probability)

Non-epistemic necessity Necessity

While the division between the family of possibility types and the family of necessity types can be made relatively straightforwardly, distinctions within each family are not so clear-cut. 3. The Balto-Finnic languages The Balto-Finnic languages form a subgroup of the Finno-Ugric language family that are spoken mostly around the eastern coast of the Baltic Sea. Although the precise number of Balto-Finnic languages varies in different descriptions (for example, Raun 1971 distinguishes only three; Décsy 1965 has five, Laanest 1975, and Viitso 1998 seven), the most widely accepted number is seven. These are: Estonian, Livonian, Votic, Finnish, Ingrian, Karelian and Veps. 1 The Balto-Finnic languages form a dialect continuum (Viitso 1998: 96), something that can be observed most clearly in the northern branch (Finnish, Karelian and Veps). Estonian, Livonian and Votic form the southern branch. Ingrian, which historically belongs to the northern branch, shares many features with the southern branch due to its intense contact with Votic. Only Estonian and Finnish are used as official state languages. Livonian, Votic and Ingrian are on the verge of extinction. The present study is based on these seven Balto-Finnic languages. A default reference is the standard form of a language. We will refer to nonstandard forms if there is no established standard language with a long

Modals in Balto-Finnic

365

literary tradition, or when dialectal patterns provide information relevant to the discussion. 4. The study of Balto-Finnic modal verbs 4.1. Data collection Of the seven Balto-Finnic languages, Finnish and Estonian are the most thoroughly studied. Although there are several studies focussing on the semantics and the historical development of modal verbs in these two languages, studies of modal verbs in other Finnic languages are sparse or nonexistent. The number of verbs included in the class of modal verbs within one language varies in different descriptions as well (Flint 1980, Kangasniemi 1992). There is certainly no generally accepted account of modal verbs in Balto-Finnic. Before presenting the analysis of modal verbs, it should be stressed that modal verbs are not the only means of expressing modality in Balto-Finnic languages. Finnish, Karelian, Votic and Ingrian have a special morphological category of potential mood to express probability or possibility, although in the last two languages it is disappearing (Ariste 1948, Laanest 1975). While Kettunen 2 (1947: 81) reported that Livonian also had a potential mood, Laanest (1975: 155) denies its existence. All of these languages have conditional and imperative moods and a number of modal adverbs. The latter are used instead of modal verbs in particular in eastern Balto-Finnic languages. For instance, in Veps the adverb tarbiž/tariž is very frequently used to express necessity. It can be used with the infinitive of a lexical verb as in (1b) or without a verbal complement, i.e. similarly to Russian nado, as illustrated in (1a): (1)

a. tariž kodi-he (Zaitseva 2001: 106) must(=ADV) home-ILLAT ‘One needs to go home.’ b. tariž mända magadamha must(=ADV) go.INF sleep.INF ‘One needs to go to sleep.’ (Deniss Kavinov: Northern Veps, p.c.).

Other types of constructions which do not explicitly contain a modal verb are also used to express modality. For instance, in Livonian the partitive

366

Petar Keyahov and Reeli Torn-Leesik

singular of the fourth infinitive in its verbal use is often employed to express necessity (Kettunen 1947: 86). (2)

Mi’nnΩm um Ɨnda-mΩst. I.DAT be.3SG give-INF.PRT ‘I must give something.’

These other means of expressing modality in Balto-Finnic will, however, not be discussed in detail here. For the purpose of gathering a working sample of modal verbs, the following materials were taken as a descriptive basis: – grammars and other available descriptions of Balto-Finnic languages; – linguistic studies focussing on any aspect of the modal system of a particular Balto-Finnic language; – bilingual dictionaries; – published texts (usually stories, memoirs, etc.); – informants: three for Karelian, one for Veps, and nine for Ingrian. As modal verbs in the Balto-Finnic do not form a clear morphosyntactic class, restrictive selectional criteria had to be applied when choosing the modal verbs for the study. These criteria were (i) the number of languages in which the verb occurs as a modal and (ii) the modal domains it covers. The application of these criteria helps to identify the core modals from idiosyncratic uses of verbs in particular languages. 4.2. The inventory of modal verbs Modal verbs in Balto-Finnic follow two patterns: a personal modal pattern and an impersonal modal pattern 3 . In the personal pattern, the verb agrees with the subject of the sentence in person and number and the subject is in the nominative. The modal verb is followed by the infinitive of a lexical verb. In the impersonal pattern, the modal verb is in a 3SG form followed by the infinitive of a lexical verb and shows no agreement with the (agentive) nominal argument. The latter is either in the genitive (or dative) or in an external local case (adessive, allative). These two patterns are illustrated in (3): (3)

a. Sinä voi-t nukkua. you.NOM can-2SG sleep.INF ‘You can sleep.’

Modals in Balto-Finnic

b. Sinun täytyy you.GEN must.3SG ‘You must sleep.’

367

nukkua. sleep.INF

The two modal patterns show variation between languages as well as variation involving the same verb within one and the same language. For instance, there can be differences in case marking involving the same verb in different languages or dialectal differences in case marking by the same verb in the same language. These differences are represented in examples (4–5). Example (4) illustrates a personal pattern in Estonian and examples (5a) and (5b) show dialectal differences in Livonian, where the former shows a personal pattern and the latter an impersonal pattern. (4)

Mina pid-i-n I must(=hold)-PST-1SG ‘I had to leave earlier.’

varem minema. earlier go.INF

(5)

a. Mina varim pid’ läem. I earlier must(=hold).PST.1SG go.INF ‘I had to leave earlier.’ (Salats dialect, Saukkonen 1965: 123) b. MinnΩn varald pid-iks läem. I.DAT early must(=hold)-COND.3SG go.INF ‘I would have to leave earlier.’ (Piza dialect, Saukkonen 1965: 123)

In Karelian and Veps, the counterpart of Estonian pidama ‘must’ can follow either pattern, whereas in Finnish, Votic and Ingrian it only follows the impersonal pattern. Comparing these two patterns in different BaltoFinnic languages, one finds that Estonian uses the personal pattern with nearly all verbs except for the verbs tulema ‘come; must’, tarvitsema ‘need’ and pruukima ‘need’, where the last two demonstrate variation between the personal and impersonal patterns. Other Balto-Finnic languages allow more verbs either to follow the impersonal pattern or to demonstrate pattern alternation. As already mentioned, Balto-Finnic modals verbs do not form a clear morphosyntactic class. In order to extract the most typical modal verbs, the criteria given in 4.1. were specified as follows: – The verb occurs as a modal in at least three Balto-Finnic languages; – The verb expresses at least two types of modality in addition to its primary (premodal) meaning at least in two languages.

368

Petar Keyahov and Reeli Torn-Leesik

Figure 1 shows that seven verbs satisfy these two criteria: voida, saada, pitää, tulla, tarvita, lie- and täytyä (the forms given are those from Standard Finnish as representative for all Balto-Finnic languages). In the Figure below, the darker the shade of grey, the greater the number of languages in which the respective modal verbs occur. The labels POSSIBILITY and NECESSITY refer to two modal domains. NECESSITY

POSSIBILITY

täytyä

in all 7 languages in t 5 languages

lie-

voida

in t 3 languages

pitää

saada

tulla tarvita Figure 1. The core modal verbs in Balto-Finnic

As can be seen from Figure 1, three verbs voida, saada and pitää form the core of this class as they occur in all languages. If the sample restriction is relaxed to five or more languages, the verbs lie- and tulla enter the group. If the restriction is relaxed even more, to three or more languages, täytyä and tarvita can be added. The vertical broken line divides the domains of possibility and necessity. The verb voida is the only verb that is exclusively used for expressing possibility. The other verbs, lie- and saada, which are used across the area as markers of possibility, cover also some types of necessity in certain restricted areas. Table 2 presents the original or premodal meanings of these modal verbs. 4 The original or premodal meaning of the modal verbs in BaltoFinnic tends to be preserved along with their modal and postmodal

Modals in Balto-Finnic

369

meanings. For example, the original meaning of the verbs saada and tulla is maintained in all languages. Table 2. The original or premodal meanings of the core modal verbs in BaltoFinnic voida saada pitää tulla täytyä tarvita lie-

‘to be able/capable’ < *‘to be strong’(Saukkonen 1966: 74–75) ‘to get’ OBJECT > ACTIVITY > SPACE > TIME > QUALITY. It seems that the modal verb tulema has retained its premodal semantics to a greater degree since it is semantically restricted to the first (and partly the second) stage while the verb pidama has developed further applications. The fact that the verb pidama is not subject to the selectional restrictions that govern tulema means that pidama is more obligatory and has higher system relevance within the class of modals. This line of reasoning is, however, misleading as these selectional restrictions should not be attributed to verbs but to constructions (cf. Hansen (to appear) for a similar situation in Slavonic). The restrictions which govern occurrences like (19b) are not attributed to a particular Estonian verb, but to the impersonal construction with a non-nominative actor argument. Thus, verbs like pidama that follow the personal pattern, which is a default construction in Estonian clausal syntax, are naturally more obligatory and have higher system relevance than the verbs following the impersonal pattern, which is much more restricted in its occurrence. 5.4. Structural scope The structural scope of a gram is “the structural size of the construction it helps to form” (Lehmann 2002: 128). The structural scope of a sign decreases with grammaticalisation. The parameter of structural scope allows a comparison between the grammaticalisation degree of verbs that follow the personal syntactic pattern and that of those following the impersonal pattern. The members of the first group could be called raising verbs as they do not play any role in the selection of the subject of the sentence (see e.g. TornLeesik 2007 for Estonian). Instead, they preserve the subject demands of their infinitival complements. In contrast, the members of the latter group are control verbs as they control the selection of the grammatical subject (Davies and Dubinsky 2004). On the basis of this, one can conclude that modal verbs that follow the personal pattern in Balto-Finnic are more grammaticalised (or auxiliarised) than those that follow the impersonal pattern. Consider the following examples from Karelian: (20) a. Mie voi-n lähtie makuamah. I can-1SG go.INF sleep.INF ‘I can go to sleep.’

384

Petar Keyahov and Reeli Torn-Leesik

b. Miu-la pitäy lähtie makuamah. I-ADE/ALL must(=hold).3SG go.INF sleep.INF ‘I must go to sleep.’ (Pekka Zaikov, p.c.) While the raising verb in (20a) does not determine the argument structure outside of the VP, the control verb in (20b) selects a nominal argument and thus its syntactic scope reaches beyond the VP-boundary. In broader terms, this means that the first verb affects argument structure only at the phrase level whereas the second verb affects argument structure at the clause level. It has been shown that raising verbs derive etymologically from control verbs (Davies and Dubinsky 2004), and this is a manifestation of a decrease in structural scope or condensation. The distinction between verbs that determine the argument structure and verbs that do not is a frequently used criterion for delimiting the class of modal auxiliaries (see e.g. Goossens 1987 for English). Until the 1970s, the Finnish tradition considered only the verbs following the personal pattern as modal (see ISK 2004: 1489 for an overview) and in the Estonian tradition this strictly syntactic criterion has been dominant even longer due to the work of Rätsep (1972: 26, 1978: 35–39). The overall tendency, however, has been towards the recognition of semantic factors such as polyfunctionality, so that the class of modal verbs has become a predominantly semantic category (c.f. ISK 2004). 5.5. Bondedness The degree of bondedness or syntagmatic cohesion is low among BaltoFinnic modal verbs. Constituents may be inserted between the modal verb and its infinitival complement and there are many cases of coordination reduction. Consider (21), from Estonian: (21) Kui klient tea-b, mida ta või-b if client know-3SG what.PRT s/he can(=be_able)-3SG ja saa-b tahta, siis on võimalik be.3SG possible and can(=get)-3SG want.INF then ka vastava-id kulu-sid analüüsida. too respective-PL.PRT expenditure-PL.PRT analyse.INF

‘When the client knows what he may and can ask for, then it is also possible to analyse the respective outlays.’

Modals in Balto-Finnic

385

No cases of cliticisation are attested, although the Eastern Balto-Finnic languages need to be studied more carefully in this respect. 5.6. Syntagmatic variability The parameter of syntagmatic variability concerns the positional mutability of a sign with respect to those constituents with which it forms a construction (Lehmann 2002: 140). All modal verbs in Balto-Finnic occur only within the predicate, but within these boundaries their positional freedom is relatively high. They can occur on either side of the infinitival complement although the pre-infinitival position is the unmarked one. 6. Parameter discomfort or a wrong approach? As Lehmann admits (2002: 143í145), some of the parameters above are very difficult to quantify. This fact and the insufficient information about the paradigmatic and syntagmatic properties of each verb in the present sample make the precise quantification of the degree of grammaticalisation impossible at this stage. GRAMMATICALIZATION

X

Y

P1(c) P2(c) P3(c) P4(c)

P2(n) P3(n) P4(n)

P5(c) P6(c)

P5(n) P6(n)

Figure 3. Ideal case of grammaticalisation

P1(n)

386

Petar Keyahov and Reeli Torn-Leesik

There are also cases of a lack of correlation between parameters. According to Lehmann (2002: 146), a normal grammaticalisation process should satisfy the following condition: “an item which is grammaticalized in a construction will occupy a point on each of the six parameters in such a way that the six points are roughly on a vertical line.” It follows that the lines for any two items would be roughly parallel (or coincide) and that no intersections should be expected. This prediction is demonstrated in Figure 3. If an item X occupies point c on each parameter and an item Y the point n on each parameter, the sums of connectors of these parameter-values should form two vertically parallel lines. Consider now the evidence given in the previous sections concerning the grammaticalisation of the verbs that follow the personal modal pattern and those that follow the impersonal pattern. There are at least three parameters for which one of these groups shows a higher or a lower degree of grammaticalisation than the other. These parameters are formal integrity, paradigmaticity and structural scope. In comparison with the verbs following the personal pattern, the verbs following the impersonal pattern have lost their formal integrity to a greater degree. They lack person/number and voice inflection and they exhibit polarity restrictions more often (see Section 5.1.2.). Due to the idiosyncratic case marking of the actor, they also reveal a higher paradigmaticity as a class (see Section 5.2.). GRAMMATICALIZATION

voida

täytyä

high formal integrity

low formal integrity

low paradigmaticity

high paradigmaticity

wide structural scope

narrow structural scope

Figure 4. The relative degree of grammaticalisation of voida and täytyä

Modals in Balto-Finnic

387

Thus, based on their formal integrity and paradigmaticity, verbs following the impersonal pattern should be considered more grammaticalised. Nevertheless, their structural scope seems to be wider than the structural scope of the verbs of the personal pattern (see Section 5.4.), which is a sign of a lower degree of grammaticalisation. This parameter discomfort could be illustrated by using such absolute notions as ‘high’ or ‘low’ degree of a property. This is done in Figure 4 for one personal and one impersonal verb, the Finnish voida and täytyä respectively. It is obvious that the parameters in the figure do not correlate as expected. There has been a certain amount of disagreement in the Finnic tradition regarding which group of verbs is more grammaticalised. Most scholars (e.g. Siro 1964; Rätsep 1978; Erelt 2003: 106; Torn-Leesik 2007) are inclined to give a status of auxiliary only to the verbs following the personal pattern (due to their narrow structural scope). Laitinen (1992: 162) however claims that the verbs that follow the impersonal pattern are more auxiliary-like, the reason being that they exhibit a number of inflectional restrictions and they do not have a prototypical agent marking. The intersecting lines in Figure 4 portray the source of disagreement well. One might suspect that the lines would become straight and parallel if one chose an entirely different approach and did not compare the degree of grammaticalisation of verbs but that of constructions. Such a study certainly deserves to be carried out, but at first glance it seems to run into problems of how to operationalise parameters like paradigmaticity when comparing our two constructions. Another way to save Lehmann’s parametric harmony is to reassess the directionality of changes. The problematic parameter seems to be the one of structural scope. As Tabor and Traugott (1998) and recently Song (2005) have shown, changes in grammaticalisation need not always involve decrease in structural scope, but may also lead to an increase in structural scope. If this is correct, and increase in scope is consistent with grammaticalisation, then all the parameters correlate in the expected way and the lines in Figure 4 become straight and parallel. 7. Effects of language contact 7.1. Lexical borrowing Balto-Finnic languages exhibit several cases of lexical borrowing in the domain of modality. Examples include the borrowing of non-finite verbal forms (e.g. Karelian dolžen < Russ. ‘be obliged’(=PTCP), see Pyöli 1996;

388

Petar Keyahov and Reeli Torn-Leesik

Sarhimaa 1999) as well as particular finite verbal forms like Karelian and Veps prišlos’ (< Russ. ‘must(=come)-PST.REFL’, see Grünthal 1941: 171 for an example). The degree to which a certain loan is integrated into the structure of a given language varies from pure cases of code-switching in which a native noun agrees with a foreign inflected verb form (e.g. Ingrian toin mǀžet ‘the other can’ = ‘Ingr. other + Russ. can-3SG’: Eva Saar p.c.) to fully integrated loans like the Estonian necessity verb pruukima (< Middle Low German brnjken). The first cases are more common in highly endangered languages like Votic, Ingrian and Livonian. 7.2. Code-copying Among the cases of code-copying, a basic distinction can be drawn between semantic and morphosyntactic calques. In the first case a functional isomorphism between the host and the target language is achieved by reorganizing semantic patterns which aim at word-by-word translation. In the second case a formal isomorphism is achieved through morphosyntactic adjustment to the structure of the host-language. German oEstonian The abundant German influence first targeted Old Written Estonian í a form created and maintained by native Germans or bilingual individuals. The most striking examples are the following. The epistemic possibility uses of saama ‘to get’ in Estonian are attested earlier than the non-epistemic uses, which does not conform to the unidirectionality ‘non-epistemic’ > ‘epistemic’ (c.f. van der Auwera and Plungian 1998: 114). These epistemic uses have, however, been explained as direct translations from the corresponding uses of the German verb werden (Habicht 2001). In Old Written Estonian, the verb tahtma ‘to want’ was used with epistemic/future meaning. This now obsolete use is most likely based on the Low German wollen-future (Habicht 2001). Both German sollen and Estonian pidama ‘to hold’ have undergone a development from ‘obligation’ to ‘hearsay’ (see Hansen 2004: 256 for sollen). According to Kask (1984: 270), the hearsay application of pidama is due to the direct influence of sollen. Swedish o Finnish Although not so overwhelming, the impact of Swedish on Finnish is comparable with that of German on Estonian.

Modals in Balto-Finnic

389

The Finnish tulla ‘to come’ has developed a future meaning probably due to the influence of the Swedish komma att construction (Laitinen 1992: 222). According to Raukko and Östman (1994: 52–53) the uses of saada ‘get’ as a marker of necessity are likely to be an influence of the Swedish verb få. Although such uses are more common in Finnish than in the other Balto-Finnic languages, the same verb can also be used as a marker of necessity in Estonian, Livonian, Karelian and Ingrian, which makes Swedish influence an unlikely explanation. We are more likely dealing with frequential copying (see Johanson: this volume) where Swedish has reinforced pre-existing patterns. There is also an interesting isomorphism between the Swedish and Finnish verbs expressing ability, which cannot be coincidental. In Swedish, the verb må has developed from participant internal possibility to the postmodal meaning ‘feel (good)’ (see van der Auwera and Plungian 1998: 105). In Finnish, the verb voida has a similar lexical meaning ‘be strong; be in good health, feel good’ as in (22), but the opposite direction of development has been suggested here, i.e. from ‘be strong; be in good health’ to participant internal possibility (Saukkonen 1966: 74). (22) Minä voi-n hyvin. I feel(=be_able)-1SG good ‘I feel good. / I am in good health.’ There are also cases of structural isomorphism between Finnish and Swedish dialects. In some Finnish dialects, the necessity verbs which normally follow the impersonal pattern are used personally, and this is ascribed to the impact of the neighbouring Swedish dialects (Laitinen 1992: 48). Russian o Karelian, Veps, Ingrian, Votic The dative case in the Russian construction mne nado/nužno ‘I-DAT need(=MOD.ADV)’ is substituted in Karelian, Veps, Ingrian and Votic with an external local case (ablative-adessive-allative) (see Pyöli: 1996: 257– 258 for the situation in Karelian). Thus, following the Russian construction with the dative, the original genitive case on the nominal argument in the impersonal pattern was replaced with an external local case in Eastern Balto-Finnic. Nevertheless, the construction with the genitive tends to be productive alongside the new construction with local case and there may be cases of functional divergence between the two, but this needs to be studied further.

390

Petar Keyahov and Reeli Torn-Leesik

This Russian model has deeper consequences for the morphosyntax of modal verbs in Eastern Balto-Finnic. The Russian dative construction is similar to the impersonal modal pattern in Balto-Finnic. Word-for-word translations have resulted in an expansion of this pattern in Karelian, Veps, Ingrian and Votic. In these languages the verbs of possibility also tend to follow the impersonal pattern, although they normally follow the personal pattern. This is not surprising considering the fact that the Russian dative occurs also with modals expressing possibility. Based on the available evidence from different Karelian dialects, it is striking that practically all core modal verbs can occur in the impersonal construction with adessiveallative case marking on the nominal argument. Examples are provided in (23): (23) a. Voi-d-go miu-la tei-l’ä can(=be_able)-3SG-Q I-ADE/ALL you-ADE/ALL üödä moata. night.PRT sleep.INF ‘Can I sleep at your place tonight.’ (Palmeos 1962: 33) b. Suaw-go lähtie sinu-l? get(=get).3SG-Q depart.INF you-ADE/ALL ‘Can you go?’ (SKJ-LD 1990: 364) c. Koska auto rikkautu, mei-lä because car brake.REFL.PST.3SG we-ADE/ALL pit-i kulkie. must(=hold)-PST.3SG walk.INF ’As the car broke down, we had to walk.’(Pekka Zaikov: Standard Karelian, p.c.) d. Häne-llä tarviþþe-nut sukeltaa. s/he-ADE need-PST.PTCP dive.INF ‘S/he had to dive.’ (Anastassia Trifonova: Olonets dialect, p.c.) e. Miu-l tuloo sanuo I-ADE/ALL must(=come).3SG say.INF ‘I must say.’ (Saukkonen 1965: 154) f. Miu-l täydyö lähtie I-ADE/ALL must(=get_filled).3SG depart.INF ‘I need to go.’ (Saukkonen 1965: 147)

Modals in Balto-Finnic

391

g. Miu-la l´ieNöu ruadua. I-ADE/ALL must(=be).3SG work.INF ‘I must work.’ (KKS III 1983: 86) Table 4 presents the spread of the impersonal modal pattern with the three most central verbs. Although the evidence from Ingrian is scarce, the situation in Ingrian is probably similar to the situation in Karelian, its closest relative. It is obvious that contact with Russian has triggered a typological drift characterised by an “impersonalisation” of the modal verb system in Eastern Balto-Finnic. 10 Table 4. Verbs used with the impersonal modal pattern 11 Western 12 voida ‘to be able’ saada ‘to get’ pitää ‘to hold’

Eastern

Livonian

Estonian

Finnish

Votic

Ingrian

Karelian

Veps

no

no

no

yes

?

yes

yes

no

no

no

yes

yes

yes

yes

yes

no

yes

yes

yes

yes

yes

The second important consequence is related to the categorial interpretation of modal verbs. In Eastern Balto-Finnic languages, the modal pitää may surface as a main verb. In those cases, pitää is not accompanied by a nonfinite form of the main verb, but is itself the only form of the predicate, as illustrated by the following example from Votic: (24) mi-l piä-b olut (Heinsoo 1990: 42) I-ADE must(=hold)-3SG beer.PRT ‘I need beer.’ However, the original structure of (24), which is preserved in Finnish and Estonian, requires an infinitive of the verb ‘to be’ after the modal. Such “elliptical” uses like the one in (24) are by no means very old. Since no auxiliary should fill the predicate slot alone, one can assume that these secondary uses manifest an increase of the lexical prominence of the modal, and could be regarded as cases of degrammaticalisation. 13 The omission of the main verb also leads to a categorial reanalysis of the modal. In all such uses, the modal follows the impersonal pattern, and thus appears

392

Petar Keyahov and Reeli Torn-Leesik

only in the form of third person singular as an invariable word. The absence of an infinitival complement makes the verbal status of the modal verb even more opaque, and as a consequence, the frozen 3SG can be reanalysed as an adverb. The reanalysis into a non-inflectional class helps word-forword translations from Russian as Russian expresses necessity mostly with adverbs (e.g. nado, neĐzja, nužno). The example in (25) comes from Lude, an intermediate dialect between Karelian and Veps: (25) a. Karelian (Lude) Midä b. Russian ýego what ‘What

sinule tebe for-you do you

pidäw? (SKJ-LD 1990: 264) nado? need(=ADV) need?’

Intense contact with Russian seems to have even more general consequences for the typology of Balto-Finnic languages. Let us consider the choice between expressing modality with modal verbs and expressing modality with adverbs. Porák (1968) noted a peculiar areal distribution of the modal markers in Slavonic. He claims that the west Slavonic languages prefer modal verbs while the east Slavonic languages prefer modal adverbs (see also van der Auwera, Schalley, and Nuyts 2005 for empirical data bearing on this observation). Where the expression of necessity is concerned, a similar areal division seems to hold for Balto-Finnic. The preference for modal adverbs over modal verbs is most striking in Veps í the easternmost language of the family. For example, the adverb tar(b)iž, whose syntax and semantics duplicate that of Russian nado, is the most frequent means of expressing non-epistemic necessity in Veps. In contrast to Slavonic, where this areal distribution is explained with a more general typological pattern, namely with the structural proximity to Standard Average European, in Balto-Finnic, this typological drift is an instance of contact-induced change. Table 5. Areal properties of the modal verbs in Balto-Finnic West

East

1. high polyfunctionality of ‘to get’ 2. verbs of possibility feature always with personal pattern 3. preference for modal verbs

1. low polyfunctionality of ‘to get’ 2. verbs of possibility feature also with impersonal pattern 3. preference for modal adverbs

Modals in Balto-Finnic

393

Table 5 summarises the areal peculiarities of the system of Balto-Finnic modal verbs. There seems to be a clear-cut west-east cline: 8. Conclusions The main aim of the article was to study Balto-Finnic modal verbs in the framework of the grammaticalisation parameters proposed in Lehmann (2002). The fact that modal verbs in Balto-Finnic do not form a coherent morphosyntactic class and that the number of modal verbs varies in various treatments of different languages made the selection of modal verbs for the study a difficult task in itself. The selection was further complicated by the fact that the majority of languages discussed here are very poorly studied as regards their modal strategies. Applying the criteria of geographical spread and modal polyfunctionality helped to identify seven verbs which appear to represent best the Balto-Finnic lexical inventory of modality. Contemporary Finnic linguistics treats modal verbs as a semantically defined class with rather heterogeneous morphosyntactic properties. In other words, modal verbs form a fuzzy morphosyntactic class with few diagnostic features that apply to all of its members. The only clear distinction that can be drawn is that between personal and impersonal patterns, which allows a distinction to be drawn between modal auxiliaries and modal verbs (Torn-Leesik 2007). The results of the study on the realisation of grammaticalisation parameters in the Balto-Finnic modal verbs point to a relatively low degree of grammaticalisation. The class of modal verbs displays a relatively low degree of semantic integrity, but high formal integrity, low paradigmaticity and high paradigmatic variability. Syntagmatically, verbs following the personal pattern have a narrower structural scope than the verbs following the impersonal pattern, while both exhibit low bondedness and a relatively high syntagmatic variability. As regards language contact, the class of modals appears to be susceptible to contact-induced change. There are many cases of both lexical loans and structural calques. As for the areas of structural affinity, three main conclusions can be drawn (which, however, are not generally acknowledged in Finnic linguistics). These are: the existence of higher degrees of polyfunctionality of the modal ‘to get’ in western Balto-Finnic languages than in eastern ones, the expansion of the impersonal pattern to verbs of possibility in eastern Balto-Finnic and the increasing preference for adverbial strategies in eastern Balto-Finnic as opposed to the verbal strategies in the western ones.

394

Petar Keyahov and Reeli Torn-Leesik

9. Acknowledgements This study was supported by grants 2568 and 5202 from the Estonian Scientific Foundation. Our gratitude goes to the language informants Prof. Pekka Zaikov (Standard Karelian), Anastassia Trifonova (Olonets Karelian), Maria Peleshenko (Olonets Karelian) and Deniss Kavinov (Veps); to Eeva Saar who interviewed the Ingrian speakers Jevdokiya Vasil’eva, Zoya Kuznetsova and Nikolai Rodionov, to Natalia Kuznetsova who interviewed the Ingrian speakers Anna Gadyaka, Lyudmila Kosolap, Lyubov’ Ilyina, Sofya Alekseyevna and Valentina Dzhalalova, and to TiitRein Viitso for his valuable information about Livonian. We are also indebted to Mati Erelt, Bernd Heine, Jim Blevins and to the reviewers for their helpful comments on different drafts of the chapter. Dictionaries and text materials InkmS = Inkeroismurteiden sanakirja. [Dictionary of the Ingrian dialects.] R. E. Nirvi (ed.). 1971. Lexica Societatis Fenno-Ugricae. 18. Helsinki: Suomalais-ugrilainen Seura. Palmeos, Paula. 1962. Karjala Valdai murrak. [The Valdai dialect of Karelian.] Tallinn: Eesti NSV Teaduste Akadeemia. KKS = Karjalan kielen sanakirja. I–VI. [Dictionary of Karelian. I–VI.] Lexica Societatis Fenno-Ugricae. 16 (1968í2005). Helsinki: Suomalais-Ugrilainen Seura. SKJ-LD = Slovar’ Karel’skogo jazyka (livvikovskij dialekt). [Dictionary of Karelian (the Olonets dialect).] 1990. Petrozavodsk: Karelia. Abbreviations: Est – Estonian, Fin – Finnish, Ingr – Ingrian, Kar – Karelian, Liv – Livonian, Russ – Russian, Vps – Veps, Vot – Votic Notes 1. Not all linguists agree about the status of Ingrian as an independent language. For instance, whereas Finnish linguists often consider Ingrian as a dialect of Karelian or even of Finnish (e.g. Kettunen 1957; Turunen 1988), Soviet FinnoUgricists treated Ingrian as an independent language (e.g. Ariste 1956). Lude is also either seen as an independent language (Itkonen 1983) or a dialect of Karelian (Turunen 1988; Saarinen 1997, Viitso 1998).

Modals in Balto-Finnic

395

2. Kettunen (1947: 81, 88) also notes that ‘nomen agentis’ can be used to express the potential in Livonian. 3. It must be emphasised that the term ‘impersonal’ here refers to the use of a 3SG verb form and a lack of agreement and should thus be kept apart from the impersonal voice found in most of the Balto-Finnic languages. The impersonal voice is expressed with specific morphological markers, which is in opposition to a system of personal forms, as described in Torn (2002). 4. The premodal meaning is given as an English translation of the modal verbs in the rest of the article. 5. However, in Standard Estonian this verb has a clear lexical meaning. 6. By doing so we alter the sense in which ‘polyfunctionality’ is understood by van der Auwera, Ammann, and Kindt (2005). We also consider polyfunctional those cases that are spread across the boundary between possibility and necessity. 7. By counting the degree of polyfunctionality on the basis of van der Auwera and Plungian’s eight modality types (see Table 1), we disregard some finer distinctions such as ‘physical vs. mental’ participant-internal modality. The coding of this distinction in Balto-Finnic must be studied independently, but the general tendency is that the verbs of possibility express physical but not mental capacity (for which a special verb is preferred), while the necessity verbs are polyfunctional across the distinction between physical and mental participant-internal necessity. 8. As noted by an anonymous reviewer, post-modal meanings do not always reflect one step further in the process of auxiliarization. An example of this will be discussed in section 7.2 under the notion of degrammaticalization. 9. See Laitinen (1993) for explanatory details on the corresponding shift in Finnish. 10. As noted by an anonymous reviewer, it is not unprecedented that borrowings can become the core of the modal system of a language. Such cases are attested in Turkic languages. 11. As previously noted, the fact that a certain verb occurs in a language with the impersonal pattern does not mean that the opposite possibility is not available. 12. The notions ‘Western’ and ‘Eastern’ do not designate any deeper genetic kinship, but geographical reality as well the fact that the Eastern group of languages, unlike the Western group, has been located within the boundaries of Russia at least ever since the Russian-Swedish war in the early 18th century. 13. ‘Degrammaticalization’ is understood here in the traditional sense as a process whereby an item becomes less grammatical and/or more lexical (see Lehmann 2002 for an overview).

396

Petar Keyahov and Reeli Torn-Leesik

References Ariste, Paul 1948 Vadja keele grammatika. [A Grammar of the Votic Language.] Tartu: Teaduslik Kirjandus. Ariste, Paul 1956 Isuri keelest. [On Ingrian.] Emakeele Seltsi Aastaraamat 2. [The Estonian Mother Tongue Society Yearbook 2.], 32–52. Bybee, Joan, Revere, Perkins, and William Pagliuca. 1994 The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: The University of Chicago Press. Davies, William D., and Stanley Dubinsky 2004 The grammar of Raising and Control: a course in syntactic argumentation. Oxford: Blackwell Publishers. Décsy, Gyula 1965 Einführung in die finnisch-ugrischen Sprachwissenschaft. Wiesbaden: Harrassowitz. Erelt, Mati 2001 Some Notes on the Grammaticalization of the Verb pidama in Estonian. In Estonian: Typological Studies V, Mati Erelt (ed.), 7–25. (Tartu Ülikooli eesti keele õppetooli toimetised 18. [Publications of the Department of Estonian of the University of Tartu 18.]) Tartu: Tartu Ülikooli Kirjastus. Erelt, Mati 2003 Syntax. In Estonian Language, Mati Erelt (ed.), 93–129. (Linguistica Uralica Supplementary Series/Volume 1) Tallinn: Estonian Academy Publishers. Flint, Aili. 1980 Semantic structure of the Finnish lexicon: Verbs of possibility and sufficiency. (Suomalaisen Kirjallisuuden Seuran Toimituksia 360. [Finnish Literature Society Series 360.]) Helsinki: Finnish Literature Society. Goossens, Louis 1987 The auxiliarization of the English modals: a Functional Grammar view. In Historical development of auxiliaries, Martin Harris, and Paolo Ramat (eds.), 111–143. Amsterdam: Benjamins. Grünthal, Villem 1941 Itämerensuomalaisten kielten yksikön nominatiivi objektin edustajana aktiivin yhteydessä. Lauseopillinen tutkimuskoe. [Nominative singular as object in active clauses in Balto-Finnic languages. A Syntactic Experiment.] Helsinki: Suomalaisen Kirjallisuuden Seura.

Modals in Balto-Finnic

397

Habicht, Külli 2001 Vanhan kirjaviron modaalirakenteista. [On modal constructions in Old Written Estonian.] In Congressus Nonus Internationalis FennoUgristarum 7.-13.8.2000 Tartu. Pars IV. Dissertationes sectionum: Linguistica I, Tõnu Seilenthal, Anu Nurk, and Triinu Palo (eds.), 268–276. Tartu: Eesti Fennougristide Komitee. Hakulinen, Auli and Marja-Leena Sorjonen 1989 Modaaliverbit spontaanissa keskustelussa. [Modal verbs in spontaneous conversation.] In Suomalaisen keskustelun keinoja I. [Means of Conversation in Finnish I.] Kieli 4 [Language 4], Auli Hakulinen (ed.), 73–97. Helsinki: Helsingin yliopiston suomen kielen laitos. Hansen, Björn 2004 Modals and the Boundaries of Grammaticalization: the Case of Russian, Polish and Serbo-Croatian. In What Makes Grammaticalization? A Look from its Fringes and its Components, Björn Wiemer, Walter Bisang, and Nikolaus Himmelmann (eds.), 245–270. (Trends in Linguistics. Studies and Monographs 158) Berlin/New York: Mouton de Gruyter. Hansen, Björn (in print) Modals (Modale Auxiliare). In Slavic Languages. An International Handbook of their Structure, History and Investigation (Handbücher zur Sprach- und Kommunikationwissenschaft), Peter Kosta, Tilman Berger, Karl Gutschmidt, and Sebastian Kempgen (eds.). Berlin/New York: Walter de Gruyter. Heine, Bernd, Ulrike Claudi, and Frederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: The Chicago University Press. Heinsoo, Heinike 1990 Modaalverbidest vadja keeles. [On modal verbs in Votic.] FennoUgristica 16, 38–45. ISK = Iso suomen kielioppi. [Comprehensive Finnish Grammar.] 2004 Helsinki: Suomalaisen Kirjallisuuden Seura. Itkonen, Terho 1983 Välikatsaus suomen kielen juuriin. [A survey of the roots of the Finnish language.] Virittäjä 87, 190–229, 349–386. Kangasniemi, Heikki 1992 Modal expressions in Finnish. Studia Fennica. Linguistica 2. Helsinki: Finnish Literature Society. Kask, Arnold 1984 Eesti murded ja kirjakeel. [Estonian Dialects and Literary Estonian.] Tallinn: Valgus. Kettunen, Lauri 1947 Hauptzüge der livischen laut- und formengeschichte. Helsinki: Suomalais-Ugrilainen Seura.

398

Petar Keyahov and Reeli Torn-Leesik

Kettunen, Lauri 1957 Isuri keel. [The Ingrian Language.] Virittäjä 61, 124–133. Kiuru, Silva 1977 Suomen kielen kieltohakuiset verbit: Murreaineistoon perustuva syntaktis-semanttinen tutkimus. [The negative polarity verbs of Finnish. A syntacto-semantic study based on dialectal data.] (Suomalaisen Kirjallisuuden Seuran Toimituksia 335. [Finnish Literature Society Series 335.]) Helsinki: Suomalaisen Kirjallisuuden Seura. Kiuru, Silva 1988 Agricola tulepi, jos henen tule. Ind. Preesensin yks. 3. persoonan muodot Mikael Agricolan kielessä. [Agricola’s tulepi, jos henen tule. The forms of third person singular present indicative in the language of Mikael Agricola.] Kieli 3. [Language 3.] Helsinki: Helsingin yliopiston suomen kielen laitos. Kuteva, Tania 1999 TAM-auxiliation, and the Avertive Category in Northeast Europe. Grammaticalization aréale et semantique cognitive: les langues fenniques et sames. Oural-Ural, no. 1, Tallinn: Eesti Keele Sihtasutus, 27–41. Laanest, Arvo 1975 Sissejuhatus läänemeresoome keeltesse. [An Introduction to BaltoFinnic Languages.] Tallinn: Eesti NSV Teaduste Akadeemia. Laitinen, Lea 1992 Välttämättömyys ja persoona. Suomen murteiden nesessiivisten rakenteiden semantiikkaa ja kielioppia. [Inevitability and person. Semantic and grammatical properties of Finnish necessive constructions.] Helsinki: Suomalaisen Kirjallisuuden Seura. Laitinen, Lea 1993 Nesessiivirakenne, kieliopillistuminen ja subjektivisuus. [Necessive construction, grammaticalization and subjectivity.] Virittäjä 97, 149– 170. Laitinen, Lea, and Maria Vilkuna 1993 Case-Marking in Necessive Constructions and Split Intransitivity. In Case and Other Functional Categories in Finnish Syntax, Andres Holmberg, and Urpo Nikanne (eds.), 23–48. Berlin/New York: Mouton de Gruyter. Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Metslang, Helle 1996 The Developments of the Futures in the Finno-Ugric Languages. In Estonian: Typological Studies I, Mati Erelt (ed.), 123–144. (Tartu Ülikooli eesti keele õppetooli toimetised 4. [Publications of the

Modals in Balto-Finnic

399

Department of Estonian of the University of Tartu 4.]) Tartu: Tartu Ülikooli Kirjastus. Penjam, Pille 2005 Tulema-verbi leksikaalsest ja grammatilisest kasutusest ning modaalse funktsiooni kujunemisest eesti kirjakeeles 17.–20. sajandil. [On the lexical and grammatical usage of the verb tulema and the development of its modal function in Written Estonian in the 17th– 20th centuries.] Master’s Thesis, unpublished manuscript. University of Tartu. Tartu. Penttilä, Aarni 1963 Suomen kielioppi. [Finnish Grammar.] Second edition. Porvoo: WSOY. Porák, Jaroslav 1968 Modalverbien im Tschechischen und Deutschen: In Deutschtschechische Beziehungen im Bereich der Sprache und Kultur, Bohuslav Havránek, and Rudolf Fischer (eds.), 97í101. Berlin: Akademie-Verlag. Pyöli, Raija 1996 Venäläistyvä aunuksenkarjala: kielenulkoiset ja -sisäiset indikaattorit kielenvaihtotilanteessa. [Russian influence in Olonets Karelian: the language-external and language-internal indicators in language shift situation.] Joensuu: Joensuun yliopisto. Raukko, Jarno, and Jan-Ola Östman 1994 Pragmaattinen näkökulma Itämeren kielialueeseen. [A pragmatic approach to the Baltic linguistic area.] (Publications No. 25) University of Helsinki. Department of General Linguistics Raun, Alo 1971 Essays in Finno-Ugric and Finnic Linguistics. Bloomington/The Hague: Mouton and Co. Rätsep, Huno 1972 The Syntactic Characterization of the Estonian Modal Verbs. Generatiivse grammatika grupi koosolek. Teesid. [The Meeting of Generative Grammar Group. Abstracts.] (Tartu Riiklik Ülikool. Eesti keele kateeder) Tartu: Tartu Ülikool, 26–30. Rätsep, Huno 1978 Eesti keele lihtlausete tüübid. [The types of Estonian simple sentences.] ENSV Teaduste Akadeemia Emakeele Seltsi Toimetised no 12. [Publications of Estonian Mother Tongue Society 12.] Tallinn: Valgus. Saarinen, Sirkka 1997 The Uralic Languages of Europe. In Atlas Linguarum Europae (ALE). Perspectives nouvelles en géolinguistique, Mario Alinei et al., (eds.), 289–303. Roma: Istituto Poligrafico e Zecca dello Stato.

400

Petar Keyahov and Reeli Torn-Leesik

Sarhimaa, Anneli 1999 Syntactic transfer: contact-induced change, and the evolution of the bilingual mixed codes: focus on Karelian-Russian language alternation. (Studia Fennica. Linguistica 9) Helsinki: Finnish Literature Society. Saukkonen, Pauli 1965 Itämerensuomalaisten kielten tulossijainfinitiivirakenteiden historiaa I. [A history of lative infinitive constructions in Balto-Finnic languages I.] Helsinki: Suomalais-ugrilainen seura. Saukkonen, Pauli 1966 Itämerensuomalaisten kielten tulossijainfinitiivirakenteiden historiaa II. [A history of lative infinitive contructions in Balto-Finnic languages I.] Helsinki: Suomalais-ugrilainen seura. Seuren, Pieter 2003 Verb clusters and branching directionality in German and Dutch. In Verb constructions in German and Dutch, Pieter A. M. Seuren, and Gerard Kempen (eds.), 247í296. Amsterdam/Philadelphia: Benjamins. Siro, Paavo 1964 Suomen kielen lauseoppi. [Finnish syntax.] Helsinki: Tietosanakirja Oy. Song, Jae Jung 2005 Grammaticalization and structural scope increase: possessiveclassifier-based benefactive marking in Oceanic languages. Linguistics 43/4, 795í838. Zaitseva, Maria 2001 Vepsän kielen lauseoppia. [Veps syntax.] (Mémoires de la Société Finno-Ougrienne 241) Helsinki: Suomalais-ugrilainen seura. Tabor, Whitney, and Elizabeth C. Traugott 1998 Structural scope expansion and grammaticalization. In The Limits of Grammaticalization, Anna Giacalone Ramat, and Paul G. Hopper (eds.), 229í272. Amsterdam: Benjamins. Tauli, Valter 1966 Structural tendencies in Uralic languages. (Indiana University Publications. Uralic and Altaic Series 17) The Hague: Mouton. Torn, Reeli 2002 The status of the passive in English and Estonian. In RCEAL Working papers in English and applied linguistics 7, Henriette Hendriks (ed.), 81–106. Torn-Leesik, Reeli 2007 Voice and modal verbs in Estonian. Linguistica Uralica XLIII 3, 173í186. Turunen, Aimo 1988 The Balto-Finnic Languages. In The Uralic languages: description, history and foreign influences, Denis Sinor (ed.), 58–83. (Handbuch

Modals in Balto-Finnic

401

der Orientalistik. Vol. 1. 8. Abt. Handbook of Uralic studies) Leiden: Brill Academic Publishers. Uuspõld, Ellen 1989 Modaalsusest ja modaalsest predikaadist eesti keeles. [On modality and modal predicates in Estonian.] Keel ja Kirjandus 8 [Language and Literature 8], 468í477. van der Auwera, Johan, and Vladimir Plungian 1998 Modality’s Semantic Map. Linguistic Typology 2, 125–139. van der Auwera, Johan, Andreas Ammann and Saskia Kindt 2005 Modal Polyfunctionality and Standard Average European. In Modality. Studies in Form and Function, Alex Klinge, and Henrik Høeg Müller (eds.), 247–272. London: Equinox. van der Auwera, Ewa Schalley, and Jan Nuyts 2005 Epistemic possibility in a Slavonic parallel corpus – a pilot study. In Modality in Slavonic Languages, Björn Hansen, and Petr Karlík (eds.), 201–217. (Slavolinguistica 6) München: Otto Sagner. Viitso, Tiit-Rein 1998 Finnic. In The Uralic Languages, Daniel M. Abondolo (ed.), 96– 114. London/New York: Routledge.

12. Modals in Hungarian Erika Körtvély

1. Introduction Hungarian is a Uralic language spoken by approximately 15 million people, mostly in and around Hungary in Central Europe. Hungarian has a decisive Uralic grammatical heritage, but during its history it was also influenced by non-Uralic languages from the Turkic and Indo-European language families. Hungarian is a transparent agglutinative language with very few fusional and suppletive features. Accordingly, the most generally used morphological operation of the language is suffixation. One of the most characteristic features of the syntax of Hungarian is its free word order: although some syntagms have a rigid internal structure, it is information structure that basically determines the position of the phrases in sentences. In Hungarian there are numerous ways to express modality. Various shades of possibility, necessity and volition can be expressed by modal verbs or auxiliaries, modal adjectives, modal particles, modal tags, and also by a modal affix, e.g.: 1 (1)

Holnap reggel korán kell kelnem. tomorrow morning early must get up.INF.POSS.1SG ‘I have to get up early tomorrow morning.’

(2)

Az étteremben tilos dohányozni. The restaurant.INE forbidden smoke.INF ‘Smoking is forbidden in the restaurant.’

(3)

Anna biztos moziba ment. Anna surely cinema.ILLAT go.PST.3SG ‘Anna surely went to the cinema.’

(4)

Anna, azt hiszem, elment Anna that think.1SG PREF.go.PST.3SG

már already

a the

404

Erika Körtvély

moziba. cinema.ILLAT ‘Anna, I think, has already left for the cinema.’ (5)

Éva ma megeheti Anna ebédjét Eva today eat.MOD.AFFIX Anna lunch.POSS.3SG.ACC ‘Eva may / is allowed to eat also Anna’s lunch today.’

is. too

Keeping the basic aim of the present volume in view, modality will be understood in a broad sense here, i.e. it will not be restricted to the basic logical concepts of possibility and necessity, and linguistic tools indicating the speaker’s attitude to a given fact will be referred to as modal elements. However, considering the aims of the survey again, only modal verbs, modal auxiliaries, and the modal affix -hAt ‘~may, can’ (a result of the grammaticalization of a modal auxiliary) will be dealt with in detail below. 2. The category of modals in Hungarian (extension of the category; semantics, morphology, syntax) In Hungarian, modals do not form a syntactically or morphologically uniform group. The core verbs expressing necessity, possibility and volition in Hungarian are the following: kell ‘must’, szabad ‘may, is allowed’, lehet, ‘be possible’, tud ‘can’ (6a), bír ‘can (mostly physically)’ talál ‘might’, akar ‘want’, and szeretne ‘would like’. All of these verbs have also full verb uses with the following meanings: kell ‘need’, szabad ‘free’, 2 lehet ‘be+modal affix’, tud ‘know’(6b) talál ‘find’, akar ‘want’ and szeretne 3 ‘would like’. As modals, they occur as parts of quasi-auxiliary constructions and combine with an infinitive verb form: (6)

a. Amíg dolgozol, én meg tudom while work.2SG I PREF can.DFC.1SG az ebédet. the lunch.ACC ‘While you are working, I can make the lunch.’ b. Tudom az éneket. know.1SG the song.ACC ‘I know the song (by heart).’

fĘzni cook.INF

Modals in Hungarian

405

Although all modals in Hungarian can occur in infinitival constructions, according to the prevailing analysis in Hungarian morphology (cf. Kenesei 2000), only one of them, talál ‘might’ can be considered a modal auxiliary. (Accepting Kenesei’s views, I will use the term auxiliary in the same narrow sense, i.e. auxiliaries will be understood here as syntactically dependent verbs which occur in auxiliary constructions containing an inflected verb form and a nominalized main verb, the infinitive. Additionally, in Hungarian an auxiliary cannot be the semantic predicate of the sentence, it cannot bear stress in a neutral sentence, and if its main verb is a prefixed verb, the auxiliary has to stand between the prefix and the verb, juxtaposed to the prefix. For details, see Kenesei 2000: 108.) Although not all of the verbs in the present discussion can be considered real auxiliaries in the above narrow sense (and as such they are not necessarily dependent on the infinitive they form a relatively tight syntactic and semantic unit with), for the sake of simplicity and brevity, the infinitive part of the modal construction will be referred to as the main verb here. The modals kell ‘must’, szabad ‘may’ and lehet ‘be possible’ do not take verbal agreement suffixes, hence the person and number of the subject is optionally marked either on the infinitive (7a), or is present in the sentence as a noun or pronoun in dative form (7b). (These marking strategies can appear also simultaneously (7c), but the presence of a personal pronoun always indicates emphasis.) (7)

a. Meg

kell fĘznöm must cook.INF.POSS.1SG ‘I have to make the lunch.’ PREF

b. Annának meg kell fĘzni az Anna.DAT PREF must cook.INF the ‘Anna has to make the lunch.’

az the

ebédet. lunch.ACC

ebédet. lunch.ACC

c. Annának meg kell fĘznie az ebédet. Anna.DAT PREF must cook.INF.POSS.3SG the lunch.ACC ‘Anna has to make the lunch.’ Concerning the ability of the above mentioned verbs to express more than one modal meaning, tud ‘can’, kell ‘must’, szabad ‘may’ and lehet ‘be possible’ are polyfunctional (e.g. kell ‘must’ can express dynamic or epistemic possibilities), while talál ‘might’, akar ‘want’ and szeretne ‘would

406

Erika Körtvély

like’ as modal auxiliaries are confined to indicating only one type of modality (epistemic possibility, and more intensive / less intensive volition, respectively).

3. Other means of expression of possibility, necessity and volition 3.1.1. The modal affix -hAt In Hungarian, there are three grammatical moods, the indicative, conditional and subjunctive (used also as imperative). Thus, Hungarian has not developed grammatical moods to express modality. However, it has a modal affix which can indicate both dynamic and epistemic possibilities, depending on the context: (8)

a. Ebbe a házba akárki this.ILLAT the house.ILLAT anybody bejöhet. come.in.MOD.AFFIX.3SG ‘Anybody is allowed to come into this house.’ (deontic) b. Anyukámnak van telefonja, bármikor mother.POSS.1SG.DAT be.3SG telephone.POSS.3SG, any time felhívhatom. (circumstantial) PREF.call.MOD.AFFIX.DFC.1SG ‘My mother has a telephone, I can call her any time.’ c. Anna még nem érkezett meg. Megint Anna yet no arrive.PST.3SG PREF again dugóba kerülhetett. (epistemic) jam.ILLAT get.MOD.AFFIX.PST.3SG ‘Anna has not arrived yet. She possibly got into a traffic jam again.’

The modal affix -hAt is a very productive inflectional tool to express modality in Hungarian, and, due to its semantics, when it expresses

Modals in Hungarian

407

circumstantial possibility or ability, it is interchangeable with the modal verb tud ‘can’ (although not paradigmatically). 3.1.2. Historical bases of the modal affix -hAt According to current thinking in Hungarian historical linguistics, the basis of the affix -hAt is without doubt the verb hat. Hat as a full verb can be found in the present lexicon of Hungarian as well with the meaning ‘to affect’ but as its Ugric counterparts and historical sources show, the original meaning of the verb was ‘to get further, to penetrate’. The semantics of hat developed into different but semantically related directions: 1. ‘to get under one’s influence’, to possess’, 2. ‘to be able’ 3. ‘to affect’. According to the historical grammar of Hungarian (cf. D. Bartha 1991: 102-103), the change of the verb to an affix took part in the ancient Hungarian period (896-1526 AC). Parallel to this, the verb hat in its second meaning, ‘to be able’, was used widely beside infinitives as a modal verb expressing ability also during the middle Hungarian period (1526-1772). In the last centuries the modal hat ‘to be able’ has been lost. Thus, the lexical and grammatical results of the original verb ‘to get further, to penetrate’ are a full verb with a non-modal meaning ‘to affect’, and a modal affix expressing possibility. If we look at the historical use of the affix, we can observe enrichment in its modal semantics as well: originally it could express circumstantial possibility, to which a deontic, and later an epistemic meaning was added. 3.2. Modal adjectives and adverbs, modal particles In Hungarian, modality can be expressed also by adverbs with a modal meaning, modal adjectives, and some modal particles. The most frequently used modal adjectives (indicating prohibition and permission) are tilos ‘forbidden’ and szabad ‘permitted, allowed’, but in more formal expressions adjectives such as szükségszerĦ, szükséges ‘necessary’, lehetséges ‘possible’, etc. can occur as well. The adjectives tilos ‘forbidden’ and szabad ‘allowed’ may also appear beside an infinitival main verb in predicative use, but in the case of the other adjectives expressing modal meanings this kind of use is, although not impossible, not natural.

408 (9)

Erika Körtvély

Tegnap tilos volt kimenni / kimennünk a yesterday forbidden was go out.INF go out.INF.POSS.1PL the kertbe. garden.ILLAT ‘Yesterday it was forbidden to go out to the garden. / Yesterday we were forbidden to go into the garden.’

(10) Tegnap Annának szabad volt kimenni / yesterday Anna.DAT permitted was go out.INF kimennie a kertbe. go out.INF.POSS.3SG the garden.ILLAT ‘Anna was permitted to go out to the garden yesterday.’ The person and number of the subject of the sentence can be optionally marked on the infinitive, or expressed by an overt noun or pronoun in dative form. (The presence of a personal pronoun, just like in the cases of the modal constructions with szabad ‘may’, kell ‘must’ and lehet ‘be possible’ – cf. 2 – indicates emphasis.) In Hungarian, there are numerous adverbs and particles expressing modality, e.g. valószínĦleg, feltehetĘleg, ‘probably’, talán ‘maybe’, 4 szükségszerĦen ‘necessarily’, biztos, bizonyára, biztosan ‘surely’ etc. Many elements of this group are derived from modal adjectives or verbs with meanings connected to necessity and possibility. (11) ValószínĦleg nem fogom sokat hordani probably no FUT.AUX.DFC.1SG much.ACC wear.INF az új cipĘmet, mert nagyon magas the new shoe.POSS.1SG.ACC because very much high a sarka. the heel.POSS.3SG ‘I probably won’t wear my new shoes much, because their heels are very high.’ (12) Anna megint késik. Biztos dugóba került. Anna again be late.i.3SG surely jam.ILLAT get.PST.3SG ‘Anna is late again. She must have got into a traffic jam.’ Some invariable elements, such as e.g. tán ‘maybe, even’ or bizony ‘surely, indeed’, még, már ‘~even’, etc., which have a (partly) modal meaning, are

Modals in Hungarian

409

mentioned here as modal particles. These elements, however, usually do not indicate modality on their own: they just emphasize the modal characteristics of a situation and occur often together with other modal markers. As such, they can also be considered pragmatic elements or discourse markers. Another non-inflecting word which can occur in a predicative position is muszáj ‘must’ (< German muss sein). Its meaning and usage are similar to the modal verb kell ‘must’, but they show characteristic differences semantically, syntactically, and morphologically. Although both elements express necessity, the intensity of the necessity is higher in the case of muszáj ‘must’: (13) Nagyon elfáradtam, muszáj very much get tired.PST.1SG must egy kicsit. a little.ACC ‘I am exhausted, I must rest a little.’

megpihennem PREF.have rest.INF.POSS.1SG

Positionally, as it is clear from example (13), muszáj ‘must’ does not occur between the prefix and the main verb, so it does not cause the separation of the prefix. Muszáj ‘must’, as it was mentioned above, cannot be inflected at all. Thus, muszáj has kept its original predicative status and it has to be classified in present Hungarian as a predicative particle rather than a defective verb. 3.3. Modal tags Modal tags (cf. De Haan 2006: 38), defined as syntactic constructions expressing the attitude of the speaker, are very frequently used in Hungarian. The most typical modal tags are the following: lehet, hogy ‘it may be that, maybe’; biztos, hogy ‘for sure’, elképzelhetĘ, hogy ‘it is concievable’, úgy gondolom, hogy ‘I think that’; véleményem szerint ‘in my opinion’, szerintem ‘as for me’; etc. Modal tags in Hungarian are usually epistemic in nature. 4. The grammaticalization of modals In this section the following modal verbs will be investigated: kell ‘must’, szabad ‘may, be allowed’, lehet ‘be possible’, tud ‘can’ bír ‘can (physically)’

410

Erika Körtvély

talál ‘might’, akar ‘want’, and szeretne ‘would like’. In Hungarian, as was noted above, the category of modals is not a consistent and morphologically or semantically well definable group of verbs. However, the verbs listed above can be considered as the most frequently used verbs with a modal semantics which, additionally, may appear beside an infinitive main verb. 4.1. Parameter Integrity (bleaching, erosion of the form) As already mentioned in section 2, modals in Hungarian are used also as full verbs with partly different meanings. Thus, their modal semantics are partly due to the loss or modification of their original lexical semantics. A line of change can be drawn from a concrete lexical meaning (which is in some sense pseudo-modal in nature) to a more abstract, and to some degree desemanticized meaning. (14) a. kell ‘be needed’ > ‘be necessary, must’ (both dynamic and epistemic) b. szabad ‘free’ > ‘be allowed, can’ (dynamic), c. tud ‘know’ > ‘can’ (dynamic) d. bír ‘possess, endure, (like someone) > ‘can’ (dynamic) e. lehet ‘be+MOD.AFFIX’ 5 > ‘be possible’ (both dynamic and epistemic) f. talál ‘find’ > ‘might, unexpectedly and unwillingly’ (epistemic) g. akar ‘want’ (dynamic) h. szeretne ‘would like’ (dynamic) Used as full words, kell ‘is needed’, tud ‘know’ bír ‘possess, endure, coll. like’, lehet ‘can be, may be’ (without an infinite main verb) talál ‘find’ and akar ‘want’ all have a full verbal paradigm. Szeretne ‘would like’, which is a regular conditional form of szeret ‘like’, can take all inflectional affixes which are not excluded by its conditional mood marker grammatically (i.e. it cannot have other mood markers). Szabad ‘free’, as was briefly mentioned in section 2, is basically an adjective, and it can be inflected (defectively) as a verb only in its modal use. When occurring as a full word, it has a full nominal paradigm as an adjective. A kind of morphological degeneration has affected the modal verbs kell ‘must’, 6 szabad ‘may’ and lehet ‘be possible’. 7 These verbs, although they take mood and tense markers, cannot take verbal agreement suffixes.

Modals in Hungarian

(15) a. El

411

magyaráznom : *el explain.INF.POSS.1SG PREF kelljek magyarázni (magyaráznom) must.SBJ.1SG explain.INF(POSS.1SG) ‘I need to explain.’ PREF

kelljen must.SBJ

b. El

szabadott menned : may.PST go.INF.POSS.2SG menni (menned) go.INF(POSS.2SG) ‘You were allowed to go.’ PREF

c. Ki

*el PREF

szabadtál may.PST.2SG

találnia : *ki find out.INF.POSS.3SG : PREF lehette találni (találnia) be possible.PST.DFC.3SG find out.INF(POSS.3SG) ‘It was possible for her to find it out.’ PREF

lehetett be possible.PST

Additionally, a further case of morphological degeneration can be attested in the inflectional paradigms of kell ‘must’, szabad ‘may’, lehet ‘be possible’ and talál ‘might’, namely, these verbs cannot be nominalized. They are not grammatical, for instance, as participles or action nouns, respectively: (16) a. *hazamenni kellĘ / kellés home go.INF must.IMPF.PART/ must.VN ‘needing to go home’ / ‘the act of the needing to go home’ b. *megenni szabadó / szabadás PREF.eat.INF may.IMPF.PART / may.VN ‘wanting to eat’ / ‘the act of being permitted to eat’ c. *látni lehetĘ / lehetés see.INF be possible.IMPF.PART / be possible.VN ‘being able to be seen / ‘the event of being able to be seen’ d. *felrobbanni találó / találás PREF.blow up.INF might.IMPF.PART / might.VN ‘the one who is possibly to blow up’ / ‘the event of a possible blowing up’

412

Erika Körtvély

The other modals, such as tud ‘can’, bír ‘can’, akar ‘want’ can be inflected freely and also have nominalized forms in modal constructions: (17) a. úszni tudó / tudás swim.INF can.IMPF.PART / can.VN ‘being able to swim / ‘the event of being able to swim’ b. hazamenni akaró / akarás home.go.INF want.IMPF.PART / want.VN ‘wanting to go home’ / ‘the urge to go home’ etc. Szeretne ‘would like’, because of its fixed finite mophological status cannot be nominalized. As could be seen above, all of the modals in scope were affected by semantic bleaching to some degree, and the verbs kell ‘must’, lehet ‘is possible’, szabad ‘may’ and talál ‘might’ show erosion also in their morphology. 4.2. Parameter Paradigmaticity (paradigmaticization) The Hungarian modal verbs form a group which can be characterized by semantic parameters rather than formal ones. They are not homogeneous in their phonological shape, they do not show the same morphological reduction, and, syntactically, although all of them can occur with an infinite main verb, some of them are more auxiliary-like than the others. As was shown above, some of the Hungarian modals have a defective paradigm, but they cannot be considered a group of irregular verbs. The Hungarian modal constructions can only be differentiated semantically from other syntactic verb (auxiliary) + infinitive structures. Thus, they do not form a “modal subparadigm” in the verbal paradigm. 4.3. Parameter Paradigmatic Variability (obligatorification) Although not as periphrastic constructions, i.e. not as a part of the verbal inflectional paradigm, modals are, to some degree, interchangeable with each other. If the linguistic situation is not fully specified semantically and pragmatically, the verbs expressing volition (akar ‘want’, szeretne ‘would

Modals in Hungarian

413

like’), or those expressing ability (tud ‘can’, bír ‘can’), can be interchanged. (Of course, due to the basic lexical meaning of the verbs, the expressions are not totally equivalent.) (18) a. El

akarok / szeretnék menni moziba. want.1SG would like.1SG go.INF cinema.ILLAT ‘I want / would like to go to the cinema.’ PREF

b. Fel

tudom / bírom emelni can.1SG / can.1SG lift.INF ‘I can lift the bag.’ PREF

a the

táskát. bag.ACC

Other modals, for instance, szabad ‘may’ in a deontic use, or tud ‘can’ in its circumstantial and deontic uses are compatible with the modal affix -hAt in its relevant meanings. Similarly, there is only a stylistic restriction concerning the interchangeability of the modal kell ‘must’ and the predicative modal particle muszáj ‘must’ (for details, see Kiefer 2005: 52-111). (19) a. Ki PREF

szabad may

menned a go.INF.POSS.2SG the

kertbe. garden.ILLAT

b. Kimehetsz a kertbe. PREF.go.MOD.AFFIX.2SG the garden.ILLAT ‘You may (=are allowed to) go out to the garden.’ (20) a. ErĘs vagyok, strong be.1SG

el PREF

is gyalogolhatok too walk.MOD.AFFIX.1SG

odáig. there.TERM

b. ErĘs vagyok, el is tudok gyalogolni odáig. strong be.1SG PREF too can.1SG walk.INF there.TERM ‘I am strong, I can even walk there.’ etc. The examples for the interchangeability of a paradigmatic item (the modal affix -hAt) and certain modal verbs (tud ‘can’, szabad ‘may’), as well as that of the modal particle muszáj ‘must’ and the modal verb kell ‘must’ suggest that the compatibility of these modal elements is clearly semantic and is not related to paradigmatic variability, not even in a very broad sense.

414

Erika Körtvély

4.4. Parameter Scope (condensation) To clear up the condensation level of the Hungarian modals, two characteristic features connected to the problem of structural scope will be considered here, namely, the ability of giving arguments and the ability of being semantic predicates. The first question is aimed to show the condensation level of the modals investigated, while the other targets the semantic aspect of structural scope. If the infinitive part of a modal construction can be transformed to another (not infinite verbal) complement without the loss of the original meaning of the expression, it can be considered that the inflected verb functions as a full verb (at the clause level) rather than an auxiliary (at the VP level). In the examples (21a-h), the infinitival parts of the modal constructions are changed to complement clauses. As the examples show, kell ‘must’, szabad ‘may’, tud ‘can’ bír ‘can’, akar ‘want’ and szeretne ‘would like’ are acceptable with non-infinitival complements as well, while the verbs lehet ‘be possible’ and talál ‘might’ are not. (21) a. Haza kell mennem. ~ Az kell, hogy home must go.INF.POSS.1SG that must that hazamenjek. home.go.SBJ.1SG ‘I have to go home.’ ~ ‘It is necessary that I go home.’ b. Meg szabad enni a tortát. ~ Szabad (az), hogy PREF may eat.INF the cake.ACC may that that megegyék a tortát. PREF.eat.SBJ.3PL the cake.ACC ‘The cake may be eaten.’ ~ ‘It is allowed for the cake to be eaten.’ c. Innen jól lehet látni a hegyeket. from here well is possible see.INF the hill.PL.ACC zLehet (az), hogy innen jól látják a hegyeket. 8 may that that from here well see.3PL the hill.PL.ACC ‘From here the hills can be seen well.’ d. Egyszerre tudok énekelni és táncolni. ~ Tudom simultanelously can.1SG sing.INF and dance.INF can.DFC.1SG

Modals in Hungarian

(azt), hogy egyszerre énekelek és that.ACC that simultaneously sing.1SG and ‘I can sing and dance simultaneously.’

415

táncolok. dance.1SG

e. Bírok vinni két táskát is. ~ Bírom (azt), can.1SG carry.INF two bag.ACC too can.DFC.1SG that.ACC hogy két táskát is vigyek. that two bag.ACC too carry.SBJ.1SG ‘I can carry even two bags.’ f. Ne piszkáld a bombát, mert fel talál no poke at.SBJ.2SG the bomb.ACC because PREF might robbanni. ~ *Ne piszkáld a bombát, blow up.INF no poke at.SBJ.2SG the bomb.ACC mert azt találja, hogy felrobban. because that.ACC might.DFC.3SG that blow up.3SG ‘Don’t poke at the bomb, because it might blow up.’ g. Táncolni akarok. ~ dance.INF want.1SG táncoljak. dance.SBJ.1SG ‘I want to dance.’

Azt akarom, hogy that.ACC want.DFC.1SG that

h. Énekelni szeretnék. ~ Szeretném, hogy énekeljek. sing.INF would like.1SG would like.DFC.SG that sing.SBJ.1SG ‘I would like to sing.’ ~ I would like that, that I sing.’ As the examples show, talál ‘might’ and lehet ‘be possible’ do not function at the clause level, so their structural scope is lower than that of the other modal verbs. The ability of being semantic predicates is functionally equivalent to the ability of assigning thematic roles to the main parts of the sentence such as, for instance, the subject. To illustrate the ability of giving thematic roles (e.g a subject role) I use weather-verb constructions, which are typically subjectless in Hungarian. Thus, modal verbs, which are not grammatical in combination with weather-verbs, do not avoid a subject, i.e. they are able to give thematic roles and can be semantic predicates of a sentence (cf. Kenesei 2000: 109-110.). As the examples below show, kell ‘must’, szabad ‘may’,

416

Erika Körtvély

lehet ‘be possible’, tud ‘can’, bír ‘can’, akar ‘want’ and szeretne ‘would like’ can occur as semantic predicates of a sentence, while talál ‘might’ cannot. (22) a. be

talál esteledni (Kenesei 2000: 110) might.3SG evening.fall.INF ‘the evening might fall’ PREF

b. *be

tud / bír esteledni (Kenesei 2000: 110) can.3SG can 3SG evening.fall.INF ‘evening can fall’ PREF

kell / szabad / lehet / akar 9 / szeretne PREF must may is possible want.3SG would like.3SG esteledni(e) evening fall.INF(POSS.3SG) ‘evening has to / may / is possible to / wants to / would like to fall’

c. *be

As the data suggest, the structural scope of talál ‘might’ is the lowest. It is followed by lehet ‘be possible’, and the scope of kell ‘must’, szabad ‘may’, tud ‘can’, bír ‘can’, akar ‘want’ and szeretne ‘would like’ is almost as high as that of a full verb. 4.5. Parameter Bondedness (coalescence) The syntagmatic cohesion between a modal and its main verb is usually indicated by simple juxtaposition, i.e. the level of the coalescence is relatively low. In this section the syntactic and semantic freedom of the Hungarian modals will be discussed. One of the most important syntactic symptoms indicating coalescence is the inseparability of the parts of a certain grammatical structure in an elliptic construction. As is clear from the below examples, in Hungarian all of the modal verbs in scope can be elided: (23) a. Mosogatnom és do the washing up.INF.POSS.1SG and ‘I have to do the dishes and clean.’

takarítanom kell. clean.INF.POSS.SG must

Modals in Hungarian

b. Szabad ennünk és may eat.INF.POSS.1PL and ‘We may eat and drink.’

innunk. drink.INF.POSS.1PL

c. Lehet látni és is possible see.INF and ‘One can see and hear.’

hallani. hear.INF

d. Tudok írni és can.1SG write.INF and ‘I can read and write.’

olvasni. read.INF

e. Bírok futni és can.1SG run.INF and ‘I can run and swim.’

417

úszni. swim.INF

f. Vigyázz, a bogár (még) meg talál watch out.SBJ.2SG the bug PTL PREF might.3SG csípni vagy harapni. bite.INF sting.INF or ‘Watch out, the bug might sting or bite you.’ g. Énekelni és táncolni sing.INF and dance.INF ‘I want to sing and dance.’

akarok. want.1SG

h. Szeretnék egy kicsit ülni és would like.1SG a little.ACC sit.INF and ‘I would like to sit a little and rest a little.’

pihenni. have a rest.INF

Another relevant criterion to show the syntactic bondedness of grammatical elements is the possibility of inserting material between their parts (cf. Lehmann 2002: 134). In Hungarian, as was already mentioned above, the parts of a modal construction are usually in juxtaposition. If the main verb is not prefixed, the modal verb appears beside the main verb, and if it is, the modal occurs between the prefix and the main verb. (The sequence of the modal construction will be discussed in more detail in 4.6. below.) It has to be mentioned that a modal construction, similar to an auxiliary construction, usually has only one incidence of stress which falls on the main verb. If there

418

Erika Körtvély

is a prefixed infinitive in the construction, the prefix and the modal verb have a common accent on the prefix. Because the syntactic behavior of modal structures with a prefixed or not prefixed main verb is different, they will be shown separately below. In neutral sentences, the modal verbs kell ‘must’, lehet ‘be possible’, 10 and talál ‘might’ cannot be separated from their prefixless main verbs. (24) a. Ma takarítanom kell. : * Takarítanom today clean.INF.POSS.1SG must clean up.INF.POSS.1SG ma kell. (z Takarítanom ma KELL. ‘I MUST clean today.’) today must ‘I have to clean today.’ b. Az ablakból látni lehet a tornyot. : the window.ABL see.INF is possible the tower.ACC *Az ablakból látni a tornyot lehet. the window.ABL see.INF the tower.ACC is possible ‘Out of the window one can see the tower.’ c. Ha a bomba robbani talál a bokorban, if the bomb blow up.INF might.3SG the bush.INE fuss. : *Ha a bomba robbanni a bokorban run.SBJ.2SG if the bomb blow up.INF the bush.INE talál, fuss. might.3SG run.SBJ.2SG ‘If the bomb happens to blow up in the bush, run.’ It has to be noted that in the case of talál ‘might’ the separation of the parts of the modal construction, unlike the other modals, does not lead to a focused sentence, but to an ungrammatical sentence. The modals tud ‘can’ and bír ‘can’ show variation in their syntactic behaviour. They are acceptable as modals both as unstressed (more auxiliarylike) parts of a modal construction and also as more independent verbs that bear their own stress and have a relative positional freedom in the sentence (see 27 a-b). In their unstressed use they are not separable from their main verb: (25) a. A könyvbĘl the book.ABL

minden nap every day

olvasni tudok egy kicsit. : read.INF can.1SG a little.ACC

419

Modals in Hungarian

*A könyvbĘl minden nap olvasni egy the book.ABL every day read.INF a ‘I can read a little from the book every day.’

kicsit tudok. little.ACC can.1SG

b. A táskát egész nap cipelni bírom. : *A táskát the bag.ACC whole day carry.INF can.DFC.1SG the bag.ACC cipelni egész nap bírom. carry.INF whole day can.DFC.1SG ‘I am able to carry the bag all day.’ Szabad ‘may’, akar ‘want’, and szeretne ‘would like’ seem to be the most unbound modals in Hungarian. Although they show similarities to the behaviour of tud ‘can’ and bír ‘can’, they are more natural in sentences where they have their own stress and do not follow the infinitive. However, in '[INF+MODAL] constructions, just like in the cases of the other modals, it is not possible to insert material between them and their main verb. (26) a. Tegnap jó volt, mert játszani szabadott a Yesterday good be.PST.3SG because play.INF may.PST the kertben. : *Tegnap jó volt, mert játszani garden.INE yesterday good be.PST.3SG because play.INF a kertben szabadott. the garden.INE may.PST ‘Yesterday it was nice, because playing in the garden was allowed.’ b. Játszani akartam a kertben. : play.INF want.PST.1SG the garden.INE kertben akartam. want.PST.1SG garden.INE ‘I wanted to play in the garden.’ c. Játszani szeretnék a kertben. : play.INF would like.1SG the garden.INE kertben szeretnék. would like.1SG garden.INE ‘I would like to play in the garden.’

*Játszani play.INF

*Játszani play.INF

a the

a the

420

Erika Körtvély

If the infinitive and the modal are acceptable with separate stresses, it is also acceptable to insert linguistic material between them: (27) a. A 'könyvbĘl 'minden nap 'tudok egy 'kicsit 'olvasni. the book.ABL every day can.1SG a little.ACC read.INF ‘I can (I can manage to) read a little from the book every day.’ b. A 'táskát 'egész nap 'bírom 'egy kézze 'cipelni. the bag.ACC whole day can.1SG one hand.INSTR carry.INF ‘I can carry the bag with one hand all day.’ c. 'Ma 'szaba 'csokit 'enned. today may chocolate.ACC eat.INF.POSS.2SG ‘Today you may eat chocolate.’ d. A 'könyvbĘl 'minden nap 'akarok 'pár oldalt the book.ABL every day want.1SG some page.PL ‘I want to read some pages of the book every day.’

'olvasni. read.INF

e. 'Szeretnék a 'Scalában egy' áriát 'énekelni. would like.1SG the Scala.INE an aria.ACC sing.INF ‘I would like to sing an aria in the Scala.’ As is clear from the examples, kell ‘must’ and talál ‘might’ cannot be separated from their main verb, and, as modals, they cannot bear stress in neutral sentences. Thus, they form a tighter grammatical unit with their main verb than the other modal verbs. Szabad ‘may’ lehet ‘be possible’, tud ‘can’ bír ‘can’, akar ‘want’ and szeretne ‘would like’ can have also their own stress as modals in the sentence, and in this use they accept complements inserted between the modal and the main verb. In modal + prefixed main verb constructions, the syntactic behaviour of the modal verbs is more uniform. If the main verb has a prefix, the prefix cannot be separated from the modal verb even in non-neutral sentences. (The prefixless main verbs of these modal constructions can appear anywhere in the sentence, after the prefix and the modal verb, without the loss of the original meaning). (28) a. Ma ki kell takarítanom today PREF must clean.INF.POSS.1SG

a lakást. | Ma the flat.ACC today

421

Modals in Hungarian

ki

kell a lakást takarítanom. must the flat.ACC clean.INF.POSS.1SG ‘Today I have to clean the flat.’ PREF

b. Ma meg szabad enned a today PREF may eat.INF.POSS.2SG the | Ma meg szabad a csokit today PREF may the chocolate.ACC ‘Today you may eat the chocolate.’

csokit. chocolate.ACC enned. eat.INF.POSS.2SG

c. Az ablakból meg lehet nézni a tornyot. | the window.ABL PREF is possible look.INF the tower.ACC Az ablakból meg lehet a tornyot nézni. the window.ABL PREF is possible the tower.ACC look.INF ‘Out of the window one can look at the tower.’ d. Ma el tudom olvasni a könyvet. | today PREF can.DFC.1SG read.INF the book.ACC el tudom a könyvet olvasni. PREF can.1SG the book.ACC read.INF ‘Today I can read (finish) the book.’

Ma today

bírom emelni a táskát. | Fel bírom PREF can.DFC.1SG lift.INF the bag.ACC PREF can.DFC.1SG a táskát emelni. the bag.ACC lift.INF ‘I can lift the bag.’

e. Fel

f. Vigyázz, mert még meg talál watch out.SBJ.2SG because PTL PREF might.3SG az a bogár. | […] még meg talál that the bug PTL PREF might.3SG a bogár csípni. the bug bite.INF ‘Watch out, because that bug might bite you.’

csípni bite.INF az that

g. Holnapra el akarom olvasni a tomorrow.SUBLAT PREF want.DFC.1SG read.INF the el akarom a könyvet. | Holnapra

422

Erika Körtvély

book.ACC tomorrow.SUBLAT PREF want.DFC.1SG the könyvet olvasni book.ACC read.INF ‘I want to read the book by tomorrow.’ szeretnék énekelni egy áriát. | would like.1SG sing.INF an aria.ACC szeretnék egy áriát énekelni. would like.1SG an aria.ACC sing.INF ‘I would like to sing (a whole) aria.’

h. El

PREF

El PREF

In these sentences, the prefix and the modal verb tend to form a phonological unit, and their separation leads to ungrammatical constructions: (29) a. *Ma ki a lakást kell takarítanom. today PREF the flat.ACC must clean up.INF.POSS.1SG ‘I have to clean the flat today.’ (z Ma ki a LAKÁST kell takarítanom. ‘It is the FLAT that I have to clean today.’) b. * Vigyázz, mert még meg az a bogár watch out.INF.SBJ.2SG because PTL PREF that the bug talál csípni. might.3SG bite.INF ‘Watch out, because that bug might bite you.’ etc. On the basis of the examples above, it can be seen that those verbs which are separable from their main verbs are more independent, i.e. less grammaticalized than those which are grammatical only on the right side of the main verb. In case of the prefixed main verbs the prefix tends to represent the main verb of the construction, and, differently from the non-prefixed structures, the prefix has to be immediately followed by the modal verb. The behaviour of the modals in scope suggests again that talál ‘might’ is the most auxiliary-like modal in Hungarian, kell ‘must’ is less dependent, and the others are, as far as the syntactic criteria of coalescence are concerned, relatively free. The investigation of a semantic criterion of bondedness, namely synsemanticity vs. autosemanticity (cf. Lehmann 2002: 139), leads to a very similar conclusion. The modals kell ‘must’ szabad ‘may’, lehet ‘be possible’,

Modals in Hungarian

423

tud ‘can’, bír ‘can’ akar ‘want’ and szeretne ‘would like’ can be semantic and pragmatic foci and they can also forms independent clauses, while talál ‘might’ is not acceptable in this role: (30) a. Kell ma boltba menned? – Kell. must today shop.ILLAT go.INF.POSS.2SG must ‘Do you have to go to shopping today? – (Yes,) I have to.’ b. Szabad itt dohányozni? – Szabad. May here smoke.INF may ‘Is it allowed to smoke here? (Yes,) it is.’ c. Lehet látni a hegyeket? is possible see.INF the hill.PL.ACC ‘Can the hills be seen? – (Yes,) they can.’

– Lehet. is possible

d. Tudod olvasni az írásom? can.DFC.2SG read.INF the writing.POSS.1SG ‘Can you read my writing? – (Yes,) I can.’ e. Bírod cipelni a táskám? can.DFC.2SG carry.INF the bag.POSS.1SG ‘Can you carry my bag? – (Yes,) I can.’

– Tudom. can.DFC.1SG

– Bírom. can.DFC.1SG

f. *Talál robbanni a bomba? (*Robbanni might.3SG blow up.INF the bomb blow up.INF talál a bomba? – *Talál.) might.3SG might.3SG the bomb g. Akarsz kezet mosni? – Akarok. want.2SG hand.ACC wash.INF want.1SG ‘Do you want to wash your hands? – (Yes,) I do.’ h. Szeretnél énekelni? – Szeretnék. would like.2SG sing.INF would like.1SG ‘Would you like to sing? – (Yes,) I would.’ In modal constructions formed with prefixed main verbs the build-up of the above structures is, again, different. Beside prefixed verbs the modals cannot

424

Erika Körtvély

be focused and in answers the prefixes have to be present in the answer clauses. (31) El

kell menned must go.INF.POSS.2SG El kell. (: *Kell) PREF must must etc. PREF

boltba? (*Kell elmenned boltba?) shop.ILLAT

The above examples show that although kell ‘must’ szabad ‘may’, lehet ‘be possible’, tud ‘can’, bír ‘can’, akar ‘want’ and szeretne ‘would like’ as modals are usually combined with autosemantic main verbs in infinitive, in some contexts they are able to function also without their main verb. Thus, they cannot be considered as purely synsemantic signs. The level of semantic dependency is the highest in the case of talál ‘might’ which, as a modal, is not grammatical without its main verb. 4.6. Parameter Syntagmatic Variability (fixation) The syntagmatic variability of the modal constructions has been partly discussed in 4.5 in connection with the condition of inseparability of the elements. As could be seen, in some modal constructions the parts can be separated from each other, while it is impossible in other cases. It also has been shown that the prefixes of the infinitive main verbs form a relatively tight grammatical unit with the modal verbs: the modals always occur (in juxtaposition) on their right side, they share stress with them (placed on the prefix) and they can be focused (in a semantic and pragmatic sense) only together with the prefixes. In this section first the internal positional freedom of the elements of a modal construction will be considered, and second, the position of the elements in the sentence will be investigated. The word order in Hungarian is usually analyzed as being extremely free. However, some syntagms have their own stable internal structure (for instance, the unmarked possessive construction appears always as Possessor – Possessum, and the sequence of an unmarked attributive construction is obligatorily Attributive – Attribute Bearer, etc.). As it was clear from the examples above, in neutral sentences the sequence of the modal verb and the infinitive is usually fixed. Non-prefixed modal constructions have the 'main verb + modal word order and if the main verb of the construction has a

Modals in Hungarian

425

prefix, the sequence 'prefix + modal (…) + 'main verb is grammatical. However, the modals of Hungarian do not form a unified group as far as the level of fixation is concerned either, and some of them are well formed in a 'modal + '(not prefixed) main verb construction as well (see examples 27. ae). The modals kell ‘must’ and talál ‘might’ are grammatical only if they follow the above described sequence, but the other modals in scope are acceptable also before their main verbs. However, if the main verb has a prefix, the sequence of the modal and the prefix is always fixed: the modals have to follow the prefix and be unstressed (see examples 28-29). The position of the main verb separated from the prefix is relatively free in the sentence. As it could be seen in section 4.5, the main verb can be moved apart from the prefix + modal unit. However, in neutral sentences the positioning of the main verb before the prefix and the modal is not acceptable: (32) a. *Mennem el kell a piacra. go.INF.POSS.1SG PREF must the market.SUBLAT ‘I have to go to the market.’ b. *Menni a kertbe ki go.INF the garden.ILLAT PREF ‘One may go out in the garden.’

szabad. may

c. *Az ajtót nyitni ki the door.ACC open.INF PREF ‘It is possible to open the door.’

lehet. is possible

d. *A könyvet olvasni el the book.ACC read.INF PREF ‘I can read the book.’

tudom. can.1SG

e. *Emelni a táskát fel lift.INF the bag.ACC PREF ‘You can lift the bag.’

bírod. can.2SG

f. *A bogár csípni meg the bug bite.INF PREF ‘The bug might bite you.’

talál. might.3SG

426

Erika Körtvély

g. *Mosni a kezem meg akarom. wash.INF the hand.POSS.1SG PREF want.DFC.1SG ‘I want to wash my hands.’ h. *Énekelni egy dalt el sing.INF a song.ACC PREF ‘I would like to sing a song.’

szeretnék. would like.1SG

Thus, as the examples demonstrate, the position of Hungarian modals can be considered relatively fixed. The syntagmatically most invariable modals are kell ‘must’ and talál ‘might’, while in the cases of the other modals some minor variability can be observed. 5. Conclusion As was seen above, modals are important but not exclusive linguistic tools of modality in Hungarian. Although they all express modality, modals do not form a consistent group of verbs, neither formally nor semantically: some of them can be considered as more dependent elements, and some of them as less dependent elements. Both formal and semantic criteria show that their grammaticalization status is not identical to each other, either. Having investigated the most central modals of Hungarian it can be stated that the modal verb which is grammaticalized the most is talál ‘might’: it is both syntactically and semantically dependent on its main verb and, as such, can be considered a modal auxiliary. It is followed by the less auxiliary-like kell ‘must’ and the other verbs (lehet ‘be possible’, tud ‘can’, bír ‘can’, akar ‘want’, szabad ‘may’ and szeretne ‘would like’). These latter verbs can be categorized, in Lehmann’s terms (2002: 32), as modal verbs rather than modal auxiliaries. As has been shown above, too, the group of these modals is held together only by their modal semantics and the ability of forming (a looser or tighter) grammatical unit with an infinitival main verb. Modals in Hungarian, unlike, for instance, German or English modals, do not build a formally and semantically well recognizable modal subparadigm in the verbal paradigm, but seem to be (with the exception of talál ‘might’ and perhaps kell ‘must’) relatively free modal elements at the very beginning of a possible grammaticalization process.

Modals in Hungarian

427

Notes 1. 2. 3. 4.

5.

6.

7. 8.

For the sake of brevity, the indefinite verbal agreement suffixes are understood as defaults. Where verbs are not glossed as DFC (definite) or i (-ik: medial), they are always indefinite. The verb szabad ’may’ is, as a matter of fact, an adjective with a defective verbal paradigm. The verb szeretne ‘would like’ is the regular conditional form of szeret ‘like’ used as a modal. Following the traditional categorization of Hungarian particles (non-inflecting parts of speech including interjections, answer particles and modal elements which can occur as non-organic and unstressed elements in the sentence, cf. Kugler 2001: 240), the words biztos ’surely’ and talán ’maybe’ are considered here as adverbs rather than particles. However, their shorter forms or modified forms, such as bizony ’surely, indeed’ and tán ’maybe’ are mentioned as particles below. The modal affix -hAt, as shown in 3.1, indicates dynamic and epistemic possibilities. The verb lesz ’will3SG’, however, can also express epistemic modality on its own in some contexts: Hol a ruhám? - Ott lesz a szekrényben. ‘Where is my dress? - It is surely in the wardrobe / It must be in the wardrobe.’ In the case of kell ’must’ the statement is valid only for standard Hungarian. In some dialects both the subjective and objective verbal agreement suffixes on kell are grammatical in present indicative (el kellek (/köllök) menni ’I have to go’) and also after the past tense marker (el kellettem (/köllöttem) menni ’I had to go’). However, there are no examples for modal usage of kell ’must’ where a mood marker is followed by a verbal agreement suffix. (It is attested in periphrastic constructions for past conditional such as, for instance: meg köllötted volna hallanod PREF must.PRF.o.2SG AUX.MOD hear.INF.2SG ’You should have heard it.’ For details see (ÚMTSZ 3: 179) The impossibility of attaching further mood markers on szeretne ‘would like’ is not dealt with here because it is due to the morphological build-up of the verb and not to its modal use. The sentence with the non-infinitival complement is grammatical because in this context (3SG form of the verb + subordinator) the verb lehet ’may, can’ is formally identical with the predicative part of the modal tag lehet, hogy ’maybe’ shown in section 3.3. However, the semantic difference between the clauses in (21.c) is more salient in the past tense. The sentence Innen jól lehetett látni a hegyeket ’From here the hills could be seen well.’ cannot be transformed to *lehetett (az), hogy innen jól látják / látták a hegyeket.

428

Erika Körtvély

9.

The verb akar ‘want’ in a special, quasi-evidential use is grammatical also in this context: be akar esteledni ‘it seems/feels so that the evening will fall shortly.’ In this case the verb does not express volition but it refers only to that fact that an event will obviously take place. 10. Kell ’must’ and lehet ‘be possible’ in other positions than on the right side of the main verb indicate focus.

References BenkĘ, Loránd (ed.) 1970: A magyar nyelv történeti-etimológiai szótára. Második kötet. [Etymological dictionary of Hungarian] Budapest: Akadémiai Kiadó. Bybee, Joan, Revere Perkins and William Pagliuca 1994 The evolution of grammar. Tense, aspect, and modality in the languages of the world. Chicago: Chicago University Press. D. Bartha, Katalin 1991 Az igeképzés. In A magyar nyelv történeti nyelvtana 1. Az ómagyar kor és elĘzményei, [A historical grammar of Hungarian] E. Abaffy Erzsébet (ed.), 60-103. Budapest: Akadémiai Kiadó. de Haan, Ferdinand 2006 Typological approaches to modality. In The expression of modality, William Frawley (ed.), 27-71. Berlin: de Gruyter, Hopper, Paul J. and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Horváth, Katalin 1999 Grammatikalizáció és szófejtés. Hozzászólás -hat/-het képzĘnk keletkezéséhez. In Ember és nyelv. Tanulmánykötet Keszler Borbála tiszteletére [Grammaticalization and etymology], Nóra Kugler, and Klára Lengyel (eds.), 160-164. Budapest: ELTE BTK. Kálmán, C. György, Kálmán László, Nádasdy Ádám, and Prószéky Gábor 1989 A magyar segédigék rendszere [The system of the auxiliaries in Hungarian]. Általános Nyelvészeti Tanulmányok 17: 49-103. Kenesei, István 2000 A szófajok [Parts of speech]. In Strukturális magyar nyelvtan 3. Morfológia. Ferenc Kiefer (ed.), 75-136. Budapest: Akadémiai Kiadó. Keszler, Borbála (ed.) 2000 Magyar grammatika [Hungarian grammar]. Budapest: Nemzeti Tankönyvkiadó Kiefer, Ferenc 1981 What is possible in Hungarian? Acta Linguistica Hungarica 31: 147185.

Modals in Hungarian

429

Kiefer, Ferenc 1984 A tud ige jelentéstanáról [On the semantics of the verb tud]. Magyar Nyelv 80: 144-160. Kiefer, Ferenc 1985 A -hat/-het képzĘ jelentéstanához. Az episztemikus -hat/-het [On the semantics of the affix -hat/-het. Epistemic -hat/-het]. Általános Nyelvészeti Tanulmányok 16: 131-153. Kiefer, Ferenc 1987 On defining modality. Folia Linguistica 21: 67-94. Kiefer, Ferenc 1994 Modality. In The encyclopedia of Language and Linguistics. Ronald E. Asher (ed.), 2525-2528. Oxford: Pergamon Press Kiefer, Ferenc 2005 LehetĘség és szükségszerĦség. Tanulmányok a nyelvi modalitás körébĘl [Possibility and necessity. Essays on modality.]. Budapest: Tinta Könyvkiadó. Komlósy, András 1992 Régensek és vonzatok [Predicates and arguments]. In Strukturális magyar nyelvtan 1. Szintaxis, Ferenc Kiefer (ed), 299-527. Budapest: Akadémiai Kiadó. Kugler, Nóra 2001 Próbák és szempontok a módosítószók elhatárolásához [Tests and considerations to the classification of modal words]. Magyar NyelvĘr 125: 233-241. Laczkó, Tibor 2000 Zárójelezési paradoxonok [Bracketing paradoxes]. In Strukturális magyar nyelvtan 3. Morfológia, Ferenc Kiefer (ed.), 619-761. Budapest: Akadémiai Kiadó. Lehmann, Christian 2002 Thoughts on grammaticalization. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. M. Korchmáros, Valéria 1997 Ige vagy segédige? [Verb or auxiliary?]. In Nyíri Antal kilencvenéves, László Büky (ed.), 109-124. Szeged: JATE Ross, John R. 1969 Auxiliaries as main verbs. In Studies in Philosophical Linguistics. Series One, William Todd (ed.), 77-102. Evanston: Great Expectations Press ÚMTSZ= B. LĘrinczy Éva (ed.) 1992 Új Magyar Tájszótár. Harmadik kötet [A new Hungarian dialect dictionary]. Budapest: Akadémiai Kiadó

430

Erika Körtvély

van der Auwera, Johan 1999 On the Semantic and Pragmatic Polyfunctionality of Modal Verbs. In The semantics/pragmatics interface from different points of view, K. Turner (ed.), 50-64. Oxford: Elsevier van der Auwera, Johan and Andreas Ammann with Saskia Kindt 2005 Modal polyfunctionality and Standard Average European. In Modaity. Studies in Form and Function, Alex Klinge, and Henrik Høeg Müller (eds.), 247–272. London: Equinox. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 1 (2): 79-124.

13. Mood and modality in Berber Amina Mettouchi

1. Introduction As the title of our chapter suggests, less will be said here about modals1 than about mood and modality in Berber. Indeed, in Berber, there are no modals such as those that can be found in English for instance. Instead, particles and aspect-mood markers inside the verb stem combine to convey the whole range of meanings that other languages express through modal auxiliaries or periphrastic expressions. V1-V2 structures expressing ability, volition and sometimes obligation, possibility and probability do exist, but they are elaborations on basic “particle + TAM-marked stems” structures. Our aim in this chapter is to show that this organisation of the modal system in Berber is linked to the fundamental non-temporal and deictic feature of those languages, and that the core of the modal domain revolves around the position of the speaker, with respect to whom situations manifest themselves as real or non-real. This organisation entails that the dimension of mood, in Bhat’s (1999: 63) sense (“[m]ood is concerned with the actuality of an event”) is dominant in the verbal system of Berber. This dimension has mostly been labelled “aspectual” by berberologists and afroasiaticists, but with a definition of aspect as “viewpoint aspect” that is not the commonly accepted use of the term: a focus on boundaries and intervals. The chapter is organised as follows: first we will make a few remarks on the framework of the study and the terminology used, then we will present the relevant facts about Berber morphosyntax and give a typological profile. We will then study in detail the system of a Northern variety of Berber, Taqbaylit (Kabyle). After having characterised the Kabyle system, we will present the form-function mapping of modality in other varieties of Berber, and show that although there is considerable variation, the main building block of mood and modality remains the preverbed aorist. We will conclude by claiming that the deictic nature of the irrealis particle points to a particular grammaticalisation process that does not follow the same paths

432

Amina Mettouchi

as Romance and Germanic, where the system is rooted in agent-oriented predications. 2. Preliminary remarks Bybee et al. (1994: 176) propose that grams indicating “obligation, probability, and possibility” as well as “ability and desire” are associated with modality, whereas “imperative, optative, conditional, and subordinate verb forms” are associated with mood. We will see that the forms which pertain to the domain of mood in Bybee’s sense of the term can express modality in Berber. Bhat’s less restrictive definition of mood seems to better account for Berber data (1999: 63): “There are three different parameters that are used by languages while establishing modal distinctions; these are the following: (i) a speaker’s opinion or judgement regarding the actuality of an event, (ii) kind of evidence that is available for the speaker to form this judgement, and (iii) kind of need or requirement which forces the speaker (or someone else) to get involved in an event (or to carry out an action). The first two parameters establish ‘epistemic’ (knowledge-based) moods and the third one establishes ‘deontic’ (action-based) moods …”

Berber is a language family belonging to the Afroasiatic phylum. Berber languages are spoken in the North of Africa, from Egypt to the Atlantic ocean, and from the Mediterranean to Burkina-Faso.2 In Berber, the verbal stem is completed by an obligatory personal affix which refers to the main participant of the event or state. The following examples are from Kabyle, but this morphological characteristic is valid for the whole of Berber. ye-þþa : SBJ.3M.S-eat[PFV] : ‘he ate/has eaten’ basic utterance = personal affix : ye-+ stem : þþa stem = root (þþ) + aspectual scheme (here perfective, realised -a). The verb can be followed by clitics (dative and accusative), and a proximal or distal particle. The order of clitics is rigid.

Mood and modality in Berber

(1)

433

taqcict te-fka yas ten iD. girl SBJ.3F.SG-find[PFV] DAT.3SG ACC.3M.PL PROX ‘The girl gave them to him/her.’

TAM3 configurations are based on apophonic distinctions, or gemination or prefixation of the verbal stems. The systems are quite varied, depending on the language. For instance Kabyle (Table 1) only has four basic themes, whereas Aïr Tuareg (Table 2) has six. Table 1. Taqbaylit aspectual bases (roots DD ‘come’, and KRZ, ‘plough’) Aorist

Perfective

Negative perfective

Imperfective

-ddu-

-dda-

-ddi-

-ĠĠΩddu-

-krΩz-

-krΩz-

-kriz-

-kΩrrΩz-

Table 2. Aïr Tuareg aspectual bases (roots RTK, ‘fall’, and G, ‘do’)4 Aorist

Perfective

Perfect

Neg. Perfective

Imperfective

Neg. Imperfective

-rtΩk-

-rtak-

-rtaak-

-rtek-

-raattΩk-

-rΩttΩk-

-g(u)-4

-ge/a/Ω-

-gee/aa-

-ge/a/Ω-

-taagg(u)-

-tΩgg(u)-

The proto-Berber system, as reconstructed by Cohen (1989: 155), was originally based on an opposition between what is now the aorist (A), and what is now the perfective (B). Secondary distinctions appear with the imperfective (geminated, level II), the drift of the aorist into the modal domain (level III), and the introduction of a concomitance particle lla. This reconstruction is tentative, and does not take into account the negative themes, especially the negative perfective, which is pan-Berber, but it gives an idea of one of the possibilities of evolution of the system.

434

Amina Mettouchi

Table 3. Reconstruction of the Proto-Berber aspectual system (Cohen 1989: 155), with root GM ‘draw water from a well.’5 A ı

Į

ȕ

I

Į

y-agem

II II I I V

B

y-agem y-agem

y-ugem

ye-ttagem

y-ugem

ye-ttagem

y-agem

ye-ttagem

lla ye-ttagem

V

y-agem

ye-ttagem

lla ye-ttagem

V I

y-agem ye-ttagem

lla ye-ttagem

ȕ

y-ugem y-ugem y-ugem

lla ye-ttagem

y-ugem

lla ye-ttagem

3. A detailed survey of Kabyle (Taqbaylit) Kabyle is spoken in the North of Algeria (and in some big cities like Algiers, alongside Arabic) by about four million speakers, to which another million of emigrant speakers in Europe and Canada can be added. The region is quite small but very densely populated. It is mountainous, and is separated into two dialectal areas: the Western area, and the Eastern one. Our examples are from Western Kabyle. 3.1. Typological profile of Kabyle Predicates can be verbal or non-verbal, but our focus in this chapter will be on verbal predication. The verbal system is dominated by mood and aspect, which are marked thanks to vocalic alternations, prefixation or consonant gemination. Preverbal particles are used to modify the basic moodaspectual schemes and provide specific modal and aspectual values (irrealis, progressive). As far as the noun phrase is concerned, there are no articles, and noun phrases are either in the citation (unmarked) form — the “free state” — or in the “annexed state” (marked). Independent pronouns form a special paradigm, with no state alternations. The possessor always

Mood and modality in Berber

435

follows the possessee, and adjectives are placed after nouns. Word order at the clause level is quite flexible (information-structure dependent) but Kabyle is clearly a head-initial language. All verbs are finite. The verbal system is asymmetrical with respect to negation (Mettouchi 2006): the aorist can only be negated in the optative, and there is a negative perfective as the counterpart of the perfective. Moreover, the imperfective appears in negative modal contexts. This will be studied in detail below. 3.2. Mood and modality in positive utterances The two forms encountered in this domain are the unpreverbed aorist (imperative, hortative, bare aorist), and the preverbed aorist (also sometimes the preverbed imperfective). The (preverbed or unpreverbed) aorist is very frequently used: counts on data from conversational and narrative corpora show that it appears in almost 35% of occurrences. This is due to the fact that it has a wide range of uses, which will be discussed below. The survey of all uses (see also Mettouchi 2002) is necessary to understand the scope of the modal domain in Kabyle. 3.2.1. Unpreverbed aorist The aorist stem is used in Kabyle mainly for the formation of imperatives and hortatives, with special sets of personal affixes. The following table presents the imperative set of affixes, with the stem eþþ (‘eat’, aorist). Table 4. Imperative affixes 2 singular

2 plural masculine

2 plural feminine

eþþ

eþþ-et

eþþ-emt

In the following example, the wild cat (protagonist of a traditional oral tale) addresses things and objects in his house. (2)

Ȗli

d

a

SBJ.2SG.fall[IMP.AOR]

PROX

VOC

tiqi‫ܒ‬ tacninent / drop.FS light t-ernu-‫ڲ‬ a ta‫ۊ‬nunt ugercel. SBJ.2SG-add[AOR]-2SG VOC pancake.FS bran.AS ‘Fall, light drop, and you too, bran pancake.’

436

Amina Mettouchi

The unpreverbed aorist also, rarely, appears (with the standard set of affixes) in sequence with a perfective, an imperative or an optative, to link a series of actions done or to be done. This is the case with t-ernu-‫‘ ڲ‬you add’ in example (2) above, which follows an imperative, and therefore is interpreted as an order in this context. Imperatives can also be formed on the imperfective stem, conveying insistence on the desired involvement of the hearer. By contrast, the aorist imperative ‫ۊ‬ader ‘be careful’ in example (3) below would be simply oriented towards the realisation of the process. (3)

Ȗur-ek tt‫ۊ‬adar iman-ik at-2M.SG SBJ.2SG.take.care[IMP.IPFV] self-POSS.2M.SG ‘Beware, take good care of yourself.’

3.2.2. Preverbed aorist Most of the occurrences of the aorist in the data are preverbed with the particle ad, which derives from a proximal deictic marker. It is still used as a proximal demonstrative (‘this’) in some Berber languages (Shilha for instance), but not in Kabyle. The ad+aorist form covers a wide range of meanings, and can be found either in independent or main clauses, or in complement clauses. ad+aorist in main or independent clause In main or independent clauses, the value of ad+aorist is often that of a future statement: (4)

a

d

IRR

PROX

ini-Ȗ sin wawalen say[AOR]-SBJ.1SG two words.AS af ‘les partis’ yagi. on ‘les partis’ those ‘I’ll say a few words about those (political) parties.’

In this context, the imperfective stem can appear after the irrealis preverb instead of the aorist. This adds a conative or habitual dimension to the future, or underlines the Aktionsart (with activity-type verbs).

Mood and modality in Berber

437

The preverbed aorist also has optative uses, as in the following example: (5)

a

wen

i-barek ‫ۿ‬ebbi. IRR ACC.2M.PL SBJ.3M.SG-bless[AOR]-3M.SG God.AS ‘May God bless you.’

It is also the basis of the formation of hortatives, which in Kabyle are based on the aorist stem, circumfixed by the 2nd person plural imperative suffixes and the 1st person plural standard prefix, and preverbed by ad or its variant a. The following example shows how the hortative is used. (6)

iyya-wt a come[IMP.AOR]-SBJ.2M.PL IRR lwa‫ۊ‬id. the.one ‘Come and let’s walk as one.’

n-le‫ۊۊ‬u-t SBJ.1PL-walk[AOR]-2M.PL

But the ad+aorist form is not limited to future, hortative or optative (i.e. non-present and non-past contexts). It is very often found in narratives to describe a series of actions or deeds in a habitual framework. In the following example, while leading her sisters home, the elder sister regularly utters a magical formula in order to protect herself and her sisters from the lion. (7)

a

s

id

te-nni awal agi IRR DAT.3SG PROX SBJ.3F.SG-say[AOR] word.FS those ad ye-qqel ar deffir. IRR SBJ.3M.SG-return[AOR] to behind ‘She would tell him those words, and he (the lion) would walk back.’

In similar contexts, an imperfective or a perfective could be used, with a proper habitual frame in the preceding context. The ad+aorist form, contrary to the perfective or imperfective, implies that the series of actions is not seen a posteriori, or as characterizing the protagonist, but as it unfolds, materialises.

438

Amina Mettouchi

The preverbed aorist can also be used for recipes or instructions for use. In the following example a woman is explaining how to realise a piece of witchcraft: (8)

a

ten

te-zzi-‫ڲ‬

sebİa me‫܀܀‬at seven times i tfunast-im, a s t-ini-‫”…“ ڲ‬. to cow.AS-POSS.3SG IRR DAT.3SG SBJ.2SG-say[AOR] ‘You turn the thing seven times above your cow, and say “…”.’

IRR ACC.3M.PL SBJ.2S-turn[AOR]-2SG

When they describe a tendency of the topic, proverbs are very often in the ad+aorist form. In example (9), an imperfective could be used (see example (10)), but the effect is more descriptive. In example (9), the situation described is understood as everyone’s potentiality, as something that we are all bound to do, something that stems from our nature as human beings. The translation here uses (prosodically marked) will with precisely the same nuance it has in English: not the future meaning, but the idea that this is something that occurs quite naturally, in the course of things. (9)

kul yiwen ad i-tbaİ tajadit -is. all one IRR SBJ.3M.SG-follow[AOR] ancestry.FS-POSS.3SG ‘Everyone will (= is bound to) follow the example set by their forefathers.’

(10) kul yiwen ye-ttabaİ tajadit -is. all one SBJ.3M.SG-follow[IPFV] ancestry.FS-POSS.3SG ‘Everyone follows the example set by their forefathers.’ The ad+aorist form can therefore be characterised as marking the coming into existence or the advent of a situation or action, viewed as the outcome of a potentiality. ad+aorist in complement clauses It is in this part perhaps that the comparison with modals in the languages of Europe can be most fruitful. In most cases we have a V1-V2 structure, with V2 in the ad+aorist form. Given that all verbs have a personal affix,

Mood and modality in Berber

439

and that all verbs are marked for aspect-mood, we can always check the correspondence between the personal affix of V1, and that of V2. Unlike English where a variety of modals (might, should, would…) can appear in the complement clause, Kabyle, like most Romance languages, only features the ad+aorist form, or a perfective or imperfective. The ad+aorist form, which can be considered here to be analogous to the subjunctive mood of other languages, is sometimes obligatory, and sometimes it alternates with perfective or imperfective forms. When verbs expressing obligation and possibility are used, they are in the third person masculine singular, and almost always in the perfective. Their V2 is in the ad+aorist form, or sometimes in the ad+imperfective form, and bears the appropriate personal affix. It must be noted, however, that the frequency of use of those structures is quite low among monolinguals, and higher among bi- or multilinguals, which seems to point to borrowing and analogy rather than internal grammaticalisation. (11) i-laq a s d SBJ.3M.SG-miss[PFV] IRR DAT.3SG PROX awi-Ȗ tamacahutt swadda, ur SBJ.1SG-bring[AOR] story from-below NEG bȖi-Ȗ ara. SBJ.1SG-want[NEG.PFV] NEG ‘I had (/have) to tell them the story from the beginning, (and) I didn’t (/don’t) want to.’ The verb laq means ‘to miss’, ‘to want’ and is an Arabic loanword.6 It is only used impersonally and is similar to French falloir: ‘il fallait que je leur raconte l’histoire.’ It is always followed by the ad+aorist (or ad+imperfective) form. The 3rd person masculine prefix is a rare instance of impersonal meaning, personal affixes being almost always referential in Berber. This, and the fact that ad+aorist on its own can have a deontic meaning, may point to a calque of a Latin or Romance structure. The verb mken ‘be possible’7 is also an Arabic loanword, used impersonally in Kabyle, and is also perfectly integrated in the conjugation. It can be followed by ad+aorist when the realisation of the situation or action is possible. (12) ye-mken ad i-ruh. SBJ.3M.SG-let[PFV] IRR SBJ.3M.SG-go[AOR] ‘(It’s possible that) he might go.’

440

Amina Mettouchi

When the judgement of possibility concerns a realised event or situation it is followed by a verb in the perfective or imperfective. The verb mken is however not the most often heard. Instead, the adverbial waqila ‘maybe’ is almost always used in sentence-initial position (see example 13). This shows, in our opinion, that Kabyle prefers particles and adverbials to verbs for the expression of modal concepts. (13) waqila ad i-ruh ar Tizi Wezzu. maybe IRR SBJ.3M.SG-go[AOR] to Tizi Ouzou ‘(It’s possible that) he might go to Tizi Ouzou (city).’ The next type of probability is interesting, because V1 is itself in the ad+aorist form. It can be either the verb ili ‘be’, or the verbs af, ‘find’ or as, ‘arrive’, bearing the appropriate person affix. It is followed by a V2 in the perfective. The following dialogue is about the onset of the liberation war in Algeria. A young man is explaining how his father reacted to the news. (14) - I babak, ye-sla ? - and your.dad, SBJ.3M.SG-hear[PFV] ? - ‘what about your father, has he heard (about that) ?’ - ad y-ili ye-sla, ... - IRR SBJ.3M.SG-be[AOR] SBJ.3M.SG-hear[PFV], - ‘Probably, (lit. it will be he heard : he probably heard).’ It is possible to use the following expressions in lieu of ad y-ili ye-sla: ad taf-e‫ ڲ‬yesla (lit. ‘you’ll find he heard’), and a d y-as yesla (lit. ‘it will arrive (that) he heard’). (14’) - ad t-af-e‫ڲ‬ ye-sla, ... - IRR SBJ.2M.SG-find[AOR]-SBJ.2M.SG SBJ.3M.SG-hear[PFV], - ‘Probably, (lit. you will find he heard : he probably heard).’ (14”) - a d y-as ye-sla, ... - IRR PROX SBJ.3M.SG-arrive[AOR] SBJ.3M.SG-hear[PFV], - ‘Probably, (lit. it will arrive he heard : he probably heard).’ This set of modal expressions points to an element that we consider central in the expression of modality in Kabyle: the deictic dimension. Indeed, just

Mood and modality in Berber

441

as ad, the preverbal irrealis particle, is derived from a deictic element, those expressions imply that probability is the coming into existence (verbs as, ‘arrive’, and ili, ‘be’ or af, ‘find’) of a situation that could be witnessed by the hearer as well (14’), in a joint, converging, pointing gesture. Verbs expressing ability have the same personal affixes as their V2; cf.: (15) Tameddakwelt-is te-zmer friend.FS-POSS.3SG SBJ.3F.SG-be.able[PFV] t-ini. SBJ.3F.SG-say[AOR] ‘her friend, she could tell her.’

a

s

IRR

DAT.3SG

Verbs expressing want, desire, and need are characterised by the fact that their personal affixes vary, as well as those in the complement clause. (16) te-bȖa-m a t-İiwn-em, SBJ.2M.PL-want[PFV]-2M.PL IRR SBJ.2M.PL-help[AOR] ur te-ܲ‫܀‬i-m ara amek. NEG SBJ.2M.PL-know[NEG.PFV]-2M.PL NEG how ‘you want to help, (but) you don’t know how.’ The introduction of another participant, as in ‘you want him to help’ would be expressed by: (16’) te-bȖa-m SBJ.2M.PL-want[PFV]-2M.PL ‘You want him to help.’

ad i-İiwn. IRR SBJ.3M.SG-help[AOR]

Refusal and need work on the same pattern: (17) y-ugwi a s d SBJ.3M.SG-refuse[PFV] IRR DAT.3SG PROX i-ddu. SBJ.3M.SG-come[AOR] ‘He refused to accompany him.’ (18) ye-hwe÷ SBJ.3M.SG-need[PFV] ‘He needs to go.’

ad

i-ruh.

IRR

SBJ.3M.SG-go[AOR]

442

Amina Mettouchi

Semantically, all these examples are characterised by the fact that the predicate in V2 is construed as potential, virtual. This is in keeping with the definition we gave of the ad+aorist form above. These “V1-V2” modal expressions have not attained the same degree of grammaticalisation. The following parameters can be used to classify these expressions on a cline of grammaticalisation:  if the personal affix of V1 is coreferential with the personal affix of V2, we consider V1 as less auxiliarised than if the personal affix of V1 is the third person masculine, and stands for the complement clause;  given that in Berber the NP that is coreferential with the personal affix (the ‘lexical subject’) can appear after the verb (“VS” word order), we can say: if an NP cannot appear after V1 (before V2) although it is coreferential with the V1 affix, then the structure is more auxiliarised than if that NP can appear after V1 and before V2. The cline is then as follows:  Obligation (deontic values), probability and possibility are the most grammaticalised,  desire, need and ability are the least grammaticalised. Note, however, that it is possible that periphrastic obligation, probability and possibility are actually structural borrowings, since they coexist with ad+aorist forms that are not preceded by V1, but rather by particles, adverbials or thetic-deictic expressions (‘you will find that…’, ‘it will arrive that…’, ‘it will be that…’, etc.). ad+aorist in concessive and purposive clauses Those clauses are introduced by concessive or purposive markers. We will show two examples, one with xas (‘even if’, ‘although’), the other with akken (‘so that’, ‘in order to’). (19) xas a t regm-en, even.if IRR ACC.3M.S insult[AOR]-SBJ.3M.PL awal ur t id i-ttali. word.FS NEG ACC.3M.S PROX SBJ.3M.S-climb[IPFV] ‘Even if they insult him, he doesn’t (/won’t) say a word.’ (20) wid those

ur

k

ne-ssin NEG ACC.2M.S NEG.PTCP-know[NEG.PFV]

ara POST.NEG

Mood and modality in Berber

443

ad

‫܀‬essi-n dg-ek asekkud SBJ.3M.PL-screw[AOR] in-2M.SG look.FS akken ad cfu-n fell-ak. so IRR SBJ.3M.PL-remember[AOR] on-2M.SG ‘Those who don’t know you will examine you thoroughly so that they can remember you.’ IRR

With the concessive structure, a perfective or imperfective can be used instead of the ad+aorist form, if the concessive clause is actualised: (19’) xas regm-en t/ even.if SBJ.3M.PL-insult[AOR] ACC.3M.SG awal ur t id y-uli. word.FS NEG ACC.3M.SG PROX SBJ.3M.SG-climb[PFV] ‘Although they insulted him, he didn’t say a word.’ 3.2.2.4. ad+aorist in admonitive and permissive utterances Finally, ad+aorist appears after imperatives in admonitive and permissive utterances. (21) ‫ۊ‬ader ad te-Ȗli-‫ڲ‬. [SBJ.2SG]be.careful[IMP.AOR] IRR SBJ.2SG-fall[AOR]-SBJ.2SG ‘Be careful or you’ll fall!’ (lit. ‘be careful (not) to fall’) (22) anef as ad [SBJ.2SG]leave[IMP.AOR] DAT.3SG IRR ‘Let him go.’

i-ruh. SBJ.2SG-fall[AOR]

The relationship between the imperative ‫ۊ‬ader ‘beware’ and V2 is very close: the negative semantic component contained in the verb ‫ۊ‬ader has to be taken into account for the interpretation of ad te-Ȗli-‫ڲ‬, ‘you’ll fall’. Similarly, the realisation of ad i-ru‫‘ ۊ‬he will go’ depends on that of the command anef=as, ‘leave him’. If perfective or imperfective forms were used instead of ad+aorist, the meaning, as well as the prosodic pattern, would be radically altered:

444

Amina Mettouchi

(21’) ‫ۊ‬ader / te-Ȗli d. [SBJ.2SG]be.careful[IMP.AOR] /SBJ.3F.SG-fall[PFV] PROX ‘Be careful, she fell !’ (22’) anef as / i-ruh. [SBJ.2SG]leave[IMP.AOR] DAT.3SG / SBJ.2SG-fall[AOR] ‘Let him, he left.’ 3.2.2.5. ara+aorist in relative clauses The marker ad does not appear in relative clauses; instead, a specific particle, ara, which is of nominal origin,8 functions as irrealis marker in relative clauses. Following this particle, the aorist appears with the same values as after ad in independent, main and complement clauses. (23) tame‫ܒܒ‬ut ara ye-siwel. woman.FS REL.IRR SBJ.3M.SG-call[AOR] ‘the woman that he will call.’ Ara is mood-dependent, but does not depend on number, gender or person. It appears before the aorist, and sometimes the imperfective. Another particle, i, appears before the perfective and the imperfective (realis). (23’) tame‫ܒܒ‬ut i ye-sawel. woman.FS REL.REAL SBJ.3M.SG-call[PFV] ‘the woman that he called.’ Temporal-conditional clauses referring to modal planes of the type described in examples (8), (9) and (10) are also formed with ara. (24) m’ara bdu-n taglizit-nnsen when.REL.IRR SBJ.3M.PL-begin[AOR] English-POSS.3M.PL awal ur t-fehhm-e‫ڲ‬. word NEG SBJ.2SG-understand[IPFV]-2SG ‘When(ever) they begin their gibberish, you don’t understand a word.’

Mood and modality in Berber

445

Other markers Other markers can be combined with ad+aorist to express modal values. The first such marker is the word wissen, ‘perhaps’, the outcome of an amalgamation process between wi ‘who’ and ye-ssn-en ‘knowing’. This word is reinforced in (25) by ad y-ili ‘possibly’, ‘probably’, already presented under section 3.2.2.2., example (14). (25) imi te-ssusem, wissen since SBJ.3F.SG-be.quiet[PFV] who.knows ad y-ili hwi-Ȗ as. IRR SBJ.3M.SG-be[AOR] please[PFV]-SBJ.1SG DAT.3SG ‘if she stopped talking, perhaps it was (because) she was attracted to me.’ The second marker to be discussed here is a complex construction. In (26), the verb ini, ‘say’ is in the ad+aorist form, and is complemented both by a complement clause, and by a dative clitic, which has a reflexive meaning when used with verb ini. This structure, quite close to French on dirait ‘one would say’ or Italian si direbbe, is the most frequent for the expression of interpretational judgement. (26) a

s

t-ini-‫ڲ‬ aggur IRR DAT.3SG SBJ.2SG-say[AOR]-2SG month.FS ur ‫ܒܒ‬ise-Ȗ. NEG sleep[NEG.PFV]-SBJ.1SG ‘It seemed (you’d think) I hadn’t slept for a month.’

3.2.3. Conclusion on modality in positive utterances This survey of the various uses of the bare aorist stem and the ad+aorist form in positive modal utterances shows that: í the bare aorist rarely appears as such, except in imperatives. Its value is simply to posit a relation between a subject and a verb, the precise nature of which is determined either by its insertion in the imperative paradigm, or by its dependency with regard to a preceding verb.9  the ad+aorist form is very frequent. The (former deictic) particle ad plays the role of a semantic linker between the sub-specified predication

446

Amina Mettouchi

construed by the aorist, and the speaker’s situation of utterance or viewpoint.10 It covers the whole domain of modals, subjunctive and infinitive in English. It can be employed as past, present or future, and is therefore unspecified for tense. Its aspectual value is mostly perfective, since for habitual or conative predications, the ad+imperfective form is sometimes preferred. Its modal value is that of a potential in the course of, or aiming at, actualisation. Depending on the context, either the potential domain or its actualisation is given prominence. The dimension of assessment linked to the speaker’s judgement is always present, so ad+aorist is never a realis marker. 3.3. Mood and modality in negative utterances In Kabyle, the aorist is excluded from negative contexts, except for the negative optative: (27) a wer d y-uȖal! OPT.NEG PROX SBJ.3M.SG-come.back[AOR] ‘May he not come back!’ The expression a wer is probably derived from the combination of irrealis particle ad and negative particle ur. The prohibitive is built of the combination of the imperfective stem with the imperative set of affixes, preceded by the negative particle ur. The negative particle is the same for all verbal11 negative utterances. (28) Anf-et as ur Leave[IMP.AOR]-SBJ.2PL DAT.3SG NEG xeddem-t acemma. SBJ.2PL-do[IMP.IPFV]-2PL thing.FS ‘Leave him alone, don’t do anything.’ The negative hortative has the same form as the negative future (see below): a preverbal negative particle followed by the imperfective stem with the standard first person plural prefix. It is here followed by the postverbal “negative”12 particle ara.

Mood and modality in Berber

(29) ur NEG

ne-ttru‫ۊ‬u

ara!

SBJ.1PL-go[IPFV]

NEG

447

!

‘Let’s not go!’ Otherwise, perfective utterances are negated thanks to particle ur and the use of the negative perfective form, which has a basic counterfactual value13, as in example (20). The negative perfective can be considered as a modal form, though it hasn’t been viewed as such in Berber studies, which consider it simply as the negative “counterpart” of the perfective, and therefore an aspect. The negative “counterpart”14 of positive utterances in the ad+aorist (30), and in the imperfective (30’), is the imperfective (30’’). (30) ad

y-ali ar SBJ.3M.SG-climb[AOR] to ‘He will climb to the village.’ IRR

taddart. village.AS

(30’) ye-ttali ar taddart. SBJ.3M.SG-climb[IPFV] to village.AS ‘He (usually) climbs to the village’ (also: ‘he is climbing to the village’). (30’’) ur

ar taddart. to village.AS ‘He will not climb to the village’ (also: ‘he is not climbing to the village’, or ‘he doesn’t (habitually) climb to the village’). NEG

ye-ttali

ara

SBJ.3M.SG-climb[IPFV]

NEG

This phenomenon points to a close interaction between aspect and mood in Kabyle (Mettouchi 1996 and to appear). The negative domain has its own mapping, organised around the notion of counterfactuality. We have shown that in Kabyle the negative domain only contains forms whose semantic values underline the confrontation of the contents of the predication to the actual situation assessed, or the confrontation of competing viewpoints on that situation (Mettouchi 2000, 2006). To conclude, mood (in the wide sense of the term) and modality are rooted in the system of verbal oppositions, deemed “aspectual” by berberologists, but which we would tend to call mood-aspectual. No “modals” emerge as such, even if some verbs expressing need, ability, etc. are used in modal expressions. Moreover, for the deontic domain as well as

448

Amina Mettouchi

for the domain of possibility and probability, the use of ad+aorist, associated in some cases with adverbial particles or markers, is more frequent than the use of expressions of the type ‘it is possible/necessary that…’. One can propose to represent the building-blocks of modal expressions as in the following schema:  first level: the stem is associated with either the imperative, the hortative, or with standard affixes;  second level: this construction is either preverbed or remains unpreverbed;  third level: the ad+aorist construction is used as a complement clause in modal structures involving a V1 expressing obligation, possibility, probability, want, need etc. It can also be used, as in our examples revolving around probability, to modify a complement clause in the perfective or imperfective. Stem (aor or imperf)

imperative

standard

hortative

pers.aff - stem

pers.aff - stem

ad + pers.aff - stem

pers.aff – stem (modal V)

+

ara + pers.aff - stem

ad + pers.aff - stem

Figure 1. The building-blocks of modality in Kabyle

a wer + pers.aff stem

Mood and modality in Berber

449

Contrary to French where the subjunctive is clearly a subordination mood, the ad+aorist form is quite autonomous in Kabyle. It appears in main and complement clauses (the variant ara+aorist appears in relative clauses), and is motivated mainly by its semantics. It looks at the way the link between the sub-specified predication conveyed by the aorist and the speaker’s situation of utterance or viewpoint is construed, namely as a potentiality in the course of, or aiming at, actualisation. 4. Modal configurations in other Berber languages Not all Berber languages have the minimalistic modal configuration of Kabyle, with a polyvalent ad+aorist form for almost the whole of the irrealis domain. Actually, each system should be studied in itself, in considerable detail. There is no space to do so here, and we would only like to give an overview of the various configurations, in the following varieties:15 Tachelhit, Figuig Berber, Ahaggar Tuareg, Siwi Berber, in order to characterise the role of mood and modality in each language. The imperative is attested in all varieties, and based on the aorist stem. The imperfective imperative may be used, as in Kabyle, for specialised effects. The hortative is always based on the aorist stem, but its formation differs in the various languages: in Tachelhit it involves the suffixation of the (first person plural) dative clitic16–aȖ to the stem, itself associated with imperative affixes: if the imperative is second person singular, then the hortative is addressed to one person; if the imperative is second person plural, then the hortative is addressed to a group. In Figuig Berber the construction also involves the aorist stem, the suffixation to the stem of the dative clitic –axdd (first person plural dative clitic, probably amalgamated with the proximal particle dd), and the imperative affixes. Kossmann (1997) also mentions an adhortative whose structure is similar to that of Kabyle: a first person plural standard (i.e., non-imperative) prefix, a feminine or masculine suffix, and the preverbed particle ad. The prohibitive in Tachelhit is based on the aorist stem, with the standard second person affixes, preverbed by the sequence “irrealis + negative marker” (ad ur). In Figuig Berber the prohibitive is based as in Kabyle on the imperfective stem, with the imperative affixes, and preceded by the negative marker ur. In Ahaggar Tuareg, the prohibitive is built on either the negative perfective or the negative imperfective stem, with the

450

Amina Mettouchi

standard second person affixes, preceded by the negative marker ur. Siwi Berber uses the imperfective stem and the standard affixes, preceded by the negation la (a loanword from Arabic). The bare aorist has different values in the various Berber languages. In Tachelhit it is still quite frequently used in narratives17, with a perfective value, whereas in other dialects it is limited to the modal domain. The semantic domain of the ad+aorist18 form (which we could label the “virtual”) is different depending on the emergence or not of “future tenses”. Indeed in Tachelhit, for instance, the emergence of a temporal future based on the verb ‘want’ has limited the ad+aorist form to the expression of potentiality (and the optative). In some Tachelhit varieties (Hebaz 1979) there is a secondary distinction between neutral (temporal) future (particle rad), and “certain” future (particle ddad). The rad+aorist form comes from the grammaticalisation of i-ra ‘he wants’ + ad+aorist. The ddad+aorist form comes from the grammaticalisation of i-dda ‘he comes’ + ad+aorist. We therefore have a threefold opposition between prospective (ad+aorist), neutral future, and “certain” future. Only the ad+aorist form has a range comparable to that of Kabyle,19 and can refer to present or past. The two other forms are more temporal in that they are non-present and non-past. Leguil (1982) uses this argument to claim that Tachelhit is introducing the parameter of tense in its system. Similarly, in Figuig a future, described by Kossmann (1997) as expressing certainty or will, has appeared. Its form, sad+aorist, comes from the grammaticalisation of i-xes ‘he wants’ joined with the ad+aorist form. Interestingly, its negation does not imply the use of the imperfective (preceded by the verbal negation ur), as for the ad+aorist form, but the preverbing of a negative marker normally used for non-verbal sentences to the sad+aorist form. In Siwi Berber, there is no mention of a “future split”. The ga+aorist form is negated simply by the preverbing of the negative marker la, with no change of TAM stem. From the survey presented here, we can conclude that each system has its own organisation, even if some forms are clearly modal for the whole of Berber: the aorist, preceded or not by particle “ad” (or its equivalent “ga”/“ed”). The personal affix paradigm is the primary factor in the distinction between the potential-virtual (including future and optative), and the hortative, imperative and prohibitive. Finally, from the available data, it is very difficult to determine whether periphrastic expressions of obligation, probability or possibility are widely used or not. The texts we had access to tend to show that, as in Kabyle,

Mood and modality in Berber

451

these are alternative formulations, the ad+aorist form being flexible enough to accommodate a number of readings. 5. Conclusion on mood, modality, viewpoint and stativity What can we conclude on the expression of mood and modality in Berber? First the extension of the use of forms such as ad+aorist suggests that mood20 as a component of predication is a central category in the grammar of Berber. We would even argue that it is dominant, as the realis/irrealis distinction plays a role at various levels of the language. For instance, in some Berber languages, like Kabyle, there are two “relators” for relative clauses, one is realis, and the other irrealis. We have shown (Mettouchi 2000) that for Kabyle the semantic value of the perfective was actually that of a realis marker. Tense is obviously not relevant, and this has been noted by berberologists from the start, but aspect, in the interval/boundaries sense of the term, is not central either. Rather, what is interpreted as completeness and incompleteness is actually more of a factual/counterfactual distinction (see Mettouchi 2000, 2006). The asymmetry of the system with respect to negation, and the existence of specific negative themes is another argument in favour of mood-dominance in Berber. Secondly, the morphological material used, namely particles, implies two things:  as particles, and not verbs, they do not give information on agentparticipation.  as deictic particles, they show that at least in the first (pan-Berber) stage modal values were construed as the potentiality of the manifestation of a situation in the speaker’s sphere. This type of modal construal is linked to other features of the verbal system of Berber, revolving around the notion of stativity. We have shown this notion to be central in Kabyle (Mettouchi 2004), in which a high number of verbs have a basic stative reading that is not resultative, and in which another category is composed of ambitransitives. The rarity of occurrence of the passive derivation, and the frequency of the causative one (which is actually centrally transitivizing, and changes stative predications into dynamic ones) are arguments in favour of an underlying, original stativity of the system. Deixis is also fundamental in Berber: apart from the deictic origin of ad, the proximal particle plays a very important role at various

452

Amina Mettouchi

levels of predication: its proximal uses (direction towards the speaker) only represent a small part of its semantic range: it is also a benefactive marker, a resultative one, it is linked to thetic verbs, and expresses manifestation in the speaker’s sphere (Mettouchi 1998). We propose to link the stativity of the verbal system to mooddominance and the role of deixis, by claiming that predication in Berber, and particularly so in Kabyle, is speaker-oriented, not agent-oriented (Mettouchi to appear). Lehmann’s treatment of the grammaticalisation of modals and mood as the derivation of modal verbs or auxiliaries from full verbs (2002: 24-25) does not account for languages such as Berber, where such processes are peripheral and partly contact-induced. However, the Berber ad+aorist may not be all that distant from the English modals: like ad, modals are not inflected for person, like ad, they act directly on the verbal stem. The English bare infinitive has some common features with the aorist. If we do not take into account the etymology of modals, we can consider that they form a set of irrealis preverbal particles with specialised uses. Romance modal verbs are more liable to express agent-oriented modality, because they inflect for person. However, when “dummy” il ‘he/it’ and the middle voice are used, the orientation of the predication is shifted away from the protagonists and towards the speaker’s viewpoint, as in, for instance, the distinction between il peut venir (‘he can come’: ‘he is allowed to come’, ‘he is able to come’ and ‘he may come’) and il se peut qu’il vienne (‘it is possible that he might come’). Notes 1. I thank the organisers of the workshop on ‘Modals in the languages of Europe’ at the Valencia SLE meeting for inviting me to present data on a non-European language. I believe as they do that the study of languages surrounding the European area, and Kabyle is one of them, may help to grasp by comparison the specificities of the expression of modality in the languages of Europe. 2. For an overview of Berber languages, see Chaker and Mettouchi (2005). 3. Tense–Aspect–Mood is the usual name for those types of markers, but it is important to note that Berber languages are tenseless. 4. The data are from Galand (2002 (1974): 125–127). The vowel /u/ appears in forms where there is no personal suffix; the vowels e/a/ԥ alternate depending on the personal affix.

Mood and modality in Berber

453

5. A: unmarked form expressing the subject-predicate relationship; B: marked form expressing a limitation of the subject-predicate relationship; lla: preverb expressing concomitance, simultaneity. Į: unmarked; ȕ: marked; ı: ancient A forms that have drifted into the modal domain. 6. In Kabyle around 30% of the lexicon is borrowed from Arabic, but totally integrated morphosyntactically. 7. Another verb, zmer, is used for ability/capacity (see example (15)). It is not impersonal, and is of Berber origin. 8. It can be reconstructed as the grammaticalisation of a noun meaning “thing”. 9. In this, we are in full agreement with previous studies on the aorist in Berber (Galand 1977, 1987 and 2003, Prasse 1972, Penchoen 1973 and Chaker 1997). 10. See Mettouchi (2002) for a more detailed study of the ad+aorist form. 11. Only the non-verbal negations are different (Mettouchi 1996, 2006). 12. For a detailed study of the uses of this postverbal particle, which does not always appear in negative utterances, see Mettouchi (2001). In that paper, we have argued that the irrealis ara found in relative clauses (see 3.2.2.5.) and the postverbal negator ara are actually the outcome of two grammaticalisation paths starting with the same item, the indefinite ara (‘thing’). 13. Especially when it appears after a hypothetical marker, cf. Picard (1957) and Mettouchi (2000). 14. This formulation is rather broad and doesn’t account for the deep pragmatic difference between the negative and the positive domains. I have addressed this question elsewhere, for instance Mettouchi (1996, 2006), but will not dwell on it here, for lack of space. 15. Figuig Berber (Kossmann 1997) is spoken in oases in the East of Morocco, near the Algerian border, around Oujda. Tachelhit (Hebaz (1979), completed with Galand 1994 and Aspinion 1953) is spoken in Southern Morocco. Ahaggar Tuareg (Cortade 1969) is spoken in the Hoggar mountains of Southern Algeria, in the Sahara, and Siwi (Laoust 1931) is spoken in the Egyptian oasis of Siwa, west of Cairo towards the Libyan border. 16. Hebaz (1979: 14) considers this a dual marker. 17. “Alors que les emplois modaux de l’aoriste sans particule, surtout après impératif, sont fréquents dans les trois langues considérées, les emplois aspectuels, et plus particulièrement ceux qui confèrent à l’aoriste la valeur d’un accompli, ne sont courants que dans le sud et le centre du Maroc” (Galand 2002 (1987): 265). 18. ed+aorist in Ahaggar, ga+aorist in Siwi. 19. It expresses prospective values (will, obligation, necessity), optative ones, introduces complement clauses, concessive clauses, purposive clauses. 20. In Bhat’s (1999) sense, cf. section 1.

454

Amina Mettouchi

References Aspinion, Robert 1953 Apprenons le berbère: initiation aux dialectes chleuhs. Rabat: Félix Moncho. Bhat, Shankara 1999 The Prominence of Tense, Aspect and Mood. AmsterdamPhiladelphia: Benjamins. Bybee, Joan, Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. University of Chicago Press. Chaker, Salem 1997 Quelques faits de grammaticalisation dans le système verbal berbère. In Mémoires de la SLP, 103-121. Paris: Klincksieck. Chaker, Salem, and Amina Mettouchi 2005 The Berber Language. In Encyclopedia of Language and Linguistics 2nd Edition, Keith Brown (ed. in chief), Languages of the World section, Sarah Ogilvie (ed.), vol.1, Elsevier (édition en ligne 2006). Cohen, David 1989 L’Aspect verbal. Paris: PUF. Cortade, Jean-Marie 1969 Essai de grammaire Touareg (dialecte de l’Ahaggar). Université d’Alger, institut de Recherches Sahariennes. Galand, Lionel 1974 Présentation d’un parler: le Touareg de l’Aïr. Petites Sœurs de Jésus. Contes Touaregs de l’Aïr, 13-41. Paris: SELAF. Reprinted in Etudes de linguistique berbère, Lionel Galand, 117-147. Leuven-Paris: Peeters, 2002. Galand, Lionel 1977 Continuité et renouvellement d’un système verbal: le cas du berbère. BSLP 72 (1): 275–303. Galand, Lionel 1987 Les emplois de l’aoriste sans particule en berbère. In Proceedings of the 4th International Hamito-Semitic Congress, Marburg, 20-22 September, 1983, Herrmann Jungraithmayr, and Walter Müller (eds.), 361-379. Amsterdam-Philadelphia: Benjamins. Reprinted in Etudes de linguistique berbère, Lionel Galand, 259-272. LeuvenParis: Peeters, 2002. Galand, Lionel 2002 Etudes de linguistique berbère. Leuven-Paris: Peeters.

Mood and modality in Berber

455

Galand, Lionel 2003 L’Aoriste berbère, l’aspect et les valeurs modales. In Mélanges David Cohen, Jerome Lentin, and Antoine Lonnet (eds.), 235-246. Paris: Maisonneuve et Larose. Hebaz, Boudjemaa 1979 L’Aspect en berbère tachelhit, PhD thesis, University of Paris 5. Kossmann, Maarten 1997 Grammaire du parler berbère de Figuig (Maroc oriental), MS10 Ussun AmaziȖ, Paris-Louvain: Peeters. Laoust, Emile 1931 Siwa-1, son parler. Paris: Leroux. Leguil, Alphonse 1982 La Naissance des temps en chleuh. Bulletin des Etudes Africaines de l’Inalco 3: 57-84. Mettouchi, Amina 1996 La négation dans les langues du Maghreb, Synthèse. In La Négation en berbère et en arabe maghrébin, Salem Chaker, and Dominique Caubet (eds.), 177-195. Paris: L'Harmattan. Mettouchi, Amina 1998 La Particule D en berbère (kabyle): transcatégorialité des marqueurs énonciatifs. In Proceedings of the 16th International Congress of Linguists, Paris 20-25 juillet 1997, Bernard Caron (ed.). Oxford: Pergamon. Mettouchi, Amina 1999 Négation, co-énonciation et référenciation: le marqueur ara en kabyle de l'Ouest. In Comptes-Rendus du GLECS tome XXXIII (1995-1998), 87-104. Paris: Publications des LANGUES'O. Mettouchi, Amina 2000 Accompli et négation en kabyle. In Mélanges offerts à Karl-G. Prasse: Etudes berbères et chamito-sémitiques, 281-293. ParisLouvain: Peeters. Mettouchi, Amina 2001 La grammaticalisation de ara en kabyle, négation et subordination relative. In Travaux du CerLiCO n°14, Gilles Col, and Daniel Roulland (eds.), 215-235. Rennes: P.U. Mettouchi, Amina 2002 La Forme ad+aoriste en berbère (kabyle). In Articles de Linguistique Berbère - Mémorial Vycichl, Naït-Zerrad Kamal (ed.), 335-347. Paris: L’Harmattan. Mettouchi, Amina 2004 Diathesis, Aspect and Stativity in Taqbaylit Berber. In Nouvelles Etudes berbères, Berber Studies vol. 8, Kamal Naït-Zerrad, Rainer

456

Amina Mettouchi

Vossen, and Dymitr Ibriszimow (eds.), 95-109. Köln: Rüdiger Köppe. Mettouchi, Amina 2006 Nonverbal and verbal negations in Kabyle (Berber): A typological perspective, In Studies in African Linguistic Typology, Erhard F.K. Voeltz (ed.), 263-276. Amsterdam: John Benjamins. Mettouchi, Amina to appear Asymétries aspectuelles au regard de la négation et propriétés typologiques des langues. In La Négation et les énoncés non susceptibles d'être niés, Franck Floricic, and Renée Lambert-Brétière (eds.). Paris: Presses du CNRS. Penchoen, Thomas 1973 Etude syntaxique d’un parler berbère (Aït Frah de l’Aurès). Studi Magrebini V, Napoli. Picard, André 1957 Du Prétérit intensif en berbère, mémorial André Basset. Paris: Adrien-Maisonneuve. Prasse, Karl-Gustav 1972 Manuel de grammaire Touarègue (tahaggart). Copenhague: Academisk Forlag.

14. Modality in Basque Alan R. King In memoriam Andolin Eguzkitza (1953—2004) Larry Trask (1944—2004)

1. Introduction Expressions of modality are not lacking in Basque, but it is a matter for debate whether as a whole they constitute a distinct and compact formal category within Basque’s rather complex syntax and morphology. 1 Recognisable modal meanings are conveyed using a variety of morphological, syntactic or lexical resources that are integrated in different ways into the language’s overall structure. Any approach to a study of modality in specific languages in terms relevant to linguistic typology rests on the at least implicit assumption that there exists (non-language-specifically) a relevant set of modal notions that may together constitute a semantically or functionally meaningful category, and also that there exist non-arbitrary strategies of linguistic analysis, the use of which can be motivated through certain cross-linguistic or typological principles, for the identification and characterisation of language-specific manifestations of such (sets of) notions. The study of modality in a language such as Basque in which the category, if it exists at all, affects a somewhat nebulous area in the language’s grammar forces us to focus our attention more closely on the nature of those assumptions (above all the second one) and their applicability to a wide range of languages. I must therefore begin with a few words about methodology. A natural way to set about identifying modal expressions in a language (unless their distinctive formal properties are so salient as to make their identity plainly obvious) is first of all to search for expressions of a range of cross-linguistically identifiable modal notions such as possibility, obligation and so on. These expressions must next be “filtered”, looking at their formal properties and their place in the language’s grammatical and lexical system in order to determine whether or not they represent grammatical exponents of modality and whether they are of such a nature as to help justify the concept of modality as a category. Such expressions

458

Alan R. King

could turn out to be simple lexical items with no distinguishing grammatical features worth noting; or again, the usual means of conveying (what we insist on calling) modal notions may depend on operations at a higher (more abstract) level in the language’s grammatical structure, such as mood, for which reason there may be little internal justification for identifying these specifically as modality markers. We shall next proceed to examine the grammatical properties of such modal exponents as have survived this initial scrutiny. Since one purpose of this is to verify the existence of the category of modality as part of the language’s grammatical structures and processes, we shall be particularly interested in finding out the extent to which each modal exponent is grammaticalised (i.e. incorporated as an integral element of, or “built into”, the language’s grammatical structure). Doing this requires that we look not only at the specific properties of individual modal exponents, but also at the general properties of the grammatical system of the language. However, descriptions of the grammatical properties of individual modal exponents alone will not tell us whether (or to what extent) these exponents taken together constitute a grammatical paradigm. Hence the need for a heuristic procedure taking a number of exponents (i.e. linguistic forms or constructions) as input and yielding an evaluation of the paradigmaticity of this set of exponents (or of a subset thereof) as output. What, then, is a grammatical paradigm? What criteria determine paradigmaticity? Let us define a grammatical paradigm as a range of exponents that, while forming part of the larger framework of a language’s grammatical structure, are characterised by an exclusive set of common properties. Thus two formal features of such a set of exponents need to be ascertained: their mutual symmetry (sharing of a set of properties) and specialisation vis-à-vis the overall grammatical system. In reference to the whole grammatical system within which a paradigm is embedded, we may expect formal symmetry to involve a common (a) degree and (b) type of grammaticalisation in its components, while formal specialisation of the paradigm involves some sort of differentiation of components of the paradigm from other items in the language that are assumed not to belong to this particular paradigm. A major issue needing to be addressed here is the coexistence of different, “competing” exponents of particular modal notions. Sometimes this may happen as a result of, or in connection with, processes of diachronic change. If diachrony is excluded from consideration, the chronologically simultaneous presence in the language of different

Modality in Basque

459

exponents involving contrasting grammatical properties, possibly representing successive stages in an ongoing process of change, puts us in the dilemma of having to either incorporate them all into a more complex synchronic account or settling for an arbitrarily incomplete description. Allowing references to a diachronic characterisation of the process can enrich the account by permitting observations on the implications of such change for the coherence of the paradigm and even of the category per se at a language-specific level, as well as contributing potentially valuable data for the development of grammaticalisation theory. A related problem is that of dialect variations as they affect the choice or patterning of modal exponents, as well as their uses or relative frequency. Such variations may also have a sociolinguistic or stylistic dimension and involve a contrast between literary and colloquial registers, for example. Further issues may be raised by specific alternations such as paradigm-internal irregularities and anomalies. By definition typology seeks out generalisations, and indeed there are some language-internal variations that can safely be overlooked without compromising the validity of the overall characterisation. But where the assignment of a typological property could depend critically on the dialect, register or part of a paradigm one chooses to look at or overlook, oversimplification may be incompatible with the empirical programme inspiring a sincere typological enterprise. 2. The Basque verb system Basque’s complex synthetic verb morphology incorporates multiple argument indices and tense and mood markings into a compact formal system. However, only a small number of verb lexemes have access to synthetic verb morphology. In a second, periphrastic system of conjugation (the “compound tenses”), applicable to virtually all verbs, auxiliary verbs are the bearers of the aforementioned argument indices and tense/mood categories. In these compound tenses the lexical verb adopts one of a closed set of non-finite forms often referred to as “participles” in Basque grammars. In this system, the tense and mood of the auxiliary determine components of the compound tense form’s meaning while the choice of participle correlates with another layer of TAM choices conventionally referred to in Basque grammars as “aspect”, the possible values for which are imperfect, perfect, future and “aspectless” or aorist.

460

Alan R. King

Because the vast majority of Basque verb lexemes do not have access to the first level of synthetic conjugation, the second layer described is the most highly grammaticalised manner in which such non-synthetic verbs can be “conjugated” at all. Beyond that second layer, there exist still other periphrastic formations whose relatively less evolved grammatical properties set them apart from the aforementioned compound system and which therefore may be said to occupy a position in Basque grammar comparable to that held by compound tenses in languages with a more compact central system of simple tenses. For want of any better term I shall refer to this layer of Basque conjugation as the “tertiary tenses” or “third layer” of Basque tense constructions. This layer includes, among others, the progressive or continuous (e.g., ikusten ari da ‘is seeing’, ikusten dago ‘ditto’) and resultative perfect (ikusia du ‘has seen’, ikusita dauka ‘ditto’) aspects, which will not be discussed below, and also most of Basque’s modal constructions, which will of course be discussed at length. Basque transitivity frames select, at least in morphological terms, one or more grammatical participants which, if represented by noun phrases in the clause, will be associated with one or another of three morphological cases, called absolutive, ergative and dative. Omitting dative participants from consideration here in order to simplify the discussion, verbs may be classified for present purposes at least in three main transitivity types: transitive (having an ergative argument functioning as subject and an absolutive argument as direct object), unergative (with only an absolutive argument, functioning as subject) and unaccusative (with a single ergative argument, again functioning as subject). 2 Table 1. Synthetic tenses of etorri present past hypothetic

non potential n-a-tor 1s-PRS-come ‘I’m coming’ n-en-torr-en 1s-NPRS-come-PST ‘I was coming’ (ba)-n-en-tor (if)-1s-NPRS-come ‘(if) I came’

potential n-a-tor-ke 1s-PRS-come-POT ‘I may come’ n-en-tor-ke-en 1s-NPRS-come-POT-PST ‘I would/might have come’ n-en-tor-ke 1s-NPRS-come-POT ‘I would/could come’

Modality in Basque

461

There are a total of seven morphologically simple tense/mood forms (making up the “first layer”). Leaving aside the imperative, these fit neatly into a three-by-two matrix consisting of three basic “tenses” (present, past and hypothetic) and three corresponding potential ones as illustrated in Table 1. 3 In the compound tense system (the “second layer”), the same range of simple tenses, marked on auxiliaries, combine with a choice of four participles as explained above to express imperfect, perfect, future or aorist aspect. Different auxiliaries may occur depending on two factors: (a) aspect and (b) transitivity. Table 2 shows the present-imperfect, present-perfect, present-future and present-aorist forms, with third-person-singular subjects, of the transitive verb ekarri ‘bring’, the unergative verb etorri ‘come’ and the unaccusative verb jarraitu ‘continue’. 4 There exist a number of compound verbs composed of a non-verbal element and a light verb (usually in that order). Some of these have “to have” as their corresponding light verb, a circumstance which has occasionally led to some confusion because the same verb “to have” is also the auxiliary verb used to form the non-aorist transitive or unaccusative compound tenses. Consider the verb behar izan ‘to need’, the present tense forms of which are illustrated in Table 3: in the imperfect, perfect and future “aspects” the corresponding compound tense forms of the light verb izan appear, but in the aspectually unmarked simple form, the synthetic form of the light verb is used. Table 2. Some compound tense forms Transitive imperfect ekar-tzen n-a-u bring-IMPRF 1s-PRS-have ‘X brings me’ perfect ekarrn-a-u bring-PRF 1s-PRS-have ‘X has brought me ‘

ekar-tzen h-a-u bring-IMPRF 2s-PRS-have ‘X brings you’ ekarr-i h-a-u bring-PRF 2s-PRS-have ‘X has brought you’

ekartzen du bring-IMPRF 3s.PRS.have ‘X brings Y’ ekarr-i du bring-PRF 3s.PRS.have ‘X has brought Y’

future

ekarr-i.ko n-a-u bring-FUT 1s-PRS-have ‘X will bring me’

ekarr-i.ko h-a-u bring-FUT 2s-PRS-have ‘X will bring you’

ekarr-i.ko du bring-FUT 3s.PRS.have ‘X will bring Y’

aorist

ekar n-a-za-(n) bring-AOR 1s-PRS-AUX(SBJ)

ekar h-a-za-(n) bring-AOR 2s-PRS-AUX(SBJ)

ekar d-e-za-(n) bring-AOR 3s.PRS-()-AUX(SBJ)

462

Alan R. King

Unergative

Unaccusative

imperfect etor-tzen da come-IMPRF 3s.PRS.be ‘X comes’

jarrai-tzen du continue-IMPRF 3s.PRS.have ‘X continues’

perfect

etorr-i da come-PRF 3s.PRS.be ‘X has come’

jarrai-tu du continue-PRF 3s.PRS.have ‘X has continued’

future

etorr-i.ko da come-FUT 3s.PRS.be ‘X will come’

jarrai-tu.ko du continue-FUT 3s.PRS.have ‘X will continue’

aorist

etor d-a-di-(n) come-AOR 3.PRS-PRS-AUX-(SBJ) ‘(that) X come’

jarrai d-e-za-(n) continue-AOR 3s.PRS-()-AUX-(SBJ) ‘(that) X continue’

Table 3. The compound verb behar izan simple

behar nau ‘X needs me’

behar hau ‘X needs you’

behar du ‘X needs Y’

imperfect behar izaten nau ‘X needs me’ (habitual)

behar izaten hau ‘X needs you’ (habitual)

behar izaten du ‘X needs Y’ (habitual)

perfect

behar izan nau ‘X has needed me’

behar izan hau ‘X has needed you’

behar izan du ‘X has needed Y’

future

behar izango nau (or beharko nau) ‘X will need me’

behar izango hau (or beharko hau) ‘X will need you’

behar izango du (or beharko du) ‘X will need Y’

3. A note on epistemic particles There is a small, closed set of particles which serve to make explicit certain epistemic nuances of utterances. These particles have a unique distribution in Basque: they immediately precede the finite synthetic verb form (whether main verb or auxiliary). In (1) there is no modal particle, while (25) exemplify the particles omen, ote, bide and al. (1)

Nagusia bihar dator. boss tomorrow X.comes ‘The boss is coming tomorrow.’

Modality in Basque

(2)

Nagusia bihar omen dator. boss tomorrow MP X.comes ‘They say / I hear the boss is coming tomorrow.’

(3)

Zergatik etorri ote da? why come.PRF MP AUX(X.is) ‘I wonder why he has come.’

(4)

Ezagutzen bide nau. know.IMPRF MP AUX(X.has.me) ‘It seems that he knows me.’

(5)

Nagusiak ekarri al du? boss.ERG bring.PRF MP AUX(X.has.Y) ‘Has the boss brought it?’

463

While the existence of such forms must not be forgotten, they do not appear to be centrally important to the Basque modal paradigm presented here, for which reason I shall overlook them in the rest of this presentation. 4. The expression of necessity and obligation In the transitive compound verb behar izan ‘to need’, conjugated with simple or compound tense forms of the light verb izan, the latter is made to carry the indices corresponding to the argument structure of the main predicate, whether the corresponding noun phrase arguments are overtly present (nagusiak in (6-8) and ardoa in (6)) or implicit as the object in (7) and (8): (6)

Nagusiak ardoa behar du. boss.ERG wine.ABS need X.has.Y ‘The boss needs wine.’

(7)

Nagusiak behar du. boss.ERG need X.has.Y ‘The boss needs him/her/it.’

464 (8)

Alan R. King

Nagusiak behar nau. boss.ERG need X.has.me ‘The boss needs me.’

Necessity or obligation is expressed by a tertiary construction using behar izan as a modal auxiliary: (9)

Nagusiak etorri behar du. boss.ERG come.PRF need X.has(.Y) ‘The boss has to come.’

The lexical verb component of the construction, here etorri ‘come’, takes the form of the perfect participle (which happens also to be the verb’s conventional citation form). Etorri is itself an unergative verb, but in this construction the modal auxiliary retains its own transitive/unaccusative form, so the subject of the clause, nagusiak ‘the boss’, is case-marked as ergative and indexed in the verb as the ergative subject of the sentence. The retention of the formal transitivity of this modal auxiliary may be taken as one indication that the grammaticalisation of this modal construction has not progressed so far as to divest behar izan of its inherent, lexically determined transitivity. Since the point raised here is crucial to the following analysis, we shall examine it more closely before continuing. (10) and (11) are examples of simple sentences with a verb in a compound tense (the future). (10)

Nagusia etorriko da. boss.ABS come.FUT AUX(X.is) ‘The boss will come.’

(11)

Nagusiak ekarriko nau. boss.ERG bring.FUT AUX(X.has.me) ‘The boss will bring me.’

In the Basque grammatical context, the high degree of grammaticalisation of this construction is suggested by the lack of syntactic autonomy of the tense auxiliaries (da and nau respectively), which copy the transitivity of the main verbs (unergative in (10), transitive in (11)) and index the main verbs’ sets of argument (‘X’ in (10), ‘X-me’ in (11)). In (12) and (13), on the contrary, the complete autonomy of the main verb erabaki ‘to decide’,

Modality in Basque

465

which (as shown in its auxiliary) is completely unaffected by the transitivity and ignores the object arguments of the subordinate verbs etorri and ekarri, constitutes formal evidence for the non-auxiliarisation of erabaki: (12)

(13)

Nagusiak etortzea erabaki boss.ERG come.VN.ABS decide.PRF ‘The boss has decided to come.’ Nagusiak boss.ERG

du. AUX(X.has.Y)

ni ekartzea erabaki me bring.VN.ABS decide.PRF

du / *nau. AUX(X.has.Y) / *AUX(X.has.me)

‘The boss has decided to bring me.’ The autonomy of the behar izan construction in not reflecting an unergative main verb’s intransitivity seems to indicate a low position for this modal construction on the grammaticalisation scale. However, other evidence, brought out by comparing (14) and (15), provides evidence against the full syntactic autonomy of the modal verb: (14)

Nagusiak ekarri behar du. boss.ERG bring.PRF need X.has.Y ‘The boss has to bring him/her/it.’

(15)

Nagusiak ekarri behar nau. boss.ERG bring.PRF need X.has.me ‘The boss has to bring me.’

Here we see that person indices of both the subject and the direct object of the main verb are indexed through the modal auxiliary’s light verb component. Another feature of the behar construction that distinguishes it from the construction with erabaki is also worthy of note: the non-finite form taken by a verb that is a complement of erabaki is the verbal noun, whereas that of the complement of behar is the participle. Far more verbs take complements with a verbal noun than with a participle. An alternative construction which may represent a different phase in the diachronic process of grammaticalisation of the behar izan construction is the construction with beharra (behar + the singular article -a). Unlike the

466

Alan R. King

behar izan construction, here behar is treated as a substantive participating in the full range of noun phrase phenomena. 5 Behar as a noun can thus enter into a range of syntactic contexts that are not allowed in the light-verb construction behar izan, such as qualification by an attributive adjective (16), determination by a demonstrative (17) or a quantifier (18), or use with a possessor (19) or a qualifying relative clause (20): (16) behar handi -a need big ART ‘great need’ (17) behar hau need this ‘this need’ (18) behar batzuk need some ‘some needs’ (19) nire beharr my need ‘my need’

-a ART

(20) [sortu den] beharr arise.PRF AUX.SBJ need ‘the need that has arisen’

-a ART

Substantival behar admits a complement clause with an implicit coreferential subject, as in (21) and (22): (21) Nagusiak etorri beharra du/dauka. boss.ERG come.PRF need.ABS X.has.Y ‘The boss needs to come.’ (22) Nagusiak ni ekarri boss.ERG me.ABS bring.PRF ‘The boss needs to bring me.’

beharra du/dauka. need.ABS X.has.Y

Modality in Basque

467

Such structures are developed into sentences expressing necessity or obligation by means of a frame which has the beharra phrase as the direct object of ‘to have’ (cf. French j’ai besoin de faire quelque chose). In this construction, but not in the behar izan construction, the Bizkaian and Gipuzkoan dialects spoken in the west of the Basque Country permit the use of another lexical verb for ‘have’, eduki, hence Nagusiak etorri beharra dauka. The direct object of the main verb (first person in (22)) is not indexed on the finite verb (du or dauka), whose syntactic object is the whole beharra phrase. In negative polarity contexts behar behaves like a normal common noun in taking the partitive suffix -(r)ik in this construction, but again not in the behar izan construction; thus (23), the negative counterpart of (9), differs from (24), the negative counterpart of (21): (23) Nagusiak ez du etorri behar. boss.ERG NEG X.has.Y come.PRF need (a) ‘The boss doesn’t have to come’, ‘The boss needn’t come.’ (b) ‘The boss mustn’t come.’ (24) Nagusiak ez du/dauka etorri beharr -ik. boss.ERG NEG X.has.Y come.PRF need PRT ‘The boss doesn’t have to come’, ‘The boss needn’t come.’ While there is substantial overlap in meaning between (9) and (21), in the negative there is an important difference: the potential ambiguity over the scope of negation admitted for the construction in (23) is not present in (24), where the prohibition reading is not possible. It is clear that the alternative beharra construction is relatively ungrammaticalised and lower down on a grammaticalisation scale than the behar izan construction. Some of the differences observed may simply be reflexes of the syntactic fact that behar izan is a light-verb construction (itself a form of grammaticalisation), as opposed to beharra izan/eduki which clearly is not, rather than necessarily being a significant case of the grammaticalisation of a modal construction as such. But discounting those, other pertinent signs of the grammaticalisation of the behar izan modal construction are the participle form of the verb in the complement clause (although this property is shared by the beharra construction) and the partial sacrifice of behar izan’s syntactic autonomy.

468

Alan R. King

Thus it seems that the degree of grammaticalisation of a modal expression such as behar izan can be evaluated (positively) by observing the extent to which its grammatical properties match those of more “grammatical” structures (such as the second-level compound tenses), or (negatively) of ‘non-grammatical’ subordination constructions. In this context it is interesting to note the existence of a colloquialism in the use of behar izan that consists of eliminating behar izan’s autonomous transitivity, making the construction mimic more closely the behaviour of second-layer auxiliaries in adopting a transitive or an intransitive form selected by the main verb, not by behar izan itself. As a result, rather than (9), here repeated for convenience as (25), we may hear (and less often, read) (26): (25) Nagusiak etorri behar du. boss.ERG come.PRF need X.has.Y ‘The boss has to come.’ (26) Nagusia etorri boss.ABS come.PRF ‘ditto’

behar da. need X.is

5. The expression of volition The dominant pattern observed for the behar izan construction is quite closely paralleled by the nahi izan construction expressing volition. The shared properties are: use of a light verb construction with izan ‘to have’ as light verb, which again may also be used with noun phrase complements (compare 6, 8): (27) Nagusiak ardoa nahi boss.ERG wine.ABS want ‘The boss wants wine.’ (28) Nagusiak nahi nau. boss.ERG want X.has.me ‘The boss wants me.’

du. X.has.Y

Modality in Basque

469

In its modal use, retention of the compound verb’s inherent transitivity with intransitive verb complements (cf. 9): (29) Nagusiak etorri nahi want boss.ERG come.PRF ‘The boss wants to come.’

du. X.has.Y

indexing of the main verb’s direct object on the modal auxiliary (cf. 15): (30) Nagusiak ekarri nahi boss.ERG bring.PRF want ‘The boss wants to bring me.’

nau. X.has.me

and use of the perfect participle form of the lexical verb (etorri, ekarri) when the matrix and subordinate clauses have coreferential subjects. When the subjects are not coreferential (this is common with nahi izan, less usual but possible with behar izan), either the verbal noun of the subordinate verb is used: (31) Nagusiak nik ekartzea boss.ERG I.ERG bring.VN.ABS ‘The boss wants me to bring (it).’

nahi want

du. X.has.Y

or a finite clause in subjunctive mood is employed: (32) Nagusiak nik ekar boss.ERG I.ERG bring.AOR ‘ditto’

dezadan AUX(I.AOR.Y)

nahi du. want X.has.Y

The participle is not an option in this case: (33) *Nagusiak nik ekarri boss.ERG I.ERG bring.PRF

nahi du. want X.has.Y

One colloquial tendency, largely restricted to the westernmost dialect, Bizkaian, consists of using the verbal noun even with coreferential subjects (gura izan is widely used in Bizkaian as an equivalent to nahi izan):

470

Alan R. King

(34) Nagusiak etortea gura/nahi dau. boss.ERG come.VN.ABS want X.has.Y ‘The boss wants to come.’ (= 29) Although nahi, like behar, exists as a lexical noun meaning ‘wish’, there is no tradition of using ?nahia izan/eduki in a way parallel to beharra izan/eduki, but an equivalent expression exists using the noun gogo, which also means ‘wish’. Unlike behar, gogo usually takes a genitive verbal noun in -tzeko as its clausal complement, rather than a participle (compare 21, 22, 24): (35) Nagusiak etortzeko gogoa du/dauka. boss.ERG come.VN.GEN wish.ABS X.has.Y ‘The boss wants to come’, ‘The boss fancies coming’ (36) Nagusiak ni ekartzeko gogoa du/dauka. boss.ERG me.ABS bring.VN.GEN wish.ABS X.has.Y ‘The boss wants/would like to bring me.’ (37) Nagusiak ez du/dauka etortzeko gogo -rik. boss.ERG NEG X.has.Y come.VN.GEN wish PRT ‘The boss doesn’t want to come’, ‘…doesn’t feel like coming.’ The colloquial tendency to realign behar izan as an auxiliary lacking syntactic autonomy by having its forms reflect the transitivity of the main verb (cf. 26) is partially paralleled with nahi izan, since a comparable trend can also be observed, but not in the same geographical dialect (etorri behar da is a western usage, etorri nahi da an eastern one). All in all, the differences in syntactic patterning between the necessity modal behar izan and the volition modal nahi izan are minor ones of detail, whereas the similarities, summarised in Table 4, are clearly systematic. 6

Modality in Basque

471

Table 4. Necessity and volition constructions compared Necessity modal modal form inherent transitivity type copies transitivity of main verb? takes object indices from main verb? subordinate verb form

Volition

behar izan LVC with ‘have’ transitive no yes

nahi izan LVC with ‘have’ transitive no yes

perfect participle

perfect participle

Displaying both symmetry and specialisation, these two modal constructions have some of the trappings of a grammatical paradigm. But does that paradigm also extend to the modal notion of possibility? 6. Expressions of possibility 6.1. Potential forms In the simple (first layer) tense system, the potential forms (see Table 1: natorke, nentorkeen, nentorke), when used, often express epistemic possibility: I may come, I might have come, I might come. This usage is practically obsolete in colloquial speech and generally infrequent (and stylistically marked) in the modern language. The semantic range of such forms is commonly covered, in modern colloquial western Basque, by the use of the future described in the next subsection. However, the compound (second layer) tenses that are formed with the corresponding potential forms of some auxiliaries are in common use. Not all such compound tenses express modality, however (cf. (38), expressing hypothetic mood; this form is typically found in the consequence clause of hypothetical conditional sentences such as “If I asked him, the boss would bring it”), and those which do are not necessarily epistemic in meaning (39a, ability), although they may be (39b, possibility):

472

Alan R. King

(38) Nagusiak ekarriko luke. boss.ERG bring.FUT AUX(X.POT.HYP.Y) ‘The boss would bring (it).’ (39) Nagusiak ekar dezake. boss.ERG bring.AOR AUX(X.AOR.POT.PRS.Y) (a) ‘The boss can bring (it).’ (b) ‘The boss may bring (it).’ To complicate matters further, modern dialects differ in the interpretation of (40): (40) Nagusiak ekar lezake. boss.ERG bring.AOR AUX(X.AOR.POT.HYP.Y) Gipuzkoan dialect: ‘The boss could (or can) bring it.’ Lapurdi dialect: ‘The boss would bring it.’ With these reservations, in majority western usage today, possibility is often expressed by a compound tense of the potential aorist type, as in (39) above or (41): (41) Nagusia etor daiteke. boss.ABS come.AOR AUX(X.AOR.POT.PRS) ‘The boss can/may come.’ Because such exponents of possibility have a morphological basis (consisting of the potential forms of an auxiliary with the “aorist” aspect), their range of possible tenses is more limited than those of the tertiary constructions with behar izan and nahi izan, since there is no free “aspect” slot; they also have no non-finite forms. In a sense, then, this set of exponents of possibility is “defective” in comparison to necessity and volition. The result is a suppletive paradigm in which the “missing” forms of the possibility modal are provided by a tertiary construction with ahal, to be examined below: thus nagusia etor daiteke ‘the boss can come’ but nagusia etorri ahal izango da ‘the boss will be able to come’, nagusia etorri ahal izanez gero ‘if the boss can come’, etc.

Modality in Basque

473

6.2. Epistemic use of future aspect A very common modal meaning of certain compound tenses is found in modern colloquial western dialects, when the “future aspect” is employed not to express any kind of future, but rather probability, conjecture, and in general various shades of epistemic possibility. Both (42) and (43) have at least two different readings: (a) a “more literal” meaning that interprets the future marker as such, and (b) an ‘extended meaning’, of widespread use in the varieties indicated, that is purely epistemic (comparable to English “That will be the postman”, not expressing future time). (42) Nagusiak jakingo du. boss.ERG know.FUT AUX(X.has.Y) (a) ‘The boss will know.’ (b) ‘The boss probably knows’, ‘The boss may/must know.’ (43) Nagusiak jakingo zuen. boss.ERG know.FUT AUX(X.had.Y) (a) ‘The boss would have known.’ (b) ‘The boss probably knew’, ‘The boss may/must have known.’ In (42) the (a) meaning has a future time reference and is non-modal, whereas the (b) meaning has a present time reference and expresses a conjecture (i.e., epistemic possibility or probability). In (43) the (a) meaning conveys a consequence of an unfulfilled condition (a mood category), whereas the (b) meaning has a straightforward past time reference and again expresses conjecture. 6.3. Use of ahal Central and eastern Basque dialects can freely express non-epistemic possibility using the following type of periphrastic construction: (44) Nagusia etortzen ahal da. boss.ABS come.IMPRF able X.is ‘The boss can come.’

474

Alan R. King

(45) Nagusiak ekartzen ahal du. boss.ERG bring.IMPRF able X.has.Y ‘The boss can bring him/her/it.’ (46) Nagusiak ekartzen ahal nau. boss.ERG bring.IMPRF able X.has.me ‘The boss can bring me.’ Ahal (as a morpheme expressing possibility; its other uses are irrelevant here) may have started life as a modal particle, but synchronically is best described as forming a somewhat anomalous compound verb with the light verb izan, thus ahal izan ‘to be able’. Unlike behar izan ‘to need’ and nahi izan ‘to want’, ahal izan cannot be used with a noun object, but its clausal complement may be implicit, in which case ahal izan is construed as an unaccusative, so in response to the question Nor etor daiteke? ‘Who can come?’ one could say: (47) Nagusiak ahal du. boss.ERG able X.has ‘The boss can.’ But when the complement is expressed, ahal izan copies its transitivity from the main verb and indexes its arguments, as (44-46) show. This is a partial departure from the pattern of standard usage with behar izan and nahi izan, where the modal’s inherent transitivity is maintained (Nagusiak etorri behar du ‘The boss must come’ versus Nagusia etortzen ahal da ‘The boss can come’). Another difference is the use of the -tzen or imperfect participle form to form tenses in the imperfect or future aspect (but not in other aspects), thus etortzen ahal da, etortzen ahalko da, but etorri ahal (izan) da, etc. However, in the Bizkaian and Gipuzkoan dialects ahal izan falls into line with behar izan and nahi izan in this respect by taking the perfect participle throughout, thus etorri ahal da, etorri ahalko da, etorri ahal izan da etc. It was noticed above that the use of potential forms to express possibility excludes the possibility of specifying aspect or forming nonfinite forms. Ahal izan is used in such cases, resulting in a suppletive paradigm. Consequently use of the ahal construction is most widespread when aspect is marked and in non-finite contexts, e.g.:

Modality in Basque

475

(48) Nagusia etorri ahal izaten/izan/izango da. boss.ABS come.PRF can be.IMPRF/PRF/FUT AUX(X.is) ‘The boss is habitually/has been/will be able to come.’ (49) Nagusiak ekarri ahal izaten/izan/izango nau. boss.ERG bring.PRF can have. IMPRF/PRF/FUT AUX(X.has.me) ‘The boss is habitually/has been/will be able to bring me.’ (50) Nagusia etorri ahal izatea/izateko/izanez gero boss.ABS come.PRF can have.VN.ABS/have.VN.GEN/have.in.case ‘the boss’s being able to come’ / ‘in order for the boss to be able to come’ / ‘provided the boss can come’ But can the ahal izan construction also be used with simple tenses, i.e. in the absence of aspect specification, in the western dialects (as it can in the eastern ones)? There actually seems to be a division of opinions on this. Some Basque speakers find (51) unacceptable, yet others accept it: (51) Nagusia etorri boss.ABS come.PRF ‘The boss can come.’

ahal da. able X.is

Table 5. Necessity, volition and ahal constructions compared

modal modal form inherent transitivity copies transitivity? takes object indices? subordinate form

Necessity

Volition

(Possibility)

behar izan with ‘have’ transitive no yes perfect

nahi izan LVC with ‘have’ transitive no yes perfect

ahal izan LVC with ‘have’ transitive yes yes perfect

LVC

The basis of this division may be dialectal. Tolerance of (51) is mainly limited to speakers of the Bizkaian dialect. 7 Gipuzkoan speakers generally find (51) unacceptable. 8

476

Alan R. King

Table 5 summarises the syntactic properties of the ahal izan construction (western version), side by side with the necessity and volition exponents already listed in Table 4: 7. Expressions of impossibility In eastern dialects, the ordinary mechanism of negation (with ez) is used in conjunction with the two possibility constructions Nagusia etor daiteke (41) and Nagusia etortzen ahal da (44), as in (52) and (53) or the transitive counterparts (54) and (55): (52) Nagusia ez daiteke etor. boss.ABS NEG X.AUX.POT come.AOR ‘The boss cannot come.’ (53) Nagusia ez da etortzen ahal. boss.ABS NEG X.is come.IMPRF able ‘The boss cannot come.’ (54) Nagusiak ez nazake boss.ERG NEG X.AUX.POT.me ‘The boss cannot bring me.’

ekar. bring.AOR

(55) Nagusiak ez nau ekartzen ahal. boss.ERG NEG X.has.me bring.IMPRF able ‘The boss cannot bring me.’ None of these sentences is used in the modern western dialects, which require that a different construction be used to express impossibility. This construction, found in all Basque dialects, uses the particle ezin. Basque written tradition offers a range of variant constructions, a full description of which would be worthy of a separate monograph; here I shall only mention the currently dominant patterns. Ezin may have started life as a productive lexical derivation morpheme, 9 a function still found e.g. in (56):

Modality in Basque

477

(56) ikus -ezin -a see IMPOS ART ‘invisible’ As in the case of ahal izan, however, modern usage can largely be explained by assuming the existence of a compound verb ezin izan ‘to be unable’, some aspects of the usage of which parallel those of ahal izan. Like ahal izan, in the absence of an overt complement clause ezin izan is taken to be unaccusative, so in response to the question Nor etor daiteke? ‘Who can come?’ one could say: (57) Nagusiak ezin du. boss.ERG NEG.POS X.has ‘The boss cannot.’ When a complement is expressed, the transitivity of ezin izan is adapted to that of the main verb, and the modal verb indexes the arguments of the main verb (like ahal izan). In normal word order in main clauses, ezin izan precedes the complement rather than following it. This is a characteristic negative sentence pattern found also with the general negator ez (compare 60 and 61): (58) Nagusia ezin da etorri. boss.ABS NEG.POS X.is come.PRF ‘The boss cannot come.’ (59) Nagusiak ezin nau ekarri. boss.ERG NEG.POS X.has.me bring.PRF ‘The boss cannot bring me.’ (60) Nagusia ez da etorriko. boss.ABS NEG AUX(X.is) come.FUT ‘The boss won’t come.’ (61) Nagusiak ez nau ekarriko. boss.ERG NEG AUX(X.has.me) bring.FUT ‘The boss won’t bring me.’

478

Alan R. King

This is the predominant syntactic pattern with ezin, but one innovation that appears to be spreading and which some speakers seem to have adopted as a hypercorrection consists of maintaining the inherent transitivity of ezin izan in the same conditions as behar izan and nahi izan, producing prescriptively incorrect (in the majority opinion) but not uncommon sentences like (62) instead of (58): (62) Nagusiak ezin du etorri. boss.ERG NEG.POS X.has come.PRF ‘The boss cannot come.’ Table 6 summarises all the main tertiary modal constructions according to present-day accepted usage (remembering that ahal izan is suppletive for the potential construction): Table 6. Necessity, volition, possibility and impossibility constructions compared Necessity

Volition

(Possibility)

Impossibility

behar izan LVC + ‘have’ transitive

nahi izan LVC + ‘have’ transitive

ahal izan LVC + ‘have’ transitivity

ezin izan LVC + ‘have’ transitive

no

no

yes

yes

yes

yes

yes

yes

perfect

perfect

perfect

perfect

modal modal form inherent transitivity copies transitivity? takes object indices? subordinate form

8. Other patterns While the above survey covers the patterns that are generally acknowledged to constitute the main resources of the language for the expression of modal notions, there are others that might be mentioned to complete the account. Basque is not lacking in lexical strategies (other than those already mentioned with beharra and gogoa) that may convey modality, such as the use of other generic or abstract nouns:

Modality in Basque

479

(63) Nagusiak etortzeko premia ikusten du. boss.ERG come.VN.GEN necessity.ABS see.IMPRF AUX(X.has.Y) ‘The boss perceives the need to come.’ (necessity) (64) Nagusiak etortzeko asmoa boss.ERG come.VN.GEN intention.ABS ‘The boss intends to come.’ (volition)

du/dauka. X.has.Y

(65) Nagusia etortzeko asmotan dago. X.is-in-a-state boss.ABS come.VN.GEN intention.INE ‘The boss is planning to come’, lit. ‘…is in a mind to come’ (possible) (66) Nagusiak etortzeko aukera du/dauka. boss.ERG come.VN.GEN opportunity.ABS X.has.Y ‘The boss has an option to come.’ (possibility) (67) Nagusia etortzeko moduan dago. boss.ABS come.VN.GEN manner.INE X.is-in-a-state ‘The boss is able to come’, ‘…is in a position to come’ (possibility) (68) Nagusiak ez du/dauka etortzeko modurik. boss.ERG NEG X.has.Y come.VN.GEN manner.PRT ‘The boss has no way of coming.’ (impossibility) or adjectives, e.g. (69) Derrigorrezkoa da nagusia etortzea. obligatory X.is boss.ABS come.VN.ABS ‘The boss is required to come.’ (necessity) (70) Nagusia etortzera behartuta dago. boss.ABS come.VN.ALL obliged X.is-in-a-state ‘The boss is obliged to come.’ (obligation) (71) Nagusia etortzea posible da. boss.ABS come.VN.ABS possible X.is ‘It is possible for the boss to come’, ‘The boss may possibly come.’ (possibility)

480

Alan R. King

(72) Ezinezkoa zaio nagusiari etortzea. impossible X.is.to.Z boss.DAT come.VN.ABS ‘It is impossible for the boss to come.’ (impossibility) or other sentence patterns, e.g. (73) Gerta daiteke nagusia etortzea. happen X.AUX.AOR.POT.PRS boss.ABS come.VN.ABS ‘The boss may come.’ (lit. ‘the boss’ coming may happen’) (possibility) Potentially of greater interest are what we may term “other syntactic periphrases”: grammatical constructions lacking either a specifically modal lexical item or a “dedicated” grammatical modality marker which nonetheless conventionally take on modal meanings: 10 (74) Nagusia etortzeko(a) da. boss.ABS come.VN.GEN(.ABS) X.is ‘The boss is to come’, lit. ‘the boss is of/for coming’ (obligation, volition or future) (75) Nagusiak etortzea du/dauka. boss.ERG come.VN.ABS X.has.Y ‘The boss should come’ or ‘Let the boss come!’, lit. ‘the boss has coming [understood: as a duty or option, according to context]’ (cf. 66) (obligation/possibility) (76) Nagusia etortzekotan dago. boss.ABS come.VN.GEN.INE X.is-in-a-state ‘The boss plans to come.’, lit. ‘the boss is in [understood(?): a state of mind] of coming’ (cf. 65) (volition) (77) Nagusiak ez du/dauka etortzerik. boss.ERG NEG X.has.Y come.VN.PRT ‘The boss can’t come’, lit. ‘the boss has no coming.’ (cf. 68) (impossibility)

Modality in Basque

481

9. Conclusions This chapter attempts to offer a typological diagnosis of Basque modality, a subject that is on the face of it complicated, resists large-scale generalisations and has been treated quite sparsely if at all in existing grammars. The survey began with some brief observations about the broad structure of Basque conjugation in general and the role of transitivity and argument indices in particular, and these same issues, none of which are intrinsically related to modality per se, have continued to centre our attention at almost every stage of our discussion of a wide range of modal constructions. My purpose in so doing, as explained in the introductory section, has been to provide the information necessary for an assessment of elements of symmetry and specialisation with regard to these exponents for the purpose of determining, in as systematic and non-arbitrary a way as possible, whether or not there exists such a thing as a grammatical paradigm of Basque modality; what place if any the category of modality holds in Basque grammar; and lastly, if modality is a category and if this category is represented by a grammatical paradigm, the structure and form of that paradigm insofar as such information may be of typological interest. The “modal verbs” behar izan and nahi izan (and also ahal izan and ezin izan) have a common internal structure: “have + X” where X is a lexeme that may also function as a noun. This is an “unnecessary” parallelism of substance, somewhat analogous, say, to the coincidence that the three French modal verbs devoir, pouvoir, vouloir all end in -oir – not for any obviously essential reason, yet the parallelism exists nevertheless. 11 More strikingly, the patterns of behaviour of behar izan and nahi izan with regard to transitivity and argument indexing are almost identical, and that behaviour is different from that found in most other constructions, including second-layer compound tenses and other third-layer periphrastic constructions: so we have symmetry and specialisation, as far as these two modals are concerned at any rate. We note the syntactic autonomy of the “modals” in comparison to the compound tense auxiliaries, but also that this autonomy is not so great as that of some other tertiary auxiliaries such as ari izan, and certainly less than that of a lexical verb taking a clausal complement, such as erabaki. One other shared peculiarity of behar izan and nahi izan is that both govern the perfect participle, a systemically less marked non-finite form of the main verb different to that usually found in subordinate verbs.

482

Alan R. King

Basque exponents of possibility include a periphrastic construction (with ahal) that comes close to paralleling the behar izan and nahi izan constructions without managing to be identical to them (at least in standard usage), and another major strategy (with potential auxiliaries) that is decidedly higher up on the grammaticalisation scale (integrated into the second layer), but which for that very reason is aspectually more limited than the other modal exponents and therefore enters into a suppletive relationship with the more periphrastic alternative. Moreover, considerations of dialectal variants and past and present diachronic tendencies provide some evidence for developments towards the emergence of a more compact, regularised and grammaticalised modal paradigm. 12 In conclusion, 13 although Basque expressions of modality do not all constitute a single, formally homogeneous paradigm, some of them do to a considerable degree and there appears to be a general diachronic drift towards the consolidation of such a paradigm. This fuzzy set of Basque modals is characterised by the grammatical properties listed in Table 6. The most radical departure from this prototype in the modal system as a whole is the lack of symmetry in the possibility exponents, most notably the consolidated existence of a possibility exponent at a higher level of grammaticalisation than that found in the other modal notions, though limited to affirmative possibility in the most widely spoken dialects. Notes 1. This original paper draws on many ideas found in my earlier writing on this subject as found in my doctorate thesis (King 1993), especially chapter 6 (also titled “Modality in Basque”), and other publications listed in the references. Since the present article is a synthesis, most of its observations concerning Basque grammar are well known facts references to which may be found in too many publications to make individual references either necessary or practical; only a few key representative or highly relevant works among those I have consulted during thirty years of study of the Basque language can be listed in the bibliography. For the presentation of these facts and the general conclusions drawn, I alone am of course responsible. 2. In the unergative type (so called here to emphasise the single absolutive argument, but traditionally simply called intransitive in most Basque grammars), the verb is only marked for one argument. In the unaccusative type there is also only one (subject) argument index, but the auxiliary is the same as

Modality in Basque

3.

4. 5. 6.

7. 8.

9.

483

the transitive auxiliary. Assuming a widespread morphological analysis according to which the third-person-singular absolutive marker in transitive verb forms is realized as “zero”, a morphological ambiguity arises at this point, since jaten du, for example, may be interpreted either as transitive, meaning ‘X eats Y’ (e.g. ‘she eats it’) or as unaccusative, meaning ‘X eats’ (as in ‘She eats when she’s hungry’). The abbreviations used in the glosses are listed in the glossing index of the volume. () marks the semantically empty “linking morpheme”. Some morphological information is omitted when an understanding of the discussion does not hinge on it. ‘X’, ‘Y’ and ‘Z’ in the glosses stand for any third-personsingular argument functioning as subject, direct object or indirect object respectively. For the transitive verb, first, second and third person singular direct object forms are shown. Auxiliaries follow main verbs (except in the negative, see below). This is the reason why we cite it as beharra: determination by the article is the default situation for Basque common nouns. “Modal form” refers to the modal exponent’s internal composition (LVC = light verb construction); “inherent transitivity type” to whether the modal’s default valency (which manifests itself when occurring in isolation) is transitive or intransitive; “copies transitivity of main verb?” to whether or not the modal’s inherent transitivity type gets overridden when this conflicts with that of the main verb. Changing etorri to etortzen we get (44), widely used in the east as we have noted. Matters have been complicated slightly further by the existence of a certain initiative (challenged by some, and possibly now devoid of proponents) to promote the ahal pattern of (51) as a language teaching norm at the “elementary Basque” level. The ahal construction thereby acquired a vaguely textbookish ring, over and above both traditional dialect distribution and established literary practice, rendering the introspective determination of grammaticality more problematic. This is a tentative conjecture suggested by reconstruction rather than documentation; both “grammatical” and derivational usages of ezin are attested since the beginnings of Basque written literature (Michelena, 1992: 665-679). In contrast to present-day majority usage (briefly described here), in early texts in all dialects the order ezin + X seems to have predominated over the now more widespread X + ezin, where in the case of derivational uses X stands for a verb and the resulting word is usually either an adjective meaning ‘un-Xable’ or a noun meaning ‘in-X-ability’; in its grammatical use, X would again be a verb, with the auxiliary usually following the latter in the common early usage I am referring to, i.e., ezin + VERB + AUX. Thus the direction of morpheme order change in the lexical construction appears to have been from

484

10.

11.

12.

13.

Alan R. King

ezin + X to X + ezin, but in the grammatical construction from ezin + VERB + AUX to ezin + AUX + VERB. The “older” grammatical construction not only looks more similar to the “older” lexical construction, but is also quite unlike any other (old or new) verbal periphrastic construction, whereas the “newer” grammatical construction is clearly reminiscent of, and was presumably attracted towards, the non-modal negative order illustrated in (60) and (61). The resulting periphrastic construction is used in all dialects but its use in proportion to alternative ways of expressing impossibility (see below) has fluctuated notably over time and between dialects, as is seen in the philological evidence and editors’ remarks in the dictionary entry cited (Michelena, 1992 sub ‘ezin’). It is well known from grammaticalisation studies that standard modal expressions in many languages have evolved from such syntactic schemata, cf. for example the have to construction in English; thus these may be viewed as future grammaticalisation candidates. Interesting semantic issues may also be involved, but given their complexity and lack of space I shall not enter into these. Unless it exemplifies a general drift towards greater iconicity in language that would favour abstract form-function parallelisms and result in a more or less vague trend towards intra-paradigmatic morphological homogeneity. The internal structure of the Basque modal verbs may however be functionally pertinent in a more primary way insofar as it permits simple (layer one) conjugation of the light verb; this avoids obligatory marking of aspect in the modals. In other languages we also sometimes find that modal verbs, even though formally similar in most respects to ordinary verbs, exercise certain TAM-related privileges. This observation unfortunately begs a more thorough exploration of the Basque aspect system than the scope of the present chapter permits (including real aspect per se, not just the conventional “aspect” label employed in these pages). There have even been crossovers in innovative change, where paradigm items with different properties tend to imitate each other, seemingly in a drive towards obtaining a ‘better’ paradigm. Consider the colloquialism which replaces etorri behar du ‘has to come’ with etorri behar da (cf. ezin da etorri), and that which replaces ezin da etorri ‘cannot come’ with ezin du etorri as if trying to bring the construction in line with etorri behar du. Basque speakers have had intensive contacts with speakers of many other languages since prehistory (e.g., early Celtic neighbours), through recorded history (e.g., Latin) and down to modern and contemporary times, when Basque has been greatly influenced by Gascon and French in the north, Castilian and other Iberian Romance varieties in the South. Yet there seems to be little in either the grammatical features or the phonological substance of Basque modal expressions that might easily be identified as borrowings or areal features shared with any of these contact languages.

Modality in Basque

485

References Allières, Jacque 1979 Manuel pratique de basque. Paris: Picard. Eguzkitza, Andolin 1985 Aditz modalak euskaraz [Modal verbs in Basque]. Euskera XXX, 1: 215-221. Euskaltzaindia 1987 Euskal gramatika: lehen urratsak [Basque grammar: first steps] (vol. 2). Bilbo: Euskaltzaindia. Goenaga, Patxi 1985 Nahi eta behar [Nahi and behar]. In Symbolae Ludovico Mitxelena septuagenario oblatae, Jose L. Melena (ed.). Vitoria: Universidad del País Vasco. Goenaga, Patxi 2006 Behar-en lekua euskal hiztegian eta gramatikan [The place of behar in the Basque lexicon and grammar]. In Andolin gogoan: Essays in honour of Professor Eguzkitza, Beatriz Fernández, and Itziar Laka (eds.). Universidad del País Vasco. Hualde, José Ignacio and Jon Ortiz de Urbina (eds.) 2003 A grammar of Basque. Berlin: Mouton de Gruyter. King, Alan R. 1987 Euskal aditzen deskribapenaren arazo nagusiak [Important issues in the description of Basque verbs]. In Donostiako Udako V Ikastaroak. Euskal morfosintaxia eta fonologia: eztabaida gaiak, Pello Salaburu Etxeberria (ed.), 83-102. Euskal Herriko Unibertsitatea. King, Alan R. 1993 Communicative grammar of the Basque verb (selected aspects). Doctorate thesis, University of London. King, Alan R. 1994 The Basque language: a practical introduction. Reno: University of Nevada Press. King, Alan R. 1998 Asimetría de la obligación en las expresiones modales. In Estudios de tipología lingüística, Luque Durán, Juan de Dios, and Antonio Pamies Bertrán (eds.). Universidad de Granada. Lafitte, Pierre 1978 Reprint. Grammaire basque. Navarro-labourdin littéraire (éd. revue et corrigée). Donostia: Elkar. Original edition, Baiona, 1944. Michelena, Luis 1992 Diccionario General Vasco/Orotariko Euskal Hiztegia, vol. VII. Bilbo: Euskaltzaindia.

486

Alan R. King

Trask, R.L. 1984 Trask, R.L. 1997

Synchronic and diachronic studies in the grammar of Basque. Doctorate thesis, University of London. The history of Basque. London: Routledge.

15. Modals in Turkic Lars Johanson

1. Introduction The present chapter deals with grammaticalized expressions of volition, necessity and possibility in the Turkic languages, in particular cases of contact-induced renewal of old modal categories. It is organized as follows: Section 2 provides basic information concerning the Turkic family. Section 3 deals with the expression of volition, necessity and possibility by means of bound inflectional markers reflecting advanced stages of grammaticalization. Section 4 deals with analytic (periphrastic) devices consisting of nominal or verbal predicates with nonfinite forms as complements. The devices may have emerged, or their use may have been corroborated, under foreign influence. Section 5 deals with the “subjunctive” function of modal suffixes used to signal modality in dependent clauses, probably under the influence of Persian, Slavic, Greek, etc. Section 6 summarizes the observations concerning the degrees of grammaticalization of the devices dealt with in Sections 3-5 and their use in stylistic registers. It is noted that the copied devices represent less advanced stages of grammaticalization than their respective originals in the model languages. 2. The Turkic family Turkic is first documented in sources of the 8th century AD. The oldest written languages include stages of East Old Turkic, mostly simply referred to as “Old Turkic”: varieties originating in the Turkic eastern steppe empires, e.g. Orkhon Turkic, documented in inscriptions found in the territory of today’s Mongolia, and Old Uyghur, used from the 9th century on in the territory of today’s Xinjiang. According to some scholars, “Old Turkic” also comprises Karakhanid, the first Turkic written language of the Islamic era (11th century, center Kashgar). Other older written languages include Khorezmian Turkic, Chaghatay, Ottoman Turkish, etc.

488

Lars Johanson

The modern Turkic-speaking world is comprehensive: It extends from the Southwest, Turkey and neighboring areas, to the Southeast, to Eastern Turkistan and farther into China. From here it stretches to the Northeast, via South and North Siberia up to the Arctic Ocean, and finally to the Northwest, across West Siberia and East Europe. The Turkic family presently comprises about twenty standard languages, the most important being Turkish, Azeri, Uzbek, Tatar, etc. More than a dozen Turkic languages are spoken in Europe. In the following brief sketch, special attention will be paid to CircumAnatolian Turkic, i.e. Irano-Turkic and Balkan varieties, with frequent comparisons with Siberian Turkic, i.e. Tuvan, Khakas, Altay Turkic and Yakut. Language contacts have played an essential role for the renewal of the Turkic modal expressions of volition, necessity and possibility. Turkic varieties spoken in intense communication areas of Central Asia, Iran, South Siberia and the Balkans have copied morphosyntactic features of this kind from Indo-European languages, in particular from Iranian and Slavic, thus creating structural compatibilities with the modal systems of these languages. 3. Synthetic devices Turkic languages basically express the modal categories volition, necessity and possibility synthetically by means of bound inflectional markers, suffixes. Most of these are attested in similar forms at the oldest known stage of the development of Turkic documented in the East Old Turkic inscriptions. These old markers already represent advanced stages of their respective grammaticalization processes. The expressions of volition and necessity are all of unknown origin, i.e. they cannot be traced back to independent lexical elements. Whatever the lexical sources may have been, they have already undergone the changes typical of grammaticalization: extension of occurrence, desemanticization, decategorialization and material erosion. There is no indication that these devices have been copied from other languages. The modal markers in question will be presented cursorily with some examples from older and more recent stages of Turkic. Though it is difficult to define original functions for these markers, attempts will be made to give an idea of certain basic notions from which the various usages may be derived.

Modals in Turkic

489

3.1. Volition Volition is expressed by voluntative, optative and hypothetic markers, meaning ‘it is desirable that’, etc., suggesting that the action in question be carried out. The notions include demands, requests, directives, commands, impositions, entreaties, admonitions, warnings, exhortations, proposals, recommendations, advice, encouragement, incitament, etc. They also include desiderative, precative, permissive, promissive, intentional senses of wish, hope, desire and willingness. The volitional content may be realizable or unrealizable. The devices used are not imperatives in the sense of direct commands to second persons. They do not define relations between participants and the realization of the action. They are thus not agent-oriented, i.e. objective moods that denote the will of the subject referent, but rather subjective moods. This does not, however, mean that they are necessarily speakeroriented in the sense of expressing the speaker’s own will. The desiderability may also be conceived of as impersonal, representing a general or higher will. Three types of markers signal volition: voluntative, optative and hypothetic markers. (1) Voluntative markers express a strong wish, request, command, advice or permission. They mostly suggest that the realization of the action is conceived of as dependent on the cooperation of the subject referent or another participant, i.e. they are used to encourage or incite someone to action. The meaning of strong wish gives rise to expressions of intention, promise and willingness, especially in the first person. As mentioned, it is not necessarily the speaker’s wish that is expressed. The desiderability may be conceived of as impersonal. Identity of speaker and subject referent is represented in the first person, e.g. I will/shall act = I want myself to act. Markers of this kind are found at the oldest known stages of Turkic, e.g. Old Uyghur:  1 p. sg. -(A)yIn, e.g. bar-ayïn [go-VOL.1SG] ‘I will/shall go’.  1 p. pl. -(A)lIm, e.g. bar-alïm [go-VOL.2SG] ‘we will/shall go’. This marker may have adhortative function, expressing an appeal to carry out an action together, ‘let us go’.  3 p. -zUn, e.g. bar-zïn [go-VOL.3SG] ‘he/she/it shall go’. This marker may have jussive function, expressing a command concerning a third person, ‘let him/her/it go’, but it is not an imperative according to the definition given above.

490

Lars Johanson

Modern Turkish displays rather similar suffixes: -(y)AyIm, e.g. gid-eyim [go-VOL.1SG] ‘I will/shall go, let me go’. -(y)AlIm, e.g. gid-elim [go-VOL.2SG] ‘we will/shall go, let us go’. -sIn, e.g. git-sin [go-VOL.3SG] ‘he/she/it will/shall go, let him/her/it go’. The voluntative suffixes are of unknown origin, they cannot be traced back to lexical sources. It is possible that the first-person markers are residues of old indicatives representing secondary functions of an old low-focal (general) present marker, whose primary function was renewed by a new focal present tense in vowel + r (the so-called “aorist”). The latter has now, in turn, been defocalized itself and is a non-focal present conveying modal shades of meaning, e.g. at-ar [throw-NFPRS.3SG]‘tends to throw, will throw’. For the notions of focality and defocalization see Johanson (2000: 38-39, 89-95, 99-101). (2) The optative is a more purely desiderative mood in most Turkic languages. Optative markers express wish, desire or hope, mostly without suggesting that the realization of the action is dependent on the cooperation of the subject referent or another participant. Again, what is expressed is not the will of the subject referent and not necessarily that of the speaker. Speaker and subject referent coincide in the first person, e.g. may I act = I want myself to act. Examples: the Old Uyghur marker -GAy, e.g. bar-Ȗay [go-OPT] ‘may he/she/it go’; the modern Azeri marker -(y)A, developed from -GA(y), e.g. gơd-e [go-VOL.3SG] ‘may he/she/it go’, gơde-k [go-VOL-1PL] ‘may we go’. The optative suffixes are of unknown origin; they cannot be traced back to lexical elements. Besides their primary function of ‘wish’ the optative markers may also serve to express predication, potentiality, necessity and obligation as secondary functions. The Old Uyghur optative expresses wish, necessity, etc. Since the action can only be realized in the relative future, optative markers are sometimes referred to as future markers. Ottoman Turkish possessed a complete and formally transparent optative paradigm based on the marker -(y)A and conveying “souhait” and “incitation” (Deny 1921: 416). This paradigm has almost totally vanished in Standard Turkish, probably due to the close semantic similarity to the voluntative. In the Turkic languages of the Siberian group, optatives also express deontic and epistemic necessity or probability, obligation, permission, request, etc. The

Modals in Turkic

491

primary meaning of the Yakut optative is the assessment of an action as possible (Böhtlingk 1851: 307). (3) Hypothetic markers express an imaginative mood in the sense of supposed events, and may as such convey desiderative meanings: desire, wish or hope in the sense of ‘imagine it is the case’, ‘what if it were the case?’. The marker-sA(r may be used in this way when occurring in independent sentences, e.g. Ottoman Turkish bak-sa-m [look-HYP.1SG] ‘I will look’, ‘what if I look?, kör-se-ƾiz [see-HYP-2PL] ‘may you see’, ‘what if you see?’, Uzbek kơl-sл-ƾ [come-HYP-1SG] ‘if you only would come’, ‘what if you came?’, ye-sл-k [eat-HYP-1PL] ‘what if we ate?’ (Kononov 1960: 397-398). This marker, later known as a conditional marker, was previously used as a voluntative mood. According to C. Brockelmann, this is the oldest Turkic modal form to express a wish (1954: 378). Though it is of unknown origin, there have been attempts to trace it back to some form of the verb sa- ‘to count, to think’. The desiderative use is nowadays infrequent; the marker is mostly used in dependent clauses expressing hypothetical conditions (‘if it happens be the case’). 3.2. Necessity Necessity is expressed by various markers meaning ‘it is necessary that’. The conditions motivating the necessity for the subject referent to carry out the action may be physical or social. The markers may be used to express directives that impose or propose that the action be carried out, to compel, incite or encourage to action. Expressions of necessity can develop into a sense of desire or intention. They normally also express deontic obligation in terms of moral, legal or social norms. The obligation may be strong, compulsive, in the sense of must, have to, need to, or weaker, obligative or advisory, in the sense of should, ought to. Old Uyghur displays the marker -GU, e.g. saklan-Ȗu ol [protect.oneselfNEC COP] ‘one must protect oneself’. According to A. von Gabain, it had an abstract basic meaning from which necessitative, potential and purposive meanings developed: “Aus seiner abstrakten Grundbedeutung entwickelte sich ein Müssen, Können und ein Zweck” (1959: 36). The meaning ‘one must’ is normally expressed by the extended form -GU-lUK, e.g. bil-gülük

492

Lars Johanson

[know-NEC] ‘necessary to know’. According to M. Erdal, the suffix -lUK appears to have been added to make the necessitative meaning explicit (2004: 303). Modern Azeri uses -(y)A-sI, derived from a similar old verbal noun + -sI, e.g. yaz-asï-yam [write-NEC-1SG] ‘I should write’; for Kashkay see Csató (2006: 219-210). This necessitative, often referred to as an ancient future suffix, has vanished in Turkish except for some expressions employed for cursing, e.g. kör ol-asÕ [blind become-NEC.3SG] ‘may he/she become blind’ (Lewis 1967: 115). Some languages, e.g. Turkish and Azeri, use the necessitative marker -mAlI, e.g. Turkish oku-malÕ-yÕm [read-NEC-1SG], Azeri oxu-malï-yam [read-NEC-1SG] ‘I should/must read’, Turkish ver-meli-sin [give-NEC-2SG], Azeri ve‫ޟ‬r-meli-sen [give-NEC-2SG] ‘you should/must give’. Markers such as -mAlI may also be used for necessitative expressions that do not contain any element of will, e.g. memnun ol-malÕ [satisfied beNEC.3SG] ‘he/she must be satisfied (for example, judging from his happy smile)’. 3.3. Possibility Turkic languages use certain devices to express possibility in the sense of ability, i.e. inherent modality distinctions containing no element of stance, but defining relations between a participant and the realization of the action in the sense of ‘to be able to’. The devices may be used to represent internal and external (physical or social) enabling conditions for the subject referent with respect to the action. Markers used to express ability have in general also developed modal functions in a deontic and/or epistemic sense. As deontic markers they contain an element of will, expressing an evaluation in terms of norms, i.e. stating whether the action is acceptable, permissible, unacceptable or forbidden, e.g. you can/may go now. As epistemic markers they indicate the extent to which the speaker is committed to the truth of the proposition. They may express possibility, i.e. that the proposition may possibly be true, or probability, i.e. that it is more likely to be true. Ability to carry out the action is expressed in some early East Old Turkic texts by the lexical verb u-, negated u-ma-. It is normaly expressed by postverbs consisting of a converb marker, -U or -GAlI, and a following auxiliary verb, u- ‘to be able’, bol- ‘to become, to be’, e.g. ešit-geli bol-

Modals in Turkic

493

[hear-CVB be] ‘to be able to hear’. East Middle Turkic displays hypothetic mood -sA + bol- and -U bil- with the auxiliary bil- ‘to know how to, to be able to’. Later languages use postverbs with bil- ‘to know, al- ‘to take’ or bol- ‘to become’ as auxiliary verbs, e.g. Uzbek oqi-y-ål- [read-CVB take] ‘to be able to read’, Khakas sarna-p pol- [sing-CVB become] ‘to be able to sing’, Azeri gel-e bil- [come-CVB know], Turkish gel-e-bil- [come-CVB know] (univerbation) ‘to be able to come’. Negated forms are, in some cases, less transparent, e.g. Turkish -(y)AmA- in gel-e-me- [come-CVB-NEG] ‘not to be able to come’. In Azeri, Iraq Turkic, etc. -(y)Abil- has the negation -(y)Abilme-, e.g. vơr-e bil-me-m [give-CVB know-NEG-1SG] ‘I cannot give’. Markers of ability may, as mentioned, also express deontic possibility, ‘it is acceptable/permissible that’, and epistemic possibility, i.e. express the speaker’s uncertainty about the truth-value, suggesting that the proposition may possibly be true, ‘it is probable/possible that’. This is true of markers such as Turkish -(y)Abil- (Boeschoten 1990: 84-87) and Azeri -(y)A bil-. However, -(y)Abil- in Gagauz, a Balkan language closely related to Turkish, only expresses ability (Menz 1999: 58). We will come back to analytic devices that express possibility of other kinds. Not all Turkic languages use ability markers for epistemic possibility. For this purpose, East Old Turkic and many other languages use special modal particles which convey some kind of stance, a cognitive or affective attitude towards the events described: meanings of subjective reasoning and belief, personal involvement, emotions, evaluations, judgments concerning the propositional content. Turkic is generally considered to have few adverbs expressing concepts of this kind. Qualifiers evaluating the actuality of the action as certain, probable, possible, improbable, impossible, etc. (‘really ’, ‘actually’, ‘indeed’, ‘absolutely’, ‘probably’, ‘maybe’ ‘possibly’, etc.) have very often been copied from contact languages. On the other hand, Turkic possesses native devices that form systems of grammatical stance categories, for example epistemic particles such as erki, which occurs in Old Uyghur and Karakhanid, an uninflected utterance-final presumptive particle. It may express speculation, skepticism, likelihood, etc., e.g. Men kergek er-d-im erki [I necessary be-PST.COP-1SG PTL] ‘I must have been useful, I guess I was useful’; see Johanson (2004) and (2008). According to G. Clauson, the best translation of it is a periphrasis like ‘do you suppose that ...’ or ‘it is likely that’ (1972: 223).

494

Lars Johanson

A very common grammaticalized device to express possibility is the old intraterminal (present/imperfect) aspecto-temporal item in -(V)r, often rather confusingly referred to as “aorist”. In the course of its development, it has been defocalized (see Johanson 2000: 38-39, 89-95, 99-101) to a non-focal present that has almost totally lost its indicative value and has taken over modal functions in the sense of disposition, inclination, tendency, potentiality, prospectivity, etc., e.g. Tatar kil-er [comeNFPRS.3SG] ‘he/she/it may/will/can come’, Turkish ol-ur [becomeNFPRS.3SG] ‘it is possible’. Also other grammaticalized expressions of deontic and epistemic modality are observed. In the Turkic languages of the Siberian group, markers that are formally ‘optatives’ may, as already noted, express possibility, probability, permission, etc. 3.4. Comments The modal notions dealt with so far are expressed by old inflectional suffixes, still alive in almost all Turkic languages. It is unknown whether these markers go back to constructions with auxiliary elements of lexical origin. They may have been predicates expressing, as part of the propositional content, ‘objective’ modalities, i.e. relations between a participant and the realization of the action, e.g. desire (‘to want to’), willingness (‘to be willing to’), obligation (‘to have to’), necessity (‘to need to’), ability (‘to be able to’), permission (‘to be allowed to’). Predicates of these kinds often develop into moods. Such expressions may have grammaticalized, evolving into volitional, necessitative and potential moods, gradually taking on inflectional expression. As mentioned, the voluntative, optative, hypothetic, necessitative and possibility markers are not exclusively speaker-oriented. The desiderability, necessity or possibility does not always express the speaker’s own stance, but may often be interpreted in a more general impersonal sense, e.g. ‘it is desirable, wanted, requested, conceivable, necessary, probable, possible, acceptable, permissible that’. The voluntatives and optatives may be less dependent on the speaker’s own will, the hypothetic mood less dependent on the speaker’s own imagination, the necessitative and possibility moods less dependent on the speaker’s own assessment. The moods are often also, particularly in older languages, open to different interpretations with respect to the subject, when this is not

Modals in Turkic

495

expressed explicitly. The subject may have a specific referent or be impersonal. A late remnant of this is even found in modern Turkish, where the third-person necessitative marker has personal but also impersonal uses, e.g. gel-meli [come-NEC.3SG] ‘he/she/it ought to come’ or [comeNEC] ‘it is necessary to come, one must/should come’. It is sometimes claimed that the Gagauz necessitative lacks personal suffixes, e.g. ben gelmeli ‘I must/should come’, etc. (Pokrovskaja 1964: 214). However, Astrid Menz confirms that her informants use forms with personal suffixes, e.g. gel-meli-yim [I come- NEC-1SG] (Menz 1999: 54). The East Old Turkic necessitative marker -GU is an example of a similar vagueness, since it can refer to both agents and patients, e.g. bơr-gü ‘someone who shall give’ or ‘something that shall be given, something to give’. Erdal refers to it as a “projection participle”, used for presenting projections of expectations, evaluations and intentions (2004: 303). 4. Analytic devices Turkic languages display various analytic (periphrastic) devices for expressing volition, necessity and possibility: nominal or verbal predicates with nonfinite forms as complements. The synthetically expressed moods are, as mentioned, semantically vague, e.g. open to various interpretations. The analytic devices can be used to convey more specific information. The analytic constructions basically express ‘objective’ modalities, but they have also played an essential role in the renewal of the ‘subjective’ modalities expressing volition, necessity and possibility. Language contacts have played an essential role for this renewal. It is impossible to claim that all these analytic devices have emerged under foreign influence, but their use has undoubtedly been corroborated or expanded by foreign models. The following survey will illustrate some of the analytic devices, which primarily express ‘objective’ modalities, but have also participated in the renewal of the corresponding ‘subjective’ moods. 4.1. Volition Volition can be expressed with verbs meaning ‘to want’ plus a nominal form of the lexical verb as complement. Turkish iste- ‘to want’ takes the

496

Lars Johanson

verbal nouns (infinitives) in -mAk and -mA as complements, e.g. git-mek iste- [go-VN want] ‘to want to go’, git-me-si-ni iste- [go-VN-POSS3SG-ACC want] ‘to want him/her/it to go’. In most Turkic languages the subject of the complement clause can be marked with a possessive suffix, a possible overt subject taking the genitive case, e.g. Turkish Ali’nin git-me-si-ni iste[Ali-GEN go-INF-POSS-3SG-ACC want] ‘to want Ali to go’. Even Gagauz, which tends to copy volitional patterns of the Balkan type, with extensive “infinitive loss”, prefers infinitival patterns in samesubject constructions, e.g. iste-r git-me: [want-NFPRS go-INF] ‘he/she wants to go’, iste-di git-me: [want-PST go-INF] ‘he/she wanted to go’. 4.2. Necessity Necessity can be expressed with adjectives meaning ‘needed, necessary’ plus a nominal form of the lexical verb. Old Uyghur uses the adjective kergek together with participles, -GU kergek and -mIš kergek, e.g. munï bil-miš kergek [this know-PTCP necessary] ‘one must know (have learnt) this’. The subject can be referred to with a possessive suffix, e.g. bil-miš-im kergek [know-PTCP-POSS1SG necessary]. Modern Turkish uses gerek or lazÕm, the latter ultimately of Arabic origin. The lexical verb occurs in the impersonal infinitive form in -mAk, e.g. bil-mek gerek/lazÕm [know-INF necessary] ‘it is necessary to know, one must know’. Personal forms take the verbal noun suffix -mA plus possessive suffixes, e.g. bil-me-m gerek/lazÕm [know-INF-POSS1SG necessary] ‘it is necessary for me to know, I must know’, bil-me-si gerek/lazÕm [know-INF-POSS3SG necessary] ‘it is necessary for him/her to know’. As in most Turkic languages, an overt subject stands in the genitive, e.g. Turkish Ali-’nin git-me-si gerek/lazÕm [Ali-GEN go-VNPOSS3SG necessary] ‘Ali must go’. Modern Turkish can also use the verb gerek- ‘to be necessary’, e.g. git-me-m gerek-iyor [go-VN-POSS1SG be.necessary-PRS] ‘I must go’. The gerek / lazÕm construction is very frequent in modern Turkish. In Ottoman, the subject could also be expressed by a dative constituent, e.g. biz-e bil-mek gerek [we-DAT know-INF necessary] ‘it is necessary for us to know, we must know’. The predicate gerek could also be used in the sense of ‘obliged’, e.g. bil-mek gerek-sin [know-INF necessary-2SG] ‘you must know’.

Modals in Turkic

497

Patterns of these kinds are observed in many Turkic languages. In South Siberia, Tuvan xerek ‘necessary’ combines with a participle, e.g. þaȖa: þorud-ar xerek [letter send-PTCP necessary] ‘it is necessary to send a letter’. Khakas kirek ‘necessary’ combines with the infinitive to express ‘it is necessary, one should/must/ought to’, e.g. uzi-rȖa kirek [sleep-INF necessary] ‘one must sleep’, al-arȖa kirek [take-INF necessary] ‘one must take’. A possible overt subject is in the dative, e.g. saȖa: ügren-erge kirek [you-DAT learn-INF necessary] ‘you must learn’. These analytic constructions may be corroborated by foreign influence. The last Khakas sentence just cited can be compared to Russian tebe nado uþit’sja [you-DAT necessary learn-INF]. The adjective meaning ‘necessary’ is often a borrowed item. Modern Azeri uses la:zïm and va:jïb ‘important, urgent, necessary, requisite’, both ultimately of Arabic origin. Yakut uses na:da, a copy of Russian nado ‘necessary’, e.g. üören-iexxe na:da [learnINF necessary] ‘it is necessary to learn’, suruy-uoxxa na:da [write-INF necessary] ‘one must write’; cf. Russian nado pisat’ [necessary write-INF]. A similar Yakut item is ba:r ‘existing’, e.g. ïl-ïaxxa ba:r [take-INF existing] ‘one must take’. The words na:da and ba:r can also be used as nouns meaning ‘necessity’, e.g. bar-ar na:da-m suox [go-PTCP necessity-POSS1SG nonexisting] ‘I don’t need to go’. In Gagauz, analytic constructions with la:zïm ‘necessary’ are the predominant expressions of necessity. Infinitival constructions with the lexical verb in the -mA: form are common, e.g. sor-ma: la:zïm [ask-INF necessary] ‘one must ask’. A possible overt subject is in the nominative. An alternative pattern corresponds to the Russian combinational pattern dolz¡en ‘obliged’ + postposed infinitive, e.g. la:zïm sor-ma: [necessary ask-INF] ‘one must ask’. A possible overt subject is in the dative, e.g. san-a la:zïm sor-ma: ‘you must ask’ [you-DAT necessary search-INF]. This construction is restricted to the written language. It should be noted that the so-called “infinitives” used in the Khakas, Yakut and Gagauz examples above are nominal forms of the verb provided with dative suffixes: al-ar-Ȗa, üören-iex-xe, sor-ma: (< sor-ma-ya). They have thus originally had purposive meanings reminiscent of the subjunctive functions to be dealt with under 4. Turkic varieties under strong Iranian influence use constructions reflecting Iranian patterns. Iranian languages use impersonal necessitative expressions, e.g. Persian la:zim ast, Kurdish la:zÕm e ‘it is necessary’. The situation is similar in other Turkic varieties spoken in Iran and in Iraq Turkic (Bulut 2000).

498

Lars Johanson

The analytic constructions have strongly restricted the use of the necessitative suffixes. For example, Gagauz -mAlI has a very low frequency, and Azeri has lost its corresponding necessitative mood (Cafero÷lu and Doerfer 1959: 303). 4.3. Possibility Possibility, in particular deontic possibility, is often expressed by forms of bol- ‘to become, to be’, e.g. Turkish olur, Uzbek bolлdï, Khakas polar, Tuvan bolur [be-NFPRS] ‘(it is) possible’. The lexical verb often occurs in a converb form, e.g. Tuvan kir-ip bol-ur be? [enter-CVB be-NFPRS Q] ‘is it possible to enter?’. Khakas also exhibits forms of the verb þara- ‘to be right, to be suitable’: þar-ir ‘it is possible/permissible’, ‘one may’, and þara-bas ‘it is impossible’, ‘it is not allowed’, ‘one must/should not’ together with the socalled infinitive in -(V)rGA, originally a purposive marker containing a dative suffix -GA, e.g. al-arȖa þar-ir [take-INF be.suitable-PRS] ‘one may take’, al-arȖa þara-bas [take-INF be.suitable-NEG-PRS] ‘one must not take’. A possible overt subject stands in the dative, e.g. saȖa: ‘to you’, maȖa: ‘to me’. Altay Turkic has a similar construction, e.g. taƾkïla:-rȖa Ћara-r ba? [smoke-inf be.suitable-prs q] ‘is it possible to smoke?’. This pattern is similar to Russian patterns with možno ‘it is possible, allowed’ and nel’zja ‘it is impossible, forbidden’ and with overt subjects in the dative, e.g. tebe ‘to you’, mne ‘to me’. It was, however, used early in languages that had no intense contact with Russian, e.g. Ottoman Turkish (from the 14th century on) yara-r ‘it is suitable, permissible, possible’, negated yaramaz (Clauson 1972: 956). The adjective meaning ‘possible’ may be a borrowed item such as mümkin, mümkün, ultimately of Arabic origin, e.g. Turkish gir-mek mümkün mü? [enter-INF possible Q]‘is it possible to enter?’, Uzbek kфr-iš-im muլ mkin-mi? [see-INF possible-Q] ‘can/may I see?’ (Kononov 1960: 399). Gagauz uses var ‘there is’ + nasïl/niЋe (l) ‘how’, a copy of a Bulgarian pattern containing ima kak ‘it is possible’ (ima ‘there is’, kak ‘how’), e.g. girme: var mï niЋe? [enter-INF there.is Q how] ‘is it possible to enter?’. The spoken language tends to contract var + niЋe to varïnЋa. Impossibility is expressed by yok ‘there is not’ instead of var; cf. Bulgarian njama. The

Modals in Turkic

499

question particle mI stands between var/yok and the word for ‘how’. In the written language, the lexical verb may be preposed, e.g. girme: var mï niЋe? [enter-INF there.is Q how] ‘is it possible to enter?’. Unlike -(y)Abil-, these very frequent constructions express deontic possibility. In impersonal constructions the verbal noun in -mA: is used. It is, like the Khakas socalled infinitive in -(V)rGA, originally a purposive marker containing a dative suffix. 5. Subjunctive function of modal suffixes The relevance of areality for the renewal of the modal constructions is obvious. It is necessary to look at their use in a contact-linguistic and areal perspective. An important result of intensive language contact in Turkic varieties spoken in Central Asia, Iran, the Balkan area, etc., is the combinational-semantic copying of certain morphosyntactic patterns from Indo-European languages, in particular from Iranian and Slavic. The features acquired are imitations of syntactic subordination, in which finite verb forms, marked with optative, voluntative, hypothetic or necessitative suffixes, come to be used as a syntactic “modus subjunctivus”. Markers primarily signaling mood in independent sentences, occur in dependent clauses, often in connection with junctors such as ki and kim. This use of the moods will be referred to as their subjunctive function. This contact phenomenon has been present for a millennium in certain Turkic varietes. The new constructions were first copied into Middle and Late Old Uyghur translated texts. They were further developed as a phenomenon of Iranicization under the strong influence of Persian, and are present to varying degrees in Chaghatay, Ottoman, Azeri, Uzbek, etc. Balkan Turkish displays similar patterns as a result of copying from nonTurkic languages, mainly Slavic. The patterns copied formed systems analogous to the systems of the contact languages. The new patterns changed basic syntactic structures in the affected varieties. There is a wide range of frequently used constructions with synthetic modal markers in subjunctive function. Thus, in Turkic varieties spoken in Iran, modal constructions of volition, necessity and possibility are essentially semantic-combinatory copies of Persian patterns with modal auxiliaries (KÕral 2005). Balkan Turkish differs strongly from Standard Turkish through the high frequency and wide range of these modal constructions. Subjunctive clauses in West Rumelian Turkish, East Rumelian

500

Lars Johanson

Turkish and Gagauz are calques on constructions in non-Turkic Balkan languages. West Rumelian Turkish patterns are copied from subordinate clauses in Macedonian and Albanian with subjunctive, optative, and conditional markers (Friedman 1985). Dialects of Cyprus Turkish display similar patterns. They are certainly due to old Iranian influences, but also to the corroborating force of Greek, with which the varieties have had long and intensive contacts. One result of this contact phenomenon is the preference of finite verbs forms over nonfinite verb forms. West Rumelian Turkish, East Rumelian Turkish and Gagauz tend to eliminate nonfinite forms by substituting inflected modal forms for them. This is often ascribed to typical the Balkanism which consists of absence or low frequency of infinitives. Gagauz has replaced infinitive constructions in favor of the Bulgarian model modal word + da construction with the lexical verb in the optative. The verbal noun in -mA: is, however, used in impersonal and same-subject complement clauses. As far as the markers are concerned, Chaghatay, Koman and some other older languages preferred the optative in -GAy. Modern Uyghur and Iranicized Uzbek varieties use the hypothetic marker -sA, seldom -GAy. Ottoman preferred the optative in -(y)A, though at older stages also other modal markers were used. In Balkan Turkish, the optative predominates, though the hypothetic marker can also occur, at least in southwestern Macedonia (Kakuk 1972, Mollova 1968). The use of an oblique mood for a subjunctive function is a grammatical agreement phenomenon and does not provide more information than the modal verb itself does. The choice of a subjunctive mood is often determined by the clause type and its relation to the superordinate predicate. The obligatory use in subjunctive function changes the function of the modal marker into a purely syntactic one and makes it semantically redundant, in a sense like a case governed by an adposition. In many cases, the foreign influence consists in reinforced, increased or expanded use of preexisting patterns. In the framework of the Code-copying model (see, e.g., Johanson 2002), this is regarded as a result of frequential copying, which may lead to increased or decreased occurrence of a feature. The copying may be facilitated by the existence of similar patterns in the model code.

Modals in Turkic

501

5.1. Volition Several Turkic languages have copied analytic constructions in which a modal verb meaning ‘to want’ occurs with a dependent clause with a finite predicate in a subjunctive mood. In different-subject constructions, the dependent clause can be preceded by a junctor. The constructions are thus rather similar to French patterns such as il veut qu’elle vienne ‘he wants her to come’. The subjunctive function is mainly copied from Iranian patterns. It is characteristic of the old Turkic literary languages of Central Asia and of spoken languages that have been in long and intensive contact with Persian and Kurdish. For example, Persian volitional constructions are formed with xa:stan ‘to want’ plus a finite verb in the subjunctive, e.g. mi-xa-m ber-a-m [want-PRS-1SG go-SBJV-1SG] ‘I want to go’ (colloquial; see below). Ottoman Turkish used, under Persian influence, the optative in the subjunctive function, e.g. dile-r-im [ki] ... gel-e-m [want-NFPRS-1SG [JUNCT] come-OPT-1SG] ‘I want to come’, dile-r-im [ki] ... gel-e-sin [wantNFPRS-1SG come-OPT-2SG] ‘I want you to come’ (literally ‘I want may you come’). In the spoken language, these constructions were asyndetic or employed the particle dE ‘also, and’ as a junctor. The third-person voluntative suffix -sUn was used from the 14th century on instead of the optative suffix -(y)A, e.g. buyur kim ... vơr-me-sün-ler [order-IMP.2SG JUNCT give-NEG-VOL-3PL] ‘order them not to give’ (Adamoviü 1985: 241). Colloquial Turkish registers may still use patterns of this kind, almost exclusively with a third-person subject of the dependent clause, e.g. git-sin iste-di-m [go-VOL.3SG want-PST-1SG] ‘I wanted him/her/it to go’ or istemi-yor-um git-sin [want-NEG-PRS-1SG go-VOL.3SG] ‘I don’t want him to go’. Subjunctive constructions have, however, almost completely disappeared from Standard Turkish. In Azeri varieties spoken in Iran, volitional constructions normally contain iste- ‘to want’ plus a subordinated verb in the optative, e.g. ist-ir ala [want-PRS.3SG take-OPT.3SG] ‘he/she wants to take’. The Azeri optative corresponds to the Persian subjunctive. Corresponding Persian volitional constructions are formed with the modal verb xƗstan and a following verb in the subjunctive, e.g. mi-xa-m be-ra-m (< mi-xa:h-am be-rav-am) [PRSwant-1SG go-SBJV-1SG] ‘I want to go’. The particle ki can be used as a junctor, e.g. isti-r ki [want-PRS.3SG JUNCT]. This type has ousted the construction -mA(k) + iste-, which is still in use in Northern Azerbaijan

502

Lars Johanson

(KÕral 2001, 2005). For corresponding constructions in Kashkay, see Csató (2005). Similar patterns are found in other Turkic varieties spoken in Iran. Kowalski noted that the Eynallu dialect frequently uses ise- ‘to want’ with a subjunctive construction which is certainly copied from Persian (“... die sicher dem persischen xwƗstan mit dem Subjunktiv nachgebildet ist” (1937: 66). Khorasan Turkic exhibits similar patterns, e.g. iste-r-em gel-e-m [wantNFPRS-1SG come-OPT-1SG] ‘I want to come’. Iraq Turkic prefers constructions such as ist-ir-em ki gel-e-siz [wantPRS-1SG ki come-OPT-2PL] ‘I want you to come’ (cf. Standard Turkish gelme-niz-i isti-yor-um [come-VN-POSS2PL-ACC want-PRS-1SG]), ist-ir-i yơ:sin [want-PRS.3SG eat-VOL.3SG] ‘he/she wants to eat’ (cf. Standard Turkish ye-mek isti-yor [eat-INF want-PRS.3SG]). The verbal noun in -mA is not used, and the functions of -mAk are heavily restricted (Bulut 2000). Turkish varieties of Cyprus display the same patterns, e.g. isde-r-im gide-siƾ [want-PRS-1SG go-OPT-2SG] (cf. Turkish git-me-ni isti-yor-um um [go-VN-POSS2PL-ACC want-PRS-1SG]); see Demir (2002). In Balkan Turkish, the pattern -mA plus possessive suffix has vanished in different-subject constructions, being replaced by patterns with voluntatives, optatives, conditionals or the old necessitative -(y)AsI, e.g. ben sen-i iste-r-im gel-e-sin [I you-ACC want-NFPRS-1SG come-OPT-2SG], ben seni iste-r-im gel-esi [I you-ACC want-NFPRS-1SG come-NEC] ‘I want you to come’. In all Turkish dialects of Bulgaria, these analytic patterns are also found in same-subject constructions. The Balkan Turkic language Gagauz, closely related to Turkish, has similar constructions, e.g. iste-r-im öl-e-yim [want-NFPRS-1SG die-OPT1SG] ‘I want to die’. In different-subject constructions, the lexical verb must be in the optative. The construction with -mA plus possessive suffix is completely lost in Gagauz. The frequency of the analytic constructions has increased strongly under the impact of constructions such as Bulgarian iskam da + finite verb or xotet’ þtoby + finite verb. All Turkish dialects of Bulgaria have this kind of same-subject optative construction (Németh 1965: 123; Kakuk 1960). In Gagauz as spoken in Moldova it is almost lacking, probably due to the influence of Russian (Menz 1999: 50).

Modals in Turkic

503

5.2. Necessity A similar phenomenon is the subjunctive use of voluntative, optative and hypothetic suffixes in analytic necessitative constructions. In several older and more recent Turkic languages, an adjective meaning ‘necessary’ is used together with a hypothetic marker in a kind of subjunctive function, e.g. Karakhanid -sA kerek, Tuvan -sA xerek, Uzbek -sл kerлk (Kononov 1960: 398) ‘he/she/it must’. Also in Ottoman, the hypothetic marker -sA was used, e.g. al-sa-m gerek [take-HYP-1SG necessary] ‘I should/must take’, al-sa gerek [take-HYP.3SG necessary] ‘he/she/it should/must take’. This pattern was used less and became unusual after the 18th century (Adamoviü 1985: 286). It is, however, still used in modern Turkish to express expectation, presumption or probability, e.g. bugün gel-se gerek [today come-HYP.3SG necessary] ‘he/she/it should/ought to come today’. It is difficult to decide whether the phenomenon just mentioned is contact-induced, but this is undoubtedly the case with certain patterns found in Turkic varieties under strong Iranian or Slavic impact, where an adjective meaning ‘necessary’, kergek, gerek, la:zïm, etc., is followed by a clause with the finite verb in an oblique mood. The modal markers are used like the subjunctive in the Iranian or Slavic constructions. One pattern found in Ottoman is gerek + optative, e.g. gerek var-a-sïn [necessary goOPT-2SG] ‘you should go’. Modern Azeri displays the same pattern, e.g. gerek oxu-ya-k [necessary read-OPT-1PL] ‘we must read, we need to read’. In Azeri varieties spoken in Iran, necessitative constructions are expressed by gerex ‘necessary’ plus optative, e.g. gerex ged-a-x [necessary go-OPT-1PL] ‘we must go’. This corresponds to the Persian construction with ba:yad ‘necessary’ plus a verb in the subjunctive. Instead of gerex, the adjective bå:yed (< Persian ba:yad) may be used, e.g. bå:yed oxi-ya [necessary read-OPT.3SG] ‘he/she must read’. Negation and aspectotemporal markers are added to the verb, e.g. gerex gör-m-ir-di-m [necessary see-NEG-INTRA-PST.COP-1SG] ‘I should not see’. Persian uses the negative prefix na- before ba:yad, while the verb remains in the affirmative form. This is also the case with Azeri patterns containing the copy nebå:yed (KÕral 2005). Iraq Turkic displays constructions such as gerek ơd-e-sen [necessary doOPT-2SG] ‘you must do’ and la:zïm-dï sor-u:-m [necessary-PST ask-OPT1SG] ‘I must ask’ (Bulut 2000).

504

Lars Johanson

The situation of the Balkan Turkish varieties is similar. The necessitative suffix -mAlI has been given up. Under the influence of Bulgarian and Macedonian, Balkan Turkish exhibits patterns with voluntatives and optatives in subjunctive function. We find constructions such as o la:zïm git-sin [he/she/it necessary go-VOL.3SG] ‘he/she/it must go’, sen lazïm gid-e-sin [you necessary go-OPT-2SG] ‘you must go’, la:zïmdïr þalïš-alïm [necessary-COP work-VOL.1PL] ‘we must work’. Compare the Slavic pattern treba da + subjunctive, e.g. Macedonian treba da odam ‘I must go’ (= la:zïm gid-e-yim). Turkish varieties of Cyprus exhibit patterns such as lazïm al-alïm [necessary take-VOL.1PL] ‘we must take’. Gagauz exhibits similar patterns such as o la:zïm-dï oku-sun [he/she necessary-PST.COP read-VOL.3SG] ‘he/she had to read’, in which the combinatory properties of the Bulgarian da construction with the invariable form trjabva plus finite lexical verb are copied. Gagauz -mAlI is almost totally ousted by analytic patterns with an oblique mood. 5.3. Possibility Possibility may be expressed by constructions containing a modal word and a verb in an oblique mood with subjunctive function. This is also an effect of foreign influence. For example, Uzbek may use the construction bolлdi ‘(it is) possible’ with the hypothetic mood in -sA as a subjunctive, e.g. yåz-sл-m bol-л-di [write-HYP-1SG become-PRS-3SG] ‘I can write’. Azeri varieties of Iran, use ơli-ye bil- [do-CVB know] ‘to be able to do’, followed by a verb in the optative. This construction, which can express ability and other kinds of possibility, is influenced by Persian constructions consisting of the modal verb tava:nestan plus a verb in the subjunctive, e.g. ơli-ye bil-er mašin apar-a [do-CVB know-NFPRS car drive-OPT.3SG] ‘he/she can drive a car’; cf. Persian mi-tun-e ma:šin be-bar-e [PRS-be.able-3SG car SBJV-drive.3SG]. Negation can be expressed by ơli-ye bil-me- [do-CVB know-NEG] or by the impossibility marker -(y)AmA added to ơle- ‘to do’ (KÕral 2005: 288-291). Iraq Turkic has a special periphrastic construction for impossibiliy. The verb ơt- ‘to do’ in combination with -(y)Abil- expresses ‘to be unable to’ and is followed by the lexical verb in the optative or voluntative, e.g. ơdebil-m-ir-i yiȖlл-sin [do-POS.NEG-PRS-3SG cry-VOL.3SG] ‘he/she cannot

Modals in Turkic

505

weep’, literally ‘he/she cannot that he/she shall weep’. The corresponding Iranian periphrastic constructions consist of a modal verb (Persian tava:nestan, Kurdish karin), followed by the lexical verb in the subjunctive, e.g. Kurmanji nÕ-kar-Õm bÕ-gir-Õm [NEG-be.able-1SG SBJVweep-PRS.1SG]‘I cannot weep’, literally ‘I cannot that I shall weep’ (Bulut 2000: 165-166). Balkan Turkish displays interesting copies of Slavic possibility constructions with subjunctives. In West-Rumelian Turkish, the verbal form olur ‘it becomes, (it is) possible’ corresponds to Slavic može ‘can’ and combines with clauses in which a finite lexical verb carries an optative or voluntative marker. Bulgarian and Macedonian possess constructions expressing ability and possibility, in which verbs derived from the Slavic verb *mogti, e.g. Bulgarian moga ‘to be able’, are complemented by clauses preceded by the conjunction da ‘that’ and carrying a finite lexical verb in the subjunctive mood. Nonfinite forms of the lexical verb do not occur. The corresponding West-Rumelian Turkish constructions are clearly modeled on this areal pattern. Thus olur corresponds to može da ‘it is possible’ and olur mu to može li da ‘is it possible?’. Gagauz displays possibility constructions containing a finite lexical verb in a subjunctive mood and introduced by the elements var nasïl and var niЋe (often contracted to varïnЋa) ‘there is + how’, e.g. ben varïnЋa gid-i:m [I there.is.how go-VOL.1SG] ‘I can go’. The elements used in the sense of ‘possible’ are copies of Bulgarian ima ‘there is’ and kak ‘how’. Impossibility is expressed with yok ‘there is not’, which corresponds to Bulgarian njama. The question particle mI can stand between var/yok and the following word for ‘how’, e.g. var mï nasïl gid-i:m [is there how goVOL.1SG] ‘can/may I go?’. 6. Levels of grammaticalization The phenomena dealt with above raise several questions regarding the degree of grammaticalization. On processes and parameters of grammaticalization, see Lehmann (1995), Heine and Kuteva (2002, 2005). Let us briefly summarize our findings. Stage 1. The old synthetic devices, most of which are still employed, represent an advanced degree of erosion of their material shape and of desemanticization, suggesting more general meanings open to various interpretations.

506

Lars Johanson

Stage 2. The analytic devices reflect much less advanced degrees. They hardly show any signs of formal erosion, and they can convey more precise meanings than the synthetic ones. Stage 3. The use of dependent clauses containing finite subjunctive verb forms represents a lower degree of grammaticalization than the use of nonfinite forms. In varieties where Stages 1 and 2 coexist, the analytic devices are preferred in more objective register, i.e those that require less ambiguous modes of expression. The synthetic devices are more suited for subjective registers which tolerate relatively vague modal and other attitudinal devices (see Johanson 2006). The frequency of Stage 3, analytic constructions with subjunctive moods, has decreased in some modern languages. The Persian influence during the Ottoman Turkish period has ceased in the modern Turkish period, partly due to a puristic language policy in Turkey. The Persian impact on Azeri is still much stronger. The standard language, however, is based on varieties in the northern part of the Azeri-speaking area that are less Persian-dominated. On the spoken language of Tebriz, see KÕral (2001). The present status of Balkan Turkish is too weak to allow clear statements on the diection of the development. One intricate problem concerns so-called contact-induced grammaticalization. What is the relation between originals and copies in terms of stages of grammaticalization? Grammaticalization proceeds unidirectionally from less to more grammaticalized items. As for the contact-induced renewal of Turkic modal expressions, the copies represent less advanced stages of grammaticalization than their originals. There are differences with respect to semantic, combinational and frequential properties. The copies that make use of subjunctive markers imitate Indo-European subordination, without being subordinative themselves; the dependent clauses are not embedded into a matrix clause. The use of these constructions is often contextually restricted and optional rather than obligatory. Heine and Kuteva, who take contact-induced grammaticalization to follow the same principles of grammaticalization as changes not involving language contact, note: “wherever there is sufficient evidence, it turns out that the replica construction is less grammaticalized than the corresponding model construction” (2005: 101). While grammaticalization can obviously be a result of language contact in areal contexts, the copying act itself is not a grammaticalization process.

Modals in Turkic

507

What is copied is the result of a code-internal process of the model code construction at a specific stage of its path of grammaticalization. The replica construction is less grammaticalized than the model construction. If copying would involve grammaticalization processes, we would, in these cases, be confronted with instances of reverse directionality, i.e. violations of the unidirectionality principle assumed for such processes (see Johanson forthc.). Transcription In transcriptions of morphemes, capital letters represent morphophonemes: V = vowel, I = high unrounded vowel, U = high rounded vowel, A = nonhigh unrounded vowel, G = velar g / Ȗ. Hyphens are used to indicate morpheme boundaries. Acknowledgement Work on this article, a contribution to the Special Research Project (SFB) 295 of the Deutsche Forschungsgemeinschaft, University of Mainz, was carried out while the author was Fellow-in-Residence at the Swedish Collegium for Advanced Studies. References Adamoviü, Milan 1985 Konjugationsgeschichte der türkischen Sprache. Leiden: E. J. Brill. Boeschoten, Hendrik E. 1990 Acquisition of Turkish by immigrant children. A multiple case study of Turkish children in the Netherlands aged 4 to 6. (Turcologica 6.) Wiesbaden: Harrasowitz. Böhtlingk, Otto 1851 Über die Sprache der Jakuten. Grammatik, Text und Wörterbuch. (Dr. A. Th. v. Middendorff’s Reise in den äussersten Norden und Osten Sibiriens 3.) St. Petersburg: Buchdruckerei der Kaiserlichen Akademie der Wissenschaften.

508

Lars Johanson

Brockelmann, Carl 1954 Osttürkische Grammatik der islamischen Literatursprachen Mittelasiens. Leiden: E. J. Brill. Bulut, Christiane 2000 Optative constructions in Iraqi Turkmen. In Studies on Turkish and Turkic languages, AslÕ Göksel, and Celia Kerslake (eds.), 161-169. (Turcologica 46.) Wiesbaden: Harrassowitz. Cafero÷lu, Ahmet, and Gerhard Doerfer 1959 Das Aserbeidschanische. In Philologiae turcicae fundamenta 1, Jean Deny, Kaare Grønbech, Helmuth Scheel, and Zeki Velidi Togan, (eds.), 280-307. Wiesbaden: Steiner. Clauson, Sir Gerard 1972 An etymological dictionary of pre-thirteenth-century Turkish. Oxford: Clarendon Press. Csató, Éva Á. 2005 On copying in Kashkay. In Linguistic convergence and areal diffusion. Case studies from Iranian, Semitic and Turkic, Éva Á. Csató, Bo Isaksson, and Carina Jahani (eds.), 271-283. London, New York: Routledge Curzon. Csató, Éva Á. 2006 Gunnar Jarring’s Kashkay materials. In Turkic-Iranian contact areas. Historical and linguistic aspects, Lars Johanson, and Christine Bulut (eds.), 209-225. (Turcologica 62.) Wiesbaden: Harrassowitz. Demir, Nurettin 2002 ‘Wollen’ in Zyperntürkisch. Mediterranean Language Review 14, 9-20. Deny, Jean 1921 Grammaire de la langue turque (dialecte osmanli). Paris: Éditions Ernest Leroux. Erdal, Marcel 2004 A grammar of Old Turkic. Leiden, Boston: Brill. Friedman, Victor A. 1985 Balkan Romani modality and other Balkan languages. Folia Slavica 7, 381-389. Gabain, Annemarie von 1959 Das Alttürkische. In Philologiae turcicae fundamenta 1, Jean Deny, Kaare Grønbech, Helmuth Scheel, and Zeki Velidi Togan (eds.), 2145. Wiesbaden: Steiner. Heine, Bernd, and Kuteva, Tania 2002 World lexicon of grammaticalization. Cambridge: Cambridge University Press.

Modals in Turkic

509

Heine, Bernd, and Kuteva, Tania 2005 Language contact and grammatical change. Cambridge: Cambridge University Press. Johanson Lars 2000 Viewpoint operators in European languages. In Tense and aspect in the languages of Europe, Östen Dahl (ed.), 27-187. Berlin and New York: Mouton de Gruyter. Johanson Lars 2002 Structural factors in Turkic language contacts. [With an introduction by Bernard Comrie.] London: Curzon. Johanson Lars 2004 On the Turkic origin of Hungarian igen ‘yes’. Acta Orientalia Hungarica 57, 93-104. Johanson Lars 2006 Indirective sentence types. Turkic Languages 10, 73-89. Johanson Lars 2008 Remodeling grammar: Copying, conventionalization, grammaticalization. In Language contact and contact languages, Peter Siemund, and Noemi Kintana (eds.), 61-79. (Hamburg Studies on Multilingualism) Benjamins. Johanson Lars forthc. Notes on Turkic stance particles. In Historical, areal and typological aspects of South Siberian Turkic, Marcel Erdal, and Irina Nevskaya (eds.) Wiesbaden: Harrassowitz. Kakuk, Suzanne 1960 Constructions hypotactiques dans le dialecte turc de la Bulgarie occidentale. Acta Orientalia Hungarica 11, 249-257. Kakuk, Suzanne 1972 Le dialecte turc d’Ohrid en Macédonie. Acta Orientalia Hungarica 26, 227-282. KÕral, Filiz 2001 Das gesprochene Aserbaidschanisch von Iran. Eine Studie zu den syntaktischen Einflüssen des Persischen. (Turcologica 43.) Wiesbaden: Harrassowitz. KÕral, Filiz 2005 Modal constructions in Turkic of Iran. In Linguistic convergence and areal diffusion. Case studies from Iranian, Semitic and Turkic, Éva Á. Csató, Bo Isaksson, and Carina Jahani (eds.), 285-293. London, New York: Routledge Curzon. Kononov, Andrej N. 1960 Grammatika sovremennogo uzbekskogo literaturnogo jazyka. [A Grammar of Modern Standard Uzbek] Moskva, Leningrad: Izdatel’stvo Akademii Nauk SSSR.

510

Lars Johanson

Lehmann, Christian 1995 Thoughts on grammaticalization. München: Lincom Europa. Lewis, Geoffrey L. 1967 Turkish grammar. Oxford: Clarendon Press. Menz, Astrid 1999 Gagausische Syntax. Eine Studie zum kontaktinduzierten Sprachwandel. (Turcologica 41.) Wiesbaden: Harrassowitz. Mollova, Mefkûre 1968 Parler turc de Florina. Balkansko ezikoznanie 13, 95-127. Németh, Julius 1965 Die Türken von Vidin. Sprache, Folklore, Religion. (Bibliotheca Orientalis Hungarica 10.) Budapest: Akadémiai Kiadó. Pokrovskaja, Ljudmila A. 1964 Grammatika gagauzskogo jazyka. Fonetika i morfologija. [A Grammar of Gagauz. Phonetics and Morphology] Moskva: Nauka.

16. Concluding chapter: modal constructions in the languages of Europe Björn Hansen and Ferdinand de Haan "Mögen hätt' ich schon wollen, aber dürfen hab ich mich nicht getraut." (Karl Valentin) 1. ‘Modals’ as a cross-linguistic construction type As emerges from the individual chapters there are diverging traditions in the treatment of modals in the linguistic descriptions of the individual languages. The tradition can even diverge between closely related languages such as Slovak and Polish. The term ‘modal’ is well established in Germanic linguistics and one finds special chapters about ‘modals’ or ‘modal verbs’ in every handbook of English, Danish, Dutch or German grammar. The same holds for the Balto-Finnic languages. In the grammatical descriptions of e.g. Russian and Greek, however, modals usually do not show up as a class sui generis; they are not treated as grammatical markers and are, thus, assigned to the lexicon. Most authors of our book note that it is difficult to draw a sharp line between modals and nonmodals. Kehayov and Torn-Leesik (this volume) characterise the situation in Fennistics by pointing out that ‘the number of verbs included in the class of modal verbs within one language varies in different descriptions. The present volume contains a large amount of data, which makes it possible to draw some conclusions concerning the features of modals from a cross-linguistic perspective. The point of departure for our study was the working hypothesis that modals are polyfunctional word-like expressions of modality which are in a process of grammaticalisation (see introduction). We understand the notion ‘modality’ in a relatively narrow sense, including only necessity, obligation, possibility, permission, probability and volition and excluding discussion of such topics as evidentiality and (ir)realis. In the following we are going to develop the hypothesis that modals form a cross-linguistic category identifiable by its specific semantics and its typical morphosyntactic mode of expression. We will try to show that all European languages have constructions at their

512

Björn Hansen and Ferdinand de Haan

disposal which can be captured by a cluster of gradable semantic, morphological and syntactic features. If all of these features are developed to a high degree (in one construction) we speak of a fully-fledged or central modal: A fully-fledged modal is a polyfunctional, syntactically autonomous expression of modality which shows a certain degree of grammaticalisation. ‘Polyfunctional’ is understood as covering a domain within the semantic space of modality. A fully-fledged modal functions as an operator on the predicational and/or the propositional level of the clause.

Modals represent conventionalised multidimensional pairings of form and function and can therefore be treated as a construction type in the sense of Construction Grammar (e.g. Goldberg 2003). In the following we shall use the terms modal and modal construction (henceforth ModCxn) interchangeably. The category is based on prototypicality, i.e. it is not defined by means of a set of discrete necessary and sufficient properties, but by a cluster of gradable attributes. Such a category has fuzzy boundaries and can overlap with neighbouring categories. In a prototypically structured category not every member is equally representative; there are central or core and peripheral members. The data presented in this book clearly show that modals form categories with fuzzy boundaries. Our treatment of the category modals allows for a certain degree of structural heterogeneity, i.e. in the individual languages, the category can contain structurally different elements. The idea of modals as a graded category containing various subcategories is in line with the traditional treatment in English linguistics. Note, however, that the distinction we make between ‘fully-fledged’ (or ‘central’) and ‘peripheral modals’ is not identical to the contrast ‘central vs. marginal modal’ used in English linguistics to distinguish between must and ought to, for instance, must being the central modal, ought to the marginal one (e.g. Quirk et al. 1985: 137).1 The latter distinction is based on the presence or lack of the language-specific formal features associated with modals in English (so called NICE properties and others)2. In our definition both must and ought to would be ‘central modals’, since ought to shows clear signs of grammaticalisation: it is polyfunctional (deontic and epistemic meaning) and shows the morphosyntactic behaviour of an auxiliary, although it does not share all of the morphosyntactic features of the modals of the type must and can. Whereas modality as a functional domain can safely be assumed to be a universal concept, modals as specific morphosyntactic means of expression

Modal constructions in the languages of Europe

513

of modal notions are to be treated as a typologically relevant, but not as a universal category. In our sample we found one language in which modals seem to have appeared only recently due to language contact (the Berber language Taqbaylit), and cases of languages lacking modals completely are reported in the linguistic literature. There are languages where modality is expressed by irrealis morphemes, and not by separate words. This happens in Haruai (Comrie 1991) and other Papua New Guinean languages and also to a high degree in Berber. On the other hand, modals are not a specific European category, as they are also found outside of Europe (see section 6 for details). Modals share features with neighbouring categories, i.e. with lexical expressions on the one hand, and with modal affixes on the other hand. They differ from both lexical elements and affixes in their hybrid nature: they are morphologically more or less autonomous words and fulfil grammatical functions, a property usually associated with auxiliaries. The point of departure of our considerations is the observation that modals seem to vacillate between the status of lexical and grammatical elements. Heine (1993) and Anderson (2006) try to grasp this oscillating nature of auxiliaries by locating them on a grammaticalisation chain extending from lexical verbs to fully fledged inflectional markers. Heine defines auxiliaries as “linguistic items covering some range of uses along the Verb-toT(ense)A(spect)M(odality) chain” (ibid. 70). An auxiliary “is no longer a fully lexical item, but not yet a grammatical inflection either, and it is likely to exhibit properties that are characteristic of the intermediate stages between fully lexical items and inflectional forms” (ibid, 86). According to this understanding, there are no necessary and sufficient features constituting a category of auxiliaries or modals as categories sui generis. In contrast to Heine’s and Anderson’s view, we postulate that modals constitute a sort of a focal point on a grammaticalisation chain which, however, does not go in the direction of the emergence of inflectional markers. Our data show that modals tend to become morphologically defective, but they usually do retain their word status. We found several instances of the transition into an uninflected element (e.g. Greek prepi, or Albanian mund); i.e. modals can acquire a particle-like form, but rarely agglutinate to the modified verb. There is no language in Europe with a class of modals comprising a closed set of elements. In all languages modals form classes with a small core and a ‘fuzzy’ periphery overlapping with neighbouring categories. This also holds for the Germanic languages, which are unique in that the majority of modals show a dedicated morphological and morphosyntactic marking. We find elements which

514

Björn Hansen and Ferdinand de Haan

share only some, but not all features characteristic of that class; e.g. the borderline cases of the so called ‘emerging modals’ in English (Krug 2000) or Danish turde ‘to dare’ (see the chapter by Mortelmans, Boye, and van der Auwera). Both would be treated as peripheral modals in our framework. This fuzziness of the category also explains the heterogeneous treatment of modals e.g. in the descriptions of the Balto-Finnic languages (see above). After having discussed the internal structure of the category, we can proceed to semantics. Semantic polyfunctionality is understood in relation to the universal semantic map of modality: an element is polyfunctional iff it covers more than one primitive function within the map (for the terms function and domain see de Haan 2005).3 Modals are expressions of modality which can be distinguished from purely lexical elements with modal meaning like English possibility or Hungarian szükségszerĦ ‘necessary’ by their polyfunctionality which is the result of semantic shifts typical of a grammaticalisation process. One frequently distinguishes between three types of modality: dynamic, deontic and epistemic modality. In all European languages, it is possible to distinguish polyfunctional modals from lexical content words with modal meaning, i.e., words which are not subject to an auxiliarisation process and which have only one modal meaning; compare the Polish core modal móc ‘can’ and the modal content word potrafiü ‘to be capable’ The former can express either ‘capacity’ (dynamic), ‘objective possibility’ (dynamic), ‘permission’ (deontic) or ‘perhaps’ (epistemic), while the latter is confined to ‘capacity’. Grammatical polyfunctionality may extend beyond modality, i.e. may reach into functions which following van der Auwera and Plungian (1998) can be called postmodal meanings like evidentiality or tense (e.g. Danish skal, English shall). Hence, polyfunctionality turns out to be a crucial feature distinguishing modals from lexical elements. This does not preclude modals from having additional lexical meanings; e.g. the Hungarian modal kell ‘must’ is a polyfunctional expression of modality which shows only few traces of grammaticalisation on the morphological and syntactic level and which at the same time is used in the lexical sense ‘to be needed’ (see the chapter by Körtvély). Kell still has many features of a lexical verb and should therefore be assigned to the first stages of the grammaticalisation chain which implies its status as a peripheral modal. In contrast to that, Greek bori ‘can’ is highly polyfunctional in the domain of modality and has no additional lexical meanings. These cases illustrate that modals vary in respect to their position on the content word > modal-chain. Greek bori covers later stages on the grammaticalisation chain.

Modal constructions in the languages of Europe

515

Modals may share this polyfunctionality with modal affixes like the Hungarian potential –hAt or the Latvian debitive jƗ-. What distinguishes them is the morphological form: whereas affixes are fixed to the verbal stem, modals retain a certain degree of independence in relation to the verb they modify. In this respect, they share morphological features with lexical elements (on the morphology of modals see sections 3.2 and 3.5). For capturing the function of modals in the sentence, we adopt the notion of operators as proposed in functional syntactic models. In this sense, modals are characterised as operators on the predicational and/or on the propositional level of the clause. If the modal has a dynamic or deontic meaning it works on the level of the predication and helps to locate the state of affairs in a real or imaginary world. In the case of epistemic meanings the modal functions on the higher level of the proposition as one of the means through which the speaker specifies his attitude towards the truth of the proposition. It is important to note that we do not ascribe to the view that modals imply the subordination of a verbal complement encoded in a non-finite form. Neither do we assume that modals have to carry TAM-marking and agreement with the subject (cf. also Anderson 2006 passim). As a matter of fact, the syntactic homogeneity found in English is very rare and modals in the individual languages tend to be used in various syntactic surroundings. As Elšík and Matras show in their analysis of modals in Romani dialects, modals of one and the same language can form a bewildering variety of morphosyntactic constructions.4 This languagespecific variation depends on the inventory of syntactic templates available in the individual languages. Almost all Modal Constructions show features associated with clause linkage; features typical of bi-clausal sentences are e.g. the use of non-finite forms of the lexical verb as in Germanic or the presence of a complementiser as in the languages of the Balkan area. Other modals, however, occur without complementisers and behave like serial verbs or particles. This leads us to the conclusion that modals cannot be identified by syntactic dependency relations. We shall deal with the question of syntactic and semantic scope relations in section 3.4.

516

Björn Hansen and Ferdinand de Haan

2. Distribution of modals and modal affixes in Europe There are two main ways in which modality has been grammaticalised in the languages of Europe: either as modal verbs or as modal affixes. There are other ways, such as modal particles, but they fall outside the scope of the paper, as they are rarely polyfunctional. Grammatical mood has also been omitted (except in conjunction with modal elements), as this warrants a separate study. We have included modal adverbs/adjectives (or even nouns, as in Basque behar ‘need’), as they can be polyfunctional and together with auxiliary (sometimes called ‘light’) verbs form ModCxns. This section surveys the distribution of modals, ModCxns and modal affixes. We can see a strong areal distribution here. ModCxns can be found all over Europe. Modals are of course best known from the Germanic and Romance language families, but they are found in most language families discussed in this volume, the notable exceptions being the Berber and Turkic families, as well as Basque. That is not to say that modal verbs behave the same in all languages, for there are striking differences from language to language with respect to selectional criteria and degrees of grammaticalisation. We will come back to these differences in section 3 below. Some examples of modal verbs are: (1)

English John must go to school.

(2)

French On doit attendre. one/we must.PRS.3SG wait.INF ‘We must wait.’

(3)

Maltese jista’ jkun ma can.IPFV3M.SG be.IPFV3M.SG NEG ‘I might forget’

(4)

Finnish Sinun täytyy nukkua. you.GEN must.3SG sleep.INF ‘You must sleep.’

niftakar=x remember.IPFV1SG-NEG

Modal constructions in the languages of Europe

517

ModCxns with adjectives or adverbs (or nouns) can be found in many language families as well, but they seem most prevalent (most grammaticalised) in Slavonic. They can be found in Celtic, Baltic, Basque and Albanian as well. (5)

Albanian Në rregull, mund të largohesh. in order can COMP leave:PRS.PASS.2SG ‘Okay, you can (may) leave.’

(6)

Irish B’ be-PST ‘I had to’

(7)

éigean dom necessity for.me

Russian Ivan dolžen byl rabotat’. Ivan.NOM must.SG.M be-PST work-INF ‘Ivan had to work.’

Modal affixes are prevalent in the eastern part of Europe. It is not always easy to distinguish between modal affixes and mood, but we adopt de Haan’s (2006) heuristic that modal affixes are never obligatory, while mood generally is (see also section 3.3 on paradigmatic variability for a discussion on obligatoriness). Among the languages analysed in the present study the following ones have modal affixes: a) Balto-Finnic language family: Finnish, Karelian, Votic, Ingrian, and possibly Livonian; b) Turkic language family: Turkish, Gagauz, Azeri, Uzbek, c) Hungarian and d) Latvian (see map 1). It seems that modal affixes are typical of agglutinating languages located in the Eastern part of the European linguistic area and less typical of the IndoEuropean languages. The only Indo-European language with a modal affix is Latvian which is known to have been under considerable influence from Balto-Finnic languages. However, note that the modal affix is prefixal rather than suffixal. (8)

Finnish sano-ne-t say-POT-2SG ‘You can say.’

518 (9)

Björn Hansen and Ferdinand de Haan

Turkish ben yüz-e-me-m I swim-CVB-NEG.POT-1SG ‘I can’t swim.’

(10) Hungarian Ebbe a házba akárki bejö-het. this.ILLAT the house.ILLAT anybody come.in-POT.3SG ‘Anybody is allowed to come into this house.’ (11) Latvian Tev mani jƗ-pavada. you.DAT me.ACC DEB.accompany ‘You must accompany me.’

Map 1. Areal distribution of modal affixes

Modal constructions in the languages of Europe

519

3. The grammaticalisation of modals in the languages of Europe In this section we will discuss the ways in which modals and ModCxns have been grammaticalised in the languages of Europe. We do this by discussing the parameters established by Lehmann (2002, originally 1981). There are six parameters: three paradigmatic ones and three syntagmatic ones. As we shall see, not all of these parameters are equally important, but all of them do tell us something about the grammaticalisation of modals and the nature of grammaticalisation in general. 3.1. Integrity: semantic sources and erosion The integrity of a sign is “its possession of a certain substance which allows it to maintain its identity, its distinctness from other signs, and grants it a certain prominence in contrast to other signs in the syntagm.” (Lehmann 2002:112). There are two subtypes: phonological integrity and semantic integrity. Phonological integrity concerns the reduction of phonological words and a reduction of words into affixes. This does not seem to play a big role in the area of modals: in most of the languages of Europe, main verbs that take on modal meanings do not lose their word status, nor do they develop a different phonology in their modal meanings from their main verb meanings. The only exceptions seem to be English will > ‘ll, and will not > won’t, although that is mostly used as a future marker (note that modals such as must, can and may are not reduced, but any material that is attached to these modals reduces, viz. must have > musta). Also, Breu notes that Albanian modals like mund ‘can’ are reduced in the Geg dialects. The Hungarian modal affix –hAt, which comes from the full verb hat ‘to get further’, may fit in here as well (Körtvely, this volume), but even though a main verb is turned into an affix, there does not seem to be any phonological reduction of the stem. Morphological reduction does seem to play a role, and it will be discussed in the next two sections. Apart from these cases, it seems that phonological reduction is not a factor in modals and hence that any theory that states that modal affixes are formed from modal verbs are not supported by the data from European languages. Semantic integrity deals with ways in which the linguistic sign adds new, more grammaticalised (and hence more abstract) layers of meaning and sheds old ones. This process is usually referred to as bleaching or

520

Björn Hansen and Ferdinand de Haan

desemanticisation. In the case of modals it means that a modal element adds the abstract meanings of possibility, ability and necessity (or, even more abstractly, the meanings of weak and strong deontic, dynamic and epistemic modality), and loses its original main verb meaning (referred to by van der Auwera and Plungian 1998 as the premodal meaning). This has happened with the core modals in Germanic, where, for instance, the original meaning of English may, ‘to be strong or able, to have power’, has long since been lost. However, in most of the languages surveyed, the modal meanings of possibility and necessity are added to the original, main verb meanings. This means that there is no bleaching of original meanings, but rather a layering of meaning. Such layering can be found in many European languages. One example from Balto-Finnic is the verb saada which has a premodal meaning of ‘to get’ and a variety of modal meanings. In Estonian there are five (Kehayov and Torn-Leesik, this volume, using the terminology of van der Auwera and Plungian 1998): (a) participantinternal, (b) participant-external non-deontic possibility, (c) deontic possibility, (d) epistemic possibility, (e) participant-external non-deontic necessity. Some of these are more common than others, but the example shows that it is possible for a modal to add meanings without shedding its premodal meaning. A striking case is the Hungarian affix –hAt, which comes from the (still extant) main verb hat ‘to get further’ > ‘to affect’. In a previous stage of Hungarian there was an auxiliary verb hat ‘to be able’ which is now lost and the affix can only express various types of possibility, but not, apparently, ability. This means that we are dealing with layering, but also with bleaching, although not of the premodal meaning, but of a dynamic meaning. This could be seen as a case in which an intermediate step is lost: from the grammaticalisation path “premodal > ability > possession” it is the intermediate step that is lost. The fact that modals can layer without bleaching means that the Germanic and Romance cases, in which bleaching of modals has taken place, is atypical for the languages of Europe. Cases like Hungarian and Balto-Finnic show that bleaching is not a necessary component of grammaticalisation.5 There are many ways in which main verbs acquire modal meanings. We will now look at some paths that have been attested in more than one language family (for examples see the individual chapters). í ‘want’ > ‘need’ > ‘must’. This path is attested in Hungarian, BaltoFinnic, Arabic and Romance, among others.

Modal constructions in the languages of Europe

521

í ‘possession’ > ‘must’. This can be seen in Slavonic, Germanic, Romance, Baltic, Arabic and Hungarian. í ‘strength’ > ‘power’ > ‘possibility’. This is found in Baltic, Slavonic, Germanic and Albanian. í ‘get’ > ‘possibility’. This is the category of ‘acquisitive’ modals (see van der Auwera, Kehayov, and Vittrant in print). It is attested in Irish, Balto-Finnic and Icelandic.6 í ‘know’ > ‘can’ > ‘possibility’. This is found in Hungarian and Arabic. 3.2. Paradigmaticity The parameter of Paradigmaticity refers to “the formal and semantic integration both of a paradigm as a whole and of a single subcategory into the paradigm of its generic category”. Furthermore, “the members of the paradigm [must be] linked to each other by clear-cut paradigmatic relations, especially opposition and complementarity” (Lehmann 2002: 118). So, in order to ascertain whether modal elements form a paradigm, we need to look for formal criteria for paradigmaticity. Such criteria may be of two types: morphological and (morpho)syntactic. They may be language-specific or language familyspecific, and only if there are no formal criteria available should semantic criteria be used. That is, if we cannot find morphological or morphosyntactic evidence for a distinct category of “modal” in a given language, we are left with defining the category on semantic grounds, which means that the category of modals in that language will be diverse morphologically or syntactically. For morphological criteria we can identify three possible ways to set modals apart from other verbs: they have a unique morphology, they have a defective morphology, or they have impersonal morphology. The unique morphology is attested in the Germanic languages. The core modals are derived from preterite-present verbs. Their verbal inflection is not defective in any way, but is due to historical developments in the verbal system. This has other implications as well, as the preteritepresent verbs do not require an infinitive marker, unlike other auxiliaries, see below. Defective morphology can be seen in those languages in which modals have lost their non-finite morphology. That is, modal verbs are no longer able to have infinite and participial forms. This is attested in a number of European languages, including English, of course, but also in Irish

522

Björn Hansen and Ferdinand de Haan

(McQuillan, this volume). In Albanian, the strong modal duhet has two options: either it is inflected only for tense, in which case it governs the subjunctive or it is fully inflected, and it then governs the past participle. In Baltic (Holvoet, this volume) infinitival forms do occur, but are rare, a fact that may point to an ongoing development toward the loss of non-finite morphology. The other way in which a modal may show defective morphology is by loss of morphological distinctions within a paradigm. This happens for instance in Dutch, where the second person present tense is either formed by adding a –t (as is normal for Dutch verbs), or it has a zero ending, like the other singular forms. This only happens with modal verbs, and only with the preterite-presents. This development could be seen as a kind of paradigm regularisation, or it could be linked to other factors, such as a differentiation between epistemic and deontic modality. This regularisation does not seem to happen in German or Danish. Finally, a modal may become used only impersonally, that is, there is one dedicated verb form (e.g. the third person singular) that is used for all persons (and tenses). The verb may also still be used with personal forms but with different modal meanings. This is for instance the case with Estonian pidama ‘must’, which has personal endings for deontic modality while epistemic modality is limited to the third person Conditional and Imperfective. Also, Greek boro can be used both personally and impersonally, yielding different interpretations. Only the impersonal form can express epistemic modality, for instance. (12) Greek Bori/*Borun na can-3SG/can-3PL COMP ‘They may have arrived.’

irșan. came-3PL

On the other hand, the personal forms are favoured for agent-oriented deontic modality: (13) Boris/?Bori na psifisis can-2SG/can-3SG COMP vote-2SG ‘You can vote if you are 18.’

an ise if be-2SG

18. 18

Finally, a modal may become fixed, at which point it is indistinguishable from a modal adverb (such as English maybe, from [it] may be [that] …) or the Finnish form lie- which is reanalysed as a member of a non-inflectional class.

Modal constructions in the languages of Europe

523

The difference between personal and impersonal has implications for the syntagmatic criteria as well, see section 3.4 below. As far as the (morpho)syntactic criteria for determining paradigmaticity are concerned, we are obviously dealing with language-specific criteria. Criteria that have been used to set modals apart from other (auxiliary) verbs include (for details the reader is referred to the individual chapters): í English: í Germanic: í Cont. W. Germanic í Romance: í Greek: í Balto-Finnic í General:

The NICE properties The possibility of combining modal verbs with an infinitive marker The Infinitivus Pro Participio construction (IPP) Cliticisation phenomena Constraint against co-occurrence of particles Atypical marking of the actor in impersonal constructions Loss of subcategorisation frames (which goes together with semantic bleaching see 3.1 above)

In general, in paradigms, members tend to exclude each other. However, in most cases it seems to be possible for modals to co-occur if they express a different type of modality. Lehmann (2002:118) also mentions as evidence for paradigmaticity the “sheer size of the paradigm” (although it is a “superficial aspect”). In that respect we may not be looking at paradigmaticisation, given that in many cases we are not dealing with a particularly large “paradigm”. For instance, in many languages there are few grammaticalised modals (Modern Greek, Albanian and French, to name but few). In other languages the paradigm may be larger (i.e. there are many modal verbs or ModCxns, as is the case in Germanic and Balto-Finnic), but in those cases there is free variation, something which goes against the notion of a paradigm. Lehmann (2002:120) also points out that paradigmaticity is often invoked in traditional grammars by giving a name to a category. In the present case in many grammars of the languages under discussion it is possible to find a category of “modal verbs”, which would point to a degree of paradigmaticity. However, as discussed in many of the papers in the present volume, this category is usually defined on semantic rather than paradigmatic grounds. Finally, it is noted that paradigmaticity usually implies exclusion. That is, items belonging to the same paradigm cannot co-occur. However, with

524

Björn Hansen and Ferdinand de Haan

very few exceptions (English modal verbs and the Greek particles), ModCxns can co-occur if they express a different type of modality. 3.3. Paradigmatic variability Paradigmatic variability “is the freedom with which the language user chooses a sign” (Lehmann 2002:123). In the context of ModCxns this means that the speaker faces a restriction of choice based on some criterion or another. If choice is fully restricted, we are dealing with obligatorification (Lehmann 2002:124). Paradigmatic variation does not seem to play a big role in the area of modals and ModCxns. In most cases, the choice of a ModCxn is not restricted in any way. It is normally up to the speaker which modal s/he chooses. However, there are a number of ways in which some level of restriction can be observed. There are some modals that require that the subject of the modality is animate. We find this mainly in languages of Eastern Europe, e.g. Estonian and Slavonic. The Estonian verb tule ‘must’, is incompatible with inanimate subjects, and instead the verb pea must be used. (14) Estonian a. Müts pea-b peas olema. cap must(=hold)-3SG head-INE be.INF ‘A cap must be worn.’ b. *Mütsil tule-b peas olla. cap-ADE must(=come)-3SG head-INE be.INF ‘A cap must be worn.’ (Uuspõld 1989: 475) The same is true for Russian, where the modal adverb nado is only used with animate subjects. If the subject is inanimate, then dolžen must be used. (15) Russian a. Kniga dolžna byt’ book-NOM.SG must-SG.F be-INF ‘The book should be translated.’

perevedena. translate-PTCP-SG.F

b.*Knige nado byt’ perevedenoj. book-DAT.SG must be-INF translate-PTCP-INS.SG.F ‘The book should be translated.’

Modal constructions in the languages of Europe

525

This restriction can disappear over time, of course. In early Irish, the verbs féad and caith take only human (or animate) subjects, but in Modern Irish there is no restriction on the nature of the subject. These are clearly cases of restriction of choice. There is, as can be seen, in these languages always a modal or ModCxn available to express the same level of modality if the subject is inanimate. Such restrictions do not lead to gaps in the paradigm. Within Germanic, a number of factors can be determined that restrict choice, but these do not lead to gaps in the paradigm either. In Dutch, and German, and (to a maybe lesser degree) Danish and Icelandic non-finite forms of the modals are associated with deontic modality only. If one wishes to express epistemic modality, then a finite form must be chosen (deontic modals may be finite or non-finite).7 The only place in which we can see true obligatorification is in Greek, where the particles na, as and tha have become grammaticalised. That is, in Tsangalidis’ analysis (this volume), these particles covary according to tense-aspect and modality. The modal verbs prepi and boro / bori, however, are not restricted in any way, as they can co-occur. Based on these considerations we have to agree with our contributors that paradigmatic variability is not very important in the area of modals and ModCxns. There is always a way in which modality can be expressed with any possible syntactic or semantic criterion. 3.4. Syntactic scope – condensation The structural scope of a unit is the structural size of the construction which it helps to form (Lehmann 2002: 128). From a semanto-syntactic point of view, modals are characterised as operators on the predicational and/or on the propositional level of the clause. In the following we shall show how this semantic function finds its expression on the morphosyntactic level. Our point of departure will be the assumption that ModCxns are the result of a grammaticalisation process leading to the far reaching condensation of two clauses into a single one. Our understanding of the syntactic nature of ModCxns shares features with Anderson’s approach to the syntax of auxiliary verb constructions which he defines as ‘mono-clausal structure[s] 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’ (Anderson 2006: 7). Our approach to ModCxns, however,

526

Björn Hansen and Ferdinand de Haan

differs from Anderson’s view in two significant respects. First, we claim that ModCxns are located on a continuum between bi-clausal and monoclausal structures and second, we do not assume that modals are necessarily derived from verbal elements and therefore we allow for adjectival and adverbial-like modals. In the following, we will describe the ‘structural size’ of the ModCxns in terms of the mechanisms of clause linkage. According to Lehmann (1988) tight clause linkage is based on the interlacing of elements shared by the two predicative elements which leads to the desententialisation of a clause, and the explicitness of the linking. The mechanism of interlacing is relevant if we have a look at the syntactic encoding of the subject argument within the ModCxn and at the question of which of the markers indicating agreement with the subject – i.e. number, gender etc. – are shared. The same holds for the sharing of the grammatical operators tense, aspect and mood. The second parameter concerns the overt marking of the linkage between the two predicative elements. This is usually done by connectors, in the case of ModCxns by complementisers. The data show that modals can appear in different morphosyntactic surroundings depending on the language-specific inventory of constructions. Generalizing the morphosyntactic typologies of ModCxns developed on the basis of Romani and Slavonic (this volume), we propose that ModCxns in the languages of Europe vary with respect to the following features: i) ii) iii) iv) v)

the syntactic encoding of the subject argument, assignment of the subject agreement marking to the modal and/or the main verb, TAM marking on the modal and/or the main verb, presence or absence of an auxiliary or light verb, presence or absence of a complementiser.

3.4.1. The syntactic encoding of the subject argument The encoding of the subject is a feature which depends on the inventory of subject constructions available in each language. As we cannot possibly offer even an approximate account of possible subject constructions in the languages of Europe we will restrict ourselves to some basic observations. In our understanding, ModCxns with a certain degree of grammaticalisation usually contain only one subject. We do find modal elements which allow for complex sentences with different subjects, however. These modals display a low degree of condensation and are to be located at the

Modal constructions in the languages of Europe

527

lexical end of the cline leading from elements forming fully complex sentences to mono-clausal constructions. In the following examples, each clause contains its own subject: (16) German Ich will, dass Ivan kommt. I want.1SG COMP Ivan come-3SG ‘I want Ivan to come here.’ (17) Russian Vam

nado,

þtoby ơtu

reklamu

you.DAT.PL necessary COMP this-ACC.SG ad-ACC.SG

proþitalo read-PST-N.SG

maksimal’noe koliþestvo ljudej. maximal-NOM.SG number-NOM.SG people-GEN.PL ‘What you need is that the ad should be read by a maximal number of people.’ Among the typical mono-subject constructions, we find ModCxns where the subject is coded either in the default case (Nominative), in an oblique case, or where it is deleted. In the latter two cases we are dealing with the downgrading of the subject. Compare the different instantiations of the first argument of the mono-valent verb rabotat’ ‘to work’ in Russian: (18) Russian Ivan možet rabotat’. Ivan.NOM can-3SG work-INF ‘Ivan can work.’ (19) Ivan-u možno rabotat’. Ivan-DAT possible work-INF ‘It is possible for Ivan to work.’ (20) Ø Možno rabotat’. possible work-INF ‘It is possible to work.’ Whereas languages like Russian, Lithuanian and Romani make use of all three strategies, Germanic modals do not appear in constructions with a subject marked by an oblique case.

528

Björn Hansen and Ferdinand de Haan

(21) Dutch Ze moet oponthoud gehad hebben. have.PTCP have.INF she must.PRS delay ‘She must have been delayed.’ (22) German Hans kann arbeiten. Hans.NOM can.3SG work.INF ‘Hans can work.’ It is interesting to note that the German syntactic inventory does include constructions with experiencers encoded in oblique case, but these are restricted to lexical elements and do not apply to polyfunctional modal verbs; cf.: (23) German Es ziemt sich nicht für einen Sportler it behove-3SG self not for ART.ACC.SG sportsman überheblich gegenüber anderen zu sein. arrogant towards other-DAT.PL to be.INF ‘It is not proper for a sportsman to be arrogant towards other people.’ The languages which allow for the encoding of the subject argument in oblique case differ in the cases used. Slavonic exclusively makes use of the dative, Finnish mainly uses the genitive, and Turkish and Hungarian a nominal possessive marker: (24) Finnish Sinun täytyy nukkua. you.GEN must.3SG sleep.INF ‘You must sleep.’ (25) Turkish Bil-me-m lâzÕm. know-INF-POSS.1SG must ‘I must know.’

Modal constructions in the languages of Europe

529

(26) Hungarian Holnap reggel korán kell kelnem. tomorrow morning early must get up.INF.POSS.1SG ‘I have to get up early tomorrow morning.’

Map 2. Languages lacking ModCxns downgrading the subject8 Whereas ModCxns with subjects in default cases are attested in all languages of our sample, ModCxns with downgraded subjects are found in most, but not all languages. As shown in map 2, the Germanic languages, Sorbian, Basque, Arabic and Kabyle are the only languages which seem to lack ModCxns downgrading the subject. Thus, it turns out that the Germanic languages are not very typical in their specialisation on personally constructing modals.

530

Björn Hansen and Ferdinand de Haan

3.4.2. Assignment of subject agreement to the modal and/or the main verb As ModCxns show features characteristic of condensed bi-clausal structures it is interesting to look at the subject agreement patterns. A typical ModCxn contains only one subject, but has two predicative elements which may agree with the subject in relation to gender, number, person and other categories. Apart from that, there are so-called impersonal constructions whose predicate is marked with a default morpheme expressing non-agreement. Subject-predicate-agreement can be marked in three ways: only on the modal (27a), only on the main verb (27b), or on both the modal and the main verb (27c). The latter type involves a lower degree of syntactic condensation: the two clauses do not share any agreement operators. Particularly interesting in this respect is Serbian, as it allows all three agreement patterns: (27) Serbian a. Ivan i Slobodan mora-ju radi-ti Ivan and Slobodan must-3PL work-INF b.

Ivan Ivan

i and

Slobodan Slobodan

treba da rad-e should.3SG COMP work-3PL

c.

Ivan i Slobodan mora-ju da rad-e Ivan and Slobodan must-3PL COMP work-3PL ‘Ivan and Slobodan must/should work.’

There seems to be a correlation between certain agreement marking types and semantics; thus, the type ‘agreement marking only on the main verb’ is quite frequent in epistemic usages: e.g. in Serbian and Arabic we find epistemic modals marked for the 3rd person singular. The fourth possibility already mentioned above is the complete lack of subject-predicate agreement. These are the so-called impersonal constructions: (28) Finnish Romani MƗn mote lel must take.SBJV.3SG[=INF] I.ACC ‘I have to take this medicine.’

tauva tram. this medicine.NOM

Modal constructions in the languages of Europe

531

(29) Russian Ivanu možno rabotat’. Ivan-DAT possible work-INF ‘It is possible for Ivan to work.’ (30) Italian9 Bisogna chiamare un medico. must-IND.PRS.3SG call-INF a.M.SG doctor.M.SG ‘Someone has to call a doctor.’ 3.4.3. TAM marking on the modal and/or the main verb In the traditional understanding, auxiliaries are held to carry all TAMmarkers whereas the lexical verb remains non-finite. As the data show, modals may, but need not carry the TAM-marking of the whole verbal phrase. It has to be pointed out that although TAM and subject-predicate agreement are often marked on the same element, in principle they are independent of each other; i.e. modals may carry TAM-markers without subject agreement and modalised verbs can carry subject agreement without TAM-markers, but not vice versa. This independence can again be illustrated by Serbian modals, which are characterised by a high degree of constructional variability. In example (31) the modal is in the conditional and carries the non-agreement marker of 3rd person singular neuter, whereas the lexical verb shows agreement with the first person singular subject, but has no conditional or tense marking: (31) Serbian To biTM

That COND.3SG

ja trebaloTM vas

da

I should.3SG you.ACC.PL COMP

pita-mAgr.

ask.IND-3SG

‘I should have asked you that.’ This contrasts with the following sentences where both the modal and the modalised verb agree with the subject, whereas the TAM-marking is restricted to the modal: (32) Serbian Mora-m da

id-em. vs Mora-o

must-1SG COMP go-1SG

sam

must-PST.M.SG AUX.1SG

‘I have to go.’ vs ‘I had to go.’

da

id-em

COMP go-1SG

532

Björn Hansen and Ferdinand de Haan

Constructions which split the inflectional categories of TAM and agreement between the modal and the lexical verb will be called split constructions. Following Anderson (2006) we distinguish between split patterns where TAM and agreement categories are scattered across the modal and the lexical verb and split/doubled patterns where “some categories are marked on either the auxiliary verb and/or the lexical verb alone, while others are marked on both” (ibid. 183). Parallel to the agreement patterns, there are four different instantiations of the marking of tense, aspect and mood in ModCxns. First, TAM can be marked exclusively on the modal, as for example in the Germanic and Romance languages: (33)

German Ivan kann arbeiten. Ivan.NOM can.IND.PRS.3SG work-INF ‘Ivan can work.’

(34)

Italian Bisogna chiamare un medico. must-IND.PRS.3SG call-INF a.M.SG doctor.M.SG ‘Someone has to call a doctor.’

Second, it can be marked only on the lexical verb, whereas the modal itself remains uninflected. This type can be illustrated by the following Albanian examples where the modal mund ‘can’ remains unchanged and tense is marked on the main verb by a combination of the verbal inflection with a specific complementiser (particle të vs ta): (35) Albanian Në shtëpinë in

e

tij mund të

house:ACC.SG.DEF PTL:POSS.ACC.SG.DEF

his can

COMP.SBJV

gjesh gjithfarë tabelash. find:SBJV.PRS.2SG all.kinds chart:ABL.PL.INDF ‘In his house you can find all kinds of charts.’ (36) Librin

e

botës

mund

book:ACC.SG.DET PTL:POSS.ACC.SG.DEF world:GEN.SG.DEF can

ta

shkruante

PCL:SBJV-it:ACC.SG write:IMPRF.3SG

vetëm ai. only

‘Only he could write the book of the world.’

he:NOM.SG

Modal constructions in the languages of Europe

533

Some languages have modals which syntactically behave like adverbs; i.e. they combine with fully inflected verbs without complementiser and inflect neither for TAM, nor for subject agreement;10 e.g. Slovene lahko which goes back to an adverb with the meaning ‘easily’: (37)

Slovene Tu smo lahko sreþali […]. find-PST-PL here AUX.1PL can ‘Here we could find […].’

At this point, the question arises how modals are related to discourse particles. As a matter of fact, in many languages modals have split into two elements: the modal itself and an epistemic particle (=sentence adverb). It is interesting to note that the latter lose polyfunctionality and, thus, move into the neighbouring category of discourse particles. These findings square with Ramat and Ricca (1998: 231ff) who analysed sentence adverbs in a wide range of European languages: they found that epistemic particles very often emerge either through the fusion of a modal with a second element (= univerbation) or the conversion of a modal: a) fusion of the components ‘modal.3SG’ + ‘to be’; e.g. English maybe (m may + be), French peut-être ‘perhaps’, Russian možet byt’ ‘perhaps’, Moroccan Arabic xΩ‫܈܈‬u ykuun ‘probably’; b) fusion of the components ‘modal.3SG’ + ‘to happen’; e.g. Swedish kanske, Danish måske, Dutch misschien all with the meaning ‘perhaps’; c) fusion of the components ‘modal.3SG’ + complementiser’ (‘can/must that’); e.g. Serbian/Croatian možda ‘perhaps’, valjda ‘probably’, Slovene morda ‘perhaps’; d) conversion of a modal; e.g. Polish moĪe and Romanian poate ‘perhaps’ or ‘can.3Sg’; Slovak možno ‘perhaps’ or ‘one.can’. All sentence adverbs differ from modals with adverbial morphology like Slovene lahko in their mono-functionality. A third variant of TAM-marking in ModCxns is found in Arabic where the operators are doubled: we are dealing with the juxtaposition of two finite verbal forms, both regularly marked for TAM. This construction type can be considered less condensed because both clauses display the full range of operators.

534

Björn Hansen and Ferdinand de Haan

(38) Maltese Arabic il-mara tiegƫ=i ma tista=x issajjar ART-wife of-1SG NEG can.IPFV3F.SG-NEG cook.IPFV3F.SG gƫax marid=a because ill-F ‘My wife cannot cook because she is ill’. (39) dawn kwaĪi setgƫu kienu sunetti DEM.PL almost can.PFV3PL be.PFV.3PL sonnets ‘These could almost have been sonnets.’ The fourth type of TAM-marking involves the presence of a second auxiliary or light verb. Here, the structural size of the construction, which the modal helps to form, is larger than in the cases where the ModCxn consists exclusively of the modal and the lexical verb plus its arguments. This auxiliary type will be dealt with in section 3.4.4 below. As already hinted at, TAM and agreement marking patterns in ModCxns are not evenly distributed among the European languages. The data show the Mediterranean area and South Eastern Europe to be the hotbeds of the use of finite lexical verbs. Map 3 contains four isoglosses: First, ModCxns with double-marking for both TAM and agreement are found in the analysed varieties of Arabic; second, doubled agreement marking is found in the Balkan languages in the form of split/doubled patterns, apart from the doubled pattern found in Arabic. The third isogloss shows the more general distribution of all split and split/doubled patterns, i.e. constructions where TAM is marked on the modal and the agreement on the lexical verb, irrespective of singular or double marking. These ModCxns are found in an even larger area comprising not only the mentioned varieties of Arabic and the Balkan languages, but also the Turkic languages and Hungarian. The latter two have ModCxns where the modal carries TAM-marking and the lexical verb is in a nominalised form marked with a possessive suffix for the subject. These are pure ‘split patterns’ in the sense of Anderson (2006). Finally, ModCxns with uninflected particle-like modals are mainly found on the Balkans: Albanian, Greek, Bulgarian, and Slovene.

Modal constructions in the languages of Europe

535

Map 3. Distribution of Agreement and TMA-marking in ModCxns

(ņ ņ ņ): double TMA and AGR marking (doubled pattern) (ņņņ): double AGR marking (doubled and split/doubled pattern) (- - - -): split AGR and TMA marking (split/doubled and split pattern) (ƒƒƒƒƒƒƒƒƒ): AGR and TMA marking exclusively on the lexical verb 3.4.4. Presence or absence of an auxiliary or light verb In contrast to the linguistic tradition and also in contrast to Heine (1993) and Anderson (2006) who claim that auxiliaries are generally derived from verbs, we do not assume that modals obligatorily carry verbal features. Thus, we found many instances of polyfunctional modals originally going back to other parts of speech. The elements which are not derived from verbs are usually accompanied by a second auxiliary bearing the TAM marking. This is the case with adverbial-like modals in Russian (in the

536

Björn Hansen and Ferdinand de Haan

present tense a zero element) or with nominal modals in Irish; both need the copula: (40) Russian a. Možno rabotat’. possible work-INF ‘It is possible to work.’ b. Možno bylo rabotat’. possible AUX-PST-SG.N work-INF ‘It was possible to work.’ (41) Irish Níorbh fholáir dó éirí NEG-be-PST excess for.him get.up-VN ‘It was not excess for him to get up = He had to get up’ Another construction type with a second auxiliary is represented by Basque ModCxns which contain the light ‘verb’ izan and one of the modal elements behar ‘necessity’, nahi ‘volition’ or ahal ‘possibility’. The construction has the internal structure ‘have + X’ where X is the modal lexeme that may also function as a noun. (42) Basque Nagusiak etorri behar du. boss.ERG come.PRF need X.has(.Y) ‘The boss has to come.’ 3.4.5. Presence or absence of a complementiser The last syntactic feature according to which the ModCxns in the languages of Europe vary is the explicitness of linking between the modal or the modal-auxiliary complex with the lexical verb. The linking is usually marked by a specific morphological form of the lexical verb. Some languages use infinitival forms, while others use the subjunctive marked e.g. for person and number. As a ModCxn is the result of the condensation of two clauses into a single one, some constructions contain a complementtiser originally functioning as a clause linkage marker. The use of (former) complementisers in ModCxn is typical of the Balkan region:

Modal constructions in the languages of Europe

(43) Albanian Në shtëpinë in

e

537

tij mund të

house:ACC.SG.DEF PTL:POSS.ACC.SG.DEF

his can

COMP

gjesh gjithfarë tabelash. find:SBJV.PRS.2SG all.kinds chart:ABL.PL.INDF ‘In his house you can find all kinds of charts.’ (44) Serbian Ivan i Slobodan treba da rade. Ivan and Slobodan should.3SG COMP work-3PL ‘Ivan and Slobodan should work.’ (45) Greek Boris na pijenis can-2SG/can-3SG COMP go-2SG ‘You may go now.’

tora. now

(46) Romani (Kaspiþan Xoraxane, Bulgaria) Može te džas kaj COMP go.SBJV.1PL to can ‘We can go to town.’

gav. village

The presence of a complementiser can be interpreted as a lower degree of condensation of a bi-clausal sentence into a mono-clausal one. This becomes even more evident in cases where modals allow different complementation patterns: either a non-finite form without or a finite form with a complementiser, as attested in Hungarian: (47) Hungarian Haza kell mennem. home must go.INF.POSS.1SG ‘I have to go home.’ vs (48) Az kell, hogy hazamenjek. that must COMP home.go.SBJ.1SG ‘It is necessary that I go home.’

538

Björn Hansen and Ferdinand de Haan

Here, the complementiserless construction (47) contains an untensed verbal form marked for the subject by means of a possessive suffix. In example (48), however, the construction shows more features of a complex sentence, because it contains the complementiser hogy and a fully-fledged finite form of the verb. Modals in different languages may vary in respect to the possibility to govern complement clauses introduced with a complementiser. For example, the Russian impersonal modal of necessity nado can regularly govern a complement clause introduced by the complementiser þtoby, whereas its Polish impersonal synonym naleĪy in a similar construction sounds odd: (49) Russian Nado, þtoby ljudi uznali, necessary COMP people-NOM.PL learn-PST-PL takoe svoboda. that-NOM.N freedom (50) Polish ?NaleĪy, Īeby ludzie doznali must-3SG COMP people-NOM.PL learn-PST-PL to jest wolnoĞü. that be.3SG freedom ‘The people should learn, what freedom is.’

þto what

co what

3.4.6. Conclusions We have seen that modals form a wide range of differing morphosyntactic constructions, both within one language and across languages. In this respect, the relative morphosyntactic homogeneity claimed for English seems to be rather exceptional. This diversity can be captured by a morphosyntactic typology based on the mechanisms of clause linkage among which we identified the interlacing of elements shared by the two predicative elements and the explicitness of the linking as the most crucial ones. The morphosyntactic typology is based on the features ‘encoding of the subject’, ‘agreement marking’, ‘TAM-marking’, ‘presence of an auxiliary/light verb’ and ‘presence of a complementiser’. The data from many languages reveal that modals in epistemic function tend to lose grammatical markers, thus showing a higher degree of condensation than dynamic and deontic meanings. We found many instances of epistemically

Modal constructions in the languages of Europe

539

used modals developing into uninflected discourse particles with exclusively epistemic meaning. 3.5. Bondedness Bondedness is defined as ‘the intimacy with which a sign is connected with another sign to which it bears a syntagmatic relation. The degree of bondedness of a sign varies from juxtaposition to merger, in proportion to its degree of grammaticality’ (Lehmann 2002: 131). Any increase in bondedness is called coalescence which starts as juxtaposition. The next step is the subordination of the modal element under an adjacent accent (cliticisation). Further coalescence leads to agglutination and finally ends with the grammaticalised element becoming an integral part of the morpheme modified. Recall that ModCxns are defined here as syntactically autonomous expressions of modality which implies that the parameter bondedness is used to distinguish modals from modal affixes. Thus, a modal per definitionem is characterised by a low degree of bondedness, i.e. by the simple juxtaposition with the lexical verb it modifies. In this section we shall focus on two aspects. First, we will present certain syntactic symptoms indicating the very beginning of the coalescence of juxtaposed elements which can be observed before phonological consequences like cliticisation make themselves felt. Second, we would like to discuss the relation between modals and modal affixes by looking for the very few cases of a transition of the former into the latter. It will be shown that ‘modals’ and ‘modal affixes’ do not form discrete categories, but represent two focal points on a common grammaticalisation chain. In respect to first ‘hidden’ symptoms of coalescence, our data are relatively scarce and do not present a full coverage of all modals in all languages, but do allow for some hypotheses. One of the syntactic symptoms of a beginning coalescence process is the impossibility of inserting linguistic material between the modal and the modified verb. Körtvély (this volume) notes that all modals of Hungarian allow the insertion of constituents between the modal and the lexical verb except the modal of possibility talál ‘might’. In the following example, the string ‘modified verb plus modal’ cannot be interrupted by the adjunct phrase ‘in the bush’:

540

Björn Hansen and Ferdinand de Haan

(51) Hungarian Ha a bomba robbani if

the bomb

vs (52) *Ha a bomba robbanni if

the bomb

talál

a

bokorban, fuss.

blow up.INF might.3SG the bush.INE

a bokorban talál,

blow up.INF the bush.INE

run.SBJ.2SG

fuss.

might.3.SG run.SBJ.2SG

‘If the bomb happens to blow up in the bush, run.’ Slavonic and Greek modals need not necessarily be adjacent to the modified verb and allow other elements to be inserted between them. Exceptional in this respect is Italo-Albanian where the modals have to be placed immediately before the main verb and tend to develop into proclitics (see Breu, this volume). An interesting case of coalescence is found among the new peripheral English modals11 have got to, have to and want to which in the spoken language show the contracted forms gotta, hafta and wanna as in : (53) English […] You gotta come along. (example from Krug 2000: 72) Who do you wanna succeed? (example from Krug 2000: 141) I mean, you hafta do those things to be successful anyway. (example from www) These are cases of univerbation where the modal element coalesces with the morpheme to originally belonging to the verbal form. Coalescence is also reported in cases of the transition of modals into epistemic particles as mentioned above: 1) ‘modal.3SG’ + ‘to be’ as in English maybe, 2) ‘modal.3SG’ + ‘to happen’; e.g. Dutch misschien ‘perhaps’ and 3) ‘modal.3SG’ + complementiser’ as in Serbian/Croatian možda ‘perhaps’. A purely syntactic symptom of a slow increase of bondedness is the inseparability of the two elements under coordination reduction, which can be illustrated by the German morpheme zu which shows a higher degree of bondedness than its English counterpart to, because it does not allow coordination reduction, i.e. it is more firmly attached to the infinitival verb (Lehmann 2002: 133f.): (54) German a. Er beabsichtigt das he intend-3SG the

Problem zu beschreiben problem to describe-INF

und and

Modal constructions in the languages of Europe

541

zu erklären. to explain-INF b. *Er beabsichtigt he intend-3SG erklären. explain-INF

das the

Problem problem

zu beschreiben und to describe-INF and

(55) English He intends to describe and explain the problem. In the case of modals, we are dealing with the question if clauses like X can p and X must p can be reduced to X can and must p or not. The same holds for the possibility of the deletion of the modal: X can p and X can q becoming X can p and q. Since most contributors to our language survey posit a very low degree of coalescence and do not explicitly address the question of coordination reduction, we do not have much to say about it. Suffice it to point out here that Germanic, Hungarian and Slavonic modals freely allow constructions like: (56) English Globalisation can and must change. You can change and delete data sources. A related symptom of incipient coalescence which intersects with the parameter of structural scope is the impossibility of the use of the grammaticalised element in elliptic contexts, i.e. some modals do not allow the omission of the modified verb. Thus, the Romance modals devoir, falloir (French), a trebui (Romanian), tener que, deber (Spanish) and dovere (Italian) can be used in elliptic constructions in their deontic reading, but not in their epistemic use. This contrasts with the modals haber de, haber que (Spanish), avoir à (French) and a avea de (Romanian) which cannot be used without an infinitival (Cornillie et al. this volume). From these findings, we can conclude that epistemic usages display a higher degree of coalescence than dynamic and deontic meanings (see below). Without addressing the question whether we are dealing with ellipsis or with an intransitive use of the modal we can claim that the parameter of bondedness is also relevant for the possibility of a modal to occur with non-infinitival complements. Mortelmans, Boye, and van der Auwera (this volume) show that English modals in contrast to Dutch, German and

542

Björn Hansen and Ferdinand de Haan

Danish demand an infinitival verb and are incompatible with verbless complements: (57) English *He may/can/must/wants home. ‘He may/can/must/wants to go home.’ (58) Dutch Hij mag/kan/moet/wil

naar huis.

(59) German Er darf/kann/muss/will nach Hause. home he may.PRS/can.PRS/must.PRS/will.PRS to Dutch modals are shown to have the widest range of verbless complements; they even combine with adjectival predicates not allowed in German: (60) Dutch Deze fles moet vol/leeg. this bottle must.PRS full/empty (61) German *Diese Flasche muss voll/leer. this bottle must.PRS full/empty ‘This bottle must be filled/be emptied.’ Having discussed the earliest stages of coalescence processes, we can proceed with the few instances of the transition of modals into clitics and affixes. There are some rare cases of the split of a modal into two elements: a regularly juxtaposed modal on the one hand, and a clitic or even agglutinated affix on the other. The Serbian/Croatian volitional modal hteti has additionally developed a future meaning in which case it loses a syllable and cliticises to the verbal stem or another constituent usually occupying the first syntactic slot in the clause:

Modal constructions in the languages of Europe

(62) Serbian ja hoüu pevati ovu I want-1SG sing-INF this-ACC.SG ‘I want to sing this song.’ vs (63) a. ja üu pevati ovu I FUT.1SG sing-INF this-ACC.SG b. pevaüu (peva-ti + üu) sing-FUT.1SG ‘I will sing this song.’

543

pesmu song-ACC.SG pesmu song-ACC.SG

ovu pesmu this-ACC.SG song-ACC.SG

Note that hteti cliticises exclusively in its temporal function; it does not do so in its modal meanings. This shows that modality implies a lower degree of bondedness than tense. The second attested case of a more advanced bonding between modal and lexical verb is found in Turkish, where the modal of possibility bil- can be used either as a juxtaposed modal or as a firmly attached suffix. This is a case of agglutination. It is worth pointing out that Turkish is a very typical agglutinating language with a wide range of grammatical functions expressed by affixes and less often by independent words; e.g. passive, reflexive or causative. (64) Turkish Yüz-mek bil-ir mi-sin? swim-INF can-NFPRS12 Q-2SG (65) Yüz-e-bil-ir mi-sin? swim-CVB-can-NFPRS Q-2SG ‘Are you able to swim?’ If a modal element shows further symptoms of coalescence it develops into a fully-fledged modal affix. These affixes are firmly attached to the verbal stem with which they form an accentual unit. It is noteworthy, however, that not all modal affixes historically go back to juxtaposed modals. See the discussion in section 2. We come to the conclusion that modals are inherently characterised by a low degree of coalescence. The data show that the modals in the languages of Europe in general do not show any tendency towards increasing coalescence. Modals only rarely cliticise and develop into

544

Björn Hansen and Ferdinand de Haan

affixes. Modal affixes are not typical of the European linguistic area; they are mainly found in the agglutinating languages of Eurasia. 3.6. Syntagmatic variability “The syntagmatic variability of a sign is the ease with which it can be shifted around in its context” (Lehmann 2004: 140). As pointed out by Norde (in print) syntagmatic variability and bondedness can be seen as one and the same parameter being relevant for different stages of grammaticalisation, because elements with a high degree of bondedness are inherently fixed in a certain position. In this sense, fixation occurs in the first stages of grammaticalisation and, thus, precedes the later ongoing process of coalescence. In the case of modals, syntagmatic variability concerns the positional mutability of the modal with respect to the lexical verb with which it forms a construction. If the modal can either precede or follow the lexical verb we are dealing with a low degree of fixation and the more fixed the word order becomes, the higher the degree of grammaticalisation. However, we have to add the caveat that this parameter has to be seen in the wider context of the general word order rules available in the individual languages. Therefore, the degree of fixation of a ModCxn has to be determined in comparison to similar constructions in the individual language. Among the ModCxns of our language sample we found ModCxns with diverging degrees of fixation of the modal in relation to the lexical verb. Russian, Polish, Greek and most Hungarian ModCxns are reported to allow for pragmatically triggered reversals of the unmarked word order ‘Modal precedes lexical verb’ which is an indicator for a low degree of fixation. This is illustrated by the following Russian examples, where the modal nel’zja ‘it is not allowed/it is impossible’ occupies the unmarked pre- or the marked post-verbal position: (66) Russian a. ýego nel’zja delat’ v Indii? what.GEN is.not.allowed do-INF in India-PREP ‘What are the don’ts in India?’ b. Rebenok zabolel: þego delat’ nel’zja? child get.ill-PST what.GEN do-INF is.not.allowed ‘Your child has become ill: the don’ts.’

Modal constructions in the languages of Europe

545

In contrast, Serbian/Croatian modals never appear after the main verb. They have to be positioned before the verb irrespectively of the realisation of the main verb as an infinitive or the da-construction. The same holds for the Romance modals of necessity. Cornillie et al. (this volume) point out that Latin debere could either precede or follow the infinitival verb whereas its cognates in the modern Romance languages have to be used with the infinitive to their right, which is an increase of fixation. We can conclude that ModCxns are not characterised by the coalescence of the modal with the lexical verb, but frequently show a certain increase of fixation which sets them apart from more lexical constructions. 3.7. ModCxns and asymmetries of degrees of grammaticalisation As emerges from the chapters dealing with the parameters of grammaticalisation, ModCxns show diverging degrees of grammaticalisation on the following levels: 1) differences between individual meanings of one and the same modal, 2) differences between modals of one and the same language and 3) differences between modals of different languages. When determining degrees of grammaticalisation we have to bear in mind that the point of comparison should be lexical structures of the language analysed; i.e. we have to take into consideration the inventory of morphosyntactic structures of the language in question. Ad 1: Several authors note that modals in their epistemic use tend to show a higher degree of grammaticalisation than in their dynamic or deontic meanings; e.g. an epistemic modal can be morphologically more defective as it is limited to the 3rd person form as Estonian pidama ‘must’. Germanic modals and many Romance modals of necessity can be used in elliptic constructions in their deontic reading, but not in their epistemic use, which shows that epistemic usages display a higher degree of coalescence and a narrower structural scope than dynamic and deontic meanings (see Cornillie et al and Mortelmans et al. this volume). Ad 2: Some data suggest that there is a tendency for modals of possibility to be grammaticalised before modals of necessity. This can either correlate with the pure ‘age’ of the ModCxn or the degree of grammaticalisation. According to Elšík and Matras (this volume) Early Romani can be reconstructed to have possessed two or three possibility modals and one volition modal, but no dedicated necessity modals. A similar situation is suggested for the Slavonic languages which have inherited a common modal of possibility from their ancestor language.

546

Björn Hansen and Ferdinand de Haan

Proto-Slavonic *mogti ‘to be strong’ developed into a possibility modal before the break-up of the Slavonic linguistic unity and can be considered by far the oldest modal. The dedicated modals of necessity, however, are the result of later developments thus leading to a great diversity among the Slavonic languages in this semantic domain (see Hansen 2001). Further evidence for the privileged status of possibility is found in Hungarian where the possibility expression talál is considered the most grammaticalised of all modals. Ad 3: As each parameter of grammaticalisation allows the comparison of degrees of grammaticalisation also across languages, we can see that there are languages whose inventory of ModCxns tend to be located at the lexical end of the grammaticalisation cline; e.g. East Slavonic and Hungarian ModCxns seem to be closer to lexical constructions whereas English modals have developed many specific features setting them apart from lexical verbs. A rather high degree of grammaticalisation is also found in the Balkan languages, whose modals tend to develop into uninflected particle-like elements. 4. Modals and language contact: borrowing asymmetries and areal convergence The next topic we would like to address is the behaviour of modals in situations of language contact. Until recently, there were no cross-linguistic studies available on the question of how modals in general react in language contact situations. The known borrowing scales (e.g. Thomason 2001) operate with discrete dichotomies like content words vs. function words and claim that the former are more easily transferred than the latter. Apart from that, it is generally held that nouns are more easily borrowed than verbs. The categories used in the borrowing scales do not take into consideration the hybrid nature of modals. We are, however, in a lucky position, because we can build on the work by Elšík and Matras who generalise important results of the recently published cross-linguistic studies on the ‘borrowability’ of grammatical elements carried out in the framework of the Manchester Romani Project (Elšík and Matras 2006) and the project ‘Grammatical borrowing in Cross-Linguistic Perspective’ (Matras and Sakel 2007). We will follow Matras and Sakel (Sakel 2007) who distinguish between the borrowing of matter (MAT) and the borrowing of pattern (PAT). The authors speak

Modal constructions in the languages of Europe

547

“of MAT when morphological material and its phonological shape from one language is replicated in another language. PAT describes the case where only the patterns of the other language are replicated, i.e. the organisation, distribution and mapping of grammatical or semantic meaning, while the form is not borrowed” (Sakel 2007: 15)

Elšík and Matras (this volume) and Matras (2007) claim that ‘modality is a domain that is conspicuously susceptible to structural borrowing’. As they show, the Romani modal systems are characterised by massive borrowing of matter and pattern from the second languages spoken by Romani speakers. The authors list the following MAT-borrowings for necessity: Turkish lâzÕm, Greek prepi, South Slavonic trjabva, treba(ti), mora(ti) and valjati, Romanian trebui and musai, Hungarian muszáj, Slovak musieĢ, maĢ, treba and potrebovaĢ, Polish musieü, trzeba and powinien, East Slavonic dolžen, nado and prixoditsja, German müssen and brauchen, English ought, Swedish måste and må, and Finnish pitää. MATborrowings for possibility are somewhat less frequent: Greek boro, Macedonian and Bulgarian može, Slovene lahko, Polish móc, East Slavonic moþ’ and možno, German dürfen, Italian potere, and Finnish voida. Due to the dialectal diversity and the multitude of language contacts, these Romani data are highly relevant also to other languages and allow for some generalisations. Apart from that, the findings based on Romani are corroborated by the data compiled in the book Matras and Sakel (2007) which contains descriptions of grammatical borrowing in 27 languages spread over all continents of the world. Elšík and Matras (this volume) and Matras (2007: 45) claim that the borrowing behaviour of modals is highly asymmetrical; i.e. some modal categories are more likely to be borrowed than others. The overall likelihood of modals to be affected by borrowing is expressed by the following implicational hierarchy: necessity > possibility > volition Necessity appears at the top of the implicative scale. It is the most frequently borrowed semantic category and possibility and volition are not borrowed unless necessity is borrowed too (Elšík and Matras this volume). Although the authors of the individual chapters of our book were not explicitly asked to give a contact-linguistic account of the respective modal system, we received a large amount of data which allow for some generalisations about borrowability. As it turns out, our data clearly confirm Matras’ borrowing scale: from the some fifty languages of Europe

548

Björn Hansen and Ferdinand de Haan

(apart from the varieties of Romani) around one half has reportedly borrowed at least one modal from another language. Table 1 contains a – non-exhaustive – list of MAT-borrowings reported of in the individual chapters: Table 1. MAT-borrowings of modals Borrowing language

Borrowed from language

Modal

Semantics

Estonian Karelian Karelian, Veps Polish Czech Slovak Upper Sorbian Lower Sorbian Ukrainian Belorussian Latvian Hungarian Romanian Turkish

Middle High German Russian Russian German German German German German German via Polish German via Polish Balto-Finnic German South-Slavonic Arabic Arabic Arabic Arabic Arabic Arabic Russian Arabic Arabic Greek

pruukima dolžen prišlos’ musieü muset musiet’ dyrbjeü musaĞ musyty music’ vajadzƝt muszáj a trebui lâzÕm mümkün la:zïm va:jïb la:zïm mumkin na:da laq mken préps mborés epitrépsetΩ

¬NEC NEC NEC NEC NEC NEC NEC NEC NEC NEC NEC NEC NEC NEC POSS NEC NEC NEC POSS NEC NEC POSS NEC POSS POSS

Azeri Gagauz Uzbek13 Yakut Berber Arvanitika (Greek Albanian)

In the vast majority of cases, the languages borrow one modal of necessity; in some cases, i.e. Turkish, Azeri, Berber and Greek Albanian, we find MAT-borrowings for both necessity and possibility, which is well compatible with Matras’ borrowing hierarchy; the scale predicts that modals of possibility can be MAT-borrowings if modals of necessity are borrowed too. In the view of the rare borrowings in the semantic field of possibility it is not surprising that we found even fewer instances of MAT-

Modal constructions in the languages of Europe

549

borrowing of volitional modals. Among the authors of our book only Elšík and Matras (this volume) report of two cases: the verb wånt-(a)s- from English in Welsh Romani and the verb hoþ- from Slovene in Dolenjski Romani of Slovenia. No further MAT-borrowings of volitional modals are attested. Other cases of contact-induced replacements of volitional modals are PAT-borrowings like Balkan Romani which uses the verb mang- ‘to ask, to demand’ copying Bulgarian iskam to express volition. MATborrowing of modals seems to be quite normal in contact situations when a language with a lower degree of prestige comes into contact with a culturally dominant language, which is associated with high prestige functions as the language of education. Such a high prestige language is German which used to be a widespread language of education in large parts of Central and Eastern Europe. Therefore, the German modal müssen has become so popular in Central and Eastern Europe that it has í partly mediated by Polish í found its way into six Slavonic languages and Hungarian. A similar sociolinguistic role was played by Arabic which has considerably influenced the modal systems of the Turkic and Berber languages and by Russian which as the dominating language of Russia has been the model language for the Eastern branch of the Balto-Finnic (e.g. Karelian and Veps) and the Siberian Turkic languages like Yakut. Apart from the fact that neighbouring languages can share structural features we have to assume similarities in the text frequency of morphosyntactically similar ModCxns. Van der Auwera, Schalley and Nuyts (2005) investigate the Slavonic translational equivalents for English auxiliaries and adverbs of epistemic modality (may, might, could, maybe, perhaps), as they occur in Harry Potter and the chamber of secrets. The focus is on whether verbs and adverbs are translated by elements of the same word class. The authors show that the Slavonic translations differ substantially and systematically according to the parameter of distance from the Germanic linguistic area. 5. Modals and Euroversals / SAE-features This section offers some thoughts on the findings of the previous sections in relation to the question of “Standard Average European” (SAE, Haspelmath 2001). Specifically we ask whether modal verbs and ModCxns are as widespread in the rest of the world as they are in Europe. We contrast this with modal affixes, which seem more prevalent outside

550

Björn Hansen and Ferdinand de Haan

Europe. Haspelmath (2001: 1493) gives four criteria for determining whether a given feature is part of SAE: a) The great majority of (core) European languages possesses it. b) The geographically adjacent languages lack it. c) The Eastern Indo-European languages lack it (i.e., Armenian, IndoIranian and Indo-Aryan). d) The feature is not found in the majority of the world’s languages. We have shown that criterion (a) is met: the vast majority of the languages of Europe possess modal verbs and ModCxns. They are found in all language families in Europe, from Celtic in the west to Finno-Ugric in the East, albeit with varying degrees of grammaticalisation (see section 3). In contrast, only a few possess modal affixes, a feature which appears to be more prevalent outside of Europe (see map 1). Criterion (b) is a little harder to analyse, as it is not quite clear how literally the notion “geographically adjacent” should be taken. Haspelmath (2001: 1493) refers to “Celtic in the west, Turkic, eastern Uralic, AbkhazAdyghean and Nakh-Daghestanian in the east, and perhaps Afro-Asiatic in the south” and this does not quite correspond to our findings. For one, Celtic and (eastern) Uralic definitely fit in with the SAE model of modal verbs. The same can be said for Arabic. The varieties of Arabic discussed in Vanhove, Miller, and Caubet (this volume) all make use of modal verbs. Haspelmath (ibid.) mentions that Maltese exhibits a number of SAE features that are lacking in other varieties of Arabic (among them participial passives and a dative external possessor), but all varieties of Arabic discussed here have some system of modal verbs. The precise modal inventory differs from language to language, but that is a situation found in many of the core language families of SAE. Basque is a language that usually does not participate in the SAE Sprachbund, and it does seem that its system of modality is more akin to non-European languages. It does, however, have some ModCxns, such as behar izan ‘need’, consisting of the noun behar ‘need’ and the auxiliary izan. This is not too unlike the Slavonic constructions of the Russian type nado / nužno byt’. Given this state of affairs, it is unlikely that criterion (b) is true for modals. According to criterion (c) of SAE-ness, modals should not, or rarely, occur in the eastern Indo-European languages. However, it would appear that modals can and do occur in these languages. In Hindi (Kachru 2006: 82-4) there is a category of modal auxiliaries, although it is not a homogeneous category. For instance, the modal verb sΩkna ‘can, be able’

Modal constructions in the languages of Europe

551

takes the full range of tense-aspect affixes and the subject is in the Nominative. The verb pΩѵna ‘have to, must’ also takes the normal tenseaspect morphemes, but it takes a Dative subject. The modal cahiye ‘should’ also takes a Dative subject, but it is invariant. In short, it is just as difficult to describe a category of ‘modal verb’ in Hindi as it is in many core SAE families. Similar remarks may be made of Indo-Aryan languages in general (see Masica 1991: 374ff for an extensive discussion) and also of IndoIranian (Persian, for instance, has modal verbs as well). This means that criterion (c) is not met: modals are found in the non-European IndoEuropean languages, apparently very similar to modals in the languages of Europe and with very similar grammaticalisation paths. Criterion (d) concerns the presence of modals in the rest of the world. Because we are not aware of an exhaustive survey of modals around the world, we can only make some preliminary observations. We will concentrate on modal verbs, because we have very little information on the status of ModCxns world-wide. Nevertheless, it seems that we have enough information to note that criterion (d) is not met, either. Modal verbs do not seem to be widespread outside of Indo-European with the exception of E. and S.E. Asia. In languages from this area, we do find systems of modal verbs that are reminiscent of European systems. Languages with modal verbs in this area include Mandarin (Hsieh 2005), Cantonese (Matthews and Yip 1994), Khmer (de Haan 1997), and Thai (Iwasaki and Ingkaphirom 2005). Outside of the S.E. Asian area we can find languages with modal verbs, but it seems to be rarer. One such language is Yoruba (Adéwӑlé 1990, de Haan 1996). In many instances, the attested modal verbs in these languages are polyfunctional. It is certainly true for the languages described in the previous paragraph. For instance, Thai has the following systems of modals (from Iwasaki and Ingkaphirom 2005: 133): Table 2. Thai modals Epistemic

Deontic

Gloss

tэըƾ cam-pen khuan nâa

tэըƾ

must necessary should should may may

(khuan) nâa khoƾ àat

552

Björn Hansen and Ferdinand de Haan

(67) Thai a. khun cà tэըƾ maa dûay ná you CM must come also PP ‘You must come too, all right?’ b. kháw tэըƾ pen mэիэ he must COP doctor ‘He must be a doctor for sure.’

(134) nѓը- nѓըѓ surely (138)

lΩΩy PP

As can be seen from the table, most of the strong modal verbs (those denoting ‘must’ and ‘should’) are polyfunctional, while the weak modals (denoting ‘may’) are epistemic only. As mentioned, the weak modals are not polyfunctional, but are epistemic only. However, there are modal elements, classified as ‘potential’ by Iwasaki and Ingkaphirom, that occupy a place near the end of a sentence. One of which, dây (from the verb dây ‘get, obtain’), is polyfunctional: it can express epistemic (68a), dynamic, and deontic (68b) modality. (68) a. phrûƾ-níi chûay maa thíi bâan dây máy tomorrow help come at house POT Q ‘Can you come to my house tomorrow?’ (350) b. khҁҁ aw alay maa khuy kэը dây LK take something come chat LP POT ‘I mean you can bring anything to talk about.’ (351) Polyfunctionality is not just limited to modal verbs. As shown in van der Auwera, Ammann and Kindt (2005: 253) it also occurs with modal affixes (for example, the W. Greenlandic affix -ssa) and particles (the Ainu particle kuni). Van der Auwera and Ammann (2005) show that the situation is relatively complex outside Europe, as far as polyfunctionality is concerned. In their study of 207 languages they found that the notion of polyfunctionality is sensitive to the distinction between strong and weak modality. Their breakdown is as follows: í polyfunctionality in both weak and strong modality 36 í polyfunctionality in either weak or strong modality 66 í no polyfunctionality 105 However, most of the languages that use the same modal elements for epistemic and deontic modality in both weak and strong modality (26 out

Modal constructions in the languages of Europe

553

of 36) are either in Europe or just outside Europe (this includes Turkish and Modern Hebrew). The remaining 10 languages are mainly in Asia (including Mandarin Chinese and Thai, but also Tagalog and the Papua New Guinea language Yimas) with only one language in sub-Saharan Africa (Yoruba) and none in the Americas (if we don’t count West Greenlandic). This points to a strong areal influence. The 66 languages that have polyfunctionality in either weak or strong modality (but not both) are more evenly spread. They can be found on every continent, so an areal claim is harder to make here. However, it also includes such languages as Icelandic, Welsh and Albanian as having only partial polyfunctionality, something which is not quite substantiated in the present volume. The 105 languages in which there is no polyfunctionality at all are also more or less evenly spread among the world’s languages. It is interesting to note that the only languages in Europe in the survey that show no overlap are Saami and Votic, two Balto-Finnic languages (but see Kehayov and Torn-Leesik, this volume). This means that criterion (d) is not met either: there are languages outside of Europe that have modal verbs. The only possible difference between European and non-European languages is that modal verbs outside of Europe tend to occur in languages with very little, if any, inflectional morphology, such as Thai, Khmer, Yoruba and Mandarin Chinese. As the European languages are still highly inflectional, this may point to the one true difference between the SAE languages and the languages outside Europe. It could be argued that this difference may save criterion (d), but in the present paper we have chosen to accept the fact that criterion (d) is not met. This means that there is one criterion for SAE-hood that is met (modals occur in most languages of Europe), but that there are three that are not met. From this, we have no choice but to conclude that the feature of [+ModCxn] is not part of SAE. In the terminology of Heine and Kuteva (2006), [+ModCxn] is not a Euroversal. Finally, even though modal constructions are not unique to the European area as a whole, but nevertheless there are areal issues on a smaller scale. Indeed, as shown in the previous section, such areal issues play important roles in the shaping of the modal systems of various languages. Looking for Euroversals does not mean that one also abandons the search for areal influences.

554

Björn Hansen and Ferdinand de Haan

6. Conclusion In this article we outlined the various types of modals and ModCxns in the languages of Europe. We argued that in the individual languages modals can be identified on the basis of semantic polyfunctionality and the syntactic status of an operator on the predicational and/or propositional level of the clause. ModCxns are found in all languages of the area with some very few exceptions. The boundary between ModCxns and lexical constructions, however, is difficult to draw. Modals derive from lexical elements, mainly from verbs, but in some cases also from nominal or adverbial elements, and show specific symptoms of grammaticalisation setting them apart from their lexical sources. It turns out that only in Germanic ModCxns have developed a dedicated morphological paradigm and a dedicated morpho-syntactic marking. In all other cases we are dealing with a gradual transition from the first construction type to the other one. We tried to describe and measure the degree of grammaticalisation of ModCxns by applying the ‘traditional Lehmann parameters’ which cover both paradigmatic and syntactic aspects of the linguistic sign. Mortelmans, Boye, and van der Auwera (this volume) point out that the Lehmann parameters function as eye-openers leading to observations that might otherwise pass unnoticed: “Seemingly arbitrary facts concerning the form and behaviour of the core modals can be integrated in a larger framework and understood accordingly” (ibid.). The parameters turn out to be a powerful tool for spotting features setting modals apart from lexical elements, because most í if not all í of these features reflect the unidirectionality of grammaticalisation processes. The analyses, however, also reveal some deficiencies of Lehmann’s parameter approach. Whereas certain features can easily be compared across languages like e.g. the degree of polyfunctionality, others are much less suitable for straightforward comparison. As Mortelmans, Boye, and van der Auwera convincingly argue some of the parameters cannot be applied without taking the entire structure of the verbal system of a particular language into account. Thus, the degree of fixation of a modal within a ModCxn depends on the word order rules valid for the whole verbal system. Therefore, it is for example difficult to determine the different degree of fixation of ModCxns in a language with a syntactically fixed word order with a language with so-called free word order. We also saw that there is an asymmetry in the borrowing of modals. Modals of necessity are much easier to borrow than those of possibility and

Modal constructions in the languages of Europe

555

volition. This fact, coupled with the findings of the grammaticalisation parameters, may cause us to rethink the notion of the category of modality as a whole. Based (mainly) on the English modals we have become accustomed to viewing modality as a system in which modal notions, such as necessity and possibility, are easily definable in terms of one another. Given these asymmetries, it seems likely that taking such notions as basis for a cross-linguistic typological view of modality as a whole will only yield an incomplete picture. We looked at the question of whether the situation found in Europe is unique or whether similar patterns can be discerned outside Europe as well. Based on accepted criteria we found that Europe is not unique and that similar patterns are found for instance in Southeast Asia. Finally, we hope to have shown that the languages of Europe are very diverse in their treatment of modal notions. If this diversity is reflected in future studies on modality, we consider this book to have served its purpose. Notes 1. 2.

3. 4. 5. 6. 7. 8.

Neither does the extension of the class of ‘peripheral modals’ coincide with the one of ‘semi-auxiliaries’ (Quirk et al. 1985: 137) or ‘quasi-modals’ (Traugott 1997: 193). N-egation: modals allow negation contraction (mustn’t vs *hopen’t), Inversion without do-periphrasis (Can we go? vs *Hope we to go?), C-ode (John can swim, so can Bill) and E-mphasis (*Yes, I DÒ can come. vs Yes, I DÒ hope to come.) (Quirk et al. 1985: 137). For a discussion of modal polyfunctionality see van der Auwera, Ammann and Kindt 2006. This is due to the large number of languages that Romani has come into contact with and the large area in which different varieties of Romani are spoken. This is not to say that there cannot be other differences between main and auxiliary verbs. In fact, many verbs have a different syntactic behavior between their main and auxiliary verb reading. It is also found in the Indo-Aryan languages, see Masica (1991). There may be exceptions to this rule, see Mortelmans, Boye, and van der Auwera (this volume). Due to the complexity of data the map does not include Romani. The reader is referred to the corresponding article by Elšík and Matras (this volume).

556 9. 10. 11. 12. 13.

Björn Hansen and Ferdinand de Haan Italian bisognare is to be treated as a peripheral modal because it is restricted to different shades of participant-external necessity. Anderson (2006: 116-144) calls this type LEX-headed Auxiliary Verb Construction. Krug (2000) uses the term ‘emerging modals’; see also Bolinger (1980). NFPRS – Non-focal present or non-episodic present Although not mentioned in the article, Uzbek has also borrowed a modal of necessity (lozim).

References Adéwӑlé, Fémi 1990 'must': Analysis of a Yoruba modal verb. Journal of West African Languages 20, 73–82. Anderson, Gregory D. S. 2006 Auxiliary Verb Constructions. Oxford: Oxford University Press. Bolinger, Dwight 1980 Wanna and the gradience of auxiliaries. In Wege zur Universalienforschung, Gunter Brettschneider, and Christian Lehmann (eds.), 292-299. Tübingen; Gunter Narr. Bybee, Joan, and Östen Dahl 1989. The creation of tense and aspect systems in the languages of the world. Studies in Language 13, 51-103. Comrie, Bernard 1991 Modality without Modals. In Problems in the Modality of Natural Language, Piotr Kakietek (ed.), 29–33. Opole: WyĪsza szkoáa pedagogiczna. de Haan, Ferdinand 1996 The development of the Yoruba modal system. Afrika und Übersee 79, 85-109.

de Haan, Ferdinand 1997 Khmer and the theory of modality. Papers in Southeast Asian Languages 16, 47-66.

de Haan, Ferdinand 2005

Modality in Slavic and semantic maps. In Modality in Slavonic languages. New perspectives (Slavolinguistica 6), Björn Hansen, and Petr Karlík (eds.), 3-25. München: Otto Sagner.

de Haan, Ferdinand 2006

Typological approaches to modality. In The Expression of Modality, William Frawley (ed.), 27-69. Berlin: Mouton de Gruyter.

Modal constructions in the languages of Europe

557

Elšík, Viktor, and Yaron Matras 2007 Markedness and language change: The Romani sample. Berlin: Mouton de Gruyter. Goldberg, Adele E. 2003 Constructions: A new theoretical approach to language. Trends in Cognitive Sciences, 7. 219–224 Hansen, Björn 2001 Das Modalauxiliar im Slavischen. Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen. München: Otto Sagner. Haspelmath, Martin. 2001 The European linguistic area: Standard Average European. In Language Typology and Language Universals. An international Handbook Vol. 2 (Handbücher zur Sprach- und Kommunikationswissenschaft 20/2), Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds.), 1492-1510. Berlin: Mouton de Gruyter. Heine, Bernd 1993 Auxiliaries. Cognitive forces and grammaticalization. Oxford: Oxford University Press. Heine, Bernd and Tania Kuteva 2006 The changing languages of Europe. Oxford: Oxford University Press. Hsieh, Chia-Ling 2005 Modal Verbs and Modal Adverbs in Chinese: An Investigation into the Semantic Source. UST Working papers in Linguistics 1: 31-58. Iwasaki, Shoichi, and Preeya Ingkaphirom 2005 A Reference Grammar of Thai. Cambridge: Cambridge University Press. Kachru, Yamuna 2005 Hindi. Amsterdam: Benjamins. Krug, Manfred 2000 Emerging English modals. A corpus-based study of Grammaticalization. Berlin: Mouton de Gruyter. Lehmann, Christian 1988. Towards a typology of clause linkage. In Clause Combining in Grammar and Discourse, John Haiman, and Sandra A. Thompson (eds.), 181–227. Amsterdam/Philadelphia: Benjamins. Lehmann, Christian 2002 Thoughts on Grammaticalization. A programmatic Sketch. Second, revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9.

558

Björn Hansen and Ferdinand de Haan

Masica, Colin 1991 The Indo-Aryan Languages. Cambridge: Cambridge University Press. Matras, Yaron 2007 The borrowability of structural categories. In Grammatical Borrowing in Cross-linguistic Perspective, Yaron Matras, and Jeanette Sakel (eds.), 31–75. Berlin: Mouton de Gruyter. Matras, Yaron, and Jeanette Sakel (eds.) 2007 Grammatical Borrowing in Cross-linguistic Perspective. Berlin: Mouton de Gruyter. Matthews, Stephen and Virginia Yip 1994 Cantonese: A Comprehensive Grammar. London í New York: Routledge. Norde, Muriel (in prep.) Three common misunderstandings about degrammaticalization. In What's new in grammaticalization?, Ekkehard König, Elke Gehweiler, and Katharina Stathi (eds.) Palmer, Frank Robert 2001 Reprint. Mood and Modality. 2nd ed. (Cambridge Textbooks in Linguistics) Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Quirk, Randolph, Sidney Greenbaum, Geoffrey N. Leech, and Jan Svartvik 1985 A comprehensive grammar of the English language. London: Longman. Ramat, Paolo, and Davide Ricca 1998 Sentence adverbs in the languages of Europe. In Adverbial constructions in the Languages of Europe, Johan van der Auwera (ed.), 187277. Berlin: Mouton de Gruyter. Sakel, Jeanette 2007 Types of loans: Matter and pattern. In Grammatical Borrowing in Cross-linguistic Perspective, Yaron Matras, and Jeanette Sakel (eds.), 15-31. Berlin: Mouton de Gruyter. Thomason, Sarah G. 2001 Language contact. An introduction. Edinburgh, Edinburgh University Press. Traugott, Elizabeth C. 1997 Subjectification and the development of epistemic meaning: The case of promise and threaten. In Modality in Germanic languages: Historical and comparative perspectives, Toril Swan, and Olaf J. Westvik (eds.), 185-210. Berlin: Mouton de Gruyter. van der Auwera, Johan, and Andreas Ammann 2005 Overlap between Situational and Epistemic Modal Marking. In The World Atlas of Language Structures Online, Martin Haspelmath, Matthew S. Dryer, David Gil, and Bernard Comrie, (eds.), Munich:

Modal constructions in the languages of Europe

559

Max Planck Digital Library, chapter 76 (Available online at http://wals.info/feature/76 Accessed on 2008-08-07) van der Auwera, Johan, and Andreas Ammann with Saskia Kindt 2005 Modal polyfunctionality and Standard Average European. In Modality. Studies in Form and Function, Alex Klinge, and Henrik Høeg Müller (eds.), 247–272. London: Equinox. van der Auwera, Johan, Petar Kehayov and Alice Vittrant (in print) Modality’s semantic map revisited: acquisitive modals. In Crosslinguistic Studies of Tense, Aspect, and Modality, Lotte Hogeweg, Helen De Hoop, and Andrej Malchukov (eds.). Amsterdam: Benjamins. van der Auwera, Johan, and Vladimir Plungian 1998 Modality’s semantic map. Linguistic Typology 2, 79–124. van der Auwera, Johan, Ewa Schalley, and Jan Nuyts 2005 Epistemic possibility in s Slavonic parallel corpus – a pilot study. In Modality in Slavonic languages. New perspectives, Björn Hansen, and Petr Karlík (eds.), 3-25. München: Otto Sagner

Subject index

ability 62, 79, 89, 148, 243, 245, 248, 281, 317, 326, 344-346, 351, 363-364, 389, 407, 413, 431-432, 441-442, 447, 471, 492-494, 504, 520 adhortative 449, 489 adjective 25, 37, 50, 71, 74, 79-80, 82, 84, 87, 168, 170-171, 174, 190, 274, 299, 337, 344, 403, 407-408, 410, 479, 496-498, 503, 516-517, 526, 542 admirative (mood) 229 admonitive 443, 489 adverb 24, 35, 45, 48, 56, 71-72, 9697, 168, 170-171, 175-176, 191, 200, 204, 223-225, 308, 238, 330, 334, 341, 342-343, 346, 349, 351, 353, 355, 365, 381, 392, 393, 407-408, 440, 442, 493, 516-517, 522, 524, 526, 533, 535, 549, 554 affix 3, 13, 77, 96, 152, 160, 202, 403-404, 406-407, 410, 413, 432, 435, 436, 439-442, 446, 448-450, 513, 515-520, 539, 542-544, 549-552 agentivity 380, 116, 278, 306, 314316, 366, 679-380 agent-oriented (modality) 99, 144145, 148, 159, 325-326, 432, 452, 489, 522 agglutination 186, 202, 253, 267, 292, 332, 355, 513, 517, 539, 542-544 agreement 53, 60, 143, 171-177, 193, 271, 366, 395, 405, 410, 500, 515, 526, 530-535, 538 aktionsart 252, 282-283, 436

analytic 169-170, 181, 248, 267, 274, 278, 280, 295, 297, 487, 493, 495, 497-498, 501-504, 506 areal convergence 546 argument positions 16, 24, 28, 31, 52, 77, 83, 94-95, 97, 154, 155, 169, 171-172, 174-178, 182183, 184, 190, 193, 208-209, 217-218, 220-221, 224, 311, 313, 366, 377, 380, 383-384, 389-390, 460, 463-465, 474, 477, 526-531 aspect 71, 125-126, 131, 139, 142, 149, 152, 157, 159, 168, 171, 177, 181-182, 185-186, 229, 231, 347, 252, 271, 326-327, 344, 378, 431-434, 446-447, 459, 461, 472-475, 482, 494, 503, 525-526, 532, 551 attrition 13-14, 108, 120, 149-150, 222, 370 auxiliarisation 169, 220, 221, 225, 374, 376, 378, 514 auxiliary 11, 14, 22-23, 25, 27-30, 42, 57-58, 71, 78, 81, 87, 95-96, 98-99, 113, 120-126, 129-131, 140, 142-143, 150, 160, 170171, 173-177, 185, 192-193, 199-204, 213, 220-222, 239, 244, 246-247, 249, 251-253, 274, 276-277, 285, 295, 325328, 335-336, 343, 354-357, 369, 374, 384, 387, 391, 393, 403-405, 412, 414, 417-418, 422, 426, 431, 452, 459, 461, 462, 464-465, 469-472, 481, 493-494, 499, 512-514, 516,

562

Subject index

520, 523, 525, 526, 531-532, 534-538, 550 avertive 375-376, 380 Balkanism 260, 500 bi-clausal 515, 526, 530, 537 bleaching 3, 83, 108, 110, 147, 179180, 370-376, 410-412, 519521, 523 bondedness 2-3, 12-13, 96, 98, 129, 155, 179, 185-186, 224, 253, 370, 384, 393, 416-417, 539543 borrowability scale 312-313, 546549 borrowing 6, 180, 192, 201, 267268, 270, 278-281, 284, 288, 291, 294, 297, 301-302, 304309, 312-314, 387-392, 439, 442, 546-549 calque 299, 303, 306, 327, 334, 337, 388, 393, 439, 500 capability 81, 169, 248, 281, 306, 314, 351 capacity 81, 243, 328, 330, 336-338, 342-346, 354, 356, 364, 514 case 84, 110, 168, 173-175, 203, 229, 267, 272-273, 277, 291, 367, 377, 386, 389, 390-391, 464, 496, 527-529 certainty 71, 107, 110, 146, 333, 341, 350, 450 clause linkage 515, 526-538 clitic 120-123, 156, 233, 236, 249, 253, 276, 308-309, 312, 432, 445, 449, 540, 542-543 clitic-climbing 128 cliticisation 99, 120-123, 157, 186, 385, 523, 539 coalescence 13, 96, 155, 185, 312, 416-417, 422, 539-545 code-copying 388, 500

cohesion 2-3, 11-12, 120, 129, 151, 155, 179, 384, 416 command 443, 489 complement 22, 24, 27, 34-36, 45, 46-48, 50, 56, 58-59, 73-75, 7778, 85, 93-99, 112, 123-125, 128, 131, 140, 156, 182, 184185, 214, 221-223, 245,250251, 253, 267, 269-270, 286, 327, 365, 379, 383-385, 392, 414, 420, 438, 442, 444-445, 448, 465-470, 474, 477, 481, 487, 495-496 dativ~ 116, 217-218 complement clause 28, 36, 45, 48, 56, 93, 218, 267, 270, 414, 436, 438-439, 441-442, 444, 448, 466-467, 477, 496, 538 complementiser 113, 115, 184, 248, 275-278, 280, 286, 289-290, 297, 309-312, 331, 515, 526, 532-533, 536-538, 540, 542 concessive 149, 158, 442, 443 condensation 13, 35, 96, 98, 153, 184, 311, 384, 414-415, 525539 conditional 3, 24, 34, 71, 73, 75, 78, 81, 89, 91, 94, 96, 98, 149, 181, 202, 212, 297, 365, 380, 406, 410, 432, 444, 471, 491, 500, 502, 522, 531 connector 286, 526 construction grammar 512 contact-induced 268, 281-282, 288, 291, 305, 306, 392-393, 452, 487, 503, 506, 549 content expression 76-77, 80-81, 98, 169, 514-515, 546 coordination reduction 156, 384, 540-541 control verb 100, 217, 219, 222, 383384

Subject index copula 71, 74, 81-82, 85-86, 89, 201, 274, 276, 280-281, 286-291, 299, 302, 306-309, 536 counterfactivity 89, 211-212, 376, 447, 451 cross-reference 269, 274, 287 debitive 200-204, 209, 211, 215-216, 219, 223-224, 515 decategorialisation 87, 97, 99, 171, 220, 488 defectiveness 43, 71, 74, 78, 88, 9799, 220, 296, 330, 378, 409, 410, 412, 472, 513, 521, 522, 545 degrammaticalisation 26, 121, 251, 391 deictic 431, 436, 440-442, 445, 451452 demand 204, 278, 326, 489 deontic (modality) 3, 15, 16, 23-24, 26, 31, 44, 73, 76, 83, 107, 109, 111, 113, 115, 117, 119, 123, 127-132, 144, 148, 169, 179, 188-190, 192, 199-200, 204211, 214-220, 223-224, 231232, 245-246, 252, 255, 282, 284, 287, 307, 326, 344, 350, 363, 364, 372-373, 407, 413, 432, 439, 442, 447, 490-494, 498-499, 512, 514-515, 520, 522, 525, 538, 541, 545, 551552 desemanticisation 83, 93, 108-110, 114, 118-120, 126, 131, 147, 150, 122, 410, 488, 505, 520 desententialisation 526 desiderative 185, 489, 490-491 desire 79, 251, 278, 306, 326, 358, 432, 436, 441-442, 489-491, 494 directionality 154, 376, 387-388, 507, 554

563

directive 205, 489, 494 discourse marker/particle 409, 533, 539 doubled pattern 532, 534-535 dynamic (modality) 3, 15, 16, 19, 23, 24, 26, 31, 34, 76, 83, 99, 169, 179, 188-190, 199-200, 204, 206-209, 213-217, 219-220, 223-224, 231, 241, 245, 246, 252, 255, 364, 369-370, 405406, 410, 451, 514-515, 520, 538, 541, 545, 552 ellipsis 26, 29, 46, 59, 391, 416, 541, 545 enunciative 325-326 epistemic (modality) 3, 11, 14-16, 19-20, 23-29, 31-35, 37-59, 61, 71-77, 90, 96-97, 99-100, 107, 109-111, 114-115, 117, 119, 126, 129-131, 142-146, 149150, 152, 154, 157, 169, 179180, 188-191, 199, 204, 210214, 217-220, 224, 237, 239240, 245, 247-248, 252, 255256, 298, 307-308, 312, 316, 325, 326, 328-330, 333-336, 338-339, 341-343, 346, 348, 350-357, 363-364, 369, 371373, 378, 380-381, 388, 405407, 409-410, 432, 462, 471, 473, 490, 492-494, 512, 514515, 520, 522, 525, 530, 533, 538-541, 545, 549, 551-552 ~non-finiteness gap 59 erosion 83, 99, 131, 147, 149, 179, 249-250, 292, 376, 410, 412, 488, 505-506, 519-521 Euroversal 549-553 event modality 146, 199, 208-210, 212-213, 219 eventuality 329, 334, 336, 343, 346, 352, 354

564

Subject index

evidential 3, 16, 19, 32, 39-40, 50, 170, 181, 199, 214, 256, 374, 380, 511, 514 exhortation 489 existential construction 200 feasibility 329, 336, 343, 354 fixation 13-14, 96, 157, 187, 424425, 544-545, 554 fossilisation 311, 378, 380, 381 future 141, 146, 149-150, 157, 181182, 186-187, 223, 229, 230231, 239-240, 242, 244, 247249, 251, 253, 255-256, 297, 308-309, 312, 316, 332, 336, 339, 346, 374, 388, 436-438, 446, 450, 459, 461-462, 464, 471, 473, 474, 480, 490, 492, 519, 542, 555 fuzziness 167, 192, 380, 393, 482, 512-514 genitive 72, 82, 84, 94, 229, 366, 389, 470, 496, 528 grammaticalisation 2-3 ~path 73, 83-87, 119, 143, 146147, 250, 267, 314-316, 507, 520, 551 ~chain 99, 121, 125, 224, 325, 354-357, 513-515, 539 ~hierarchy 312-313, 547-549 ~parameter 2-3, 11-16, 20, 23, 26, 27, 37, 59-60, 82-83, 87, 93, 94, 96-98, 107-108, 110, 120, 124, 126-127, 130, 139, 146-147, 151-157, 178-187, 224, 250-253, 363, 370, 383, 385-387, 393, 410, 412-417, 424, 519-546, 549, 554-555 habitual 72, 89, 336, 436-437, 446 hearsay 16, 170, 388 hortative 205

impending action 181, 375 imperative 3, 18, 29, 42, 48, 54, 71, 78, 88, 140, 141, 142, 152, 171, 181, 184, 199, 205, 223, 246, 365, 377, 406, 432, 435-437, 443, 445-446, 448-450, 461, 489 imperfect 29, 231-233, 235, 242, 247-248, 252, 270-271, 279, 295-297, 309, 374, 376, 480, 459, 461-462, 474, 494 imperfective 140, 329-330, 332, 334, 336, 338-341, 344-354, 357, 433, 435-440, 443-444, 446450, 522 impersonal 72-73, 77, 79, 82, 85, 98, 110, 115, 117-118, 122, 125, 142-144, 148-151, 157, 159, 182, 190-192, 200-203, 206, 208-209, 216-220, 222, 224, 248, 250-251, 269-278, 285, 292, 296-299, 310-311, 340, 366-370, 377-378, 383, 386393, 489, 522-523, 538 implicational hierarchy 547 imposition 489 impossibility 214, 233, 241, 250, 252, 254-255, 307, 476-478, 498, 504-505 inanimate 116, 153, 182, 202, 331, 348-349, 524-525 indicative (mood) 52 infinitive 11, 17, 21-30, 32, 34-35, 37, 39, 42, 44, 46-58, 72, 77, 94-95, 99, 108-113, 116-120, 122-125, 128-131, 150, 174175, 182, 184-187, 190, 192, 200, 202-203, 210-211, 214, 217-218, 222, 229, 245-249, 254-255, 270, 276, 365-366, 370, 377, 379, 380, 382-385, 391-392, 404-405, 407-408, 410, 412, 414, 418-420, 424,

Subject index 426, 446, 496-500, 521-523, 536, 540-542, 545 bare~ 12, 14, 20, 41, 45, 452 Infinitivus Pro Participio construction (IPP) 21-22, 2829,33, 38, 43, 523 integrity 2-3, 12-13, 83, 108, 147150, 179-181, 222, 250-251, 306, 370-380, 386-387, 393, 410-412, 519-521 intentional 380, 489 interdiction 331-332 interlacing 526, 538 interpretive use 204-206, 223-224 intersubjective (modality) 325-326, 328, 330-332, 334-341, 343344, 346-357 IPP see Infinitivus Pro Participio construction irrealis 34, 94, 202, 211-212, 223, 431, 434, 436, 441, 444, 446, 449, 451-452, 513 jussive 246, 377, 489 juxtaposition 155, 327, 416-417, 424, 533, 539 language contact 6, 184, 191, 247, 256, 334, 355, 363, 387-392, 488, 495, 499, 506, 513, 546549 layering 111, 520 LEX-headed Auxiliary Verb 556 light verb construction 466-468 likelihood 268, 278, 286, 312, 371, 493, 547 linking morpheme 229, 483 loanword 268, 278-279, 284-286, 294-295, 297, 439, 450 locative subject 272, 283 MAT-borrowing 546-549

565

mediopassive 229, 231, 238, 242245, 247, 249-252, 255 metaphor 83, 109, 313, 382 modal ~affixes 3, 152, 202, 403-407, 413, 513, 516-519, 539, 543544, 549-550, 552 ~auxiliaries 87, 96, 126, 169, 202, 213, 221, 327, 330, 331332, 334, 339-340, 353, 404406, 426, 431, 464-465, 469, 499, 536, 550 central~ 11-16, 19, 33, 50, 52, 231, 426, 512 ~ construction 1, 6, 119, 167, 171, 173-178, 193, 200, 205, 211, 219, 248, 268, 269-277, 289, 300, 310, 405, 408, 412, 414, 417-418, 424, 460, 464, 471, 478, 481, 499, 511-560 emerging~ 11, 514 ~particle 139-142, 145-147, 152, 155-157, 204-206, 223, 231, 241, 244, 250, 255-256, 349, 351, 403, 407, 409, 413, 462.463, 474, 493, 516 peripheral~ 170, 512, 514, 555 ~source 26 ~tags 403, 409 uninflected~ 235, 241-242, 246, 248, 250, 253-256, 274-275, 277, 297, 311-312 modality 1-2, 363-364 mono-clausal 525, 527, 537 mood 30, 71-72, 81, 84, 89, 91, 93, 139-141, 148-149, 152, 158, 160, 171, 173-174, 177, 182, 199, 201-202, 204, 221, 223, 229, 241, 246, 297, 326, 365, 377-378, 380-381, 406, 410, 431-457, 458-459, 461, 469, 471, 473, 489-491, 493-495,

566

Subject index

498-501, 503-506, 516-517, 526, 532 morphological degeneration 147, 150, 370, 376, 410-411 necessitive 200-202, 206, 219, 224 necessity epistemic~ 50, 97, 146, 191, 298, 312, 364, 373, 490 non-deontic~ 364, 373, 520 non-epistemic~ 364, 370, 392 participant-internal~ 209, 292, 305-306, 311, 395 negation 2, 7, 74, 82, 85, 88, 91-94, 98, 116, 129, 157, 208-209, 214-216, 221-222, 224, 233234, 236-238, 240-241, 243244, 246, 248, 250, 253-255, 268, 276, 279-280, 289, 300305, 337, 344, 346, 369, 375, 378, 381, 433, 435, 443, 446450, 476, 492-493, 498, 503504 double~ 234 ~placement strategy 238, 255 negative ~polarity 17, 30, 207, 308, 467 NICE properties 1-2, 512, 523 object doubling 229, 236 obligation 3, 80, 89, 107-112, 117, 126, 131, 200, 235, 246, 287, 331-332, 336, 340-341, 343, 347, 349-350, 354, 356-357, 364, 388, 431, 439, 442, 457, 463-464, 491, 494, 511 weak~74, 288, 293-294, 306 obligatorification 13, 93, 98, 152, 182, 311, 378, 382, 412, 524 obligatoriness 3, 32, 38, 93, 152, 182-184, 223, 252, 383, 432, 506, 517

optative 3, 229, 246, 432, 435-437, 446, 450, 489-490, 494, 499502, 504-505 paragmaticity/paradigmaticisation 23, 12-14, 20-22, 27-28, 32-33, 37-38, 41-42, 47, 52, 57, 59-60, 87-88, 120, 122, 124, 151, 179, 181-182, 222, 251, 370, 380381, 386-387, 393, 412, 458, 521, 523 participant-external 16, 79-80, 82, 127, 143, 145, 181, 199, 209, 281, 283, 290, 293, 305, 307, 310, 313, 363-364, 369, 372373, 520 participant-internal 80, 127, 143144, 148, 180, 209, 269, 281, 292, 294, 305-307, 310-311, 363, 372-373, 389 participle 107, 113, 115, 117-118, 125, 128-130, 201, 214, 223, 242-244, 248-249, 251-253, 255, 337, 344, 351, 354-355, 411, 459, 461, 464-465, 467, 469-471, 474, 481, 495-497, 522 particle 76-77, 84, 89, 94, 139-142, 145-149, 151-153, 155-159, 204-206, 223-225, 229, 231, 233, 236-237, 239-241, 244, 246-250, 252-256, 305, 327328, 337, 341-343, 346, 349, 351, 353, 355-357, 381, 403, 407-409, 413, 427, 431-434, 436, 440-442, 444-452, 462463, 474, 476, 493, 499, 501, 505, 513, 515-516, 523-525, 532-534, 539-540, 546, 552 partitive 80, 365, 377, 467 passive 18, 25, 42, 44, 48, 71, 84, 107, 117-118, 183, 201, 213-

Subject index 214, 219-220, 223, 242, 244, 349, 351, 378, 451, 543, 550 PAT-borrowing 549 perception verb 22, 33, 185, 354-355 perfect 21-23, 27-29, 34, 153, 233, 239, 247, 378, 381, 433, 459462, 464, 471, 474-475, 478 perfective 71, 109, 140, 142, 152153, 171, 329, 332, 334, 338, 341, 344-347, 350, 352, 432435, 437, 439-440, 443-444, 446-448, 450-451 periphrastic construction 24, 129, 268, 326-327, 331, 334-335, 341, 343, 348, 451, 412, 473, 481-482, 504-505 permission 3, 17, 92, 169, 232, 246, 282-284, 286, 306, 310, 314, 316, 326, 328, 336, 338, 341, 343-344, 351, 354-356, 364, 407, 489-490, 494, 511, 514 permissive 443, 489 polyfunctional 3, 57, 72, 75, 77, 83, 97, 100, 110, 145-146, 159, 169-170, 177-179, 192, 206, 231, 281, 371-376, 384, 392393, 405, 511-512, 514-516, 535, 551-554 possessive construction 200, 288, 424 possibility non-epistemic~ 364, 473 dynamic~ 209, 231, 245, 257 non-deontic~ 364, 372-373, 520 post-modal 98, 147-149, 157-158, 181, 186, 312, 368, 374, 376, 380, 389, 514 potential mood 202, 365, 381, 494 pragmatic inference 287 pre-modal 14, 18, 149, 218, 367-369, 376, 383, 520 preterite-presents 11-12, 17, 20-21, 29-30, 32, 39-44, 48-50, 52-54,

567

57, 61, 521-522 preverb 94, 336, 344, 355, 357, 431, 434-438, 441, 446, 448-450 privileged syntactic argument 171172, 176, 193 PRO subject 217 probability 17, 107, 109-110, 180, 294, 326, 329, 333-336, 338339, 343, 348, 350-352, 354, 431-432, 440-442, 448, 450, 473, 490, 492, 494, 503, 511 pro-drop 100, 139, 185 prohibitive 181, 446, 449, 450 prohibition 140, 236, 407, 467 propositional modality 146, 199 prospectivity 450, 494 purposive 442-443, 491, 498-499 raising verb 55, 58, 213, 217, 219, 222, 383-384 reflexive 202, 253, 278, 291-293, 299, 306, 310, 312, 314, 445, 543 replica grammaticalisation 268, 278, 506-507 request 205, 489-490, 494 root modality 11, 326 scope 2-3, 12-14, 24-28, 35-37, 4448, 56, 58-59, 94-96, 98, 127129, 153, 179, 184-185, 219, 223, 234, 236-238, 240, 244, 246, 248, 253, 300, 310-311, 327, 370, 383-384, 386-387, 393, 414-416, 515, 525, 541, 545 selectional restriction 93, 182, 193, 221, 366, 383, 393, 516 semantic map 1, 143, 170, 179, 307, 514 semi-auxiliary 120, 140, 142, 257 sentence adverb 168, 533 speaker-oriented 142, 148, 452, 494

568

Subject index

split pattern 532, 534-535 Sprachbund 184, 192, 229, 550 Standard Average European (SAE) 4, 100, 392, 549-553 subcategorisation frame 123, 127, 131, 523 subject dative~ 52, 172-173, 175-176, 180, 182, 190, 193, 208209, 217, 272, 551 nominative~ 173-177, 180, 182, 190, 193, 202, 221, 272, 277 subjective 131, 325, 489, 493, 495 subjunctive 3, 52, 71, 115-116, 119, 140-141, 148, 231, 233-234, 236, 241-256, 269-271, 286288, 406, 439, 446, 449, 469, 487, 497, 499-506, 522, 536 supine 111-112, 115, 125, 130 suppletion 74, 81, 89, 91-92, 98, 258, 302, 304, 374, 381, 472, 474, 478 TAM / Tense-Aspect-Modality 125, 185, 27-272, 274-275, 277, 280, 285-286, 289-290, 295-298, 310-312, 327, 328, 344, 350, 352-353, 355-356, 378, 431, 433, 450, 459, 515, 525-526, 531-535, 551 text frequency 12, 26, 32, 40, 43, 50, 94, 98, 191, 274, 357, 410, 439, 500, 549 transitivity 24, 95, 128, 460-461, 463-465, 468, 470-471, 474478, 481 unaccusative 460-462, 464, 474, 477 unergative 460-462, 464-465 univerbation 493, 533, 540

variability intraparadigmatic~ 14, 23, 126127, 382 syntagmatic~ 2-3, 12-14, 96, 130, 157-158, 179, 187188, 223, 253, 370, 385, 393, 424-425, 544-545 transparadigmatic~ 126, 382 verb lexical~ 14, 21, 72, 78, 127, 139, 143, 146, 150, 168169, 171, 280, 284, 303, 306, 310-311, 349, 365366, 370, 379, 459, 464, 467, 469, 481, 492, 495498, 504-505, 513-515, 525, 531-532, 534-536, 539, 543-545 verbal noun 72, 76-78, 84, 86-89, 94-95, 99, 171, 248, 349, 465, 469-470, 496, 499-500, 502 voice 139-140, 142, 152, 370, 377378, 386, 452 volition 1, 3, 15, 153, 157, 184-186, 199, 209, 231, 238-240, 246, 248-254, 256, 268, 278-279, 291-294, 301-302, 306-308, 310, 312-316, 331, 347, 349, 354, 356-357, 403-404, 412, 431, 468-470, 475, 478, 487, 489-491, 495-496, 499, 501, 536, 542-543, 547, 549 voluntative 489-491, 494, 499, 501504 VP 154, 384, 414 weather verb 415 weight 2-3, 11-13, 108, 127, 179 word order 14, 71, 96, 217, 221, 223, 267, 403, 424, 435, 442, 477, 544, 554

Languages and language varieties index

Abkhaz-Adyghean 550 Ainu 552 Albanian 4, 229-266, 292, 309, 500, 513, 517, 519, 521-523, 532, 534, 537, 540, 548, 553 Altay Turkic 488, 498 Arabic 5, 13, 325-361, 434, 439, 450, 453, 496-498, 520-521, 529-530, 533-534, 548-551 Arbrisht Albanian 247-248 Armenian 278, 550 Austrian Lovari 290, 296 Auvergne Manuš 302 Azeri 5, 488, 490, 492-493, 497-499, 503-504, 506, 517, 548 Balkan language 229, 244, 254, 308, 493, 500, 534, 546 Balkan Romani 549 Balkan Turkish 499-500, 502, 504506 Baltic 4-5, 199-228, 364, 517, 521522 Balto-Finnic 5, 363-401, 511, 514, 517, 520-521, 523, 548-549, 553 Basque 4-6, 457-486, 516-517, 529, 536, 550 Belorussian 5, 167, 178, 180, 188190, 192, 548 Berber 5-6, 431-456, 513, 516, 548549 Berber Figuig 449 Bizkaian 467, 469, 474-475 Bosnian 167, 180-181, 186-189, 191 Bulgarian 5, 167-168, 170, 178, 180, 187-192, 229, 249, 278, 284, 297-298, 309, 311, 498, 500, 502, 504-505, 534, 547, 549

Burgenland Roman 305 Cantonese 551 ýáry 275, 288-289, 297 Celtic 5-6, 71, 89, 517, 550 Cerovo Cocomanja 298, 301 Chaghatay 487, 499-500 Circum-Anatolian Turkic 488 Connacht 72-74, 89 Crimean Romani 274-275, 281, 284285, 292, 299, 306 Croatian 5, 157, 167, 180-181, 186189, 191, 533, 540, 542, 545 Croatian of Burgenland 167 Cyprus Turkish 500 Czech 5, 168, 170-171, 178, 180, 188-189, 192, 548 Danish 5, 11-14, 16-17, 22, 24, 28, 30, 32, 39-48, 50, 53-56, 59-60, 511, 514, 522, 525, 533, 542 Dobšiná Romani 293 Dolenjski Romani 276, 278, 286, 295, 549 Dutch 5, 12-14, 17-29, 30-43, 45-46, 48, 50, 56, 59-60, 370, 511, 522, 525, 528, 533, 540-542 East Slavonic 167, 392, 546-547 Eastern Arabic 325 Egyptian Arabic 5, 344-354 English 1-2, 4-5, 11-17, 19-21, 24, 30, 32, 35, 59, 72, 74, 80, 98, 100, 109, 120, 150, 191, 210, 216, 220-222, 278, 288, 291, 294, 301, 327, 333, 384, 426, 431, 438-439, 446, 452, 473, 511-512, 514-516, 519-524,

570

Languages and language varieties index

533, 538, 540-542, 546-547, 549, 555 Estonian 5, 201, 364, 367, 369-384, 388-391, 394, 520, 522, 524, 545, 548 Eynallu dialect 502 Faroese 53 Finnish 5, 284, 294, 297, 363-393, 516-517, 522, 528, 547 Finnish Romani 270, 284, 297-298, 304, 530 Finno-Ugric 4, 6, 364, 374, 381, 550 French 5, 107-109, 111, 113, 116117, 119, 121-125, 128, 130, 211, 439, 445, 449, 467, 481, 501, 516, 523, 533, 541 Gabor 281 Gagauz 5, 493, 495-498, 500, 502, 504-505, 517, 548 German 2, 5, 11-14, 16-17, 22, 24, 28-39, 43, 45, 48, 56-60, 180, 191-192, 205, 212, 226, 270, 282, 284, 286, 294, 297, 301, 305, 409, 522, 525, 527-528, 532, 540-542, 547-549 Low~ 369, 388 Germanic 1, 4-6, 11-69, 120, 170, 180, 191, 193, 210, 432, 511, 513, 515-516, 520-521, 523, 525, 528-529, 532, 541, 545, 549, 554 Geg Albanian 230, 246-249, 254255 Gipuzkoan 467, 472, 474-475 Greek 4-5, 139-163, 229, 248, 251, 267, 282, 284, 291, 294, 297, 304, 308, 487, 500, 511, 513515, 522-525, 534, 537, 540, 544, 547-548 Hameln Sinti 269-270, 287-290, 305

Haruai 513 Hindi 550-551 Hungarian 5, 202, 281-282, 294, 297, 312, 403-430, 514-515, 517, 521, 528-529, 534, 537, 539-541, 544, 546-549 Hungarian Sinti 282 Icelandic 5, 12-13, 16, 22, 32, 48, 50-59, 521, 525, 553 Illyrian 229 Ingrian 5, 364-367, 369, 372, 374, 378, 388-391, 394, 517 Iranian 5, 278, 488, 497, 499-501, 503, 505, 550-551 Irano-Turkic 488 Irish 5, 71-105, 517, 521, 525, 536 Early Modern~ 72, 81, 86, 89, 95 Middle~ 72, 82, 84, 86, 94-95 Old~ 72-75, 81, 84-86, 94-95 Italian 5, 22, 107-132, 247, 284, 327, 332, 334, 445, 531, 532, 541, 547 Italo-Albanian 230, 247-248, 250, 252-253, 255-256, 540 Jordanian 325, 344 Kabyle 5, 431-437, 439, 440, 446452, 529 Kalamata 298-299 Kalderaš 280, 295, 307 Kalderaš Taikon 307 Karakhanid 487, 493, 503 Karelian 5, 364-367, 369, 372, 374375, 383, 387-392, 394, 517, 548-549 Kashubian 167 Kaspiþan Xoraxane 275-276, 279, 537 Kԥrԥmidarja 303, 309 Khakas 488, 493, 497-499

Languages and language varieties index Khmer 551, 553 Khorezmian Turkic 487 Kiev Servy 304 Klenovec Rumungro 278-279, 282 Kohila 279, 283, 294-295 Kotka 281 Koman 500 Kubanskie Servy 273 Kumanovo Kovaþki 292 Kurdish 497, 501, 505 Latin 84, 86, 108, 111, 116, 127, 130, 191, 200, 220, 230, 251, 439, 513, 545 Latvian 5, 199-225, 280, 515, 517518, 548 Old~ 200 Levantine Arabic 5, 326, 344-345, 355, 358 Lithuanian 5, 199-225, 280, 297, 304, 528 Old~ 200-201 Livonian 5, 201, 364-365, 367, 371372, 374, 388-389, 391, 394, 517 Lower Sorbian 5, 167, 178, 180-181, 188-189, 192, 548 Luþivná 293 Lude 392

571

North Slavonic 190, 192 Norwegian 53 (Old) Church Slavonic 178 Palestinian 325, 344 Parakalamos 271, 284, 304 Pazardžik Malokonare 301-302 Persian 251, 279, 487, 497, 499, 501-506, 551 Piedmontese Sinti 278, 280, 303 Piteúti Kalderaš 307 Podhradie 274 Polish 5, 167-168, 171, 176, 178, 180-182, 184, 186-192, 205210, 219, 284, 294, 511, 514, 547-549 Proto-Baltic 199 Proto-Berber 433-434

Macedonian 167, 178, 192, 229, 284, 295-296, 298, 500, 504-505, 547 Maghribi Arabic 325 Maltese 5, 325-361, 516, 534, 550 Mandarin 551, 553 Montana Kalajdži 309 Moroccan Arabic 5, 325, 336-343, 533 Munster 72-74, 83, 86, 89, 91

Rakitovo Yerli 285 Romance 4-6, 11, 107-137, 168, 210, 251, 432, 439, 452, 516, 520521, 523, 532, 541, 545 Romani 4-6, 267-322, 515, 526, 528, 537, 545-549 Romanian 5, 107-108, 111-113, 115, 119, 122, 124-126, 128, 130, 294-295, 297, 308-309, 533, 541, 547-548 Rumelian East 499-500 Russian 5, 167-168, 170-172, 174175, 178, 180-183, 186-190, 192, 200, 202-203, 208-209, 217, 219, 221, 251, 276, 284, 299-300, 304, 306, 365, 389392, 394, 497-498, 502, 511, 517, 524, 527-528, 531, 533, 535-536, 538, 544, 548-550 Ruthenian 167

Nakh-Daghestanian 550 New Indo-Aryan 267

Saami 553 Scandinavian 12, 53

572

Languages and language varieties index

Selice Rumungro 270, 289, 301 Sepeþides 286-287 Serbian (Serbian/Croatian) 5, 157, 167-168, 172, 174-177, 180181, 183-192, 294-295, 297, 530-531, 533, 537, 540, 542543, 545 Servy Ghympeny 272, 291 Shilha 436 Siberian Turkic 488, 549 Sicilian 327 Slavonic 2, 4-6, 11, 80, 93, 97-99, 143, 149, 152, 157, 167-197, 200-201, 210-211, 216, 224, 382-383, 392, 517, 521, 524, 526, 528, 540-541, 545-550 Slavošovce 297-298 Sliven Muzikanta 271, 296, 298 Slovak 5, 167, 169-170, 178, 180, 188-190, 192, 278, 288, 294, 297, 511, 533, 547-548 Slovene (Slovenian) 5, 167, 177178, 180, 188-191, 204, 206, 276, 278, 284, 286, 295, 533, 534, 547, 549 South Slavonic 174-175, 191-192, 547-548 Southern Dutch 17 Spanish 5, 107-130, 200, 541 Swedish 53, 294, 297-298, 374, 388389, 533, 547 Syrian 325, 344 Tachelhit 449-450 Tagalog 553 Taqbaylit 431, 433-434, 513 Tatar 488, 494 Thai 84, 86, 551-553 Thracian 229 Tosk Albanian 230, 242, 245-249, 254 Tuareg Ahaggar 449 Tuareg Aïr 433

Turkic 5-6, 487-510, 516-517, 534, 549-550 Turkish 5, 276, 294, 312, 487-488, 490-496, 498-506, 517-518, 528, 543, 547-548, 553 Ottoman~ 487, 490-491, 498, 501, 506 Tuvan 488, 497-498, 503 Ub Gurbet 294-295 Ukrainian 5, 167, 178, 180, 188, 190-192, 283, 548 Ulster dialect 89, 91-92 Upper Sorbian 167, 178, 180, 188189, 192, 548 Uralic 5, 192, 403, 550 Ursari 280, 303, 309 Uyghur 5, 500 Old~ 487, 489-490, 493, 496, 499 Uzbek 5, 488, 491, 493, 498-500, 503-504, 517, 548 Vһlþi Dol 298 Veps 5, 364-370, 372, 374, 388-392, 394, 548-549 Vidin Kalajdži 297 Votic 5, 364-374, 378, 388-391, 394, 517, 553 Welsh Romani 278, 286-288, 294, 297, 301, 549 West Greenlandic 553 West Iranian 278 West Rumelian Turkish 499-500, 505 West Slavonic 167, 191-192, 392 Xandžary 283 Yakut 488, 491, 497, 548-549 Yimas 553 Yoruba 551, 553