On Comitatives and Related Categories: A Typological Study with Special Focus on the Languages of Europe 9783110197648, 9783110185874

Citation preview

On Comitatives and Related Categories

≥

Empirical Approaches to Language Typology 33

Editors Georg Bossong Bernard Comrie Yaron Matras

Mouton de Gruyter Berlin · New York

On Comitatives and Related Categories A Typological Study with Special Focus on the Languages of Europe by Thomas Stolz Cornelia Stroh Aina Urdze

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

Stolz, Thomas. On comitatives and related categories : a typological study with special focus on the languages of Europe / by Thomas Stolz, Cornelia Stroh, Aina Urdze. p. cm. ⫺ (Empirical approaches to language typology ; 33) Includes bibliographical references and index. ISBN-13: 978-3-11-018587-4 (cloth : alk. paper) ISBN-10: 3-11-018587-3 (cloth : alk. paper) 1. Grammar, Comparative and general ⫺ Case. 2. Grammar, Comparative and general ⫺ Grammaticalization. 3. Markedness (Linguistics) 4. Typology (Linguistics) 5. Europe ⫺ Languages. I. Stroh, Cornelia. II. Urdze, Aina. III. Title. P240.6.S76 2006 4151.5⫺dc22 2006032106

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

ISBN-13: 978-3-11-018587-4 ISBN-10: 3-11-018587-3 ISSN 0933-761X © Copyright 2006 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.

Preface and acknowledgements

This book is the outcome of more than a decade (1993–2006) of at times exceedingly hard work. In the beginning, the idea was to finish off our topic quickly in a joint article. However, relatively soon it became clear to us that Comitatives and related categories are a subject matter which cannot be dealt with satisfactorily within the limits imposed by the general principles of journal articles. Moreover, the whole issue proved to be so intricately interesting from the linguistic point of view that we decided to devote a book-length typological study to it. We set out to collect the necessary empirical data in 1996 when we received the first of altogether three consecutive grants by the Deutsche Forschungsgemeinschaft (DFG) within the Förderungsschwerpunkt Sprachtypologie (1996–2002). This generous long-term financial support enabled us to develop our own line of reasoning in a predominantly inductive manner, i.e. starting from a relative small set of theoretical axioms and a huge amount of concrete language data drawn from a wide variety of languages. At the same time, suggestions put forward by external reviewers of our project proposals impelled us more than once to change track slightly – and thus from 1998 onwards, the project became a decidedly European affair as we were encouraged to give prominence to a much smaller number of sample languages (which originally counted about 320 languages from all over the world). With a view to doing areal linguistics properly, we also applied the most basic principles of corpus linguistics by way of comparing languages on the basis of one and the same sample text, namely the original and the translations of Antoine de Saint-Exupéry’s Le Petit Prince. Clearly, this involved extensive nonlinguistic handy work by all members of the project team as a parallel corpus had to be created in a computer-compatible way. In the course of this preparatory work, we also learned to appreciate the advantages of textbased language comparison. Therefore, much of the data presented throughout this book stem from direct analyses of original texts. This additional task notwithstanding, we have published two dozens of articles and other work (dutifully mentioned in the references) related to the topic of this monograph. This was our principle strategy to reduce the size of the then book-to-be because potential side lines and potentially difficult-tointegrate chapters could be relegated to separate publications. The material we collected and the hypotheses we developed are so rich that at the time of

vi

Preface and acknowledgements

our writing this preface, we are seriously considering to present the (voluminous) statistical side of our research results as a book of its own. Likewise, the occupation with the negative counterpart of the Comitative, the Abessive, has come to the fore again only recently after a long period of negligence and chances are that this topic will keep us busy for some time. Thus, for some of the issues on which we do not dwell in this book followup studies are already in preparation. In a way, we are looking forward to the day when we are no longer haunted by the ghosts of Comitatives, Instrumentals and Abessives no matter how close friends we and they have become over the years. In the process of data collection, we went hunting for translations of Le Petit Prince all over Europe. Some of the translations turned out to be rare items and many were already out of print. With a view to working with a sizeable parallel corpus, we had to spend literally hundreds of Euros from our private purse to acquire these books. The German embassies in a variety of countries (Armenia, Azerbaijan, Belarus, Croatia, Moldavia, Slovakia, South Africa) and the Alliance Française in Malta, the Fryske Akademie in Ljouerts, and other culture-political institutions kindly assisted us in our search for the translations. Unfortunately, the book-hunt took some time and the translations were not simultaneously available. We thus, had to work on the individual texts separately which meant that some analyses could be applied to a particular translation only two years after the first of the series had already been subjected to the same procedure. After funding ceased in 2002, we decided to cut our further research short at that point and turn our findings into a book. Exactly when we were hard pressed to finish our work in order to apply for funding of a new project on a completely different topic, ten of eleven chapters of the manuscript disappeared from our hard disc overnight. This was the most depressing moment of the entire project especially because for a couple of days we were not sure whether backups had been made of the most recent versions. Luckily, some not so recent files could be found within a reasonable span of time and we successfully reconstructed part of the lost paragraphs. October 2003 will always be remembered as the time of maximal frustration – whereas December 2003 marks the culmination point of the project as we eventually added the last full stop to our draft version on which this book is based. For a long time, we considered our book to be without competitor. However, after the completion of the manuscript version early in 2004, we became aware of the fact that we were mistaken as Christian Lehmann and his project team produced a manuscript dealing with a similar topic. Since

Preface and acknowledgements vii

we had access to the monographic article by Lehmann, Shin, and Verhoeven (2005) only during the period in which our own book was under review for EALT, we did not want to postpone the date of publication of Comitatives and related categories any further by discussing the findings of our colleagues from Erfurt in detail. Owing to the fact that we could not find any major issue of dissent in the two approaches, we decided to make do with a general reference to the Erfurt paper as the most recent typological contribution to the study of Comitatives without going into the details. We reserve the discussion of Lehmann, Shin, and Verhoeven (2005) for one of our planned follow-up studies. Likewise, we like to mention the unpublished Habilitationsschrift by Annette Endruschat on Komitativität und Verben: durch ‘mit’ eingeleitete präpositionale Objekte in den romanischen Sprachen (forthcoming which was brought to our knowledge only when we were already giving the final touches to our monograph. As Endruschat draws extensively on various publications of ours in order to develop a theoretical framework for her analyses, it can be taken for granted that the approach resembles the one applied in this book. Thus, we have opted for mentioning Endruschat’s work only in passing – again as a recent contribution to Comitative research. Hopefully, no other pertinent major publication has escaped our notice. Note that the publications listed in the references do not exhaust the bibliography on linguistic work related to Comitatives. This book has benefited from the help offered to us by a countless number of people. First of all, we are grateful to the other (partly also former short-term) members of our project team at the University of Bremen: Suphi Abdülhayo÷lu, Sahin Asik, Cafer Balci, Aline Barthélemy, Maria Fontdevila, Thomas Gebel, Sabine Gorsemann, Kemal Güler, Traude Gugeler, Werner Hackbarth, Kai Herkströter, Sonja Kettler, Tamar Khizanishvili, Nataliya Levkovych, Tobias Mahlow, Susanne Müller, Ute Rotermund, Anna Sabater, Nino Sakwarelidse, Daniela Schuto, and Oxana Schwarz. We are also indebted to the many colleagues world-wide who readily gave us access to their publications, helped us find the translations of our sample text and/or filled in the questionnaire we used in the initial phase of the project: Nina Afanasjeva, Jarmo Alatalo, Saulius Ambrazas, Evi Annamalai, Thomas Balke, Laimute Balode, Dror Ben-Arie, Mayrene Bentley, Winfried Boeder, Norbert Boretzky, Sonja Bosch, Jürgen Broschart, Gisela Bruche-Schulz, Pierre Cadiot, Éva Ágnes Csató, Oliver Cromm, Mary Dalrymple, Aleksandra Derganc, Mark Donohue, Werner Drossard, Stefan Dyla, Annette Endruschat, Marcel Erdal, Mati Erelt,

viii Preface and acknowledgements Marta Fernández-Villanueva, Shadi Ganjadi, Martin Haspelmath, Irmeli Helin, Gerhard Hentschel, Michael Herslund, Armin Hetzer, Jarich Hoekstra, Wim Honselaar, George Huttar, Birgit Igla, Yoshihiko Ikegami, Lars Johanson, George Jones, Elzbieta Kaczorowska, Maria KoptjevskajaTamm, Debbie Krause, Diana Krull, Tuula Laakkonen, Hagen Langer, Mei Larjavaara, Jean LeDû, Rudolf Leger, Shlomo Lerman, Joachim Letsch, Intira Limsongprot, Silvia Luraghi, Timur Maisak, Stella Markantonatou, Elena Maslova, Kazuto Matsumura, Thomas Menzel, Marianne Mithun, Edith Moravcsik, Claire Moyse-Faurie, Nicole Müller, Aleksander Murzaku, Rogier Nieuweboer, Alberto Nocentini, Ole Norling-Christensen, Kemal Olfazer, Alwiya Omar, Chris Padro i Cirera, Enrique Palancar, Mikael Parkvall, Asya Pereltsvaig, John Phillips, Acrisio Pires, Bernard Pottier, Aina Razimagomedova, Richard Renault, Marco Rocha, Karmele Rotaetxe Amusategi, Valerij Sagovjev, Merja Salo, Agnes Sándor, HansJürgen Sasse, Christoph Schröder, Gunter Senft, Fritz Serzisko, Anna SĘrés, Valerij Solvjev, Xulio Sousa Fernández, Leon Stassen, Christel Stolz, Chuluu Ujiyediin, Tim van Baar, Catharina van Klincken, Willy van Langendonck, Yann Vincent, Suzanne Wash, Bernhard Wälchli, Urs Widmer, Hajo Zarachadet. Selected aspects of our research project were presented to academic audiences in Berlin, Bochum, Bremen, Essen, Günzburg, Nijmegen, Potsdam, and San Francisco. We like to express our gratitude to all our discussants. Andreas Ammann, Bernard Comrie, Ray Fabri, Bernd Heine and Elisabeth Verhoeven read the manuscript version of this book (or parts thereof) and made valuable comments which have left their mark in the printed book. Alan Aydelott took pains to make our nonnative English conform to the rules of English grammar and style. The publishing house Mouton de Gruyter and again Bernard Comrie deserve a heartfelt word of thanks for accepting our study for publication in the EALT series. We emphasise that everything we say in this book – and also how we say it – falls exclusively under our own responsibility. Thomas Stolz, Cornelia Stroh, and Aina Urdze Bremen, February, 2006

Contents

Preface and acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v List of abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii Part A. The long and winding road leading from intuition via problems to comitatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. Some problems with form and function . . . . . . . . . . . . . . . . . . . . . . 5 3. Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 4. Categories, definitions and terminology . . . . . . . . . . . . . . . . . . . . 15 4.1. The situation and its ingredients . . . . . . . . . . . . . . . . . . . . . . 16 4.1.1. Accompaniment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4.1.2. Instrumental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.2. The prototype and its derivations . . . . . . . . . . . . . . . . . . . . . . 25 4.2.1. Orientation and roles . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.2.2. Below the micro-role level . . . . . . . . . . . . . . . . . . . . . . 41 4.3. The primary marker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 5. Methodology and theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 6. Corpus/Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 7. Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 Part B. What happens when a universal blows up? Metaphor, syncretism, markedness . . . . . . . . . . . . . . . . . . . . . . . . . . 98 8. The Companion-Metaphor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 8.1. Global statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 8.1.1. Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 8.1.2. Phyla . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 9. Syncretism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 9.1. Agentivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 9.2. Coordination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 9.3. Spatial relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 9.4. Possession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 9.5. Semantic map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 10. Systematicity of interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 10.1. WITHOUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 10.2. Formal likeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

x

Contents

10.3. Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4. Boundedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.5. Zero-marking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.6. Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.7 Markedness and grammaticalisation . . . . . . . . . . . . . . . . . .

172 177 181 187 191

Part C. Europe: A continent where many things appear to be the same but turn out to be different under the looking-glass . . . . . . 196 11. A corpus-based analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 11.1. The Little Prince . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 11.1.1. General characteristics . . . . . . . . . . . . . . . . . . . . . . . . 200 11.1.2. Relators and their frequencies . . . . . . . . . . . . . . . . . . 204 11.1.3. Positive vs. negative relators . . . . . . . . . . . . . . . . . . . 217 11.2. On being (dis)similar in Europe . . . . . . . . . . . . . . . . . . . . . 222 11.2.1. The similarity co-efficient . . . . . . . . . . . . . . . . . . . . . 222 11.2.2. Dancing with girls, seeing with one’s heart and other shared contexts . . . . . . . . . . . . . . . . . . . . . 243 11.2.3. Where European languages disagree . . . . . . . . . . . . . 273 12. Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282 12.1. A-Type: Maltese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282 12.1.1. Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282 12.1.2. The functional domain of bi . . . . . . . . . . . . . . . . . . . . 285 12.1.3. The functional domain of ma’ . . . . . . . . . . . . . . . . . . 294 12.1.4. Summing up the Maltese situation . . . . . . . . . . . . . . . 305 12.2. B-Type: Latvian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307 12.2.1. Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307 12.2.2. The functional domain of ar . . . . . . . . . . . . . . . . . . . 310 12.2.3. Bare instrumental vs. PP . . . . . . . . . . . . . . . . . . . . . . 318 12.3. C-Type: Icelandic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322 12.3.1. Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322 12.3.2. The functional domain of með + DATIVE . . . . . . . . . . 323 12.3.3. The functional domain of með +ACCUSATIVE . . . . . . . 338 12.3.4. Summing up the Icelandic situation . . . . . . . . . . . . . . 346 12.4. Across the types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 Part D. Something better change! Origins, life-cycle, contacts: The dynamics of comitatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357 13. Grammaticalisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357 13.1. On sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357

Contents

xi

13.2. Going almost full cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . 14. Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14.1. Overt borrowing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14.2. Contact-induced typological change . . . . . . . . . . . . . . . . . 15. This is the end... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

363 370 371 376 387

Appendix I Sample languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396 Appendix II Genetic affiliation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422 Appendix III Questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447 Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508 Index of languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536 Index of authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542 Index of subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548

List of abbreviations

ABS ABL ACC ACT ADHORT ADJ ADV AFF AG AL ALL ANT AOR ASS ASSOC AUG AUX BEN CAUS CIRC CL COM COMBI COMPL CONDIT CONNECT CONT CONTR CONV CORR DAT DEB DEC

absolutive ablative accusative actor adhortative adjective adverb affective agent alienable allative anterior aorist assertive associative augmentative auxiliary benefactive causative circumstative class comitative combinative completive conditional connective continuative contrastive converb correlative dative debitive declarative

DEF DEM DET DIM DIR

DO DU DUMMY DUR DYN ELA EMPH ERG EXCL EXI EZAFE F FOC FUT GEN HAB HON ILL IMMED IMP IMPERF IMPS INAL INCEP INCL INCOMPL INDEF INDIC

definit demonstrative determiner diminutive directional direct object dual dummy durative dynamic elative emphasizer ergative exclusive existential Ezafe feminine focus future genitive habitual honorific illative immediative imperative imperfective impersonal inalienable inceptive inclusive incomplementive indefinite indicative

Abbreviations INESS INF INS INTENS INTERR INTRANS

IO IRR IT LIG LOC M MED

MR N NARR NEG NMLZ NONPAST NOM NTR NUM OBJ OBL ORN PART PASS PAST PAT PERF PL POSS POT PREP PRES PRET PREV PROG

inessive infinitive instrumental intensifier interrogative intransitive indirect object irrealis itive ligature locative masculine mediative Modus Relativus nomen narrative negator nominalizer non past nominative neuter numeral object oblique ornative partitive passive past tense patient perfective plural possessor potential preposition present preterit preverb progressive

PROP PTCPL QUOT

ƒ RECIP RED REFL REL SG STAT SUBJ SUBJUNC SUBORD SUPER SUPESS

T/A TOP TRANS UNDER

UQ UTR VEN

VN 1 2 3

proprietive participle quotative relator reciprocal reduplication reflexive relative singular stative subject subjunctive subordinated superlative superessive tense/aspect topical transitive undergoer unified quantity utrum ventive verb noun first person second person third person

xiii

Part A The long and winding road leading from intuition via problems to comitatives

1. Introduction All languages of the world provide ways and means to express relations of instrumentality and accompaniment – no matter whether this is achieved by employing highly grammaticalised and formally condensed expressions or apparent circumlocutions of sorts. In German, for instance, the preposition mit ‘with’ is used to encode both relations. Consider example (A1). (A1)

German (Moor 1994: 75) [mit dem Jungen auf einem Arm und one:DAT.M arm and with DET.DAT.M boy:DAT on anderen der Einkaufstüte] PP im DET.DAT.F shopping_bag in:DET.DAT.M other:DAT.M stieß sie die Autotür [mit with push.PRET.3SG she DET.ACC.F car_door unnötiger Heftigkeit]PP [mit dem violence with DET.DAT.M unnecessary:DAT.F Fuß] PP zu foot close ‘Carrying the boy on one arm and the shopping bag on the other, she kicked the door of the car shut with unnecessary violence.’

This sentence contains three occurrences of PPs headed by the preposition mit ‘with’. The relations encoded by this preposition, however, are semantically not identical. The PP mit dem Jungen (…und der Einkaufstüte) ‘with the boy (…and the shopping bag)’ can be classified as an instance of Accompaniment, the PP mit unnötiger Heftigkeit ‘with unnecessary violence’ is an example of Modal and the third PP, mit dem Fuß ‘with the foot’ instantiates Instrumental. Simplifying, we can characterise the relations involved as follows:

2

Introduction

– Accompaniment: Two participants of the same predicate fulfil the same role. – Instrumental: One participant serves as the tool with which another participant carries out an action. – Modal: an adverbial modification of a predicate. For linguists and especially for typologists and universals researchers, it is interesting to know whether such relations are associated with each other in other languages beyond the individual language (in this case, German) – and, if they are, whether this is a universal pattern or an option which only certain languages take. Our book is devoted to providing an answer to these questions. Beginning with Chapter 2 (and a modified repetition of example [A1]), we will discuss this and other similar, and less similar, cases in considerable detail. Our conventions for the presentation of examples and other important technical and terminological points are explained in Chapters 2–7. More elaborate definitions of the categories involved are given in Chapter 4. Before we look more closely at the data, we give insights into the theoretical convictions to which we adhere. General-comparative linguistics investigates the full range of structural diversity and unity of human languages in order to pave the ground for a comprehensive account of the phenomenology of what is possible in language and what is not. To this end, large amounts of data have to be viewed from a crosslinguistic or a typological perspective.1 General-comparative linguistics is the necessary prerequisite for general-comparative grammar (Lehmann 1984, 1989a), as it provides the empirical component for the latter. Our idea of general-comparative linguistics owes a lot to the concept of Empirische Universalienforschung (Bossong 1985). To our mind, general-comparative linguistics is Empirische Universalienforschung and Empirische Typologie at the same time, pace Seiler (2000: 28; 2004: 3–4). In addition, we are also indebted to many of the ideas put forward by Sasse (1988) as to the problems one inadvertently creates by applying supposedly universal categories to individual languages which themselves have not formed part of the sample of languages whose structures served as the basis for generalisations. The idea of studying Comitatives2 on a broader scale came to our minds when we read Seiler’s seminal papers on the topic of concomitance (Seiler 1974a–c). Further inspiration has come from the research conducted by the Cologne UNITYP-crew on the dimension of participation (Seiler 1988; Seiler and Premper 1991). The most important reference for much of our argumentation (especially in Section 1.3) is the

Introduction

3

pertinent publications of the UNITYP-offspring situated in Erfurt (Lehmann, Shin, and Verhoeven 2000; Lehmann and Shin 2005). However, Comitatives represent entire distribution profiles of Comitatives, i.e., they comprise – participant relations – inter-participant relations – intra-participant relations and various other relations as well, and thus cannot be equated with the classic thematic role of COMITATIVE (Lehmann, Shin, and Verhoeven 2000) nor with CONCOMITANCE (Seiler 1974a–c, Lehmann and Shin 2005).3 To our mind, Comitatives are closely associated with these areas but also transcend them as the discussion in the subsequent chapters will amply demonstrate. Comitatives are neither cases (= comitatives or Comitatives ) nor thematic roles (= COMITATIVES) in the usual sense. Rather, they are (semantic/conceptual) networks whose foundations may ultimately be similar to the networks postulated in psycholinguistically-minded connectionism (Berg 2001: 59). Owing to the fact that this study itself is not concerned with psycholinguistic matters but with typological issues, the answer to the question what status the networks actually have must be set aside for follow-up studies dedicated to theory. There is a certain dialectics behind our reasoning: One has to discuss comitatives, Comitatives and COMITATIVES as well as instrumentals, Instrumentals and INSTRUMENTALS in order to come to grips with Comitatives. This is what determines much of our argumentation throughout this book. Our approach has clear functionalist leanings (Croft 1995) and is overwhelmingly empirical-inductive. It does not necessarily presuppose any particular formal model of grammar and/or language (cf. Chapter 5 below) although we consider the framework(s) of grammaticalisation theory (Heine, Claudi, and Hünnemeyer 1991; Lehmann 1995) the most fertile for our endeavour. A full-blown theory of language may be developed only later on after our knowledge of the observable and derivable facts has increased substantially. Thus, one of the major goals of general-comparative linguistic research consists in the presentation of new or hitherto neglected data (be they connected to often discussed problems or evidence of something formerly unknown). General-comparative linguistics cannot fulfil its tasks by simply reiterating and re-evaluating isolated examples which have been handed down to us over many generations of mostly theoreticallyminded linguists. It is our contention that there is first and foremost an ur-

4

Introduction

gent need for an inventory of linguistic phenomena although we do not deny that even the most data-oriented approach always – as a general epiphenomenon of any scientific activity – starts from perhaps only implicit theoretical premises. Likewise, we do not cast doubt on the general value of proper theoretical-linguistic approaches. Rather, we take it for granted that theoretical deduction and empirical induction complement each other and that a faithful picture of what human language really looks like can only be achieved by a combination of both sides – and thus there will be ample opportunity for the integration of more theory-oriented ideas after the completion of this study in follow-up investigations. For the time being, we adopt a deliberately eclectic approach by drawing on various theories, models, analyses and hypotheses with diverse backgrounds. This book-length study of Comitatives and related categories is a contribution to empirical general-comparative linguistics. However, it is by no means restricted to a simple list of what we have found in our sample languages. As a documentation of our empirical findings, it contains at the same time a thorough evaluation in terms of linguistic typology and universals research. Owing to the fact that we are doing typology, this study aims at classifying languages according to the ways they organise Comitatives and related categories, and owing to the fact that we are doing universals research, this study tries to capture the invariant traits which define Comitatives independent of the idiosyncrasies of individual languages. Step by step, the importance of the subject at hand for general (and not only general-comparative) linguistics will become clearer in the remainder of Part A, where we introduce the basic notions, concepts and conventions in due course. In Chapter 2, we approach the topic of our investigation from a short presentation of some pertinent examples which circumscribe the general problems to be solved in the subsequent chapters. Chapter 3 is devoted to a presentation of the basic goals of this investigation. The definitions of our basic notions and the categories we focus on are provided in Chapter 4, where we also explain our terminology and other technicalities. Chapter 5 is devoted to methodological issues, including an outline of the history of research on Comitatives and related issues, with a short state-of-the-art report of current contributions to the subject at hand. This is followed by Chapter 6, where we describe our sample and the empirical basis used for our analyses and statistics. In Chapter 7, we sketch the structure of the remaining chapters of the book.

Some problems with form and function

5

2. Some problems with form and function With a view to facilitate comprehension, we approach the problems to be scrutinised in more detail later on, on an intuitive basis, in a manner of speaking. Let us assume, for a start, that the term “Comitative” is the label for the grammaticalised expression used to encode ACCOMPANIMENT relations – a short-cut working definition which we shall to elaborate upon in Chapter 4. The German preposition mit ‘with’4, like its English counterpart with (Croft 1990: 9–10)5, is a pertinent example of an expression of this kind, as it is used to combine two NPs of the same clause which represent two participants6 in a situation with one participant accompanying the other (cf. Chapter 4 below). German, however, is a language whose Comitative lacks a distinct encoding of its own, because the preposition functioning as the marker of the Comitative serves a wide variety of other grammatical purposes as well. Irrespective of the fact that various influential theoreticians assume that, within a clause, any semantic case can only be assigned once, (each NP bears a distinct case of its own7), examples like (A1) are anything but rare.8 We repeat (A1) as (A1’) here in slightly modified form. The example is drawn from the German translation of a rather popular English detective novel. In this sentence, there are three occurrences of the preposition mit ‘with’ which governs the morphological dative and functions here as the head of altogether four different PPs: The sentence-initial mit has scope over the coordinated NPs dem Jungen PP und der Einkaufstüte PP and thus there is no need to repeat the preposition after the conjunction und ‘and’ (which would have produced a grammatically acceptable structure, nevertheless). (A1’)

German (Moor 1994: 75) mit dem Jungen auf einem Arm und one:DAT.M arm and X/Y DET.DAT.M boy:DAT on der Einkaufstüte im anderen DET.DAT.F shopping_bag in:DET.DAT.M other:DAT.M stieß sie die Autotür mit X/Y push.PRET.3SG she DET.ACC.F car_door unnötiger Heftigkeit mit dem Fuß zu X/Y DET.DAT.M foot close unnecessary:DAT.F violence ‘Carrying the boy on one arm and the shopping bag on the other, she kicked the door of the car closed with unnecessary violence.’

6

Some problems with form and function

This triple overt occurrence of mit attests to the well-known and hotly debated form-function mismatch which is so characteristic of human languages. Ever since the early days of our discipline, linguists have been intrigued by the fact that supposedly different contents may be expressed by identical formal means (Van Valin 1993: 43 and Chapter 5 below).9 We agree with those linguists who assume that identity on the expression side is not just a funny coincidence with no importance whatsoever for linguistics. If the phenomenon of formal identity had not been taken seriously, modern approaches such as grammaticalization theory (Lehmann 1995) would never have developed. There is, however, no consensus among scholars as to what counts as an instance of formal identity and what linguistic units are actually involved (Traugott and Heine 1991: 2–3). For the time being, we adopt a rather crude preliminary procedure, presupposing that the distribution of a grammaticalised expression (i.e. the free or bound morpheme) delimits the extension of a category whose exact linguistic status still has to be determined.10 The intended meaning of our frequently employed expression that ‘something is marked for Comitative/Instrumental’ is that this ‘something bears the same primary relator that is also used to encode the Comitative proper in accompaniment situations’. This provisional method can be applied to our German example above. If we include the mit blocked under conjunction, we observe that, in (A1’), there are at least three different adverbial relations for which mit indiscriminately serves as a marker, cf. (A2): ACCOMPANIMENT occurs twice – while animate and inanimate COMPANIONS, MANNER and INSTRUMENTAL occur once each.11 (A2)

Three functions – one form [NP dem Jungen]…] adverbial: ACCOMPANIMENT [NP der Einkaufstüte]…] adverbial: ACCOMPANIMENT

sie stieß die Autotür zu [PP mit [NP unnötiger Heftigkeit]] adverbial: MANNER [NP dem Fuß]] adverbial: INSTRUMENT

Stylistically, the sentence may not sound absolutely well-formed to every native speaker of German (because of the repeated employment of one and the same marker, which is considered bad style at least in the written register). Notwithstanding these purely stylistic reservations, the sentence is

Some problems with form and function

7

grammatically correct and is fully in line with colloquial spoken German, where stylistic considerations are less restrictive. Note also that the accumulation of mit-tokens in this sentence is not induced by the English original, where with occurs just once.12 The multi-purpose nature of mit is of course hardly surprising as we find similar constellations in a variety of other languages where the translation equivalents of mit display an almost equally wide or even wider range of functions. Consider the Swedish examples in (A3), for instance, where the preposition med ‘with’ is used in much the same way as its etymological cognate mit. However, shared etymology is not necessarily an indicator of functional identity, cf. Part C. (A3) (A3.1)

(A3.2)

(A3.3)

Swedish13 ACCOMPANIMENT (Montan and

Rosenqvist 1982: 71) på söndag-ar-na bruk-ar Maria ät-a on sunday-PL-DEF.UTR.PL use.to-PRES Maria eat-INF middag med sin familj titt-a på TV med TV X/Y lunch X/Y POSS.3 family look-INF on kamrat-er-na och gå ut med hund-en buddy-PL-DEF.UTR.PL and go.INF out X/Y dog-DEF.UTR ‘On Sundays Maria eats lunch with her family, watches TV with her friends and walks the dog.’ INSTRUMENT (Montan and Rosenqvist 1982: 68) ska jag skriv-a med blyertspenna shall I write-INF X/Y pencil eller med kulspetspenna? or X/Y ball-pen ‘Shall I write with a pencil or a ball-pen?’ MANNER (LPP Swedish X 45) över allting svara-de kung-en over everything answer-PRET king-DET.UTR med upphöjt lugn calm X/Y majestic:NTR ‘Over everything, said the king with majestic calmness.’

As in German, one and the same Swedish morpheme covers ACCOMPANIMENT, MANNER and INSTRUMENTAL relations in the realm of adverbial adjuncts. On the basis of a larger Swedish corpus than the one we have made use of for the present purpose, examples of several functionally distinct instances of med within one and the same sentence might have shown up, too.

8

Some problems with form and function

As a matter of fact, the homologies between German mit and Swedish med go far beyond the functional domains mentioned above. On superficial inspection, the few differences there are between the two prepositions hardly seem to matter at all. It is therefore tempting to assume that the indisputable similarities can be explained in terms of linguistic genealogy, as both languages are closely related members of the Germanic phylum. The genealogical argument becomes less convincing as soon as we consider languages like Baka, a Ubangi language spoken in Cameroon. There cannot be any doubt that Baka is too far off from the two Germanic languages – not only in genealogical terms (Indo-European vs. Niger-Kordofanian) but also geographically (Europe vs. Africa) and typologically ([mildly] inflecting vs. [largely] isolating). Nevertheless, the preposition t8 ‘with’ conforms to the pattern known from German and Swedish because it is employed for ACCOMPANIMENT, INSTRUMENTAL and MANNER relations, cf. (A4). (A4) (A4.1)

(A4.2)

(A4.3)

Baka ACCOMPANIMENT (Kilian-Hatz 1995:

124)

tE ngaà tE jO~ wOtO~ pass.IMP X/Y I X/Y there ‘Come with me over there!’ INSTRUMENT (Kilian-Hatz 1995: 159) kO~nO wà tE ngbala ma à IMPERF chop.PRES fire_wood X/Y machete I ‘I am chopping the fire-wood with a machete.’ MANNER (Kilian-Hatz 1995: 132) /a gO tE wolo sèkò X/Y quickness chimpanzee 3SG.NARR go ‘The chimpanzee moved quickly.’

As the discussion in Part B will amply demonstrate, the pattern familiar from German, Swedish and many of their closest kin also occurs elsewhere on the globe. This observation may give rise to a change of opinion: If the use of one and the same morpheme for three different functions recurs in languages which are in no way related to each other, then this unexpected likeness might reflect something more general, perhaps a linguistic universal. The languages employ the same patterns because this is determined by some independent factor preventing structural variation: In a manner of speaking, German, Swedish and Baka cannot help being similar, because there simply is no choice. This would yield a one-to-many mapping relation between form and function as schematised in (A5). On this basis, one could

Some problems with form and function

9

also question the validity of postulating a three-way distinction of functions in the first place. The prepositions could either be syncretistic (meaning that despite their identity on the expression side, there are nevertheless three distinct categories) or homophonous (meaning that there are three segmentally identical prepositions – one for each function and thus one-toone mapping of form and function applies). However, this latter solution is especially unsatisfactory, cf. Part B. (A5)

One-to-many mapping relation

Language German Swedish Baka

Form mit med

t8 ACCOMPANIMENT

INSTRUMENT

MANNER

Function

This universalistic interpretation of the facts cannot be upheld – at least not in its radical form. First of all, example (A4.1) contains two instances of t8 , of which only one of which is in line with the usages made of mit and med in German and Swedish, respectively. The Germanic languages and Baka too use their prepositions to introduce the COMPANION in an ACCOMPANIMENT relation. In addition, Baka uses tE as a marker of directionality in combination with a local adverb: tE jO~ ‘towards there (= thither)’. A similar usage of the two Germanic prepositions is ungrammatical. Thus, the functional domains of the morphemes which encode the Comitative in the three languages are not exactly the same; they just overlap in an admittedly sizeable segment. Furthermore, many other languages deviate from the above pattern in various ways (which will be discussed in some detail in Part B). Maori, Hungarian and Classical Nahuatl provide interesting examples, cf. (A6)–(A8). For the sake of the argument, we only give a somewhat simplified account of the situation in these languages. (A6) (A6.1)

Maori ACCOMPANIMENT (Bauer

1993: 285) ka tuu ngaa tamaahine i te mataihi katau T/A stand DET.PL daughter at DET house_front right o te marae me tooraaua kookaa me Hinemaurea GEN DET court X SG.GEN.3DU mother X Hinemaurea ‘His daughters stood at the right of the front of the house, in the court, with their mother, Hinemaurea.’

10 Some problems with form and function (A6.2)

(A6.3)

INSTRUMENT (Bauer

1993: 283) ka haehae i ngaa ringa ki te mataa T/A cut DO DET.PL arm Y DET obsidian_flake ‘She cut her arms with obsidian flakes.’ MANNER (Bauer 1993: 290) i harihari ia i tooraaua tuutataki-tanga SG.GEN.3.DU meet-NMLZ T/A happy he at ‘He greeted her with joy (= he was glad at their meeting).’

In the Austronesian language Maori, ACCOMPANIMENT and INSTRUMENT relations are encoded by two distinct prepositions, namely me and ki. Neither of the two can replace the other. In addition, there is no overt marker for the MANNER relation, which is normally expressed by simple verbal modifiers or a separate predication (Bauer 1993: 290) as in (A6.3). Thus, Maori is characterised by a kind of one-to-one mapping relation of form and function (as far as the three categories under scrutiny are concerned). (A7) (A7.1)

Hungarian ACCOMPANIMENT (Lee

1990: 128) fiam-mal játszik DET son:POSS.1SG-X/Y play ‘He is playing with my son.’ INSTRUMENT (Lee 1990: 129) ceruzá-val írtunk pencil-X/Y write:1PL ‘We are writing with a pencil.’ MANNER (Lee 1990: 134) Balogh Tamás uram fanyar-an mosolygott Balogh Thomas Mr unpleasant-SUPESS smile:PRET.3SG ‘Mr Thomas Balogh smiled unpleasantly.’ a

(A7.2)

(A7.3)

In contradistinction to the languages discussed above, Hungarian employs bound morphology to encode cases and various categories. The suffix -vel/ -val is used for ACCOMPANIMENT and INSTRUMENT relations alike.14 Its vowel changes according to Hungarian vowel harmony, and the initial consonantal segment usually assimilates to the final consonant of the (morphologically simple or complex) stem to which the suffix is attached. Functionally, Hungarian conforms to the picture we saw in German, Swedish and Baka. However, MANNER relations require different case markers (although the use of -vel/-val is also marginally possible [Lee 1990: 69]): in (A7.3) the superessive -(V)n occurs, but there is almost a full paradigm of

Some problems with form and function 11

other cases which are more frequently employed for the purpose of marking relations (Lee 1990: 134–136). In a way, Hungarian is intermediate between the languages represented in (A5) and Maori: It allows for oneto-many mapping relations, although to a lesser extent. One-to-many mapping exists side by side with one-to-one mapping. MANNER

(A8) (A8.1)

(A8.2)

(A8.3)

Classical Nahuatl ACCOMPANIMENT (Launey 1981:

125) mǀztla huƗl-la-z in Pedro tomorrow hither-go-FUT DET Pedro Ư-huƗn in Malin-tzin POSS.3SG-X DET Maria-DIM ‘Tomorrow Pedro will come together with Maria.’ INSTRUMENT (Launey 1981: 128) cuahui-tl ic ǀ-nƝch-cocô wood-ABS Y ANT-OBJ.1SG-harm:PL ‘They have harmed me with a stick.’ MANNER (Launey 1981: 128) cual-li ic xi-nemi Y IMP-live good-ABS ‘Live decently!’

The same observation holds for Classical Nahuatl – the only difference being that in this Amerindian language, instrument and manner relations receive identical expressions (the postposition ic) whereas ACCOMPANI15 MENT requires the obligatorily possessed -huƗn as a relator. These morphemes are by no means interchangeable. Again, one-to-many mapping applies to only two of the three categories. However, the categories which receive identical expression are not the ones we encountered in Hungarian. At this point, we can also add Latin to our short survey (Kircher 1985: 146–147; Kühner and Stegmann 1955: 379–380 and 507): The inflectional ablative is used among other things for INSTRUMENTAL relations (gladiǀ interficƯ ‘to be killed by a sword’), and the preposition cum, which governs the inflectional ablative, is used for ACCOMPANIMENT relations, especially when the COMPANION ranks high on the animacy hierarchy (cum militibus ‘with the soldiers’). Moreover, this preposition is also the first choice when it comes to encoding MANNER relations (cum dƯligentiƗ ‘carefully’), although the pure ablative suffices in set phrases (injre ‘rightfully’) and is optional if the MANNER NP is syntactically heavy (mƗgnƗ dƯligentiƗ ‘with much carefulness’ [but also: mƗgnƗ cum dƯligentiƗ]).

12 Some problems with form and function Table (A9) is a synopsis of the above discussion of languages which do not fully conform to the German, Swedish or Baka model. Grey shading marks instances of syncretism. For better understanding, we repeat the Swedish data here. (A9)

Different mapping relations

Language Maori Hungarian Classical Nahuatl Latin Swedish

Form me -vel/-val -huƗn cum + ABL med ACCOMPANIMENT

ki -vel/-val ic med

0 -(V)n ic cum + ABL med

INSTRUMENTAL

MANNER

ABL

Function

If we combine (A5) and (A9), we immediately recognise that the full range of logically possible patterns is attested, namely in five constellations: There are languages which morphologically lump together all of the three relations and, at the same time, we find languages which keep them strictly apart. In between these two extremes, there are three patterns – each with a different pair of categories sharing the same morphological expression. This simple list does not tell us whether any of the patterns is crosslinguistically preferred or dispreferred. It therefore may give rise to the idea that full arbitrariness is at work and that there is thus absolutely nothing to be gained from an investigation into form-function relations of the kind reviewed here. In a field where anything might happen, predictability is next to zero, and therefore no scientific insights can be gained. We do not subscribe to this pessimistic interpretation because among other things it is based on too small a fragment of language structures. An assessment can only be successful if a suitably large section of the structural phenomenology has been scrutinised. We contend that the very concept of the linguistic sign is at stake if one discards the systematicity of form-function relations on the basis of such fragmentary data alone. As a matter of fact, the different solutions found in the above seven languages are not random, but depend upon a variety of other factors not included in the tripartite paradigm of ACCOMPANIMENT, INSTRUMENTAL and MANNER relations. The five patterns of syncretism (including absence of syncretism) are indicative of linguistic types, i.e., these patterns tend to occur in harmony with other structural solutions (especially the organisation of syncretism beyond the above three categories). Thus, we assume that there are conditions which determine the choice of the appropriate pattern in a given language. With a view to identifying and

Goals 13

properly evaluating these factors, it is of the utmost importance to get a better idea of the crosslinguistic distribution and internal makeup of syncretistic patterns involving or revolving about the Comitative. In addition, there are many other important tasks closely connected to the evaluation of the determining factors. Chapter 3 is dedicated to a presentation of those tasks which we tackle in this book. 3. Goals16 First of all, we intend this study to demonstrate that Comitatives are indeed worth investigating. Accordingly, we accumulate evidence from as many languages as possible for the grammatical relevance of Comitatives. This is mainly done in Parts B–C. On functional grounds, it is legitimate to assume that every language must provide ways and means to express ACCOMPANIMENT. However, there are no a priori reasons for these means to be highly grammaticalised. One could easily imagine that languages resort to an open reservoir of circumscriptions and periphrases without any standardised procedure of encoding a Comitative at all. Much to our own surprise, we have never encountered any language whose expressions of ACCOMPANIMENT indisputably fall outside the realm of grammar (in the sense of regular rulebased and/or construction-framed morphosyntactic patterns).17 As the examples in Chapter 2 suggest, the close association of Comitatives with what is conventionally called “grammar” does not imply that they must be provided with an absolutely distinct encoding of their own. Therefore, another important assignment for this study is to show that Comitatives exist at all, be it in individual sample languages or in a universal perspective. With a view to proving our point (namely, that Comitatives are there, even though one cannot see them on superficial inspection), we have developed a model of the linguistic sign which assumes a multilayered structure and recursive embedding of linguistic signs – to be discussed in Chapter 5. At the same time, we put forward a model of functional implicitness based on the recurrence of syncretistic patterns which also cover expressions of ACCOMPANIMENT. We focus on this question especially in Part B. By way of crosslinguistic comparison, we are able to draw so-called cognitive maps which, on the one hand, display the neighbourhood relationships of the Comitative and associated categories, and, on the other, indirectly corroborate that Comitatives are still present where grammar fails to single them out by distinct formal means. Given the

14 Goals universality of the Comitative, we have to check whether or not Comitatives behave similarly everywhere. As a matter of fact, the COMITATIVE in one language is not automatically identical with the COMITATIVE in another. We therefore set out to describe at least the most important aspects of the structural and functional phenomenology of Comitatives (yielding a kind of grammar of Comitatives with a morphological, a syntactic and a semantic component). This catalogue of features provides an adequate basis for identifying classes of languages which are characterised by a substantial degree of similarity (= types). In addition, this procedure enables us to test our sample languages for genealogical, areal and typological parameters. Are there any genealogical, areal or typological correlates which dictate how the Comitative behaves, structurally and functionally, in a given language? Moreover, we can also address the following issue: If we know that certain languages converge as to the list of features of their Comitatives, to what extent do these languages make similar use of this category? With a view to answering this question, we are forced to complement the matrix-typological method with a text-based comparison of languages (cf. Part C). We also develop a quantification procedure in order to determine the degree of similarity between the languages of our European sub-sample. Comitatives are not isolated from other categories. There is an intriguing interaction between Comitatives and various other categories – especially the Instrumental. We will outline these relationships and explain the moving forces behind the preferences and dispreferences of the Comitative regarding how it associates with certain categories when it comes to forming syncretistic chains. Furthermore, we describe the systematicity that regulates the relationship that holds between the Comitative and the Instrumental. For this purpose, we adopt a markedness-theoretical model. Closely connected to the issue of the markedness relations of the Comitative and comparable categories is the dynamic aspect: The expressions of Comitatives are not static, time-stable phenomena; rather, they have a past, present and future. Thus, we devote ample space to the discussion of grammaticalization channels, thereby identifying the (lexico-semantic) origins of typical expressions of Comitatives and their further fate in the ongoing process of grammaticalization (including a sideways glance at language contact). In all these areas, the notions of prototype and continuum have proven themselves to be immensely helpful if one aims at tidying up what would otherwise look like a mess. Additionally, both the prototype and the continuum are necessary and handy tools to come to grips with the

Categories, definitions and terminology 15

status of many subcategories that cover the space between prototypical Comitatives, prototypical Instrumentals and various other categories. Some languages stand out from the rest of the world’s languages because of their peculiar way of handling the Comitative. Incidentally, these are exactly those languages with which linguists are most familiar, namely the languages spoken in Europe. Owing to the fact that such European peculiarities have been mistaken for the general linguistic standard, not merely so, but quite pronouncedly even in research on the Comitative, we inspect Europe more closely in order to see whether or not the languages of Europe form a solid block of marked cases. In addition, we compare these languages on the macro-level and on the micro-level and thus are able to determine to what extent there is continent-wide convergence. This areallinguistic perspective is complemented from the diachronic point of view by our attempt to trace the diffusion of typological features on the geolinguistic map of Europe. The research project at the basis of this study has not come out of the blue. There is, of course, a rather long history of linguistic controversies about the Comitative. This history has to be documented at least in those parts that are immediately relevant for our own typological undertaking. Likewise, our results and findings have to be explicitly connected to the ongoing discussion about theory and practice of typology and universals research. Only in this way can we assign the Comitative as a research topic its proper place in general-comparative linguistics. 4. Categories, definitions and terminology In the preceding chapters, we have made use of a small number of terms which are intuitively clear from their role in traditional and received modern linguistics. These terms presuppose notions which are crucial to understand. Therefore, this chapter aims at introducing all the necessary definitions. We start from the most general and proceed down to the particular, i.e., the first question asked is ‘What is meant by ACCOMPANIMENT?’ and the second one ‘What morphological shape can the relator have?’

16 Categories, definitions and terminology 4.1. The situation and its ingredients Utterances inter alia serve the purpose of describing so-called states of affairs (in German: Sachverhaltsdarstellung) in the guise of declarative sentences (Seiler 1988: 114–118). More precisely, the state of affairs is not simply part of real-world ontology with the linguistic expression being a faithful copy of it. Rather, the linguistic expression shapes the state of affairs, gives contours to it or even creates it in the first place.18 This idea is deeply rooted in constructivism – the philosophical movement with which we share a number of basic convictions (without making us full-blown representatives of this school of thought). These shared convictions are also largely in line with the thoughts expressed in the programmatic chapters of the publications by members of the UNITYP-project (e.g. Seiler 2000: 15). The process of creating a state of affairs via the imposition of linguistic structure may of course be influenced by human perception and other nonlinguistic factors. However, the particulars of this process are only of minor importance to the concrete subject at hand (although they should not be underestimated in a general theory of language). What should be borne in mind, however, is the fact that language structure may require us to treat something as an instance of a certain state of affairs, although, in extralinguistic reality, the situation which our utterance refers to has nothing in common with other situations classified as exemplars of the same state of affairs. In addition, other situations which may ontologically resemble one another to a large degree can be allocated to different classes of states of affairs according to the regulations of a given language or the creativity of its speakers. There is no denying that we are facing some kind of relativism which is, perhaps, only rather mild – otherwise we would have to abandon for good the idea that the expression side of linguistic signs is an indicator of anything (semantic networks, for instance). The latter point is a very important distinguishing factor of our own approach: Many scholars working in the field of participation, semantic roles, etc., assume – though most of them only tacitly – that states-of-affair require a propositionally and syntagmatically complete (i.e. primary) predication for their description. However, we want to go beyond this boundary by also including secondary predications – among them adnominal phrases. The use of certain grammatical means does not always respect the boundaries between full and small clause. For methodological reasons (cf. Chapter 5), we will also consider what Lehmann, Shin, and Verhoeven (2000: 22–29) call interparticipant relations (Interpartizipantenrelationen).19

The situation and its ingredients 17

4.1.1. Accompaniment The Comitative is especially interesting in terms of general-comparative linguistics because it does not lend itself to easy classification. There are, of course, numerous suggestions as to where exactly the Comitative should find its proper place in the landscape of say, case-relations. Owing to the fact that the COMITATIVE is of a more complex nature than many of the other wellknown categories, attempts to treat the COMITATIVE in an analogous fashion to its supposed fellow-categories unavoidably face serious problems. These problems are such that one is tempted to claim that either the COMITATIVE does not belong to the same paradigm/level of abstraction as the RECIPIENT or the like, or that it is something completely different. This extraordinary character of the COMITATIVE calls for a detailed presentation of its traits. COMITATIVES are inconceivable without ACCOMPANIMENT relations. ACCOMPANIMENT relations are based on situations in which at least two protagonists are involved: These protagonists are the participants of the state of affairs depicted by an appropriate utterance. This situation model is patterned after the one described for local relations in Lehmann (1992: 627–630). Ideally, there is a tripartite internal logical form which comprises the two participants, namely the ACCOMPANEE20 and the COMPANION, and a relator which establishes the relation that holds between the two participants, cf. the binary schema in (A10).21 Note that the relations under scrutiny are (universal) conceptual-semantic ones which could best be termed roles. Their translation into morphosyntactic relations is language-(type)specific and belongs to a different level of abstraction: An ACCOMPANEE is an ACCOMPANEE, no matter whether the NP which represents it in a given clause is the subject, object or an adjunct of some kind. We repeat that there is absolutely no need for a one-to-one correspondence of roles and relations. (A10)

The ACCOMPANIMENT situation ACCOMPANEE

RELATOR

COMPANION

a

ƒ

b

Despite the fact that (A10) does not look to be too complicated a formula, the precise spelling of the structure is an intricate business, as the above formula only captures the nuclear skeleton of a state of affairs whose description normally requires a number of additional elements (Stolz 2003a: 214–218), among them the nuclear predicate, cf. below. The following is a list of some observations to be made in connection with (A10).

18 Categories, definitions and terminology – First of all, in a given utterance that shapes a situation as a state of affairs which qualifies as ACCOMPANIMENT not all of the three components of (A10) have to be overtly present. – Secondly, the linguistic expressions representing the two participants as well as the relator may combine in one word. – Thirdly, the relator (and to a lesser degree the two participants as well) may belong to very different morpheme/word classes and distribution classes even within one and the same language. – Fourthly, the linearization and topology of the components in the actual utterance vary according to the regulations of individual languages. – Fifthly, there is ample opportunity for multiple symbolizations of participants and relators in one and the same utterance. Many of these points can be easily exemplified with examples drawn from German. The following description gives an account of what is possible in an individual language. Whether or not the observed phenomena lend themselves to generalisations is a completely different matter, one which we will return to below (Section 4.2). In any case, German is a practical starting point, because the three components of formula (A10) tend to be represented separately by different syntactic words. However, there are also exceptions to this rule, as (A11.2) demonstrates. (A11) Absence/Incorporation of participants (A11.1) German: no ACCOMPANEE, COMPANION = mobile prefix COMPANION ƒ Peter kommt mit Peter comes X/Y ‘Peter is coming along.’ (A11.2) German: relator and COMPANION incorporated in the lexical verb ACCOMPANEE ƒ/COMPANION der Präsident wird zum Palast eskortiert DET president be:PASS.3SG to:DET.DAT palace escorted ‘The president is being escorted to the palace.’ In (A11.2), the relator is semantically incorporated in the transitive verb eskortieren ‘to escort’. The NP representing the ACCOMPANEE in a situation whose nucleus is a verb like eskortieren is the direct object of this verb in the active and accordingly the subject in the passive. One may even dare to go one step further and claim that the COMPANION is incorporated into the semantics of eskortieren as well because the verb itself presupposes the existence of an escort of some kind. In (A11.1), the mobile verb-prefix mit-

The situation and its ingredients 19

is detached from the finite verb (infinitive mitkommen ‘to come along’).22 The sentence is grammatically and pragmatically well-formed. In German, verbs bearing the mobile prefix mit- generally allow the explicit identification of the ACCOMPANEE to be suppressed. If the ACCOMPANEE must be mentioned, the relator appears twice:23 As a preposition introducing the NP representing the ACCOMPANEE and clause-finally where the mobile prefix usually winds up in main clauses with finite lexical verb.24 Note that this double occurrence of mit distinguishes sentences with rhematic ACCOMPA25 NEE from those with a rhematic COMPANION, cf. (A12). (A12) German (A12.1) rhematic ACCOMPANEE, relator = preposition and mobile prefix COMPANION ƒ ACCOMPANEE ƒ Peter kommt mit Robert mit Peter comes X/Y Robert X/Y prefix] [V [PREP NP]rheme ‘Peter accompanies Robert.’ (A12.2) rhematic COMPANION, relator = preposition ACCOMPANEE ƒ COMPANION Peter kommt mit Robert Peter comes X/Y Robert [PREP NP]rheme ‘Peter takes Robert along.’ If the COMPANION is the rheme, then the verbal prefix is ruled out and thus there is only one occurrence of the relator as in (A12.2). As soon as the second relator is added, the reading of the sentence involves a complete reversal of the roles of the NPs: The COMPANION of (A12.2) corresponds to the ACCOMPANEE in (A12.1), and the ACCOMPANEE of (A12.2) corresponds to the COMPANION in (A12.1). For obvious reasons, we concentrate on examples of complete patterns, i.e., we work with syntactic units in which all three of the components of formula (A10) are systematically represented by morphosyntactic means.26 Thus, cases of semantic incorporation (nota bene: not material incorporation!) of one or more of the members of the ACCOMPANIMENT situation in any other element of the utterance are excluded from our study. Sentences like (A11.2), i.e. those with a verb whose semantics alone presupposes accompaniment, fall outside the scope of our investigation. Moreover, it is indispensable for the relator to be a proper grammeme, i.e., a morpheme with clearly discernible grammatical functions (on the problem of derivational morphology, cf. Part B).

20 Categories, definitions and terminology So far, we have looked mostly at cases where the NP representing one of the two participants forms part of the VP (as an adverbial adjunct, for instance), i.e., where there is a higher predicator which, in traditional terms of dependency, governs both the bare NP and the PP. In a way, the PPs headed by mit in the above German sentences could be interpreted as a kind of secondary predication (of existence) or as a small clause, roughly paraphrasable as ‘and COMPANION is there [in the same place], too’ (Miller 1985: 96; Stolz 1996: 8–9).27 In a recent survey of secondary predication, Schultze-Berndt and Himmelmann (2004: 87–88) mention the fact that socalled depictives are often marked by morphemes which are otherwise Comitative and/or Instrumental relators. Thus, there is evidently something to the idea that mit-phrases are additional predications. In this and only in this perspective (for a different interpretation, cf. [A32]), the relation between ACCOMPANEE and COMPANION could be considered an indirect one, because the co-presence of the COMPANION is stated in connection to the event, action or state expressed by the lexical verb or nuclear predication.28 The adverbial adjunct extends this predication and thus belongs to the higher predication as an embedded part (as shown in [A13]). (A13)

German ACCOMPANEE NUCPRED ƒ COMPANION Z Erich [tanzt ([[mit Erna] PP] adjunct) (Walzer)]extended predicate Eric dances (X/Y Erna) (Waltz) ---------- primary predication ---------------+++ secondary predication +++ ‘Eric is dancing (the Waltz) (with Erna).’

This supposed closer connection of the COMPANION to the nuclear predication can, however, be challenged. If we employ the nominalization test, it turns out that the resulting structures do not sound especially felicitous with transitive verbs, cf. (A14).29 The nominalised nuclear predicate is marked by underlining. (A14) German nominalizations (A14.1) complex intransitive verb/predicate tanzen ‘to dance’ ACCOMPANEE NUCPRED ƒ COMPANION Erichs Tanzen mit Erna war holperig dancing X/Y Erna was clumsy Eric:GEN ‘Eric’s dancing with Erna was clumsy.’

The situation and its ingredients 21

(A14.2) transitive verb/predicate Zaun anstreichen ‘to paint a fence’ ACCOMPANEE NUCPRED ƒ COMPANION Erichs Zaunanstreichen mit Erna dauerte lange painting_the_fence X/Y Erna took long Eric:GEN ‘Eric and Erna took a long time painting the fence.’ (A14.2) is particularly awkward, as it almost invites an INSTRUMENTAL reading (cf. below) in which the COMPANION Erna would be turned into the TOOL or MATERIAL used for the painting of the fence. In order to prevent this oddness, the preposition mit could be disambiguated by the addition of the adverb zusammen ‘together’ – a technique that we will return to in Part B. Zusammen requires that the participants joint by mit (directly or indirectly) are on roughly equal terms, meaning that they must be (more or less) at the same rank on the animacy hierarchy and must belong to the same macrorole.30 Moreover, the optional use of zusammen suggests that there is indeed a closer relationship between the two participants than between the COMPANION and the nuclear predicate. This is also reflected by the use of the coordinating conjunction and in the English translation of (A14.2). This leads us to the problem of natural conjunction (McNally 1993), which we shall only touch upon here (for a more extensive discussion, cf. Part B). For the time being, it is sufficient to bear in mind that COORDINATION and COMITATIVE are closely connected to each other (Stassen 2000: 18–21). Owing to the fact that neither German nor English provides good and direct examples, we look at the Baltic language Latvian – a language that we will return to in more detail later. In Latvian, the association of COMITATIVE and COORDINATION with each other is easier to recognize. This language possesses a highly frequent coordinating conjunction un ‘and’ (borrowed from [Low] German und) and a preposition ar ‘with’, which governs the accusative in the singular and the dative in the plural (dative government in the plural is compulsory for all Latvian prepositions). However, if two human participants are co-agents in a given situation, the use of un is disfavoured. In its stead, Latvian resorts to the preposition ar, cf. (A15).31 (A15)

Latvian (Doòuleja 247) ACCOMPANEE ƒ COMPANION un Nelda ar Rnjdolfu ïoti nozƯmƯgi X/Y Rudolf:ACC very significantly and Nelda:NOM paskatƯjƗs uz ErnestƯni PREV:look:PRET:REFL.3 on Ernestine:ACC ‘And Nelda and Rudolf looked very knowingly at Ernestine.’

22 Categories, definitions and terminology Un could replace ar and still yield a grammatical structure – this time with the COMPANION NP in the nominative (Rnjdolfs).32 However, this generalisation of un would violate stylistic constraints. However, the relation between ACCOMPANEE and COMPANION can also be indisputably direct: German mit-phrases also occur functioning of adnominal attributes, cf. (A16). Notwithstanding the fact that the attribute is clearly a constituent of the NP this time, the mit-phrase can still be interpreted as a predication in its own right. However, the attributive cases do not state co-presence in relation to nuclear predication, but rather connect more closely to the NP representing the ACCOMPANEE. In (A16), ACCOMPANEE and COMPANION are represented by NPs which form together an immediate constituent of the VP (again, appropriately marked by the use of the coordinating conjunction and in the English translation). In contradistinction, the two NPs in (A14) do not form a higher syntactic unit – although one could easily paraphrase the sentence as ‘Eric and Erna dance the Waltz together’. (A16)

German ACCOMPANEE ƒ Q NUCPRED Erich fotografiert //[die Mutter /[mit X/Y Eric photographs DET mother dem Kind]PP/attribute]NP//direct object DET:DAT child ‘Eric is taking a photo of mother and child.’

In (A16), the adnominal attribute mit dem Kind is not extractable unless we intend to give the sentence a different reading. If the PP is moved, there are only two sites where it could possibly wind up, namely either in the immediate post-verbal position or in the immediate pre-verbal one – in the latter case, the subject NP would automatically be re-positioned to the slot to the right of the finite verb. No matter where mit dem Kind moves, it would always produce a structure that invites an interpretation according to which the AGENT Erich and the child form a group. Thus, we would have a sentence equivalent to English Erich and the child take a picture of the mother, meaning: the COMPANION of the ACCOMPANEE in PATIENT function becomes the COMPANION of the ACCOMPANEE in AGENT function. In sum, the secondary predication which states the co-presence of the COMPANION in the above examples can come in two different varieties which make use of one and the same relator in German (but cf. Section 4.2): One is adverbal, the other adnominal. The adverbal one modifies the

The situation and its ingredients 23

nuclear predication, since it forms part of the VP, but it is also semantically related to the other participant. The adnominal one modifies the NP representing the ACCOMPANEE. Only in the latter case do we have a headmodifier relation between the NP representing the ACCOMPANEE and the one representing the COMPANION. We encounter similar phenomena when the INSTRUMENTAL relation is taken into consideration too. 4.1.2. Instrumental Very often, Comitatives and Instrumentals are considered two sides of the same coin. This idea of a very tight relationship between the two categories dominates the linguistic discussion on our subject matter, past and present, and is also in the foreground of Parts B–C. Therefore, we include a characterisation of the INSTRUMENTAL situation right here. The Instrumental is the grammaticalised expression of an INSTRUMENTAL situation. Both ACCOMPANIMENT and INSTRUMENTAL situations allow for nuclear predicates of various degrees of transitivity. Thus, we encounter COMITATIVES in clauses whose nucleus is an intransitive, a transitive or a ditransitive lexical verb. The same holds for the INSTRUMENTAL, although there is a higher degree of association with typically transitive verbs, which prompts us to postulate a more complex formula for the INSTRUMENTAL (as a matter of fact, the two situations do not differ categorically as to their compatibility with transitivity!). This formula is more complex, as it contains the two binary relations cƒd + c,dƒe, which, in combination, yield the complete structure, cf. (A17).33 With a view to avoiding confusion, the role of the participant that employs an INSTRUMENT to carry out a certain action is labelled the USER (instead of AGENT). TOOL is the name given to the role assigned to the INSTRUMENT. The choice of this label is motivated by considerations of prototypicality (cf. Section 4.2). (A17)

The INSTRUMENTAL situation USER

RELATOR

TOOL

RELATOR PATIENT

c

ƒ

d

ƒ

e

NUCPRED verb

German mit-phrases of the INSTRUMENTAL type are adverbal, as in (A18), which is a modified version of (A16) above: We have replaced the adnominal COMITATIVE with an optional INSTRUMENTAL adjunct representing the TOOL. The explicit mention of the TOOL is optional for various reasons – one

24 Categories, definitions and terminology being the fact that the verb fotografieren ‘to photograph’ semantically incorporates the prototypical TOOL used for such purposes, namely a camera. (A18)

German USER

NUCPRED fotografiert photographs

PATIENT

Erich [die Mutter]direct object Eric DET mother TOOL ƒ [mit der Kamera]adjunct X/Y DET:DAT camera ‘Eric is taking a photo of the mother with the camera.’ The sequence of words in (A18), however, is ambiguous, because two very different readings are possible: Not only is it possible to interpret (A18) as an instance of an instrumental situation (most likely the first option for the majority of German native speakers) where the camera is the TOOL used to take the photo(s), but it is also legitimate to analyse (A18) as involving a kind of COMITATIVE (more precisely: ORNATIVE/CONFECTIVE, cf. 4.2): In this case the PP mit der Kamera is again an adnominal attribute modifying die Mutter. In this version, the USER Erich would take a picture of a mother who herself has a camera on her (or is associated with a certain camera known from contextual information). Erich’s TOOL would then be incorporated in the verb, and it would not be identical with the camera explicitly mentioned in (A18). The INSTRUMENTAL reading is exempt from ambiguity only if the mit-phrase is moved to a different position: When immediately adjacent to the verb, be it to its left or to its right (both positions being associated with focus), mit der Kamera allows only one interpretation, namely that of a TOOL. The potential ambiguity already suggests that ACCOMPANIMENT and INSTRUMENTAL situations may have something in common, that they may be linked to each other in such a way that it becomes hard to keep them apart. As the expressions for both are formally identical in German anyway, this observation only seems to state the obvious. Yet, not all languages are like German. Another point where COMITATIVES and INSTRUMENTALS show parallel behaviour is in their use as adnominal attributes. In (A19), the verbal structure mit dem Hammer schlagen ‘to hit with the hammer’ is the basis for the nominalization der Schlag/das Schlagen mit dem Hammer marked by an asterisk, as being based on a nuclear predicate. The internal PP of this subject NP functions as an adnominal attribute. This modifying relation with its INSTRUMENTAL semantics together with the head-modifier relationship is an inheritance from the above mentioned verbal structure.

The prototype and its derivations 25

(A19)

German ƒ TOOL NUCPRED [der Schlag /[mit dem Hammer] PP/adjunct] NP DET blow X/Y DET:DAT hammer PATIENT NUCPRED traf ihn unerwartet hit him unexpectedly ‘The blow with the hammer hit him unexpectedly.’

The difference between adnominal COMITATIVES and adnominal INSTRUMENTALS is one of relationality: The INSTRUMENTAL is characterised by a very close association to the nuclear predication underlying the nominalization, whereas the adnominal COMITATIVE displays a much closer association of the two participants to one another, although this does not rule out the possibility of a weaker connection to the nuclear predicate as well. We try to capture this difference schematically in (A20). (A20)

The different leanings of COMITATIVE and INSTRUMENTAL

COMITATIVE ACCOMPANEE

COMPANION

Å------- participants ----------------(-------- predicate ---------Æ) leaning towards ------- predicate ---------Æ TOOL NUCPRED INSTRUMENTAL

4.2. The prototype and its derivations The notion of the prototype belongs to the common vocabulary of many linguists of different scientific convictions no matter to what extent they actually use the term and with what implications. Lakoff’s (1987: 5–154) widely known study is the most influential application of this psychological notion to proper linguistic subject matters. Meanwhile, the concept of the prototype has proved to be a very handy tool, especially for typological and crosslinguistic investigations (Croft 2001: 102–104), including those which are dedicated to issues of grammaticalization (Heine, Claudi, and Hünnemeyer 1991: 39–64). For the time being, it suffices to observe that, to our mind, prototypicality need not be synonymous with high token frequency. In a way, we have also tacitly employed a kind of prototype concept in the

26 Categories, definitions and terminology examples discussed so far. In the light of the massive evidence in support of the idea that language structure has a strong anthropocentric tinge34, it is absolutely legitimate to start from a prototype that relies heavily on features of the [+human] kind, cf. (A21). (A21)

The human factor in the prototypes

ACCOMPANIMENT ACCOMPANEE ƒ

COMPANION

[+human]

[+human]

[+human] USER

[¬human] ƒ

TOOL INSTRUMENTAL

NUCPRED [+movement]

[+action] NUCPRED

In practically all of these (except the alternative reading of [A18]), the ACand the USER rank high on the animacy hierarchy and bear the feature [+human]; the same holds for the NPs representing the COMPANIONS in our examples of COMITATIVES. On the other hand, the TOOL is of course considered an inanimate material object in our examples of INSTRUMENTAL relations. For the nuclear predicate, we very often employed movement verbs. This solution is in line with the idea of Lehmann, Shin, and Verhoeven (2000: 6), that “Instrument und Komitativ sind prototypisch Begleiter des Agens in dynamischen Situationen” (for the postulated connection to the AGENT role, cf. Part B, Section 9.1). The prototype is, however, only a prototype, and thus does by no means exhaust the full range of possibilities. Before we address this crucial issue, we summarise the prototypical traits of the two participants:

COMPANEE

ACCOMPANEE: The ACCOMPANEE is the more prominent participant of the ACCOMPANIMENT situation in the sense that the ACCOMPANEE is construed as being more directly involved in the action, event, and situation described by the lexical verb. The prototypical ACCOMPANEE bears the feature [+human]. The expression representing the ACCOMPANEE may be a noun, a pronoun, or a pronominal affix. It may also be discontinuously marked in several slots. COMPANION: The COMPANION is the less prominent participant of the ACCOMPANIMENT situation in the sense that the COMPANION is construed as only indirectly involved in the action, event, or situation described by the lexical verb. The COMPANION participates only via its association with

The prototype and its derivations 27

the ACCOMPANEE. As the ACCOMPANEE, the prototypical COMPANION has the feature [+human]. The expression representing the ACCOMPANEE may be a noun, a pronoun, or a pronominal affix. It may also be discontinuously marked in several slots. The question arises whether our choice of certain semantic features has to be necessarily valid for all contexts in which the relator appears as the constant element. The answer is no, as the two prototypes represented in (A21) only have the high animacy of ACCOMPANEE and USER in common – beside the employment of the same relator. Given the strong ties that are said to exist between ACCOMPANIMENT and INSTRUMENTAL situations, we have to abandon the restriction to human participants in order to come to a better and full understanding, not only of the interconnectedness of the two situation types, but also of the nature of the COMITATIVE itself. The distributional properties of the relator morpheme will lead the way. 4.2.1. Orientation and roles Semantic roles are conceived of as largely independent of syntactic relations. To mention just one well known case, the agent role is dissociated from the subject relation just as the subject is in no way tied to the semantic role of AGENT (Sasse 1982: 269–272). There is no one-to-one mapping relation between semantic roles and fundamental syntactic relations although there may exist preferences of association between them, be it in individual languages or in a crosslinguistic perspective. This dissociation is also valid for the COMITATIVE and the INSTRUMENTAL. First of all, both the ACCOMPANEE and the COMPANION can be represented by subject or object NPs. Accordingly, the morphosyntax of a given utterance may be either ACCOMPANEE-oriented or COMPANION-oriented.35 This is undoubtedly true of clauses whose nuclear predicate is itself a verb of ACCOMPANIMENT or the like. For Croft (1991: 178), however, there is an obligatorily strong association of the COMITATIVE with the subject when it defines the former in the following way: Comitative: An entity that participates in a causal chain at the same point and in the same role as the subject of the main verb. It is likely that the comitative role also requires that the subject be the initiator of an act of volitional causation […].

28 Categories, definitions and terminology As the fundamental grammatical relation of the subject has been explicitly mentioned, we shall have a look at the syntax of comitatives first before we consider their semantic role more closely. It is doubtful that Croft’s definition can be upheld without modifications if we look beyond his own examples (Stolz 1993: 10 note 22). Sentence (A16) above is a case in point: the mit-phrase contains the COMPANION of an ACCOMPANEE that itself is the direct object of the main verb. There is no syntactic bond with the subject of the same clause, nor is there any volitionality on the part of the COMPANION involved as far as the causal chain is concerned. Given the fact that cases of object or oblique-oriented COMITATIVES cannot simply be discarded as supposed exceptions, we have to decide what to do with Croft’s definition. There are of course several options (some of which are logically interconnected): – reserve the term COMITATIVE exclusively for those cases which fit the definition, – introduce a distinct term for those cases where subject-orientation does not apply, – demonstrate that the problematic cases belong to a completely different class of phenomena, – give up the idea of subject-orientation altogether, etc. Indeed, Croft (1991: 237–239) also mentions a thematic role ASSOCIATIVE – without elaborating on it, however. ASSOCIATIVE or SOCIATIVE36 is a frequently encountered term used for NP-internal relations between an ACCOMPANEE and a COMPANION (cf. [A32]). One could try and reformulate the above definition of the COMITATIVE in such a way that ASSOCIATIVE becomes the cover term for those constellations that are in disagreement with the original definition.37 It remains to be seen whether anything could actually be gained by this ad hoc solution, (cf. Chapter 5). We strongly believe that the most one can achieve in this way is to show that COMITATIVE and ASSOCIATIVE have too much in common to be strictly kept apart, especially since formal syncretism including both categories is widespread and frequent. We therefore argue that it makes more sense to look at the COMITATIVE from a broader perspective in order to understand the conceptual ties that (we are about to show to) exist between the thematic roles on the clause level and the relations that hold within NPs. As a matter of fact, the dissociation of proper syntax and semantics does not stop here. Admittedly, the phrase marked for Comitative is very often an adjunct or adverbial phrase modifying the verb – but this highly frequent

The prototype and its derivations 29

pattern is by far not the only one. Consider the examples in (A22)–(A24) where the complement38 of the verbs corresponding to the English to marry somebody is marked for Comitative in three languages of very different genealogical, areal and typological affiliations (Stolz 1993: 23; 1996: 21– 22). There is no overt element in the English translations that corresponds to the relator morpheme in the original sentences. (A22)

Saami (Bartens 1989: 95) NUCPRED son náitala he marries ‘He is marrying Siri.’ ACCOMPANEE

COMPANION-ƒ

Siri-in Siri-X/Y

(A23)

Indonesian (Kwee 1981: 123) ACCOMPANEE NUCPRED ƒ COMPANION ia kawin dengan se-orang janda he marry X/Y one-CL.human widow ‘He married a widow.’

(A24)

Turkish (Ersen-Rasch 1980: 78) NUCPRED-ACCOMPANEE Ahmet-le evlen-di-m Ahmet-X/Y marry-PAST-1SG ‘I have married Ahmet.’ COMPANION-ƒ

In contradistinction to many of our previous examples, the nuclear predicate is no longer a verb of movement (which is normally intransitive and thus not a good candidate for requiring an additional NP [discounting locative complements for the sake of the argument]). In numerous societies, marrying is not a symmetrical reciprocal act, but a combination of two distinct sub-acts distributed unevenly over the participants. For the sake of the argument, we neglect these cultural differences for the time being, though. The situation frame of the event of marrying nevertheless involves two participants who cooperate: The action is reciprocal, as the one marries the other and vice versa. This reciprocity does not, however, translate into an absolute balance of power of the participants: the situation is depicted as oriented towards the participant whose NP functions as subject of the clause. Nevertheless, these cases still conform to Croft’s definition of the COMITATIVE given above, because the NPs introduced by a comitative marker can be ascribed similar semantic properties as the ones ascribed to the subject NP. In all three languages, the relator is the same as the one that would be used to introduce the COMPANION in situations whose nuclear

30 Categories, definitions and terminology predicate is a movement verb (and the same relator is also used for the marking of INSTRUMENTAL relations).39 If we go by the criterion of formal identity (cf. Chapter 5), we have to accept as fact that the relator morpheme has a distribution which reaches far beyond the prototypical constellation. Not only do static verbs or nuclear predications replace dynamic ones: the ACCOMPANEE and the COMPANION can no longer be taken too literally, or the terms become simple labels for the participants involved in a situation whose description requires the use of the same relator that is also employed in case of prototypical ACCOMPANIMENT relations. In (A25) from Estonian, the NP representing the COMPANION depends on the adjectival predicate nominal ühevanune ‘of the same age’, which forms a discontinuous nuclear predicate with the copula. (A25)

Estonian (Hasselblatt 1992: 96) NUCPRED1 COMPANION-ƒ isa on ema-ga father is mother-X/Y ‘Father is as old as Mother.’ ACCOMPANEE

NUCPRED2 ühe-vanune one-aged

In this case, the NP emaga ‘with the mother’ is compulsory, since removing it from the sentence would produce an ungrammatical torso. The use of the morphological comitative40 in this construction is one of the options that speakers of Estonian have to express the equative – and it is obligatory with all predicative adjectives bearing the prefix ühe- ‘one-/same-’ (Hasselblatt 1992: 96). The two participants are depicted as being on equal terms – a constellation that fits in well with the reciprocity mentioned above. This reciprocal nature of the relation that holds between ACCOMPANEE and COMPANION does not, however, extend to all contexts in which the same relator is employed. In Malaccan Creole Portuguese, there is an object marker ku (< Portuguese com ‘with’41) which is extensively (but not exclusively) used with object NPs which rank high on the animacy hierarchy (Baxter 1985: 156), cf. (A26). (A26)

Malaccan Creole Portuguese (Baxter 1985: 150) eli ja dali ku John PERF hit X/Y John he ‘He hit John.’

Ku has also retained the functions of marking the COMITATIVE and the INSTRUMENTAL in the appropriate grammatical contexts, cf. (A27).

The prototype and its derivations 31

(A27) Malaccan Creole Portuguese (A27.1) ACCOMPANIMENT (Baxter 1985: 115) ACCOMPANEE ƒ COMPANION yo sa papa ta bai mar ku yo sa kanyóng GEN elder_brother I GEN father PROG go sea X/Y I ‘My father is going fishing with my elder brother.’ (A27.2) INSTRUMENTAL (Baxter 1985: 162) USER ƒ TOOL eli ja kotrá aké kandri ku faka that meat X/Y knife he PERF cut ‘He cut the meat with a knife.’ Object marking is a function that cannot be accounted for in terms of reciprocity, as it occurs also in combinations with transitive verbs whose semantics clearly imply a one-sided directionality (Baxter 1985: 151–153), thus precluding an egalitarian status of the two participants, as in (A28).42 (A28)

Malaccan Creole Portuguese (Baxter 1985: 157) aké tempu sa jenti midu ku deus GEN people fear X/Y God that time ‘People of those times feared God.’

In the light of this evidence, it becomes necessary to question whether or not it is legitimate at all to compare the various uses which relators like Malaccan Creole Portuguese ku are put to in individual languages and crosslinguistically. Before we face this crucial issue, we inspect further evidence from the realm of inter-participant relations, namely the adnominal uses of phrases marked for Comitative. We now enter the territory of those relations for which the term ASSOCIATIVE would most probably have been used in Croft’s model (cf. above). In adnominal cases, the syntactic relation between the constituents representing ACCOMPANEE and COMPANION, respectively, is of the usual headmodifier type. Example (A29) from Saami documents, an interesting case where the comitative phrase modifies a pronoun in the genitive which is itself the modifier of a possessed noun. (A29)

Saami (Bartens 1989: 94 note 2) ACCOMPANEE

COMPANION-ƒ

dat lea munno Biera-in this is my Peter-X/Y ‘This is the house of me and Peter.’

goahti house

32 Categories, definitions and terminology There is a complex possessor comprising two NPs, the second one of which lacks any relational marking that would link it to the possessed noun. In lieu of a genitive affix, we find the comitative suffix, which connects one possessor to the other. The pronoun of the first person singular, however, is inflected for the genitive and thus bears the relational marker for the two possessors as a complex unit (almost yielding a kind of group inflection). The best way to translate the relator morpheme is once again by a coordinating conjunction. As with the cases discussed in the preceding paragraphs, the two NPs joined by the comitative relator have the same semantic role: both are possessors. This equality of roles is not directly reflected by morphosyntax, though. What is more interesting, however, is the fact that the complex NP that the attributive COMITATIVE is embedded in is free to occur in any syntactic relation. This relation hierarchically overrules the internal relation holding between ACCOMPANEE and COMPANION. This higher-order relation is identical to the one which affects the head of the NP – and thus the entire NP including the modifier. Consider the two sentences (A30)–(A31) taken from the Romanian translation of Le Petit Prince.43 (A30)

Romanian (LPP Romanian V, 14) ACCOMPANEE ƒ COMPANION ideea cu turma de elefanĠi X/Y herd:DEF of elephant:PL idea:DEF îl făcu pe micul prinĠ să rîdă him make ACC little:DEF prince SUBORD laugh:3SG.SUBJUNC ‘The idea of the herd of elephants made the Little Prince laugh.’

(A31)

Romanian (LPP Romanian I, 18) oamenii mari m-au povăĠuit să man:PL:DEF big:PL me-have:3PL advise:PTCPL SUBORD le las încolo them let there ACCOMPANEE ƒ COMPANION de desene cu úerpi boa X/Y snake:PL boa of picture:PL ‘The grown-ups have advised me to let them alone with pictures of boas.’

In (A30), the complex NP idea cu turma de elefanĠi is the higher grammatical subject of the causative construction whose nuclear predicate is the finite verb făcu ‘they made’. In (A31), on the other hand, the equally complex NP desene cu úerpi boa is a constituent of the VP with the status of a

The prototype and its derivations 33

complement or adjunct. Both NPs contain an adnominal PP headed by cu ‘with’. These PPs cannot be considered to be in a semantically reciprocal relation with the nominal head of the complex NP. With a little bit of good will, both attributive NPs can be analysed as being in a part-whole relationship with their respective heads. Turma de elefanĠi is the most important aspect of the simile the narrator of the story uses to describe a situation to the Little Prince. Likewise, úerpi boa are the only visible entities reproduced on the picture which the narrator of the story used to impress the adults. In terms of syntactic relations to the main verb, the elephants and the boas are only indirectly involved in the situations described by (A30)–(A31), whereas the idea and the pictures qualify as proper participants of the nuclear predication. However, what amuses the Little Prince is the elephants (standing on each other’s back) and what fails to scare the adults is the boa (resembling a hat more than a snake). Thus, the involvement of the respective entities in the situation cannot be sweepingly denied. Given this hierarchical order of the more directly involved head noun and the indirectly involved adnominal attribute, it is tempting to analyse the entire NPs as representing one complex but unitary participant which has internal structure inasmuch as there is an intra-participant relationship between its constituents. At least on superficial inspection, the modifier participates in the description of the state-of-affairs only by virtue of its relation to the head. Lehmann, Shin, and Verhoeven (2000: 29) focus on interparticipant relations proper. However, as they also include possessorpossessee relations, (i.e. adnominal attributes) in their study, we consider it justified to extend the domain of participation to intra-participant relations in general as well. Intra-participant relations establish a third level of participation, namely the most remote one in terms of conceptual distance to the nuclear predication as the centrepiece of the situation, cf. (A32).44 The question remains to be answered whether the relations on all three levels obey the same principles. In (A32.2), the property of being bald headed is perhaps important for the identification of the referent of the subject NP. However, the bald head itself cannot be considered a full-blown participant of the nuclear predicate fährt. Indeed, it stands in a modifying (adnominal, i.e. intra-participant) relationship to the true participant der Mann. Notwithstanding its marginal status in terms of conceptual closeness to the nucleus of the situation, the attribute is still part of the description of the state of affairs and thus cannot simply be neglected on the grounds of its only mediated involvement in the situation.45

34 Categories, definitions and terminology (A32) The three levels of relations (A32.1) The levels separated I level

participant

II level

participant

predicate

participant relation

inter-participant relation predicate

participant

III level

participant

intra-participant relation predicate

participant

(A32.2) Relation networks in a situation USER/ ƒ ORNATIVE ACCOMPANEE

NUCPRED ƒ

MEANS OF TRANSPORTATION

[der Mann [mit der Glatze]] fährt [mit seinem Sportwagen] the man X/Y the bald head drive:3SG X/Y his:DAT sports_car ‘The bald headed man is driving in his sports car.’ mit der Glatze

ORNATIVE ACCOMPANEE

der Mann

AGENS

fährt

USER

NUCPRED mit seinem Sportwagen MEANS OF TRANSPORTATION

The prototype and its derivations 35

In terms of syntactic relations, NPs marked for Comitative (and, in case of syncretism, Instrumental) can be either arguments, adjuncts, objects, complements or attributes. They may also be a sub-constituent of a subject NP. Thus, it is clear that there is no purely syntactically motivated definition of Comitatives and Instrumentals. If syntax is not decisive, perhaps semantics is. A closer look at the semantic roles usually assigned to syntactic units reveals, however, that not all of the problems can be solved in a satisfactory way. The most serious problem is a very fundamental one, as it affects the interplay of form and function. Given that what we have been calling COMITATIVES (and INSTRUMENTALS) is attested in all kinds of NPs used in different syntactic relations, it comes as no surprise that our COMITATIVES (and INSTRUMENTALS) also occur in a wide range of semantic roles. In the framework of Role and Reference Grammar, Van Valin and Wilkins (1996: 306) distinguish three macro-roles to which they add a level of micro-roles. Macro-roles and micro-roles are connected to one another systematically, cf. (A33). We adopt the diagram with minor modifications from Lehmann, Shin, and Verhoeven (2000: 19). (A33)

Micro-roles and macro-roles

Micro-roles

Macro-roles

AGENT FORCE COMITATIVE

ACTOR

INSTRUMENT EXPERIENCER EMITTENT RECIPIENT

INDIRECT

BENEFICIARY SYMPATHETICUS SOURCE LOCATION

UNDERGOER

GOAL THEME PATIENT

The two roles which interest us most – COMITATIVE and INSTRUMENT – belong to the same macro-role, namely ACTOR. Being classified as ACTOR is tantamount to being active in a situation. Note that in contradistinction to many other micro-roles which are listed below them, COMITATIVE and INSTRUMENTAL associate with only one macro-role. The order of micro-roles

36 Categories, definitions and terminology in the above figure is not random but reflects their internal conceptual closeness or distance in a kind of topological metaphor. What is more, Van Valin (1993: 43) assumes that the use of identical means of expression is not only indicative of neighbourhood relations on the level of micro-roles. He also suggests that formal syncretism is determined by association to the same macro-role. Where there are identical markers, some kind of conceptual bond must exist. Since identity of formal expression is not restricted to the microroles enumerated in (A33) but occurs frequently with many other categories, there is no a priori reason to restrict a description of Comitatives to COMITATIVES, i.e., to the usual thematic roles proper. Notwithstanding their outsider status, categories other than the classic thematic roles must also be connected to their partners in syncretism by conceptual association. For the sake of the argument, we take issue with Van Valin’s (1993) hypothesis, according to which formal syncretism remains within the boundaries induced by the extension of the macro-role. For a start, consider some evidence in favour of this idea. In the Celtic language Breton, the relator gant ‘with’ is not only used in instances where the terms COMITATIVE and INSTRUMENTAL seem appropriate, but, is also used as a marker of the AGENT in passive constructions (Press 1986: 161), cf. (A34). (A34) Breton (A34.1) INSTRUMENTAL and passive AGENT (LPP Breton IV, 9) ar steredennig-se n-eo bet spurmantet nemet ur DET asteroid-DEM NEG-is been seen except one wech gant ar bellsellerez e 1909 gant ur in 1909 X/Y a time X/Y DET telescope steredoniour turk astronomer Turkish ‘This asteroid was seen just once through a telescope by a Turkish astronomer.’ (A34.2) ACCOMPANIMENT (LPP Breton IV, 50) c’hwec’h vloaz ‘zo endeo ez eo aet six year is already that is gone va mignon gant e zañvad my friend X/Y his sheep ‘It has already been six years since my friend went away with his sheep.’ The comitative marker is also used to encode the (passive) AGENT. This syncretistic pattern corroborates Van Valin’s prediction: Three micro-roles

The prototype and its derivations 37

(AGENT, COMITATIVE and INSTRUMENT) of the same macro-role ACTOR share their morphological expressions. However, the range of functions of the preposition ku in Malaccan Creole Portuguese cannot be accommodated with what the theory expects: Sentences (A26)–(A28) above document syncretism between COMITATIVE, INSTRUMENTAL and THEME/PATIENT, i.e., formal likeness of micro-roles which belong to two different macro-roles, namely ACTOR and UNDERGOER. According to the association lines in (A33), ACTOR and UNDERGOER never overlap as far as COMITATIVE and INSTRUMENTAL on the one hand and THEME and PATIENT on the other are concerned.46 Put differently, a constellation of the Malaccan Creole Portuguese type would not only be beyond the predictability of Van Valin’s model, but it also runs counter to all (model-internal) expectations. The same holds for the spatial microroles roles subsumed under the macro-role of UNDERGOER. According to the same putative restriction, syncretistic patterns involving COMITATIVE/INSTRUMENT and SOURCE/LOCATION/GOAL should be ruled out. However, the examples under (A35) show that this restriction is hardly tenable in this absolute form (Stolz 1997a: 144–147; 2001a: 341). (A35) Syncretism of spatial micro-roles and COMITATIVE/INSTRUMENT (A35.1) Irish (Ó Sidhail 1985: 105): LOCATION tá droim Chait leis an mballa wall be stand Cáit X/Y:3SG.M DET ‘Cáit is standing (with her back) against the wall.’ (A35.2) Malayo (Amaya 1989: 60): GOAL maestra-ga gumush-ina une-ka ihkuela-mba bring-3 school-Y teacher:F-ERG child-PL:ABS ‘The teacher takes the children to the school.’ What comes first to mind is of course the syncretism of INSTRUMENT and in Latin, where both micro-roles are encoded by the inflectional ablative. Consider domƯ sum ‘I am at home’, domum eǀ ‘I go home’ and domǀ veniǀ ‘I come from home’. The noun domus ‘house’ occurs in three different case forms, namely the residual locative in -Ư when combined with the static copula verb essere ‘to be’, the directional accusative in -um when combined with the motion verb Ưre ‘to go’, and the ablative in -ǀ when combined with the motion verb venire ‘to come’. Thus, the different case forms reflect the different semantic relations that hold between the locative adverbial and the main predicate: LOCATION, GOAL and SOURCE. The ablative fulfils the latter function (traditionally called ablativus separativus/ SOURCE

38 Categories, definitions and terminology originis).47 Syncretism of LOCATIVE and INSTRUMENT is especially frequent, whereas the COMITATIVE partakes in syncretistic patterns involving LOCATIVE only if it has the same encoding as the INSTRUMENT (Stolz 2001a). Example (A35.1) from the Celtic language Irish48 reflects this condition. Much the same can be said for the syncretistic compatibility of COMITATIVE and GOAL: A combination of this kind always includes the expression of the INSTRUMENT also. If no syncretism of COMITATIVE and INSTRUMENT applies, only the latter can have the same expression as GOAL. This case is exemplified by (A35.2) from the Amerindian language Malayo spoken in Colombia.49 Moreover, Wilkins (1989) argues that in the Australian language Arrarnta (attributed to the Pama-Nyungan phylum), the morphological ablative, whose primary function is to denote SOURCE, also displays functions which are typically associated with the COMITATIVE, cf. (A36). (A36) Arrarnta (A36.1) ACCOMPANIMENT (Wilkins 1989: 186) Les Kathy-nge petye-rne Les Kathy-ABL come-PAST.IMMED ‘Les just arrived with Kathy.’ (A36.2) SOURCE (Wilkins 1989: 185) re pmere-nge lhe-ke go-PAST.COMPL he camp-ABL ‘He went from the camp to the creek.’

lhere-werne creek_bed-ALL

In example (A36.1), Kathy is the instigator or controlling participant, whereas Les appears to be less actively involved in the situation described. According to Wilkins (1989: 186), a translation that would reflect this difference in control exerted by the two participants more closely would read either It was Kathy’s idea that she and Les came or Kathy drove Les to the place. The functional range of the ablative does not follow the same-macrorole constraint which would normally exclude the identity of formal means not only for SOURCE and COMITATIVE but also for SOURCE and AGENT. If the ablative in (A36.1) marks the higher agentivity/volitionality (qua control) of one of the participants, the morphological syncretism is at odds with the idea that ACTOR and UNDERGOER do not overlap in those micro-roles that imply either high(est) or no agentivity at all. Moreover, this aspect of asymmetrical control distinguishes the ablative from the regular morphological comitative in Arrarnta50, cf. (A37).

The prototype and its derivations 39

(A37)

Arrarnta (Wilkins 1989: 211) pwepelye irrpenge-larlenge arnerre-le tadpole fish-X rock_hole-Y ne-me kwaty-iperre rain-after be-NONPAST.PROG ‘After the rain, tadpoles, along with fish, live in the rockholes.’

According to Wilkins (1989: 211), the comitative suffix -larlenge does not convey any information as to differential involvement of the participants. The suffix is attached to the NP, which is not in focus (lack of comitative marking therefore is indicative of focus). The comitative is used when both participants have equal status: -larlenge does not ascribe control to one of the participants. The ablative is used to mark the participant that exerts a higher degree of control, whereas a third case category is used to identify the participant which is controlled (Wilkins 1989: 212). This third category is the so-called proprietive, cf. the minimal-triplet in (A38). (A38) Arrarnta (Wilkins 1989: 212–213) (A38.1) Ablative: marked NP exerts control Rosie Margaret-nge lhe-ke lhere-werne Rosie Margaret-ABL go-PAST.COMPL creek_bed-ALL ‘Rosie went along to the creek with Margaret (whose idea it was to go there).’ (A38.2) Proprietive: marked NP is controlled Rosie Margaret-kerte lhe-ke lhere-werne Rosie Margaret-PROP go-PAST.COMPL creek_bed-ALL ‘Rosie went to the creek with Margaret (coming along).’ (A38.3) Comitative: unmarked NP is in focus Rosie Margaret-larlenge lhe-ke lhere-werne Rosie Margaret-X go-PAST.COMPL creek_bed-ALL ‘Rosie went to the creek together with Margaret.’ The differential ascription of control in ACCOMPANIMENT situations constitutes a problem for the prototype approach because the latter does not specify whether there is symmetry or asymmetry of control among the participants. Is the prototypical scenario egalitarian or hierarchical? Furthermore, these differences in control are also suggestive of different micro-roles (or even macro-roles): One of the participants may be more active than the other, and thus only the former qualifies as full-blown AGENT. The other one (being taken along, lacking initiative or volition), on the other hand, lacks some of the crucial features. Nevertheless, both participants in the

40 Categories, definitions and terminology three versions of (A38) pass as ACTORS – though the one without control also displays traits of an UNDERGOER. This combination of the seemingly incompatible is one of the aspects which make Comitatives a very promising field of research. Again in Arrarnta, the proprietive is also frequently employed in adnominal attribution. Temporary possession, corporal and other properties as well, such as objects carried along, etc. can be encoded by a NP marked for proprietive (Wilkins 1989: 192–198), cf. (A39). (A39)

Arrarnta (Wilkins 1989: 193) artwe angkwerre urrpetye-m-urrpetye-kerte re he man sister few-UQ-few-PROP yanhe petye-me there come-NONPAST.PROG ‘The man with (= who has) six sisters is coming just over there.’51

Discounting the possessive case (Wilkins 1989: 198–202), this function is unique to the proprietive – the remaining two case categories which otherwise classify as Comitatives do not occur in adnominal constructions. This restriction is absolutely in line with the fact that adnominal attributes introduce secondary properties associated with a participant. These properties have no independent status and should therefore be considered subordinate to, if not controlled by, the participant they are ascribed to. The ablative cannot serve this purpose, as it marks the superordinate participant (a marking strategy that would cause conflicts with further relational marking of the head noun). Likewise, the comitative itself is not appropriate for encoding relationships as the one in (A39) because this case category assigns equal status to both participants. Thus, the proprietive is the best choice for intra-participant relations of this kind. What the examples from Arrarnta suggest is the following: – there may be several Comitatives with distinct functions in one and the same language, – there may be restrictions on their admission to attributive and/or predicative constructions, – the relators specify the exact spelling out of the relation that holds between ACCOMPANEE and COMPANION, – none of the various Comitatives must be syncretistic with the Instrumental. Languages differ from one another as to how many Comitatives (and Instrumentals, for that matter) they tolerate, what functions they assign to

The prototype and its derivations 41

them, whether any of them is subject to distributional restrictions in adnominal and/or predicative contexts, and what syncretistic patterns are admissible. Independent of the solutions that individual languages use, the simple fact that many of them – and not only those of Australia and adjacent areas where the patterns described for Arrarnta are indeed very widespread – make use of more than just one Comitative suggests that the caserole hierarchy (A33) does not tell the whole story. There is evidence for at least one more level whose proper place is below the present micro-level. 4.2.2. Below the micro-role level The terms COMITATIVE and INSTRUMENTAL cover a wide range of additional, more specific relations. From what we know about SAE languages such as German, there appears to be no need to distinguish more than the two ‘classic’ thematic micro-roles, because solid evidence for the existence of further subdivisions is hard to come by, as direct primary markers are missing. There are, however, structural facts outside the realm of morphology which are suggestive of more implicit distinctions, cf. Chapter 5. Furthermore, language comparison reveals that where German morphologically subsumes these 2nd order micro-roles under 1st order micro-roles, other languages keep them strictly apart by formal means. For convenience, we start from a survey of contexts in which German mit is presently used. For this purpose, consider the list of examples in (A40). (A40)

German: a (simplified) paradigm of mit

No.

Example

Relation

(A40.1)

Agnes trinkt mit Werner Kaffee. ‘Agnes is drinking coffee together with Werner.’ Agnes unterhält sich mit Werner. ‘Agnes is chatting with Werner.’ Agnes geht mit ihrer Tochter spazieren. ‘Agnes is going for a walk with her daughter.’ Agnes geht mit ihrem Hund spazieren. ‘Agnes is walking her dog.’ Agnes geht mit dem Regenschirm nach draußen. ‘Agnes goes out with her umbrella.’ Agnes kommt mit roten Augen vom Friedhof zurück. ‘Agnes returns from the cemetery, red-eyed.’ Agnes trinkt immer Kaffee mit Milch. ‘Agnes always drinks coffee with milk.’

CO-OPERATIVE

(A40.2) (A40.3) (A40.4) (A40.5) (A40.6) (A40.7)

RECIPROCAL ACTIVE COMITATIVE/ HUMAN COMPANION PASSIVE COMITATIVE/ ANIMATE COMPANION CONFECTIVE/ INANIMATE COMPANION ORNATIVE/ TEMPORARY PROPERTY COMBINATION

42 Categories, definitions and terminology No.

Example

Relation

(A40.8)

Die Agnes mit den braunen Augen wohnt woanders. ‘The brown-eyed Agnes is living somewhere else.’ Die Agnes mit dem Porsche hat keinen Führerschein. ‘The Agnes with the Porsche has no driving licence.’ Agnes terrorisiert mit ihren Kindern die Nachbarschaft. ‘Agnes terrorises the neighbourhood with her children.’ Agnes schreibt den Brief mit der linken Hand. ‘Agnes is writing the letter with her left hand.’ Agnes kommt mit dem Bus vom Friedhof zurück. ‘Agens returns from the cemetery by bus.’ Agnes baut ein Haus mit Legosteinen. ‘Agnes is building a house using lego bricks.’ Agnes schlägt das Fenster mit dem Hammer ein. ‘Agnes smashes the window with the hammer.’

PART-WHOLE/

(A40.9) (A40.10) (A40.11) (A40.12) (A40.13) (A40.14)

PERMANENT PROPERTY POSSESSION HUMAN INSTRUMENT BODY PART INSTRUMENT MEANS OF TRANSPORTATION MATERIAL TOOL

There are altogether fourteen different contexts in which the preposition mit occurs. None of the examples is stylistically marked. As a matter of fact, this list could easily be extended. The following discussion will demonstrate that (A40) covers only part of the distribution profile of the relator.52 The selection and the order of relations in the above list, however, are by no means random. At this point, a short characterisation of what is meant by the labels used under the heading relation is in order because these relations will keep us busy in the subsequent paragraphs and beyond: – CO-OPERATIVE: Both participants are actively involved in the event, share the same macro-role and act together in order to achieve a common goal; their actions affect the same PATIENT (if present). – RECIPROCAL: Both participants are actively and passively involved in the event, share the same two macro-roles and act in such a way that the action of each of them affects the other. – ACTIVE COMITATIVE/HUMAN COMPANION: Both participants are actively involved in the event but one of them is assigned a higher degree of control, cf. CO-OPERATIVE for additional features. – PASSIVE COMITATIVE/ANIMATE COMPANION: The non-human animate COMPANION lacks control, whereas the human ACCOMPANEE exerts control; cf. CO-OPERATIVE for additional features. – CONFECTIVE/INANIMATE COMPANION: An animate ACCOMPANEE carries along or transports a concrete object; the inanimate COMPANION has a different macro-role than the animate ACCOMPANEE and it in no way affects any further participant; a typical instance of an in-

The prototype and its derivations 43

tra-participant relation (adnominal attribution); the ACCOMPANEE controls the COMPANION. ORNATIVE/TEMPORARY PROPERTY: An ephemeral bodily property of – an animate participant; this property has no effect on further participants, but normally escapes control by the ACCOMPANEE; a typical instance of an intra-participant relation (adnominal attribution). COMBINATION: Two prototypically inanimate entities occur in more or – less conventionalised (but dissolvable) combinations which may be asymmetrical (cf. PART-WHOLE below), and thus one of the entities is seen as super-ordinate to the other (= a weaker kind of control); a typical instance of an intra-participant relation (adnominal attribution). PART-WHOLE/PERMANENT PROPERTY: Two entities (not necessarily of – identical animacy) are permanently associated with each other; the one cannot normally be conceived without the other; typically, one entity is an integral part of the other (= a weaker kind of control); a typical instance of an intra-participant relation (adnominal attribution). POSSESSION: A human participant owns a prototypically inanimate – entity and thus exerts control over it; possession by means of Comitative markers can be of the attributive kind as well as the predicative. HUMAN (OR ANIMATE) INSTRUMENT: The ACCOMPANEE/USER causes – animate participants to act to achieve a certain goal; the subordinated animate participants are similar to causees in causative constructions;53 the ACCOMPANEE/USER exerts control over the ANIMATE INSTRUMENT and thus has a different micro-role. 54 BODY PART INSTRUMENT: The USER carries out an action employ– ing his/her body parts alone; usually the body part is controlled by the USER; volitionality and agentivity are reserved for the USER. MEANS OF TRANSPORTATION: The USER employs a vehicle to move – from one place to another; the vehicle may be under the control of the USER or of a third party (a chauffeur, for instance) and lacks autonomy. MATERIAL: The USER employs certain substances or other to achieve – a certain goal; these substances are controlled by the USER; they may be used to effect or affect an additional participant. TOOL: The user employs a concrete instrument in order to carry out a – given action; this instrument may be specifically designed for the purpose at hand; the tool is under the control of the user. This list does not exhaust the possibilities. In many cases, various degrees of animacy, asymmetrical control, permanence, etc. could be distinguished. Likewise, the distinction of concrete vs. abstract plays an important role, as

44 Categories, definitions and terminology we will have ample opportunity to notice in the remainder of this chapter. For the time being, we make do with the above inventory in order to have a fixed point of departure for our discussion. As there is no co-variation of form and function in (A40), mit is insensitive to any of the contextual features, be they nuclear predications or participants. It combines freely with animate and inanimate participants and is attested with verbs of a wide range of classes. Thus, we lack formal evidence supporting the claim that there are any categorical distinctions below the micro-role level. The ubiquity of mit even casts doubt on the validity of the distinction of COMITATIVE and INSTRUMENTAL on the established micro-role level, as the same primary relator is employed for both functions, and thus it becomes difficult to assume two distinct form-function pairings. These difficulties notwithstanding, it is feasible to demonstrate that the relations enumerated in (A40) are more than just descriptive context specifications. Two strategies are of help in this respect: – language-internal tests: substitution and reinforcement, – language comparison: contrasting distribution profiles. The former strategy requires that we accept secondary marking devices, i.e., we have to abandon, at least temporarily, the restriction to primary markers. The latter strategy presupposes that relations are conceived of as universal or language-independent concepts which are always there, although they may lack a distinct formal expression in a given language. Both strategies have their merits – and also the inevitable disadvantages. As pointed out in the discussion of example (A14.2) above, there is a rather straightforward way to distinguish Comitative uses of mit from the Instrumental ones. Only the former allow the reinforcement of the preposition by the adverb zusammen ‘together’. This reinforcement is absolutely fine with (A40.1)–(A40.4). In addition, it is also acceptable with (A40.7) and (A40.10). In the remaining eight contexts, however, zusammen is ruled out completely. With the exception of (A40.7), those contexts which are susceptible to reinforcement by zusammen involve two animate participants: Both the ACCOMPANEE and the COMPANION are either human or, in the case of the COMPANION, non-human animate. In (A40.7), none of the participants is animate. Therefore, it cannot be high animacy that triggers reinforcement. Zusammen requires the two participants to be of equal animacy, involvement and volitionality. If it is not for special poetic/comic effects, the adverb is excluded from combinations of animate and inanimate participants. Accordingly, zusammen employed in example (A40.4) pro-

The prototype and its derivations 45

motes the dog to the status of an egalitarian partner of Agnes: Both the dog and Agnes go for a walk together but neither of them is solely in control of the event. This promotion of the non-human animate gives an unusual touch to the example. The reading of the remaining sentences for which the use of zusammen has been classified as acceptable is also more or less altered. This alteration is most obvious with (A40.7) and (A40.10). In the case of the two liquids, coffee and milk, which share the property of being both inanimate substances, the micro-role of PATIENT and the absolute lack of volitionality are in different relations to each other depending on whether the simple preposition or the reinforced variant is used. Mit alone invites the interpretation that the milk is added to the coffee (most probably in such a way that the coffee-component of the mixture is still dominant), whereas zusammen mit sounds most plausible when the two liquids are served in and drunk from two separate containers. This is again an instance of promotion, as the milk is viewed as a substance independent of the coffee. As to (A40.10), the most striking changes are triggered by the addition of zusammen: Without the reinforcement, the sentence allows for several readings, one of which depicts Agnes as the instigator of or person responsible for the actions of her children. She may let them act or even make them act, but she is not personally part of their pranks. In this way, the children can be considered the HUMAN INSTRUMENTS with whom Agnes tries to exert influence on her neighbours. As soon as we add zusammen, however, the first reading is that Agnes herself is an active member of the kids’ gang. Thus, the term HUMAN INSTRUMENT would no longer be appropriate because (A40.10) turns into something similar to (A40.1)–(A40.2). The compatibility/incompatibility of the various instances of mit with zusammen thus divides (A40) into three classes: First, those cases which do not allow reinforcement by zusammen (= [A40.5–6], [A40.8–9], [A40.11–14]), second, those cases where reinforcement is acceptable but a major change of meaning is involved (= [A40.4], [A40.7], [A40.10]), and third, those cases where the semantic changes induced by zusammen are less radical (= [A40.1–3]). With a view to checking the sub-divisions of (A40) resulting from the zusammen-test, we briefly consider the compatibility of a variety of other reinforcements and substitutes of mit, namely mit Hilfe von55 ‘with the help of’, mittels ‘by (means of)’, durch ‘by, through’, samt ‘together/along with’, and nebst ‘along with’. The text frequency of these elements is far below the one of simple mit especially the latter two are relatively rare, with samt even having a certain archaic flair. Mit Hilfe von, mittels and nebst sound stilted and somewhat bookish, reeking of a technical or admin-

46 Categories, definitions and terminology istrative style. Only durch is stylistically unmarked. Table (A41) accounts for the general acceptability of these elements in the various contexts with acceptability ranging from + (= full) to 0 (= with reservations) to ¬ (= inadmissible); the question mark ? marks those cases where there remain further uncertainties as to their acceptability. Afterwards, we will see whether the substitutions/reinforcements yield readings different from the original sentences. For simplicity’s sake, we disregard the morphosyntactic re-adjustments triggered by the various substitutions. (A41) Sentence (A40.1) (A40.2) (A40.3) (A40.4) (A40.5) (A40.6) (A40.7) (A40.8) (A40.9) (A40.10) (A40.11) (A40.12) (A40.13) (A40.14)

Substituting mit samt ? ? ? 0 0 ¬ ? ¬ ¬ 0 ¬ ¬ ¬ ¬

nebst ? ? ? ? ¬ ¬ ? ¬ ¬ 0 ¬ ¬ ¬ ¬

mit Hilfe von 0 ? ? ? ? ¬ ¬ ¬ ¬ + ? ? ? ?

mittels ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¬ ? ? ? ? ?

durch ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¬ + ¬ ¬ ¬ ¬

The number of unambiguous plusses is surprisingly small: There are just two cases that straightforwardly allow substitution. Incidentally, both plusses are associated with the HUMAN INSTRUMENT relation: Mit Hilfe von and durch replace simple mit in (A40.10) without any change of meaning. Moreover, both substitutes of mit only allow the INSTRUMENTAL reading – in contradistinction to mit, they are unambiguous. With 41 out of 70 possible cases, the minuses clearly dominate. The remaining 27 cases cover the lower end of the acceptability scale. None of these cases is absolutely free of at least a little bit of awkwardness. Where admissible at all, samt and nebst invite readings according to which the participants are not necessarily in one and the same event.56 Likewise, reciprocity cannot be expressed by either of the two relators.57 Reinforcement by zusammen, samt and nebst is largely restricted to contexts in which the participants display identical animacy, volitionality and involvement. Mit Hilfe von is the only substitute which combines with animate and inanimate participants, whereas mittels is

The prototype and its derivations 47

normally excluded from combinations with animate participants. Wherever mit Hilfe von is used it conveys an INSTRUMENTAL reading. Thus, the COMPANION NP depending on mit Hilfe von automatically becomes a HUMAN/ 58 ANIMATE INSTRUMENT in (A40.1)–(A40.4) and (A40.10). Accidental accessoires like the umbrella in (A40.5) are promoted to INSTRUMENTS by mit Hilfe von.59 At this point one may ask whether mittels and durch are good for anything at all. As a matter of fact, they are perfectly acceptable in contexts not included in (A40), namely ABSTRACT INSTRUMENTS. In combinations with abstract instruments, mittels and durch are largely interchangeable.60 These facts suggest that there is indeed a divide between more Comitative-like and more Instrumental-like contexts which determine the use of secondary markers. None of the secondary markers has exactly the same distribution as the primary marker mit, i.e., partial synonymy is the best one can get. The above results are largely corroborated by an additional test, namely the coordination test. The preposition mit can be replaced by the coordinating conjunction und ‘and’ in exactly those contexts where both samt and nebst are at least potentially admissible. In (A40.1–3) and (A40.7), und is perfectly fine, whereas (A40.4) and (A40.7) are less acceptable: In the former case, the use of und would again promote the dog to the rank of controlling participant, and in the latter case, the instrumental reading is excluded under coordination. Wherever und substitutes for mit, there are two possible interpretations: Either the two participants are involved to the same degree in one and the same event or they participate each in a different event of the same type.61 CO-OPERATION and RECIPROCITY are no longer the only options, meaning that und and mit are partially synonymous at best.62 There is, of course, more to the relations which connect mit to its potential substitutes. First of all, the case of mit vs. und reflects the basic distinction of subordination vs. coordination. When two participants of equal status are combined by und, the order in which the participants occur can usually be inverted without changing the meaning. Furthermore, coordinated participants are assigned the same semantic role and syntactic relation. If the latter is the subject, then the presence of und obligatorily goes along with plural agreement on the finite verb. If one wants to postulate an inter-participant relation between the two coordinated participants, EQUALITY is the only reasonable suggestion, as COMITATIVE is ruled out by the fact that und does not automatically imply involvement in the same event. This is completely different in the case of mit. This preposition marks an asymmetrical relation between two participants, one of whom is subordi-

48 Categories, definitions and terminology nate to the other within the spatio-temporal frame of the same situation. Mit always implies that the participants are involved in the same event – and it at least invites the reading that one of the participants has more control or dominates the other. The difference between coordination and subordination is nicely reflected by German morphosyntax. A kind of iconicity applies, in the sense that the use of mit demotes the subordinate participant, which is assigned the dative case, and thus cannot determine number agreement on the verb, which remains the prerogative of the NP representing the superordinate participant. Mit-phrases are also positionally relatively free, i.e., they may not always be adjacent to the superordinate NP, whereas NPs coordinated by und are always immediate syntactic neighbours. The potential for non-adjacency of the NPs representing ACCOMPANEE and COMPANION, respectively, is indicative of the peripheral status of the participants introduced by mit. The above discussion yields an intricate picture of paradigmatic networks dominated by partial functional equivalence. On superficial inspection, this complex agglomeration of relations may be misunderstood as counter-evidence to our claim that there is a system behind Comitatives. However, the system can only become visible if we assume different levels on which primary and secondary relators operate. As a matter of fact, mit and its various substitutes should not be lumped together: The preposition mit is a morpheme of the highest order, as it cannot be further subsumed under the distribution profile of a higher-ranking means of expression. This is exactly what happens to the mit-substitutes, as they all can be replaced by mit. This also partially explains why these substitutes are considered secondary, as opposed to the primary marker mit. Thus, the relations in (A40) can be counted out as categories on a par with the one circumscribed by the distribution profile of mit. They simply do not belong to the same level: They are readings associated with mit, but are not categorical competitors of this preposition. Nevertheless, we still need to know whether the above facts are just German idiosyncrasies or reflect more general, languageindependent principles. In Chapter 2 above, various languages were mentioned which require more than one relator to cover the distribution profile of German mit, i.e., there is no single general translation equivalent of mit, but several contextspecific translation equivalents. The contexts given in (A40) already provide a suitable (though far from comprehensive) list of potential cut-off points for the distribution of such context-specific translation equivalents. First of all, languages differ as to whether they allow the same relators to be used in NPinternal and NP-external environments. German mit is not sensitive to the

The prototype and its derivations 49

distinction of adnominal and adverbial usage. Some languages share this characteristic with German: Among these we find Estonian whose inflectional comitative marker -ga displays almost the same distribution profile as German mit. Accordingly, the suffix -ga occurs in ACCOMPANIMENT situations of various readings as well as in proper INSTRUMENTAL situations, cf. (A25). Moreover, it is also used in adnominal attribution, cf. (A42). (A42) Estonian (Hasselblatt 1992: 72) (A42.1) ORNATIVE habeme-ga mees beard-X/Y man ‘bearded man’ (A42.2) COMBINATION sõin jääti-st moosi-ga ice_cream-ELA jam-X/Y eat:PAST:1SG ‘I ate ice cream (topped) with jam.’ (A42.3) MATERIAL sõin moosi-ga jääti-st jam-X/Y ice_cream-ELA eat:PAST:1SG ‘I ate ice cream made of jam.’ The position of the NP marked for comitative varies from post-head in (A42.2) to pre-head in (A42.3) with the latter being the regular position for non-sentential attributes. This syntactic difference goes along with a difference in meaning, namely COMBINATION vs. MATERIAL (or INGREDIENT/ 63 PART-WHOLE). However, even for languages which employ one and the same relator for all of the predicative cases in (A40), it is by no means compulsory to extend the functional domain of this relator to the adnominal cases. Consider example (A43) from Turkish. (A43) Turkish (Ersen-Rasch 1980: 79) bu mayo-lu kız-ı gördün mü this swim_suit-PROP girl-ACC see:PAST:2SG INTERR ‘Have you seen this girl with (= in/wearing) the swimsuit?’ Turkish employs the postposition ile and its bound allomorphs -le/-la for almost the entire range of contexts listed in (A40), cf. (A24) and footnotes above. In adnominal attribution, however, a different bound morpheme is used, namely -lu/-li/-lı/-lü (which is sometimes called comitative in the pertinent literature, although we prefer the label proprietive). Disregarding possible etymological ties between ile/-le/-la and -lu/-li/-lı/-lü, we are deal-

50 Categories, definitions and terminology ing with two synchronically distinct morphemes. One is specialised for adnominal contexts: The other one occurs only in predicative ones. In languages which keep the grammatical expressions for Comitative and Instrumental formally distinct, we encounter a variety of patterns. These patterns demonstrate that it is not only the presumably syntactically motivated distinction of adnominal and adverbial categories that contribute to the internal differentiation of the paradigm of contexts presented in (A40). Paraguayan Guaraní is a case in point: This Amerindian language makes use of three relators whose functional domains do not overlap, cf. (A44). (A44) Guaraní (A44.1) INSTRUMENTAL (Gauto Bejarano 1990: 116) aikytƭ pe so’o che kysé-pe this meat my knife-Y 1SG:cut ‘I cut this meat with my knife.’ (A44.2) ACCOMPANIMENT (Gauto Bejarano 1990: 115) Maria oho tupao-pe Sabina ndive X Maria 3SG:go church-Y Sabina ‘Maria is going to church with Sabina.’ (A44.3) COMBINATION (Gauto Bejarano 1990: 114) ha’u kamby mbuja-pe reheve ha kambykyrakue bread-Y COMBI and butter consume:1SG milk ‘I consume milk together with bread and butter.’ (A44.4) MEANS OF TRANSPORTATION (Gauto Bejarano 1990: 111) ha’e oho mba’eyrú-pe he:go car-Y ‘He is going by car.’ (A44.5) BODY PART (Gauto Bejarano 1990: 111) ha’e oho ipý-pe he:go foot-Y ‘He is walking (on foot).’ Different postpositions are used as markers of the INSTRUMENTAL, COMITATIVE and ASSOCIATIVE. The relator -pe is more closely bound to the noun without, however, becoming a full-blown case affix. This -pe is a multipurpose morpheme with a variety of spatial functions (as in [A44.2]). It occasionally also marks the PATIENT of certain verbs (as in [A44.3]). Moreover, -pe covers big part of the range of instrumentality, as it is used for TOOLS proper (as in [A44.1]), MEANS OF TRANSPORTATION (as in [A44.4]) and BODY PART INSTRUMENTS (as in [A44.5]). In contradistinction to -pe, the

The prototype and its derivations 51

other two relators are free postpositions. The use of ndive is explicitly restricted to combinations of human participants (A44.2), whereas reheve is excluded from this context. According to Gauto Bejarano (1990: 110), reheve is only admissible in combinations of two inanimate entities (A44.3). Thus, ndive is a true Comitative marker and reheve is reminiscent of what is elsewhere labelled an Associative marker. However, nothing is said about non-human animate beings in ACCOMPANIMENT situations. If we take Gauto Bejarano’s rules literally, there is a gap64, because the distributions of ndive and reheve do not cover the whole range of possible participant combinations. In the absence of any indication as to the treatment of non-human animate COMPANIONS in our source, we assume that this is perhaps contested territory where ndive and reheve compete and -pe is excluded, as it presupposes asymmetrical control of the participants along with different animacy. In standard Mongolian, two morphological cases – comitative (encoded by -tVj) and instrumental (encoded by -(g)V:r65) – are distinguished. Only the former one is admitted in adnominal attribution, cf. (A45). (A45) Mongolian (Vietze 1988: 65–66) (A45.1) ACCOMPANIMENT bagš suragþ-taj irlee teacher pupil-X come ‘The teacher has come with the pupil.’ (A45.2) Predicative possession ene ojuutan olon nom-toj this student many book-X ‘This student has a lot of books.’ (A45.3) Attributive possession exner-tej xyn woman-X man ‘married man (lit. man with a woman)’ (A45.4) Container (COMBINATION/PART-WHOLE) tavag-taj xool plate-X something_to_eat ‘a plate with something to eat’ (A45.5) TOOL xarandaa-gaar biþix pencil-Y write ‘to write with a pencil’

52 Categories, definitions and terminology (A45.6) MEANS OF TRANSPORTATION mašin-aar javax car-Y drive ‘to go by car’ (A45.7) MATERIAL mod-oor širee xijx wood-Y table make ‘to make a table out of wood’ The comitative marker covers the functions (A40.1)–(A40.9), which the container-content relation must be added to, as a special variety of the PARTWHOLE or COMBINATION relation, cf. (A45.4). The comitative marker is also used for possession in both its forms, namely attributively and predicatively, cf. (A45.2–3). The remaining functions fall within the distribution domain of the instrumental marker. In Kalmyk, a distant relative of Mongolian, there is even a tripartition: In addition to the morphological instrumental (ending in -ar/-är)66 and the comitative (marked by -la/-lä), Kalmyk makes use of a third case most often called associative (represented by -ta/ -tä), cf. (A46). (A46) Kalmyk (A46.1) Instrumental: MEANS OF TRANSPORTATION (Benzing 1985: 63) tramvay-ar yovx tram-Y go ‘go by tram’ (A46.2) Instrumental: HUMAN INSTRUMENT (Benzing 1985: 63) ax-ar belg ilgäz brother-Y present sent:PAST ‘He sent a present via his brother.’ (A46.3) Comitative (Benzing 1985: 68) Sanž Bembä-lä balhs oržana Sanži Bemba-X town enter:PROG:PRES ‘Sanži and Bemba are going into town.’ (A46.4) Associative: Attributive possession (Benzing 1985: 152) mör-tä kün ger-tä kün man house-ASSOC man horse-ASSOC ‘rider (lit. man with a horse)’ ‘married man (lit. man with a house)’ (A46.5) Associative: Predicative possession (Benzing 1985: 153) dörvn xan arvn xoyr kötþ-tä four Khan twelve lackey-ASSOC

The prototype and its derivations 53

huþn zurhan alvt-ta thirty-six slave-ASSOC ‘The four Khans have twelve lackeys, 36 slaves.’ (A46.6) Associative: Co-operative (Benzing 1985: 153) Boris Sanž-ta škol-ïn xurg kecxäv meeting organise:PERF Boris Sanži-ASSOC school-GEN ‘Boris and Sanži have organised a school-meeting together.’ The comitative67 proper and the instrumental are banned from adnominal contexts.68 The comitative seems to be restricted to combinations of two human participants. The instrumental is also used for HUMAN INSTRUMENTS (cf. [A46.2]). Adnominally, we only encounter examples of the associative. In addition, the same case is used in predicative contexts, be it for predicative possession as in (A46.5) or for paradigm cases of cooperation as in (A46.6). Thus, the associative competes with the comitative in contexts where two human participants act together as ACCOMPANEE and COMPANION. It is not clear from the descriptions at hand whether the two competing cases are interchangeable (Bläsing 2003: 236). The cases discussed so far all reflect a neat separation of the contexts (A40.10–14) from the remainder of the above paradigm of contexts. There is a tendency to keep those contexts which are closer to the prototype of the Comitative apart from those which associate more with the Instrumental prototype. In accordance with the conventions exposed in footnote 8, we use X plus a numerical index in the transmorphemisations for those categories which are closer to comitativity and Y plus a numerical index for those which are closer to instrumentality. The numerical index reflects the number of formal distinctions, i.e., we have not allotted fixed indices to certain categories. However, the division into a Comitative and an Instrumental subdomain is not always of the same kind. Both varieties of modern Armenian are indicative of a different pattern. Western and Eastern Armenian have an inflectional instrumental -ov which also serves the purpose of encoding the Instrumental in all its shapes. Furthermore, there is a postposition hed/het which is used for the Comitative, cf. (A47). (A47) (Western) Armenian (Sakayan 2000: 138–139) (A47.1) Instrumental: TOOL hac6 tanag-ov g6 gdren knife-Y PRES cut:3PL bread:DEF ‘They cut bread with a knife.’

54 Categories, definitions and terminology (A47.2) Instrumental: MEANS OF TRANSPORTATION amen mart þi sirer otanav-ov džamportel travel:INF each man NEG love:PTCPL plane-Y ‘Not everybody likes to travel by plane.’ (A47.3) Instrumental: COMBINATION surdž6 gat-ov g6 xmem milk-Y PRES drink:1SG coffee:DEF ‘I drink coffee with milk.’ (A47.4) Postposition: ACCOMPANIMENT pžišk-i-n hed e-kar AUG-come:PTCPL doctor-ACC-DEF X ‘Did you come with the doctor?’ (A47.5) Instrumental: HUMAN INSTRUMENT miajn pžišk-ov þes a÷eknar NEG:2.SG heal:PTCPL only doctor-Y ‘You cannot become well just by (seeing) a doctor.’ (A47.6) Postposition: COMBINATION þuk-i-n hed džermag kini g6 xmen white wine PRES drink:3.PL fish-ACC-DEF Y ‘With fish they drink white wine.’ The fact that the instrumental is used with TOOLS and MEANS OF TRANSPORTATION (A47.1–2) is hardly a surprise. What makes Armenian stand out is the use of the instrumental in adnominal contexts, as in (A47.3), where the combination of two liquid substances is mentioned. However, the instrumental can also be used in PART-WHOLE relations and also when (permanent) properties are ascribed to human participants, as in (A48) taken from Eastern Armenian. (A48) (Eastern) Armenian (Kozintseva 1995: 10) (A48.1) PART-WHOLE ms-ov karkandak meat-Y pie ‘meat pie’ (A48.2) ORNATIVE snorhk’-ov kin talent-Y woman ‘talented woman’ The instrumental is also the case to be used with HUMAN INSTRUMENTS (A47.5). The postposition hed/het, however, is only used in contexts where

The protype and its derivations 55

two relatively autonomous entities/participants are involved. This condition applies to the usual ACCOMPANIMENT situation with a human ACCOMPANEE and a human COMPANION in (A47.4). Somewhat unexpectedly, hed/het occurs also as the relator for the combination of inanimate substances. However, in contradistinction to coffee and milk in (A47.3), fish and white wine, as mentioned in (A47.6), are separate items on the menu, in a manner of speaking. They form a much looser combination than the two liquids which mix to yield a special kind of hot drink. Of the five languages discussed so far, none display identical distribution of their respective relators. Nevertheless, there is convergence among these languages as to which major distinctions are made. Table (A49) surveys their distribution profiles (to a small extent also based on conjecture where concrete evidence was not available). Different nuances of grey shading and white are used to mark the distribution of individual relators. (A49)

Distribution patterns compared I

Relation (A40.1) (A40.2) (A40.3) (A40.4) (A40.5) (A40.6) (A40.7) (A40.8) (A40.9) (A40.10) (A40.11) (A40.12) (A40.13) (A40.14)

Turkish ile ile ile ile ile -li ile -li ile -li ile -li ile -li ile ile ile ile ile

Guaraní Mongolian ndive -tVj ndive -tVj ndive reheve -tVj reheve -tVj reheve -tVj reheve -tVj reheve -tVj reheve -tVj reheve -tVj -pe -(g)V:r -pe -(g)V:r -pe -(g)V:r -pe -(g)V:r -pe -(g)V:r

Kalmyk -lä -tä -lä -tä -tä -tä -tä -tä -tä -tä -tä -är -är -är -är -är

Armenian hed hed hed hed -ov hed -ov hed -ov hed -ov -ov -ov -ov -ov -ov -ov -ov

Irrespective of some minor problems, there is evidence for a potential dividing line running between (A40.9) and (A40.10), as four of the languages either switch from one relator to another (Guaraní, Mongolian, Kalmyk) or cease using one of two options (Turkish). This virtual line coincides with the distinction of instrumentality and co-presence/association: From (A40.10) to (A40.14), one of the participants functions a means with which to carry out an action. In all other contexts, the participants are either in the same space or are closely associated with each other without serving as a means for anything.

56 Categories, definitions and terminology In numerous languages, we indeed encounter a strong tendency to distinguish by formal means more than one category for the contexts (A40.1)– (A40.9), whereas the remainder of the context paradigm appears to be less susceptible to further differentiation (but cf. below) (Stolz 1997b). Animacy is one of the determining factors – though by far not the only one. In the Tungusic language Even, four comitatives are attested (three of which are not included in the native language teaching material of the 1950s [Benzing 1955: 64]).69 The paradigm in (A50) reflects not only the basic distinction of animate vs. inanimate but also three different degrees of animacy.70 It is important to note that the four markers are distributed on an overlap basis and not according to categorial discreteness: Their respective functional domain intersects with the ones of their paradigmatic neighbours, meaning that there are no markers which are exclusively dedicated to the symbolisation of one and only one animacy degree. The distinctions are therefore of an indirect nature. (A50)

Even animacy hierarchy (Stolz 1997b: 524)

Category

animate

human Comitative IV pronouns proper nouns Comitative III proper nouns Comitative II Comitative I

inanimate animal

common nouns common nouns common nouns common nouns

common nouns common nouns common nouns common nouns

All comitatives can be used with human common nouns. However, other noun types are subject to restrictions as to which comitative marker can or must be used. The further the functional domain of a comitative marker extends to the left in (A50), the more its extension to the right is limited, meaning that higher and highest degrees of animacy disfavour applicability of the same morpheme to NPs of low and lowest animacy. We repeat that animacy is not the sole factor that determines how many distinct Comitatives a language employs. There is in fact a sizeable list of other factors, be they related to animacy or not, which have a say in the internal differentiation of comitatives. For the sake of brevity, we make do with discussing only a small selection of these factors. The description is based on Stolz (1997b). –

Number For the Papuan language Tairora, two comitative morphemes are reported whose distribution is (largely) number-dependent:71 -nti is used exclusively in the singular, whereas -hampata is the only

The prototype and its derivations 57

option for COMPANIONS in the plural (although the same morpheme may also be used in the singular), cf. (A51).72 (A51) Tairora (Vincent 1973: 54) (A51.1) Singular bainti-bano bairi-nti-ro ani-ro dog-X.SG-3SG come-3SG man-SUBJ ‘The man came with the dog.’ (A51.2) Plural bi-ka-hampata ni those-PL-X.PL walk.1SG ‘I went about with them.’ – Alienability In the areal vicinity of Tairora, the Austronesian language Saliba is spoken. Saliba also displays a bipartite system of comitative markers. However, this time the underlying motivation for the formal distinction is not based on number, but on possession.73 For inalienably possessed items as COMPANIONS, the simple preposition ma is used, whereas alienability of possessees requires the more complex preposition maya, cf. (A52).74 The latter most probably goes back to a combination of two morphemes, namely, the preposition ma and an erstwhile alienability marker -ya. (A52) Saliba (Mosel 1994: 16) (A52.1) Inalienable75 ma natu-na sine se lau koya-i child-3SG woman they go garden-LOC X.INAL ‘The woman went with her child to the garden.’ (A52.2) Alienable tau-bada maya-na nigwa knife man-old X.AL-3SG ‘the old man with the knife’ – Static/dynamic In the Uto-Aztecan language Luiseño, we encounter three comitative morphemes76, namely, the neutral/general -man in addition to -éš, which is restricted to dynamic situations (= ACCOMPANEE and COMPANION move) and -to, which occurs only in static situations (= ACCOMPANEE and COMPANION are immobile) (Kroeber and Grace 1960: 98), cf. (A53). The neutral/general morpheme is the

58 Categories, definitions and terminology

(A53) (A53.1)

(A53.2)

(A53.3)

–

(A54) (A54.1)

(A54.2)

only one that occurs both on common nouns and pronouns. The two semantically more specialised morphemes are restricted to combinations with pronominal prefixes. Luiseño General/neutral (Kroeber and Grace 1960: 183) poi si’wik po-kwe:ro po-nki-man 3SG-hide 3SG-ear-X.GEN it skin.PAST po-yu-man po’-aw-man 3SG-horn-X.GEN 3SG-head-X.GEN ‘It skinned its hide along with its ears, head and horns.’ Dynamic (Kroeber and Grace 1960: 216) mamanok no-éš o-hatíax-pi 2SG-go_off-INCEP wanting 1SG-X.DYN ‘I want that you go off together with me.’ Static (Kroeber and Grace 1960: 207) po-to pe’a’wiš po-šS aki 3SG-wife 3SG-X.STAT and live:PAST ‘And his wife lived with him.’ Control In yet another Papuan language – Awiyaana – two comitatives are distinguished formally:77 The one marked by -kwada is labelled active comitative and the other marked by -yen is the so-called passive comitative, cf. (A54). Awiyaana (McKaughan and Marks 1973: 186) Active poi-kwada kwaasin kwet-i man:ABS come-3SG pig-X.ACT ‘The pig is coming along with the man.’ Passive poi-yen kwaasin kwet-i man:ABS come-3SG pig-X.PASS ‘The pig is brought along by a man.’ According to the description by McKaughan and Marks (1973), the active-comitative suffix is said to attach to the COMPANION NP, whereas the passive-comitative marker is hosted by the NP representing the ACCOMPANEE. In both cases, the NP that is not marked for comitative remains in the uninflected stem-form. Incidentally, however, this stem-form is identical with the absolutive, i.e., with the case required for the central participant of intransi-

The prototype and its derivations 59

tive verbs (such as the movement verb kwet- ‘come’). Thus, the nuclear predication in (A54.1–2) is essentially the same, namely, ‘a man is coming’, meaning that in both cases, kwaasin ‘man’ is the ACCOMPANEE. What distinguishes the two sentences, however, is the degree of autonomy of the COMPANION: In (A54.1), the pig has a higher degree of control than in (A54.2). Nevertheless, the highest degree of control is in both cases exerted by the human ACCOMPANEE. –

ACTOR/UNDERGOER

Closely related to the previous distinction is the one with which macro-role assignment is achieved via comitative markers. According to Drossard (1991: 449), the Philippinian language Tagalog employs different focus affixes78 on the verb to distinguish COMPANION-orientation (= UNDERGOER-orientation) from ACCOMPANEE-orientation (= ACTOR-orientation). The affixes are kfor the former and -akipag for the latter.79 If both participants are considered absolutely equal, a third focus marker is employed and coordination applies, cf. (A55) where underlining marks the NP(s) in focus. (A55) Tagalog (Drossard 1984: 85) (A55.1) COMPANION oriented k-in-a-usap ng binata’ [ang dalaga]focus X.UNDER-PERF-LIG-talk LOC boy DEF girl ‘The boy talked with/to the girl.’ (A55.2) Actor oriented dalaga n-akipag-usap [ang binata’]focus sa PERF-X.ACTOR-talk DEF boy LOC girl ‘The boy talked with/to the girl.’ (A55.3) Coordination: actor focus on both participants n-ag-usap [ang binata’ at dalaga]focus PERF-ACTOR-talk DEF boy and girl ‘The boy and the girl talked to each other.’ –

Permanent vs. temporary – motivated vs. incidental The last distinction to be reviewed here does not lend itself to an easy classification. The terms foreground/background put forward by Stolz (1997b: 522–523) probably do not capture the relevant functions adequately. Chukot – like other languages of the same macrophylum – distinguishes two inflectional comitatives80, both

60 Categories, definitions and terminology of which are marked morphologically by circumfixes, namely, comitative I ge-…-(t)e and comitative II ga-…-ma, cf. (A56).81 (A56) Chukot (Kämpfe and Volodin 1995: 53–54) (A56.1) Comitative I a’aþek ge-milger-e boy X1-rifle-X1 ‘the boy with the gun’ (A56.2) Comitative II n6 tosq6 cat-gË ga-melgar-ma a’aþek X2-rifle-X2 boy exit-PERF ‘The boy ran outside with the gun.’ The two comitatives have slightly different functions. According to Worth (1963: 360–361), the difference is one of status: Comitative I applies to cases where both participants are viewed as being of similar status and their cooccurrence is not incidental. Comitative II (called sub-comitative by Worth), however, is used if one of the two participants is clearly sub-ordinated and their co-presence is due to chance. This interpretation is in line with Skorik (1961: 170), who associates comitative I with coordination and comitative II with subordination (Kämpfe and Volodin 1995: 54). It is somewhat difficult to accept (A56.1) – a clear case of an ORNATIVE ([A56.2] being an example of a confective) – as an instance of coordination.82 To our mind, the difference between the two Chukot examples is one of permanent vs. temporary or motivated vs. incidental association: The rifle in (A56.1) is the attribute that helps identify the boy: In (A56.2), however, the rifle is something that incidentally happens to be there in the same place at the same time as the boy. The parameters which determine the further sub-categorisation of Comitatives are less disparate as the might appear on superficial inspection. ACTOR/UNDERGOER and control are of course closely related to each other. In the remaining cases, it could be argued that there is a basic distinction of prototypical properties of human participants vs. non-human participants behind the oppositions. The number distinction keeps individuals apart from groups/collectives, and the static/dynamic pair might be seen as a reflection of the higher dynamicity of animates as opposed to static inanimate entities. It is more difficult to integrate the permanent/temporary distinction in this general schema, which is remotely connected to animacy and/or empathy criteria.

The prototype and its derivations 61

In all of the above sketches, diversification only applies to comitativelike categories. However, this does by no means imply that instrumentallike categories are necessarily homogenous. Quite to the contrary, we encounter similar phenomena in the realm of instrumentality as well. Differentiation may apply to both extremes of the paradigm (A40) at the same time, meaning that the existence of several comitatives does in no way preclude the coexistence of several instrumentals and vice versa. Consider the case of Finnish first. In this Uralic language, there is the regular number-insensitive inflectional comitative -ine-, which is obligatory followed by a possessor suffix referring to the ACCOMPANEE. The postposition kanssa, which governs the genitive, competes with the morphological comitative and seems to be gaining the upper hand as it has been taking over many functions formerly fulfilled by the inflectional case (Nau 1995: 133).83 The functional domains of the comitative suffix and the postposition overlap considerably without being absolutely identical, cf. (A57). (A57) Finnish (A57.1) ACCOMPANIMENT (Sulkala and Karjalainen 1992: 225) hän tuli koir-ine-en / koira-n kanssa dog-X1-POSS.3 / dog-GEN X2 he come:IMPERF ‘He came with his dog.’ (A57.2) CONFECTIVE (Sulkala and Karjalainen 1992: 226) sie-llä oli mies valkois-en paida-n kanssa there-Y1 be:IMPERF man white-GEN shirt-GEN X2 ‘There was a man there who had a white shirt with him.’ (A57.3) ORNATIVE/PART-WHOLE (Karlsson 1978: 133) Rauma on mukava kaupunki vanho-ine Rauma is nice town old-X1 talo-ine-en ja kape-ine katu-ine-en narrow-X1 street-X1-POSS.3 house-X1-POSS.3 and ‘Rauma is a nice town with its old houses and narrow streets.’ What distinguishes the inflectional comitative from the postpositional construction is the fact that the former usually invites a reading according to which the two participants are more intimately associated to one another. Kanssa, on the other hand, tends to be used in contexts where the association between the two participants is of a looser kind. In general, kanssa is not always the primary option when it comes to attribution: The postposition may be used to mark CONFECTIVE relations as in (A57.2). For the more intimate relation of ORNATIVE, there is a variety of other expressions to

62 Categories, definitions and terminology choose from (spatial cases, compound adjectives, instructive case84, etc. [Sulkala and Karjalainen 1992: 226]). One of the possibilities for adnominal attribution is the inflectional comitative. We interpret the obligatory use of the possessor inflection alongside the comitative case on the COMPANION noun as an instance of iconicity (or perhaps only indexicality), which proves that ACCOMPANEE and COMPANION form a close unit. However, the use of the inflectional comitative for the purpose of nominal attribution has been the subject of a fierce debate among Finnish specialists – one party being of the opinion that what looks like an attribute on the surface is only an adverbial in disguise.85 This dispute is of course well-founded, as it seems that many of the NPs marked for comitative are diffuse as to their exact syntactic status. Nevertheless, there can be no doubt that the comitative also functions as a marker of adnominal attributes (Nau 1995: 133). Instrumentality is primarily associated with the inflectional adessive -lla/-llä – a multi-purpose case with a wide range of spatial functions (largely corresponding to English at and on) and other functions. The adessive is used for tools and is one of the possible options in other instrumental contexts except MATERIAL.86 It is also employed for predicative possession, provided the possessor is animate.87 In addition, there is the residual inflectional instructive ending in -in for both numbers, whose major function consists in marking BODY PART instruments88, cf. (A58). (A58) Finnish (A58.1) BODY PART INSTRUMENT: instructive (Sulkala and Karjalainen 1992: 224) te-i-n sen palja-in käs-in hand-Y2 make-IMPERF-1SG it:ACC bare-Y2 ‘I did it with my bare hands.’ (A58.2) TOOL: adessive (Sulkala and Karjalainen 1992: 224) te-i-n leivä-n konee-lla bread-ACC machine-Y1 make-IMPERF-1SG ‘I made the bread with a machine.’ (A58.3) Possession: adessive (Sulkala and Karjalainen 1992: 227) tytö-llä on kissa girl-Y1 is cat ‘The girl has a cat.’ (A58.4) MEANS OF TRANSPORTATION (Nau 1995: 134) matkusta-n Suom-een laiva-lla Finland-ILL ship-Y1 travel-1SG ‘I travel to Finland by ship.’

The prototype and its derivations 63

Admittedly, the current Finnish system is in transition, i.e., the coexistence of several marking strategies is partly occasioned by the reluctance of normative grammarians to cancel the inflectional comitative and the instructive from the inventory of regular case forms of the Finnish nouns. Nevertheless, the competition between, on the one hand, -ine- and kanssa and -lla/-llä and -in on the other is a feature of the written register of Finnish. On the Comitative pole, the distinctive criterion is the intimacy of the relation (translatable into permanent vs. temporary, etc.). In the realm of instrumentality, BODY PART INSTRUMENTS, behave differently (at least optionally) from other kinds of instruments. HUMAN INSTRUMENTS too tend to employ a distinct encoding (which is still based on the adessive). Thus, animacy is directly and indirectly (= BODY PART INSTRUMENTS) involved in this distinction. In Dani, there are three ways to express instrumentality89, namely, the multi-purpose post-clitic -(o)pa (also used for a variety of spatial categories corresponding to English on, at), the post-clitic -(n)en (likewise used for spatial concepts which are associated with source) and zero-marking, cf. (A59). (A59) Dani (A59.1) TOOL (Bromley 1981: 81) Henayppv posije-pa lipiloko wathe Henaipu axe-Y1 lacerating hit:3SG ‘He lacerated Henaipu with an axe.’ (A59.2) BODY PART INSTRUMENT (Bromley 1981: 86) an n-ekki-nen wam POSS.1SG-hand-Y2 pig I watoko ithiki killing net:1SG:3SG:DISTANT.PAST ‘I netted her with pigs I killed with my own hands.’ (A59.3) Weapon (Bromley 1981: 109) tok pisinit-neeikhe arrow shoot-do:PAST:SG/1SG:GOAL ‘He shot at me with an arrow.’ BODY PART INSTRUMENTS (A59.2) receive a marking distinct from the one used for TOOLS (and MATERIAL) (A59.1). The class of BODY PART INSTRUMENTS extends to a number of other items one normally would have difficulty associating with the notion of body part. Helep ‘stone’ is a case in point. Bromley (1981: 167–168) points out that stones are assigned a certain degree of animacy among the Dani speech-community and speculates that this might be the reason why stones are treated as BODY PART INSTRUMENTS. Unfortunately, he does not provide examples for HUMAN INSTRUMENTS be-

64 Categories, definitions and terminology cause his argumentation invites the interpretation that ANIMATE/HUMAN INalso require the encoding with -(n)en. With weapons as instruments (A59.3), a different construction is used. The syntactic status of the NP representing the weapon is doubtful: it is either a direct object of a verb with an incorporated pronominal GOAL or “the weapon, if mentioned, may be interpreted to occupy an instrument role” (Bromley 1981: 109). In French, BODY PART INSTRUMENTS and MEANS OF TRANSPORTATION are areas where the usual preposition avec ‘with’ is no longer the only possible option. The rules for choosing the correct relator are rather intricate – the more so if one tries to account also for the variation of marking strategies in the realm of adnominal attribution. For our interpretation we draw mainly from Cadiot (1990, 1991, 1993), who has dedicated several papers to this and related issues. As a matter of fact, avec has a large number of competitors, namely, à ‘on, at’, de ‘of, from’, en ‘in’, par ‘by, through’, etc., which cannot simply be considered secondary markers because their interaction is not an optional matter of style but is rule-governed, cf. the long but still only selective list of examples in (A60). (A60) French (A60.1) ACCOMPANIMENT (Les noces barbares 53) c-est bien ta fille qui vient le this-is well your daughter REL come DET dimanche avec ta femme Sunday X/Y your woman ‘It is your daughter who comes with your wife on Sundays, isn’t she?’ (A60.2) TOOL I (Les noces barbares 36) Ludo se mit à gratter scratch:INF Ludo REFL.3 put:PAST Y4 au mur avec un vieux clou X/Y a old nail Y4:DET wall ‘Ludo began to scratch on the wall with an old nail.’ (A60.3) TOOL II (Les noces barbares 67) il égalisa l-écume DET-foam he level_out:PAST à la petite cuiller small:F spoon Y4 DET:F ‘He distributed the foam with the coffee-spoon.’ (A60.4) MEANS OF TRANSPORTATION I (Les noces barbares 78) souvent l-après-midi Madame Blanchard DET-afternoon Mrs Blanchard often STRUMENTS

The prototype and its derivations 65

se

rendait par car aux Buissonnets go:IMPERF Y2 bus Y4:DET.PL Buissonnets ‘Often, in the afternoon, Madame Blanchard went by bus to Les Buissonnets.’ (A60.5) MEANS OF TRANSPORTATION II (Les noces barbares 101) c-était une promenade en auto outing Y3 car this-be:IMPERF a ‘This was an outing by car.’ (A60.6) MEANS OF TRANSPORTATION III (Klein and Kleineidam 1983: 176) Nicolas est arrivé à cheval Nicolas is arrived Y4 horse ‘Nicolas has arrived on horse-back.’ (A60.7) BODY PART INSTRUMENT I (Les noces barbares 40) de la tête il envoyait de petits coups send:IMPERF Y5 little hit Y5 DET.F head he boudeurs contre la cloison wall sulky against DET.F ‘He gave little sulky punches with his head against the wall.’ (A60.8) BODY PART INSTRUMENT II (Les noces barbares 83) elle soupesait la gourmette à deux doigts DET.F chain Y4 two finger she weigh:IMPERF ‘She weighed the chain with two fingers.’ (A60.9) BODY PART INSTRUMENT III (Klein and Kleineidam 1983: 175) mange avec ta fourchette, pas avec les doigts NEG X/Y DET.PL fingers eat X/Y your fork ‘Eat with your fork, not with your fingers!’ (A60.10) ORNATIVE (permanent) (Les noce barbares 57) l-homme aux huit doigts DET-man Y4:DET:PL eight fingers ‘the man with (the) eight fingers’ (A60.11) ORNATIVE (temporary) (Les noces barbares 55) une longue table avec une nappe blanche table-cloth white.F a long.F table X/Y a ‘a long table with a white table-cloth’ (A60.12) CONFECTIVE (Micmac 105) j-en revins avec un torchon a cloth I-thence return:PAST X/Y ‘I came back from there with a cloth.’ REFL.3

66 Categories, definitions and terminology The distribution of the various prepositions across functions is not random. It is determined by several parameters, which in turn interact with each other and thus yield a complex network of distinctions. Irrespective of the fact that avec may be used in most of the contexts listed in (A40), it is by no means the default solution. The major criterion for choosing avec is the referentiality and relative autonomy of the participant in question. It is regularly employed if the participant is a concrete animate or inanimate entity.90 Avec yields to à as soon as the focus of the description of the state of affairs is not so much on the concrete entities which serve as TOOLS or COMPANIONS but rather when it emphasises the technique or methodology which is employed to achieve a certain goal. Thus, à is used to denote MANNER with the understanding that MANNER includes the use of an instrument prototypically associated with certain kinds of activities whereas avec identifies/highlights a concrete (and maybe unexpected) instrument. The connection of à with the prototype is also reflected by its use in adnominal attribution, where à marks permanent or expected COMBINATIONS, whereas avec is responsible for the encoding of temporary COMBINATIONS. In the realm of BODY PART INSTRUMENTS and MEANS OF TRANSPORTATION, the position of avec is rather weak. In these areas, avec is almost never the first option. De and à are more frequent for BODY PART INSTRUMENTS (including metonymically associated concepts such as one’s voice, etc.) – especially if prototypicality applies and the description is oriented more towards specifying manner than a reference object. Avec returns to the stage when the BODY PART INSTRUMENT is again unexpected or further specified by attributes of its own. MEANS OF TRANSPORTATION mostly require other prepositions: – en is used to encode the manner in which the action involving a prototypical vehicle is carried out; in addition, en is also used to encode material91; – par is mostly employed in combination in contexts with means of public transportation; moreover, it is also used to encode ABSTRACT 92 INSTRUMENTS in the widest sense of the term ; – à is partly a substitute for en but is compulsory when it comes to using animals as MEANS OF TRANSPORTATION, etc.; – other spatial prepositions such as, e.g., sur ‘on’ and dans ‘in’ are used instead of en if the actual concrete means is in the foreground and not the MANNER (the same applies to avec); the choice of the appropriate preposition depends on the shape of the MEANS OF TRANSPORTATION: Sur is used with those which provide some kind of open

The prototype and its derivations 67

surface on which to travel, and dans is used with MEANS OF TRANS93 PORTATION that are considered to be like containers (Klein and Kleineidam 1983: 176). Discounting the possibility of employing more specific spatial relators for concrete MEANS OF TRANSPORTATION in French, we can summarise the discussion of data from Finnish, Dani and French in Table (A61) (cf. also [A40] and [A49]).94 Grey shading marks the distribution of those relators which are also used to encode TOOL-relations. Boldface is used for the expressions which involve this marker but form larger syntagms. (A61)

Distribution patterns compared II

Relation (A40.1) (A40.2) (A40.3) (A40.4) (A40.5) (A40.6) (A40.7) (A40.8) (A40.9) (A40.10) (A40.11) (A40.12) (A40.13) (A40.14)

Finnish kanssa kanssa kanssa kanssa kanssa -ine -ine -ssa/-ssä -in1 -lla/-llä -ssa/-ssä N-GEN avu-lla -lla/-llä -in2 -lla/-llä -ta/-tä -sta/stä -lla/-llä -ine-ine-ine-ine-

Dani eppetak eppetak

inom inom inom inom inom ?

eppetak

inom ? -mekke ? -(n)en -(o)pa -mekke -(o)pa -(o)pa Ø

French avec avec avec avec avec à avec à avec à avec à avec avec (l’aide de) de avec en par avec à de en à par avec

What follows from the comparison of the three languages is the insight that instrumentality is no less susceptible to further categorical distinctions than comitativity. The area between the contexts (A40.10) and (A40.14) even seems to be peculiarly sensitive to over-differentiation. The grey shading used in (A61) does not cover large contiguous areas, as opposed to Table (A49), where the highlighted cells form extended chains. The three languages are characterised by largely individual solutions. It is nevertheless possible to detect some leitmotifs. Some of the markers which are used for TOOL are also used for MEANS OF TRANSPORTATION. French par behaves similar to Dani -(o)pa and Finnish -lla/-llä in the sense that the preposition is not used beyond (A40.9) upwards.95 Thus, typically instrumental relators can be distinguished from other relators. All this corroborates our contention that 2nd order thematic relations are far less homogeneous as their usual subsumption under 1st order micro roles suggests. There is a dividing line

68 Categories, definitions and terminology between more comitative-like and more instrumental-like relations that speaks in favour of a general distinction of COMITATIVE vs. INSTRUMENT. However, the territory in between the two extremes is divided differentially, and thus the domain of what counts as an instance of COMITATIVE in one language may be radically different (= much smaller or much wider) than that of another language. The same applies to the INSTRUMENT. In addition, formal criteria support the idea that there is enough space between COMITATIVE and INSTRUMENT for additional categories – with equal rolestatus among the 1st order micro-roles. Furthermore, it is again formal criteria which prove that the inclusion of adnominal structures is necessary, because the distribution of the relators indicates that this is indeed territory often contested between Comitatives and Instrumentals. As we put so much emphasis on the formal means used to express the relations, it is in order at this point to have a closer look at the relators themselves. 4.3. The primary marker The previous chapters have demonstrated that one can easily run the risk of getting lost in the wealth of data that even a single language has in store. There are so many partly competing expressions which are more or less tightly connected with comitativity and/or instrumentality that it is next to impossible to control them all adequately. This gives rise to the question of whether all expressions are equally relevant to our purpose. We concede that, given the necessary prerequisites (of sufficient time, personnel, data and energy), only a meticulous analysis of the entire spectrum of expressions guarantees a full understanding of the workings of the assumed system. However, it is not only the lack of the above mentioned precious prerequisites which impel us to refrain from a close-up microscopic analysis. For many languages, it is possible to tell stylistic and purely grammatical options apart. Some expressions are clearly privileged either by their token frequency, by their ability to replace and neutralise, or by their psycholinguistic prominence. The latter can be tested by asking native speakers for translation equivalents of relators of foreign languages: What comes to mind first is psycholinguistically prominent.96 Likewise, dictionary entries are helpful, though not always fully reliable97: What lexicographers provide as the first options for translation, or, in the case of monolingual dictionaries, what paraphrase they offer first is psycholinguistically prominent. Similarly, one may interpret the chronology in which relevant information

The primary marker 69

about Comitatives and Instrumentals is given in descriptive grammars as being indicative of the degree of prominence associated with the expressions.98 These tests have provided us with an inventory of markers which deserve the designation primary grammatical markers. A note is in order here: We adopt the concept of primary marker (German: Hauptmarker as opposed to Nebenmarker ‘secondary marker’) from morphological theory (Wurzel 1984: 93). However, it must be borne in mind that marker primacy has two dimensions, namely, the paradigmatic one and the syntagmatic one, cf. (A62). (A62)

Dimensions of the primary marker

paradigmatic mit samt nebst … …

seinem

Bruder syntagmatic

Syntagmatically, the German prepositions used for Comitatives all take a complement in the dative case. Thus, this case morphosyntactically correlates with the prepositions (though not in an exclusive one-to-one relationship) and may therefore be considered an additional index or a secondary marker. That syntagmatic secondary markers are indeed important is one of the topics of Chapter 5. In the remainder of this chapter, however, we focus on the paradigmatic aspects: The paradigmatically primary marker is the one which ousts all other options available for the same slot because of its high degree of grammaticalisation. The following is a tentative definition of the primary marker as we understand it. Primary marker: This relator is a highly grammaticalised and functionally dedicated grammeme99 whose function is to connect the ACCOMPANEE to the COMPANION either directly or via mediation by the lexical verb of the same clause. The relator itself may have predicative power beyond establishing the relation between the two participants: It may specify the precise character of this relation. Relators may belong to a wide range of morpheme classes from bound to free morphemes. The relator morpheme may be discontinuous and hosted by one of the NPs representing the relevant participants or by other syntactic words (such as verbs) present in the clause.

70 Categories, definitions and terminology This definition notwithstanding, there are still pitfalls which make it difficult to rely exclusively on descriptive sources. Sometimes the terminology – traditional or newly created – used by descriptive grammarians and other linguists may be misleading (Stolz and Stroh 2001). Not everything that bears the label comitative or instrumental corresponds to our Comitatives and Instrumentals. In the Uralic language Mari, for instance, object (and subject) NPs may be marked for comitative if a group of similar entities is involved. The meaning conveyed by the so-called comitative suffix is one of totality, i.e., the principal function of the morphological comitative is neither the one of an object marker nor that of a Comitative, cf. the suffix -Be glossed ‘?’ in (A63). (A63)

(Eastern) Mari (Alhoniemi 1993: 59) n}6 l traktor-B e üð}5 ramaš-o lak o iktarat steer:3PL four tractor-? woman-PL ‘All four tractors are driven by women (lit. the four tractors together, women steer [them]).’

What motivates the use of the label comitative for the object NP in the above example is most probably the similarity of the status the component parts of the NP have on the conceptual level. However, this does not make (A63) an example of a Comitative. -Be is more of a number-marking affix.100 The restrictions formulated in the definition of the primary marker do not prescribe any specific morpheme status for the relator. As a matter of fact, the relator can come in a variety of shapes, as can be gathered at least partly from the examples we have reviewed so far – which, however, do not exhaust the morphological phenomenology of relators. Below, we survey a variety of exemplary cases of coding strategies with no attempt at comprehensiveness.101 Note that we have not found any convincing cases of Comitatives/Instrumental exclusively encoded by suprasegmental means. – Affixation The relator can be any kind of affix. Most commonly – because of the famous suffixation preference – we encounter suffixes or enclitics as in Japanese, where the loosely attached relators -to for Comitative and -de for Instrumental have nominal hosts, cf. (A64).102 (A64) Japanese (A64.1) comitative (Nishina 2001: 347) Taroo-wa Hanako-to kooen-ni it-ta Taroo-TOP Hanako-X park-LOC go-PAST ‘Taroo went to the park with Hanako.’

The primary marker 71

(A64.2) instrumental (Nishina 2001: 353) kagi-de doa-o ake-ta open-PAST key-Y door-ACC ‘I opened the door with a key.’ In Totonac, a Totonac-Tepehuan language of Mexico, the relators occur as prefixes on verbs: ta’:- is used for Comitative and li is used for Instrumental. These prefixes occur in the slot immediately preceding the lexical morpheme, cf. (A65). (A65) Totonac (Levy 1990: 120) (A65.1) ACCOMPANIMENT ta’:-min-lh ki-amigo my-friend X-come-COMPL ‘He came with my friend.’ (A65.2) TOOL ni: ka-li:-nik-ti ki’wi xa:lu NEG IMP-Y-hit-2.COMPL wood pot ‘Don’t hit the pot with the stick!’ More rarely, circumfixes are used. Itelmen, a close relative of Chukot (cf. [A56] above), employs circumfixes for the two comitatives while the instrumental makes do with a suffix, cf. (A66). (A66) Itelmen (A66.1) comitative (= Comitative I) (Georg and Volodin 1999: 83) p’eĊç k’-iplxe-ƺ k’oƺ-en boy X-friend-X come-3SG ‘The boy came with a friend.’ (A66.2) instrumental (Georg and Volodin 1999: 80) ºkasfa-ƺ t-6mphe-çen uË tree axe-Y 1SG-cut-3SG.PAT ‘I cut the tree with an axe.’ For the similarly rare infixes, the reader is referred back to our Tagalog examples (A55) (which are not case-markers on NPs, however). – Adpositions Adpositional strategies have a wide distribution. They are of course typical of the SAE languages, but are by no means restricted to them. In the Afro-Asiatic language Somali, the adpositions generally associate with the clause-final verb and thus may be considered to constitute verbal pro-clitics rather than full-

72 Categories, definitions and terminology blown adpositions, cf. (A67). As (A67.2) shows, two adpositions may form a clitic sequence. (A67) Somali (A67.1) comitative (Saeed 1987: 185) nín-kíi ayàa-aan lá shaqeeyey FOC-I X worked man-DET ‘I worked with the man.’ (A67.2) instrumental (Saeed 1987: 191) nín-kíi bàa-uu cèel-ka xádhig kú-ká sóo saaray well-DET rope Y-from pulled man-DET FOC-I ‘I pulled the man out of the well with a rope.’ Prepositions also occur in languages outside Europe. In the Austronesian language Samoan, the typically Polynesian preposition ma covers the full range of comitativity, whereas the multi-purpose preposition i encodes various categories of instrumentality, cf. (A68). (A68) Samoan (A68.1) CONFECTIVE (Mosel and Hovdhaugen 1992: 148) ia aluatu Sina male ili-tea well go DIR Sina X DET fan-white ‘Well, Sina brought (= went there with) the white fan.’ (A68.2) TOOL (Mosel and Hovdhaugen 1992: 144) tuli le pusi i le salu DET cat Y DET broom chase ‘Chase the cat with the broom!’ Postpositions can be found in languages of all continents. The Khoisan language Nama is only one among a very large class of postpositional languages. The syncretistic postpostion /ka covers both Comitative and Instrumental. Sentence (A69) is an example of the latter. (A69)

–

Nama (Olpp 1963: 32) heib /ka go ge ne this stick:M X/Y 2PL:M IRR ‘You should hit him with this stick.’

nî

înou bi

FUT

hit

him

Serial verbs Many languages throughout South-East Asia and West Africa as well as Creoles (Byrne 1984, Lefevbre 1989) from various parts of the world employ verb serialisation for the purpose of marking

The primary marker 73

roles and relations. In Chinese, the verb gƝn ‘to follow’ is used to encode the Comitative, whereas the verb yòng ‘use’ functions as Instrumental, cf. (A70). (A70) Chinese (A70.1) comitative (Bisang 1992: 182) wǂ gƝn tƗ shuǀhuà I X he talk ‘I am talking with him.’ (A70.2) instrumental (Bisang 1992: 184) tƗ yòng kuàizi chƯfàn he Y sticks eat ‘He is eating with sticks.’ The serial-verb strategy corresponds nicely to the idea that the phenomena under scrutiny involve more than one (sub-)predication as sentences. (A70.1–2) contain two verbs each to depict a unitary state of affairs. –

Incorporation/Body part affixes Incorporation is a feature typically associated with Amerindian languages (though not exclusively with these). North-American Indian languages are also notorious for their so-called somatic affixes, which bear some likeness to incorporation proper. Nevertheless, both strategies have to be distinguished for most of the languages – at least synchronically. In these languages, verbs may host bound morphemes that are specialised to indicate the body part which is involved directly or indirectly in the situation.103 Very frequently, these affixes denote the corporal instrument with which a given activity is carried out. On various occasions, however, Mithun (1988: 448) emphasises that these affixes cannot simply be equated with instrumentals known from languages elsewhere on the globe. She convincingly demonstrates that body part affixation belongs more to the domain of derivation and therefore has to be considered as a lexical device. For Lakota, Mithun (1988: 448–449) argues that the socalled somatic prefixes of this Siouan language do not encode the body part with which an action is carried out but rather the motion that is prototypically associated with a certain body part. Thus, it is possible – and common – to paradigmatically replace

74 Categories, definitions and terminology somatic prefixes on a verb in combination with one and the same external instrumental NP, cf. (A71): (A71)

Lakota (Mithun 1988: 449) papushnapé û kaphópe slapyuhand use pullpop ‘He popped it by pushing down on it with his hand/by slamming his palm down on it/he squeezed it and popped it with his hands.’

In Takelma, however, body part prefixes on verbs also occur in contexts where they clearly correspond to genuine BODY PART IN104 STRUMENT roles (Palancar 1999: 160). Sapir (1922: 64–72) discusses the similarities and dissimilarities of noun incorporation and body part prefixation. He argues that single body part prefixes normally have concrete (= lexical) or spatial readings105, whereas in prefix pairs, the one closer to the lexical morpheme usually invites an interpretation as BODY PART INSTRUMENT (Sapir 1922: 74). In addition, there is also a general instrumental prefix wa- which may be co-referential with an external TOOL NP (BODY PART INSTRUMENTS included) (Sapir 1922: 91–92).106 Interestingly, Takelma also marks Comitatives by a verbal affix, this time by the suffix -(a)gw. This suffix covers much of the range of comitativity (Sapir 1922: 137–140), cf. (A72). (A72) Takelma (A72.1) BODY PART INSTRUMENT (Sapir 1922: 74) s.al-8Ư-lats!ag-i’8n foot-hand-touch:him-1SG ‘I touch his foot with my hand.’ (A72.2) Instrumental (Sapir 1922: 91)107 wilau wa-ts!ayag-i’8n arrow Y-shoot:him-1SG ‘I shoot him with an arrow.’ (A72.3) CONFECTIVE (Sapir 1922: 138) nãx-Ư-heel-agw-a’n pipe-hand-sing-X-1SG ‘I shall sing with pipe in hand.’

The primary marker 75

(A73)

The difference between other cases of affixation and body part prefixes in Takelma lies in the fact that in a number of cases the so-called body part affix still closely resembles the nonincorporated body part noun (except for the missing possessor marking which is obligatory for inalienably possessed items). Incorporation is a general and normal process in Takelma. Sapir (1922: 68) observes that instrument incorporation is “a rather important trait” of the language and is by no means restricted to body part expressions, cf. the following example (A73). From this observation, we conclude that incorporation deserves the designation of primary marking strategy, as the texts do not display any other strategy with the same token frequency. Takelma (Sapir 1922: 68) gwen-waya-sgǀ’ut’i neck-knife-cut_off:them ‘He cut their necks off with his knife.’ As far as we understand the description, instrument incorporation is productive, and any noun referring to an entity, be it abstract or concrete, could potentially be incorporated. Body part nouns are different in so far as they have peculiar affixal forms whose etymological connection to the corresponding absolute noun is sometimes obscured. One may therefore interpret this situation as being indicative of a special status of BODY PART INSTRUMENTS, as opposed to TOOLS. Furthermore, these two types of instruments converge because they are more or less closely associated with incorporation – a strategy which Comitatives are banned from: Animacy restrictions generally do not allow for human nouns to be incorporated (Lehmann and Shin 2005).108

–

The latter examples (A70)–(A73) exemplify various kinds of verbal strategies of encoding Comitatives and Instrumentals. Occasionally, we will deal with Comitative-Instrumental marking on verbs in the remainder of our study. However, the focus is always on NP-internal encodings. Verbal Comitative-Instrumentals require an in-depth study of their own, which we refrain from doing in order to keep the length of this monograph within reasonable limits. Except for these self-imposed restrictions, there are no compelling reasons to exclude the verbal part of the story from an investigation of Comitatives.

76 Methodology and theory 5. Methodology and theory We have now a rather detailed picture of the most important formal and functional aspects of Comitatives and Instrumentals at our disposal. The discussion of the various phenomena is suggestive of a complexity which is a challenge to any theory and model. The major question to be answered in this chapter is how we can possibly deal with this complexity without getting into serious methodological trouble, especially since recent years have shown a growing methodological uncertainty among specialists – an uncertainty that is amply reflected by a number of publications in which authors cast doubt upon the feasibility of the kind of crosslinguistic research we engage in. Therefore, this chapter is dedicated to a discussion of the most important methodological and theoretical issues. At the same time, this is also the proper place to provide a survey of the history of research on COMITATIVES and INSTRUMENTALS. For obvious reasons, we can only dwell on a small selection of pertinent titles. For a nearly complete list of publications related to our topic, we refer the readers to the two specialised bibliographies by Campe (1994)109 and Schwarz, Stroh, and Urdze (2001), which cover material published up to the year 2000.110 Discounting the handful of early studies111 – among which we find the remarkable attempt of Noreen (1923: 339–345) to provide a kind of general, onomasiologically-based categorial framework for the description of roles, relations and functions – it is legitimate to take Fillmore’s (1968) case grammar as the starting point of intensified discussion about COMITA112 TIVES and INSTRUMENTALS in contemporary linguistics. Right from the beginning, Fillmore’s ideas and uncertainty about how to deal with the Comitative have given rise to a long-lasting dispute among linguists – the major points of dissent being: – Whether the COMITATIVE should be included among the deep/semantic cases at all, if yes, – whether the COMITATIVE in its entirety should be part of the case paradigm – or only a certain segment of its domain, – or whether it should be subsumed under the INSTRUMENTAL (= COMITATIVE is a kind of INSTRUMENTAL), – in addition, if the entire domain of the COMITATIVE is accepted as part of the case paradigm, what repercussion this has on the definition of case,

Methodology and theory 77

if no, – if there is anything at all that deserves the designation COMITATIVE, if yes, where this category belongs if not among the cases and what its status is. The late 1960s and the early 1970s experienced a vogue of publications inside and outside the framework(s) of localist-minded case-grammar (Anderson 1971), which directly or indirectly address these problematic issues113, cf. for instance Lakoff (1968), Coseriu (1970), Walmsley (1971), and Buckingham (1973). Nilsen (1973: 50) is probably the most important contribution, in which it is argued that the COMITATIVE should be excluded from the list of semantic cases. Kempf (1974: 355–357) represents the position that there are indeed the two distinct case categories COMITATIVE vs. INSTRUMENTAL, whereas Lehiste (1969) ponders the idea of lumping the two together. Seiler (1974a–c), however, advocates a completely different approach, as he assumes that COMITATIVE and INSTRUMENTAL are only two facets of a higherorder principle called CONCOMITANCE. It is astonishing to see that one and the same phenomenon can be viewed in so many different ways, yielding very divergent and even contradictory results. The competition of mutually exclusive solutions is indicative of the peculiar nature of the COMITATIVE. Moreover, the disagreement among students of COMITATIVES and INSTRUMENTALS is also connected to fundamental methodological questions. The problem inventory comprises the following questions: – How do we approach categorisation in language? – Either onomasiologically (= function/content first) – or semasiologically (= form/expression first) – If onomasiologically, then – do we presuppose a universal conceptual ‘ontology’, – or do we work on the basis of individual languages and their semantics? – How do we define function and functions? – If semasiologically, then – what is form? – Do all formal means have the same value for the identification of categories? – What is the relation connecting function to form and vice versa?

78 Methodology and theory Many of these disputed points are still light-years away from being settled. The form-function dilemma has been passed down since the formative years of linguistics as a science and thus dates back to the turn of the 18th/19th century. Ever since, form-oriented and function-oriented periods have alternated in the history of linguistic thought. More recently, Lehmann (1989a) has proposed a kind of Janusian procedure for both descriptive and comparative linguistics which involves a combination of semasiology and onomasiology, i.e., two sub-grammars are provided, each of which represents one of the two competing vantage-points. Despite the fact that this is a simple, elegant, and feasible solution in practical terms as well (an idea to which we adhere to ourselves), there remains the open question of how to define and identify function and form in the first place. In case-grammar, it has been common practice to postulate semantic relations which are largely independent of specific formal realisations. These relations are often based on interpretations of real-world events and/or a logical representation thereof. Thus, extralinguistically determined relations of this kind might hold in an utterance no matter what happens on the surface level, meaning that one and the same semantic case can have a wide range of very different expressions: At the same time, identical expression is by no means an indicator of identical functions. The latter is especially relevant for the È–criterion of generative grammar by which the repeated assignment of a case in the same clause is precluded. This provision is, of course, well-founded in the particular framework of generativism. However, its application is tantamount to confounding extralinguistic “reality” with linguistic “reality”. Owing to our idea that (not only the descriptions of) states of affairs are genuinely linguistic and therefore cannot strictly be determined by ontological facts, we are unhappy with the inflexible imposition of semantic cases independent of their structural correlates. We hasten to add that we do not deny that language-independent notions can ultimately serve as tertia comparationis. Yet notwithstanding this usefulness on a higher level, handling semantic cases in the abovementioned way has two consequences: Either two or more levels of form have to be distinguished, as was formerly done in terms of surface structure vs. deep structure, or the expression side has to be demoted to the status of something that has no importance in itself. While the former solution is at odds with the current trend away from the D/S-structure distinction, the latter solution puts the concept of the linguistic sign at stake, since the signans almost vanishes. We are of the opinion that it makes no sense to do linguistics without properly accounting for the systematicity of the expres-

Methodology and theory 79

sion side. Denying expressions a say in our search for categorisations implies giving up on finding regularities in form-function correlations. There are again several potential reactions to our conclusions. One could be that the role of form-function correlations has been much too overestimated by mainstream functionalism. This is the gist of Givón’s (1995: 18–21) sweeping critique of common functionalist methodology114 – which he also denounces as being naïve (Givón 1995: 62), as it rests on what he argues to be exaggerated assumptions of iconicity (Givón 1995: 106). He goes even further by claiming that “grammar is often much richer, specific and detailed than cognition” (Givón 1995: 335), meaning that formal variation may be there without any cognitive foundation. Yet everything rests upon our understanding of function, i.e., the functions to be ascribed to some formal phenomenon. We adopt Dressler’s (1985: 264) view of function: Function is not only IN language (= grammar), but there are also functions OF language (=communicative aspects, social functions, etc.). As language cannot be reduced exclusively to grammar (as synonym of rule-governed morphosyntax), functions also have to be sought outside the realm of grammatical categories. Even pragmatic sentence perspective and the like do not exhaust the functional potential of linguistic structures. Form variations may be induced by a long list of determining factors such as, e.g., style (and related aspects of aesthetics, rhythm, and euphony), speech tempo, language games, attitudes/social markers, etc. A function may consist in the simple feeling of speakers that the structure actually chosen is the one that fits best into a given – linguistic or extralinguistic – context (whereas it may be out of place somewhere else). Only in this broader perspective does function ultimately make sense. Other notions going by the name of function are reductionist.115 Our own identification procedure owes a lot to Lazard’s (1999) proposals, which start from prototypical functions and their grammaticalised expressions and allow us to trace the further distribution of the latter over other less prototypical, but still related and completely different functions. We also follow Leiss (1997: 136) in her appeal that “[d]ie Form ist unbedingt ernst zu nehmen” (= the expression side must be taken seriously). However, the radical interpretation of form-orientation, i.e., the hypothesis that there is no such thing as total synonymy, inevitably leads to abolishing homonymy (and in a way, allomorphy as well, including free variation) for good.116 Haspelmath (2003: 215–216 footnote 4) advocates language comparison as a means to distinguish accidental homonymy from polysemy: The latter applies when crosslinguistic data demonstrate that certain pat-

80 Methodology and theory terns are recurrent phenomena, while the former is restricted to the occasional cases. This solution is fine – however, in the absence of clear criteria for deciding what is frequently recurrent and what is rare, it often remains ad hoc: Our German examples in (A40) can be shown to be associated with distinct grammatical categories in a variety of languages. Does this mean that the long list in (A40) is just an enumeration of accidental homonymies? The fact that other SAE languages share similar inventories is not a fully convincing counter-argument, as those languages from a genetically and areally defined group whose structural options could still be idiosyncratic historical accidents. Accordingly, Haspelmath (2003: 239–240) argues that crosslinguistic variation may count as evidence for separate representations, whereas the absence of variation points the opposite direction. One has to look at the homonymy-polysemy problem from the vantage point of the linguistic sign. Our basic idea of a linguistic sign coincides with the Peircean model adopted by Dressler (1985: 281) for Naturalness Theory. It has four parts: the signans (= expression), the signatum (= content), the interpreter (= producer/recipient) and the interpretant (= “the idea to which [a sign] gives rise”). If we only look at the traditional nucleus of linguistic sign definitions as a form-function pair of signans and signatum, the problem arises whether distinct expressions for the same content constitute the basis for distinct linguistic signs, and likewise whether distinct contents for one and the same expression are instances of distinct linguistic signs, cf. (A74). (A74)

Problems with the binary linguistic sign

Signans Signatum

A

Synonymy B C P

P

Homonymy A Q

R

Do we have three signs or just one sign under each heading in (A74)? The riddle cannot be answered on the basis of the binary model, as this would require us to assign asymmetrical weight to the two basic constituents: Either the expression side is the decisive element or the content side is. The situation changes if we account for the two additional components of the linguistic sign, namely, the interpreter and the interpretant, because the picture is automatically diversified. What is associated with identical content in the binary model may cause a variety of interpretants, for instance. These distinct interpretants, in turn, may not have to do with grammatical meaning in the strict sense, but rather belong to the realm of style and various levels. The interpretant is the component which allows for functions other than the purely grammatical ones to play the decisive role. However,

Methodology and theory 81

interpretants are much less accessible than grammatical functions, as they presuppose our knowledge of the interactional details which are hard to come by, especially for crosslinguistic purposes. Owing to this scarcity of reliable data, the best we can do is operate with the old binary model and complement it with additional insights about the interpretant and the interpreter wherever possible. With a view to avoiding the pitfalls of giving one of the two basic components of the linguistic sign precedence over the other, we treat examples of the above kind (A74) as more or less tightly knit associations of signs (“Zeichenverbände”).117 Under these circumstances, there is an urgent need for clarification of how much crosslinguistic variation is needed to decide whether or not we are dealing with mentally distinct entities.118 As we are not psychologists, we refrain from trying to solve this problem in this book. In spite of these remarks, we side with Haspelmath (2003: 212) and adopt the intermediate polysemist position [thereby recognising] that there are different senses or meanings attached to each gram, but these meanings are related to each other in some fashion that it needs to be specified, so that it is by no means an accident that the different senses have the same formal expression.

For the sake of methodological consistency, pace Croft (1998) and Sandra (1998), we give polysemy precedence over other interpretations as long as there are no special reasons for a different solution. We also use the crosslinguistic perspective to come to grips with the relationship between categories/functions. In doing this, we stay strictly within the bounds of linguistics. Therefore, we do not have recourse to extralinguistic i.e. psychological evidence. Note, however, that we understand our findings to be an offer for psychology-oriented specialists to build their hypotheses on. Incidentally, Givón (1995: 175–304) uses a title very similar to the motto used by Leiss (cf. above) for two of his chapters – however, the two approaches have strikingly different ideas of what the formal phenomena to be taken seriously are like. For Givón (1995: 29), it is clear that [i]n language and cognition, a category is not identified by the presence or absence of a single criterial feature. Rather, categories are defined by clustering of a number of central features […],

whereas others have no problems accepting single feature variation as evidence for categorical boundaries. The more we think about it, the more plausible the idea becomes that people are talking about (maybe not completely) different things. If one focuses on clusterings, the individual fea-

82 Methodology and theory tures which contribute to the feature bundle are relevant only in their association with those features which belong to the same cluster, i.e., the single feature itself loses importance. If one is content with single features, there is the inherent disadvantage of losing sight of preferred feature combinations. We assume that single features as well as feature clusters define categories – though of a different kind or hierarchical rank. This view is compatible with Himmelmann’s (2000: 8) typology of typologies, in which a continuum is assumed to exist between mono- and cross-constructional holistic and partial typologies. As the review of the data in the previous chapters amply demonstrates, we would inevitably miss many phenomena which are indicative of connections between various categories/functions because their expressions coincide if we adopt the clustering approach. The distribution profiles of single features are worth the while studying, because if it is true that formal syncretism is possible only or preferentially between categories which are functionally (semantically, cognitively, conceptually, etc.) similar to each other (A74), one has to check the entire distribution of an expression to identify the network of these similar categories. For an example of a completely different way of seeing things, consider the following quote from Nilsen (1973: 48–49):119 Case grammarians are uncertain as to the status of Comitative. Shroyer has run into numerous difficulties in allowing it the status of case. First, since Comitative, like Possessive, expresses a relationship between two noun phrases, he is forced to postulate a Comitative-Object case and a Comitative-Agent case. […] Assuming that a deep case is a concept that should be used for a noun phrase, rather than for a pair of noun phrases, it is probably not wise to consider such concepts as Comitative and Possessive as deep cases, but rather as surface cases. If we do this, then it is necessary to classify Comitatives and Possessives as belonging to other deep cases.

To our mind, this suggestion can hardly do justice to COMITATIVE (or POS120 SESSIVE, for that matter). As we have seen above in the Awiyaana example (A54), Comitatives may indeed be differentiated according to degrees of agentivity. Moreover, there are languages in which the distinction of subject and object forms of the Comitative is a morphological fact. In the Amerindian language Kawaiisu, the comitative postposition -meËe/-weËe is the only postposition of the language to be sensitive to the subject/object distinction, as it also has a separate so-called accusative form -maa-ko/ -waa-ko, which contains the regular accusative suffix -ko. The simple postposition encodes co-subjects121, while the morphologically complex one

Methodology and theory 83

encodes co-objects, cf. (A75). There is thus case-agreement between the participants connected by the comitative relator. (A75) Kawaiisu (Zigmond, Booth, and Munro 1990: 59) (A75.1) Co-subject momoË o-weËe pG kee-ka-dG-mG=ni taËnipGzi man woman-X see-R-NMLZ-PL=me ‘The man and the woman saw me.’ (A75.2) Co-object managi-dG=aka Ëarina-ya nG Ë G poË o-waa-ko fix-NMLZ=it flour-ACC I water-X-ACC ‘I’m mixing flour with water.’ There are obviously languages which use formal means to distinguish various Comitatives according to their grammatical relations or agentivity. This feature may not be widespread, but it is nevertheless attested more than just once. What Nilsen considers undesirable for purely theoretical reasons is a structural option of human languages. Thus it cannot be done away with easily – it must be dealt with adequately, even if this leads to a wholesale redefinition of one’s favourite categories (for instance, semantic case). Nilsen’s argument not only requires us to be partial in the sense of treating language structures differently according to the needs of the model presupposed, but it also disallows the inclusion of whole ranges of phenomena. Inter-participant and intra-participant relations have no place in his system (but not only there!). Moreover, the extension of Comitative and Instrumental relators to clause-combining cannot be adequately be accounted for in Nilsen’s model, cf. the data in (A76)–(A77) from Finnish and Oriya. (A76) Finnish (A76.1) Instrumental (Karlsson 1978: 190) hän elää kiroitta-ma-lla kirjoja book:PL:PART he lives write-INF.3-Y1 ‘He makes a living by writing books.’ (A76.2) Instructive (Karlsson 1978: 187) lapsi tuli itki-e-n kotiin cry-INF.2-Y2 home child come:PAST ‘The child came home crying.’ In Finnish, the 2nd and 3rd infinitive are inflected for a variety of cases.122 Example (A76.1) involves the morphological adessive, among whose functions the Instrumental is prominent. The second clause includes a non-finite verbal form. The two predications are connected via the case-inflection on

84 Methodology and theory the infinitive. Likewise, the inflectional instructive (otherwise mostly used for body part instruments, cf. [A58.1]) encodes the temporal relation of simultaneity. (A77)

Oriya (Neukom and Patnaik 2003: 382) b1 y 1 s 1 b1 rh-iba s1 nge s1 nge taa-r1 grow-INF X X she-GEN age 1 dhik1 dayitw1 taa up1 r-e p1 r-il-a more duty she:OBL top-LOC fall-PAST-3SG ‘With growing age more and more duties fell on her.’

In the Indo-Aryan language Oriya, the comitative postposition s1nge (reduplicated in [A77]) is also used to encode simultaneity between two predications, only one of which is finite. Crosslinguistically, Comitatives and Instrumentals are indeed relatively strong in converb-formation. The question arises whether one should generally disregard syncretisms which go beyond a certain pre-determined area of grammar. We argue that one should by no means ignore them. However, their inclusion automatically poses problems because the original area of interest is backgrounded. In a way, Nilsen is, of course, right: If we extend the perspective to the entire distribution domain of a given expression, we cease to talk about semantic cases as they had been conceived of by (early) case grammarians. We indeed talk about something else: COMITATIVES only cover a (perhaps sizeable) fragment of Comitatives – and it is the latter ones which give the more interesting insights into the nature of category relationships. As a matter of fact, COMITATIVE forms the overarching super-domain which embraces the domains of Comitative and comitative, overlaps with those of CONCOMITANCE and COMITATIVE and comprises an additional area which is covered by neither of the included domains, cf. (A78). Note that we do not assume total inclusion for the COMITATIVE and CONCOMITANCE, as these thematic roles and principles may be dissociated from specific means of expression – at least this is what one finds in mainstream research on these topics. To recapitulate: Comitatives comprise all instances in which the Comitative – the relator used for the encoding of participant relations in prototypical accompaniment situations – occurs. These occurrences are thus associated to one another in a network. The question of what categorical status this network can be assigned remains to be answered.

Methodology and theory 85

(A78)

Inclusion and overlap COMITATIVE

CONCOMITANCE

COMITATIVE

comitative

comitative

To this end, we shall first have a look at two important contributions by Schlesinger (1995) and more recently by Croft (2001). In his study of English, Schlesinger (1995: 70) proposes a higher-level case – called C-case – under which the traditional COMITATIVE and INSTRUMENTAL, plus additional relations such as MANNER, are subsumed.123 Their subsumption under a common cover is largely motivated by the formal likeness of their expressions, as they all employ so-called with-phrases. So far, the procedure employed to identify functions which are interrelated corresponds to the one adopted in this book. In addition, Schlesinger (1995: 25–26) explicitly claims that one should let oneself be guided by the formal phenomena instead of presupposing categories beforehand on purely semantic or conceptual grounds. This is called the Linguistic Relevance Principle, and linguistic relevance is said to be language specific. However, what exactly is linguistically relevant? As Schlesinger uses with-phrases to demonstrate that certain relations can be lumped together, one expects other relations to be treated in the same way. These expectations are, however, not met: Schlesinger (1995: 71–72) has a long list of relations124 for which withphrases are regularly used but have to be excluded from the C-case nevertheless – the reason being that

86 Methodology and theory adding additional features, like Location and Time, to the definition of the C-case would throw the doors open too wide. A case that includes Locative […] would have to include other Locatives as well, and perhaps many types of prepositional phrases, too. Such a mammoth case category would most probably not permit a parsimonious statement of linguistic regularities (Schlesinger 1995: 72).

At the same time, Schlesinger (1995: 72) concedes that Perhaps all or almost all with-phrases have a common meaning component, namely, that of TOGETHERNESS in space and time. […] This fact could figure in an explanation of the diachronic development of the meanings of this word [= with]. But I think it inadvisable to introduce an overall case – say, the “Associative” – subsuming the Comitative and other constructions.

Clearly, the otherwise laudable Principle of Linguistic Relevance has its limitations if the one who employs it is still guided by presuppositions about what a category should look like, be it semantically or in terms of maximal extension. Schlesinger’s handling of his own methodological principle is inconsistent at best. Either the use of with is decisive or it is absolutely irrelevant. Schlesinger’s approach, alternating between splitting and lumping, becomes even more problematic, as language comparison reveals that many of the patterns ruled out by him are indeed widespread crosslinguistically. English is thus only one among a large number of representatives of certain syncretisms worldwide. In this context, consider also the following programmatic statement of Schlesinger’s (1995: 25) A conceptual distinction is to be admitted […] if and only if it subserves the statement of some linguistic regularity.

Does the employment of a preposition qualify as (some) linguistic regularity or not? Schlesinger leaves the reader in the dark about the exact spelling out of the two crucial notions of linguistic regularity and linguistic relevance. Schlesinger seems to be toying with the idea between the lines that syntactic devices may be more important than purely morphological ones, although no explanations for this are given.125 For us, however, the use of a given relator is, of course, a linguistic regularity and is without any doubt linguistically relevant. It is perhaps even more basic than syntactic strategies, as the latter operate on the former. Our approach is not too dissimilar from the precedence given to individual so-called GRAMS in Bybee and Dahl (1989: 97). It also finds a parallel in many studies of mostly Langackerian-inspired Cognitive Linguistics, where the meaning range of individual relators/grammemes is one of the favourite topics, (cf. among many others

Methodology and theory 87

the collection of articles on the semantics of prepositions edited by Zelinsky-Wibbelt [1993]). These parallels do not imply that we work within the frameworks provided by the cited models. A more serious matter is Croft’s (2001: 29–45) critique of methodological opportunism, to use his own wording. Croft is surely right when he observes that so far nobody has done what is actually required by the formfirst approach, namely, taking the pains of carrying out a 100% complete distribution analysis.126 To his mind, the best that has been achieved so far are descriptions of more or less extended fragments of distribution profiles (which usually only suit the purposes of the linguist, whose research interests determine the limits of the distribution analysis). In an ideal world, Croft’s critique would deserve having immediate consequences. However, in the world we live in, this is not to be expected because there are also limits as to how much time, energy and manpower that can be invested in descriptive linguistic work.127 Aside from this more practical aspect, the critique is too sweepingly applied, as it does in no way account for the common principles of scientific reasoning: Extrapolation, generalisation, prognosis, prediction, etc. are all well-founded procedures – not all of them only convenient short-cuts – which allow us to use a limited amount of information as the basis for hypotheses about ‘the unknown’. Only when these hypotheses are never put to the test and their preliminary results are inadvertently promoted to supposedly accepted knowledge (via unreflected citation practices) does the situation become unbearable. Then and only then is it appropriate to speak of methodological opportunism. Admittedly, this is the case much too often – but it is nevertheless still far from being the daily norm in linguistic discussions. Thus, we feel entitled to carry through with our analysis as far as we deem it necessary and we draw our conclusions on this basis. Others may or may not find fault with them. Empirical exhaustiveness is something that can hardly be achieved outside corpus linguistics. One can try, however, to make as much as possible of empirical facts without corpus-linguistic methods. Similarly to Schlesinger (1995), Croft (2001: 59–61) considers categories (and the constructions on which they are based) to be language specific.128 This is certainly true also of our Comitatives, as it is highly improbable that the distribution profiles of Comitatives in two languages are absolutely identical: Languages are individuals (Bechert 1991).129 We are, however, not absolutely convinced that total identity is required for two Comitatives to pass as equivalents of each other. This problem dwindles down to a minimum, since Croft (2001: 94–104) uses typology to identify the areas of over-

88 Methodology and theory lap shared by his sample languages as to certain categories.130 The outcome are the well-known semantic maps which help distinguish focal areas or prototypes from peripheral areas of a given phenomenon131. In the above chapters, we have tacitly adopted this method for the purpose of demonstrating that some of the contexts in which the German preposition mit is indiscriminately employed have the status of distinct categories in other languages. One word of caution is in order: If the extent of drawing semantic maps is the identification of the common core alone, the whole argumentation is in danger of becoming circular, because the smallest denominator for the languages compared will most often coincide or at least include the prototype. As we have started our research from postulating prototypes for ACCOMPANIMENT situations and the like, this would inevitably lead to the trivial insight that Comitatives in all languages include the Comitative. Therefore, at least for our purpose, semantic maps gain relevance only where they allow statements going beyond the prototype. Outside the prototype, there is a considerable degree of probability that languages will differ from one another. Thus, it is in the peripheral or less central areas of the semantic map that we, on the one hand, learn about the affinities between categories and, on the other, find typologically relevant evidence. The periphery of the semantic map is important for research on syncretism. It is also important for typology, because the very fringes of the map reflect the sub-universal features of the subject at hand. In this book, semantic maps are provided in Parts B–C below. In earlier work by Croft (1991: 237–239), syncretism is taken to be the major piece of evidence for conceptual affinities between categories – in this case between the semantic roles that syntactic categories are based on. The various roles are classified according to their function in the causal order underlying event description (Croft 1991: 184–191). Two distinct classes of so-called subsequent and antecedent roles are distinguished (Croft 1991: 186), which are obviously look-alikes of thematic roles. Moreover, it is claimed that, all things being equal, syncretism is restricted to class-internal combinations. In (A79), we give the inventory of roles according to Croft. The grey shadings indicate pertinent cases mentioned by Croft (1991: 187): The functional domain of English with reflects a syncretistic pattern within antecedent roles, namely, INSTRUMENTAL, MANNER and COMITATIVE, whereas by covers MEANS and PASSIVE AGENT, which also belong to the same class. For subsequent roles, the English preposition to is mentioned because it combines functions of RECIPIENT and RESULT.132 Thus, in Croft’s approach, and in sharp contrast to Schlesinger (1995), individual prepositions are treated as primary and even sole markers. They

Methodology and theory 89

are given precedence over entire constructions when it comes to identifying conceptual affinities.133 Not surprisingly, we side with Croft in his pre(radical-)construction grammar period. (A79) Syncretisms within the role classes Antecedent roles

Subsequent roles

INSTRUMENTAL MANNER COMITATIVE MEANS PASSIVE AGENT ERGATIVE CAUSE

BENEFACTIVE RECIPIENT RESULT

However, even if one looks at the facts in terms of the less radical varieties of construction grammar134, there can be no doubt that the elements which form part of a construction are signs in their own right. This is so because they enter a construction since they have certain content-based properties which fit into the constructional context. These properties may be highlighted, foregrounded or re-shaped by the context – but nevertheless there needs to be something that the context can determine in the first place.135 For instance, we consider every attestation of a given relator as an instantiation of the relator as a linguistic sign (= a form-function pair, cf. Chapter 2 above), no matter in what context it shows up. Signs may be part of other signs, which, in turn, are again contained in other signs and so forth. Thus, if constructions are form-function pairs with the status of signs, then they are signs whose complex signans the signantia of other less complex signs belong to. The signata of the very same signs are also present, but in a less obvious way, if the construction has non-compositional semantics.136 We therefore propose a layered model of the linguistic sign which is based on ideas exposed in Kilian-Hatz and Stolz (1993).137 To better understand what is meant by layered, consider the Estonian examples in (A80). (A80) Estonian (Balthasar 185, 336) (A80.1) ACCOMPANIMENT ja Barber rüüpa-b koos Balthasari-ga Balthasar-X/Y and Barber drink-3SG together sügava sõõmu deep:GEN mouthful:GEN ‘And Barber takes a sip together with Balthasar.’

90 Methodology and theory (A80.2) CONFECTIVE naine astu-s koos teretuse-ga woman enter-3SG:PRET together greeting-X/Y piiruvalgu-sse border_station-ILL ‘A woman entered the station saying hello.’ (A80.3) Temporary COMBINATION minu king-ad on savi-ga koos be:3SG mud-X/Y together my shoe-PL ‘My shoes are muddy all over.’ In these three sentences, the adverb koos ‘together’ occurs alongside the inflectional case-marker -ga on the noun. The animacy of the participants varies from equal [+human] in (A80.1), differential [+human] + [inanimate] in (A80.2) and equal [inanimate] in (A80.3). In addition, the syntactic patterns vary: koos is prenominal in (A80.1–2) and postnominal in (A80.3). Both the adverb and the case exponent partake in different constructions, cf. (A81). (A81)

Constant relators in different constructions

[NP[+human]

V[transitive]

koos

NP[+human]

-ga

NP[¬human]]accompaniment

[NP[+human]

V[move]

koos

NP[+human]

-ga

NP[¬human]]confective

[NP[¬human]

Copula

NP[¬human]

-ga

koos ]combination

X/Y

‘together’

The three constructions can be seen as having meaning of their own, and thus they behave like form-function pairs, i.e. linguistic signs. The semantics of the first construction is still largely compositional. If the English translation is anything to go by, the compositionality is less clear in the second case and minimal in the last example. Postnominal koos in copulasentences is reminiscent of idiomatic expressions, since the meaning of the construction is not the sum of the meanings of its components: Two entities are not simply together (= in the same place at the same time) but interact with each other in such a way that the entities almost fuse into one. This construction also invites the reading that one of the entities is distributed all over the place, where the other entity is situated. Or else one entity is constituted by the other. These are, of course, very special meaning compo-

Methodology and theory 91

nents which do not result from adding the meanings of koos and -ga to the ones of the other constituents of the utterance. Thus, the third example at least is a bona fide case of a construction with the status of a distinct linguistic sign. However, it is still a complex sign in the sense that it contains several other less complex signs – and not only on the level of the expression side. Koos and -ga are there because of the meanings they convey, namely, co-presence and subordinate status (= COMPANION). The exact spelling out of co-presence and subordinate status is a matter of the context provided by the construction itself. Ergo, koos and -ga are form-function pairs, each of which may enter into more extended constructions that are also form-function pairs. In this perspective, the entire construction provides the outer layer of a chain of linguistic signs, whereas the minimal meaning-bearing units inside the construction represent the inner layers of this chain. These are the theoretical and methodological foundations of our own approach. They have developed step by step over the years, starting from about 1992 until the present day. Right in the beginning of our project, it became clear that previous research on Comitatives and related phenomena has suffered from the great disadvantage of being empirically biased. The vast majority of pertinent contributions to the debate are devoted to individual languages. In a way, every single publication has painted a (sometimes only slightly) different picture of Comitatives and Instrumentals because of its being language-dependent. Serious crosslinguistic research starts with Croft (1991). The publications which have grown out of our project, however, provide the first (and so far the only available) large-scale comparisons, cf. Chapter 6. From the 1990s onwards, the discussion has moved away from the perhaps futile problem of whether or not the COMITATIVE is a semantic case. Recent contributions mainly revolve around the following topics: – the COMPANION Metaphor (Lakoff and Johnson 1980) which assumes a very close conceptual association of Comitatives and Instrumentals (Stolz 1996, 1997a; Stolz, Stroh, and Urdze 2005; Stroh 1998); – the conceptual affinities of AGENTS and INSTRUMENTALS (Schlesinger 1989; Stolz 2001b; Martínez 2001; Palancar 2001, 2002); – the distinction of WITH- vs. AND-languages (Stolz 1998a; Stassen 2000); – the distinction of legitimate and illegitimate syncretistic patterns involving Comitative and/or Instrumental (Luraghi 2001a; Stolz 1997a, 2001a).

92 Corpus/Sample All these issues have in common that they focus on aspects of syncretism, which we will turn our attention to in Part B. In addition to the abovementioned topics (and a number of minor issues), the members of our project crew have addressed a number of other subjects which have hitherto been neglected in the literature: – Comitatives/Instrumentals and their negative counterparts (Stolz 1996, 1997c). – Markedness relations of Comitatives and Instrumentals (Stolz 1998b). – The diachrony of Comitatives and Instrumentals (Stolz 2001c; Stroh 1998, 1999). – Comitatives and Instrumentals in European perspective (Stolz, Stroh, and Urdze 2003; Stroh 2004). In all these research areas, the relator-oriented approach has proved successful. Parts B–E review our findings. Owing to the fact that we have reached our conclusions on the basis of a variety of different samples, it is necessary to look at our empirical foundations first before we present the results and discuss their impact on linguistic theory. 6. Corpus/Sample Among the aims of this study, the identification of typologically relevant correlations is of prime interest. However, typologists disagree about what typologically relevant correlations should look like. Consider, for this matter, the recent debate over the pros and cons of implicational universals of the classic Greenbergian type: Cysouw (2003: 100) argues for a replacement of implicational universals by tests of statistic significance, whereas Maslova (2003: 107) deems it unnecessary to abolish the notion of implicational universals, although one may distinguish between strong and statistical variants. For Dryer (2003: 127), there are some statistical tests that cannot be applied to linguistic data, and therefore phyla (= genera) instead of individual languages should be counted (Dryer 2000). Plank’s (2003: 138) contention is probably the most important one in this discussion: He claims that one has to account for every kind of exception to the rules underlying a given postulated implicational universal. At the same time, Maslova (2000a–b) has come up with the idea that stochastic models should be used in universals research.138 Given that we cannot be sure that the present distribution of types, structural

Corpus/Sample 93

features, etc. is independent of prehistoric genetic relationships between the ancestor languages of our sample languages, the simple statistic methods commonly employed in mainstream crosslinguistic research are bound to fail. Add to this the possibility of prehistoric language contacts and the result becomes even more uncertain. The methodological uneasiness grows even more as soon as the parallel discussion about the inclusion of non-standard varieties into typological samples is considered, too (Himmelmann 2000: 10– 11; Bisang 2004, who both defend the idea that there is no significant difference between standard and non-standard varieties with regard to the phenomena to be scrutinised).139 It is claimed that dialectology holds a lot in store for typology and universals research and that the generalisations based on the usual typological samples are dubious, to say the least. The dubiousness results from the fact that typological samples normally include data drawn from written standard languages with a long tradition of normative grammar, side by side with data from mostly extra-European languages which lack such traditions. It is by no means rare that a single native speaker’s ideolect of a language of an entirely oral culture is compared to the conventionalised supralectal norm of a literate society – a norm which is often the result of age-long processes and is based on the exemplary style of the most prominent poets and novelists (of the past!). Apart from this problem of comparing the incomparable, the dialectological challenge to typological conventions also calls our attention to the internal structural diversity of diatopic systems. Taken seriously, these calls for an adequate integration of dialects and non-standard varieties may eventually lead to the promotion of idiolects to the basic unit of language comparison.140 The ideas put forward by Trudgill (2004) would even make it necessary for typologists to account for social parameters when it comes to language sampling. Taken together, all these new requirements and principled methodological doubts turn sampling into a Sisyphean undertaking. It is impossible to meet all the requirements in one go. However, it may not be necessary to work with an optimal sample, as samples may vary in size and composition according to the goals of the researcher. We have recapitulated the current discussion in order to explain why we stick to the handy tool of convenience samples (Rijkhoff and Bakker 1998). We admit that there might be reasons for using a certain sample size and guarantee its unbiased internal make-up (Perkins 1989), but we also strongly believe in the strength of unsophisticated large samples (Stolz and Gugeler 2000). Except for the final three words in the quote, we follow Himmelmann (2000: 10) who observes that

94 Corpus/Sample [f]rom a research strategic point of view it seems in fact counterproductive to require that all typological hypotheses be based on well-constructed and large samples.

We do not claim that all samples must be large. But the largeness of a sample is no obstacle. Neither are we happy about using the label typology for minimal samples141, nor do we subscribe to the idea that phyla are the basic unit of samples (Dryer 1989, 2000). Comrie (1993) has pointed out that the strict avoidance of areal and genetic clustering in sample composition may prevent us from taking notice of important facts of structural variation. As a matter of fact, closely related and/or neighbouring languages are still individuals, meaning they are by no means structural clones of each other. If we want to better understand to what extent languages may vary structurally, absolutely bias-free samples are less informative than those which allow for biases. Given the fact that different samples may serve different purposes, we have made use of a variety of convenience samples during the phase of data-collection for our project. This multi-sample procedure is reflected by the results of Parts B–C in this book, whereas Part A contains material which is not strictly sample-based. We have not felt the need to level out the differences of our various samples. As a matter of fact, it has proved useful to employ large samples (of 100–320 languages) for Part B, where we investigate the global distribution of certain phenomena connected to Comitatives and their relationship to other categories in order to check the validity of a number of putative universals and universal constraints. Part B also reveals that not all areas and phyla randomly behave similar: Europe and Indo-European languages spoken in Europe stand out from the rest of the world. This has induced us to look more closely at the situation in Europe (in Part C). For this purpose, a completely different sample was called for, namely, one made up exclusively of European languages (65 languages). As we are looking at different things from different angles in the two parts, we consider it legitimate to draw upon different samples. Note also that the samples form the basis for all our statistics. Outside the statistical contexts, we freely use data from languages excluded from the samples. Another serious matter is the sources from which we obtain the data. For practical reasons, large samples do not normally allow the researcher to use verified elicited material only. Accordingly, we rely heavily on secondhand information accessible in the extant descriptive grammars and dictionaries of our sample languages in combination with pertinent linguistic

Corpus/Sample 95

publications on the topic at hand or related issues. This procedure is common practice, and we acknowledge that it is by no means free of disadvantages. Descriptive grammarians, for instance, may simply be wrong sometimes, while the typologist is at their mercy, in a manner of speaking. However, in spite of all the possible pitfalls in perusing descriptive grammars in order to hunt for information, there is also a limit to scepticism: Distrusting descriptive grammars out of principle is tantamount to making large-scale language comparison impossible. As long as there is no reliable global database of bona fide original material covering a large number of languages which is available to typologists, we must make do with what we can lay our hands on. This also holds for the smaller European sample. With the global sample, we attempt to support statements of a macroscopic nature, whereas for Europe, our interest shifts to the mesoscopic level, where many more details come in. These details are only seldom addressed in the available descriptive works. Thus, there is a need for the analysis of raw linguistic material – however, not all raw material can be employed for the purpose of language comparison. To make things worse, there are again several parameters which determine the quality of one’s raw material data base. Parallel optimisation on all of these parameters is illusory. How does one get fully compatible, spontaneously produced, high-quality, natural native-speaker language data for as many languages as possible? Aside from all the practical questions of making contact with the necessary number of native speakers and recording genuinely natural discourse, there is the major problem of guaranteeing the compatibility of the data. Multiple original texts may be difficult to compare because speakers choose to speak about topics whose description does not invite the use of the phenomenon under scrutiny. The texts may refer to completely different events and thus may become incomparable. Of course, there is always the possibility of providing a stimulus, asking for certain things or using the frog-story approach (Slobin 1985). However, in this way the naturalness of the speech production is already affected, and at the same time the texts may still hardly be comparable because much too much depends on the individual choices of the native speaker informant. Elicitation on the basis of a questionnaire has the same shortcoming of reduced naturalness in addition to the lack of contextual embedding. For these and additional reasons, we have opted for a compromise which itself is not to be mistaken for an optimal solution. We use the translations of a literary text which happens to be available in numerous languages worldwide.142 This practical solution violates several criteria, as we not only rely on written texts which may di-

96 Structure verge considerably from spoken language, but also because we make ample use of non-original material. We are aware of these disadvantages. In the absence of any better feasible approach, we see the analysis of the translations and the original, complemented by the insights gained via questionnaire as well as the information gathered from the usual linguistic descriptions as capable of forming a relatively solid basis for further hypotheses. In terms of both sampling procedure and textual bases, methodological stringency has fallen victim to feasibility. Lamentable as it is, we prefer to make statements about languages, even on possibly shaky grounds, to search for the absolutely perfect and watertight methodology. The results of our convenience-based approach are open to testing, verification or falsification. They are hypotheses. Whether they also represent the only conceivable truth about language remains to be seen in the future. The questionnaire we employed in the initial phase of our project has yielded tangible results for 84 languages. There is, however, a clear European bias (with 62 languages) which has prevented us from evaluating the questionnaire quantitatively. The main information on our topic has been gathered from sources other than the questionnaire, with the exception of eight languages (Karaim, Kazakh, Ndyuka-Trio Pidgin, Selkup, Tatar, Tetun, Tukang Besi, and Zulu). Thus, only in a very few instances will reference be made to questionnaire-based data. Nevertheless, a copy of the questionnaire is provided in the Appendix III. Our respondents are mentioned in the acknowledgments. 7. Structure We proceed henceforth in the following way: In Part B, we address the issue of the relation that holds between Comitatives and Instrumentals in the languages of the world. The keyword for this chapter is syncretism. The universal, typological, genealogical and areal aspects of syncretism as well as distinction of Comitatives, Instrumentals and various other categories are discussed on the basis of our larger samples. In Part C, we turn our attention to the European continent in order to come to grips with the areal peculiarities of the languages spoken there. The keyword for this chapter is isogloss. We look at Comitatives and Instrumentals in order to determine how far predictions about the extent of their functional domains and their overlap are possible. In addition to the evaluation of our corpus-based study within the limits of the 65-language sample, Part C contains three language case studies,

Structure

97

each of which is representative of a type attested in Europe. Part D looks at the dynamic aspects of Comitatives and Instrumentals. The conclusions are presented in Section 15, where we summarise the new vista on the COMITATIVE, discuss structural similarity as a relative notion and finally give an outlook on work to be done in the near future.

Part B What happens when a universal blows up? Metaphor, syncretism, markedness

8. The Companion-Metaphor The success story of the Companion-Metaphor began in the 1980s with the publication of Lakoff and Johnson’s (1980) bestselling book on the metaphorical basis of language and cognition. The central piece of evidence for their far-reaching hypothesis is Comitative-Instrumental syncretism – a pattern which they consider to be a universal phenomenon: With few exceptions, the following principle holds in all languages of the world: The word or grammatical device that indicates ACCOMPANIMENT also indicates INSTRUMENTALITY (Lakoff and Johnson 1980: 135).

In Part A of this book, we have discussed numerous languages which do not conform to what the authors of Metaphors we live by postulate. Tempting as it may be, these languages cannot simply be classified as the exceptions mentioned in the above quote. First of all, they are already too numerous to pass as special cases. Moreover, they cover a wide range of constellations which go far beyond the exceptions conceded by Lakoff and Johnson (1980: 135) themselves, who took notice only of serial verb constructions of the Chinese type (A69) for Instrumentals and for syncretism of AND-coordination with the Comitative (cf. Section 9.2), for which they assume the possibility of a different underlying metaphor or of “other, nonmetaphorical, ways in which form may be coherent with content” (Lakoff and Johnson 1980: 135). However, they take this potential counterevidence rather lightly and simply do not elaborate on the possible implications of the coexistence of several options for structure-building metaphors. The exceptions vanish from the discussion, and thus the universal is indirectly styled as exceptionless. More generally, their empirical basis is clearly insufficient for a study aiming at generalisations about the organisational principles of human language(s) as such. As a matter of fact, Lakoff and Johnson (1980: 134–135) take their native English not only as the starting point but also as the prototypical case of language. They at most men-

The Companion-Metaphor

99

tion French avec in passing, which they believe to function just like English with. In Part A, however, we have shown that the system of comitativity and instrumentality in French is a particularly intricate one, in which avec has to compete with a variety of other markers, cf. (A60). In spite of this serious negligence on the empirical side, Metaphors we live by is an attractive book for us. With a view to understanding the actual weight the putative universal has for the metaphor approach, it is necessary to follow the reasoning of its proponents more closely. This makes sense, as their hypotheses have proved extremely successful among functionalists (especially those interested in grammaticalisation).143 Moreover, the authors’ ideas about the relationship of Comitatives and Instrumentals have been largely (though not exclusively) accepted as fact by many of those who have been inspired by the revolutionary aspects of the new anti-objectivist approach in linguistics. Consequently, many linguists whose background is not in typology may still believe that Comitatives and Instrumentals must of necessity be encoded syncretistically because the human mind does not allow for alternatives. In the words of Lakoff and Johnson (1980: 135), “our conceptual system is structured by the metaphor AN INSTRUMENT IS A COMPANION”, meaning that Comitative-Instrumental syncretism is not arbitrary but determined by the human conceptual system, which, in turn, is shaped by the so-called Companion Metaphor. There is thus a bidirectional determination – and Comitative-Instrumental syncretism is the paradigm case of non-arbitrary, i.e. motivated form-function relations: The use of the same word to indicate INSTRUMENTALITY as well as ACCOMPANIMENT makes sense. It makes such form-content links coherent with the conceptual system of the language. […] Metaphors in the conceptual system indicate coherent and systematic relationships between concepts. The use of the same words and grammatical devices for concepts with systematic metaphorical correspondences […] is one of the ways in which the correspondences between form and meaning in a language are ‘logical’ rather than arbitrary (Lakoff and Johnson 1980: 135).

If the hypothesis had been empirically watertight and the universal had stood the test, Lakoff and Johnson’s findings had surely transformed our discipline, as the key to the workings of human knowledge systems seemed to be ultimately found.144 However, the simple truth is that the putative universal holds for no more than a quarter of the world’s languages (estimated on the basis of our largest sample). Comitative-Instrumental syncretism is clearly a minority solution. In addition, it is also areally and geneti-

100 The Companion-Metaphor cally biased, as it has its one stronghold among Indo-European languages spoken in Europe. In Section 8.1, we briefly review the statistics of the geolinguistic distribution of Comitative-Instrumental syncretism vs. its absence. In accordance with Levinson’s (1991: 1) claim that “every plausible universal that fails is actually rather interesting”, we try to make genealogical sense of the disproved universal in Section 8.2. 8.1. Global statistics Irrespective of the fact that even dedicated followers of Lakoff and Johnson have sometimes made passing remarks about languages which do not conform to the putative universal (Heine, Claudi, and Hünnemeyer 1991: 103), people have neither looked into the crosslinguistic realisation of the formal aspects of the Comitative-Instrumental relation, nor have they considered it detrimental to the metaphor-approach that the number of supposed exceptions to the universal is higher than expected. The latter disinterest in counterevidence is easily explicable, because the rather inspiring metaphorapproach would otherwise be put at stake by a discussion of the validity of some of its structure-oriented generalisations (especially as long as one does not know about the exact percentage of cases which diverge from the postulated pattern). As we will see below, the demotion of the putative universal to the status of only one option among many competitors does not automatically mean that the entire metaphor model is wrong. However, what is perhaps much more disadvantageous to typology and universals research is the fact that an absolutely mistaken idea of how languages are structured in a certain area of grammar might become linguistic doctrine simply because it is easier to repeat and reiterate than to check the relevant data on a large scale. In this section, we use – in slightly revised form – the statistics of Stolz (1997a), which is based on a sample of 323 languages.145 A slightly revised version of this earlier article is our contribution to the World Atlas of Language Structures, which serves only as an additional reference, here, since the results of both articles match (Stolz, Stroh, and Urdze 2005).146 The areal and genetic makeup of our sample is disclosed in (B1–2). As can be gathered from the absolute numbers and the matching percentages, this is indeed a convenience sample because we have not tried to create a fully balanced or unbiased sample. However, our sample contains about 5% of

Global statistics 101

the world’s languages (we presently estimate the world’s languages to amount to some 6,500), and because of its global scope and composition we consider it empirically solid. (B1)

Areal composition147

Continent Americas Asia Africa Oceania Europe Total

(B2)

Languages 78 66 65 63 51 323

Percentage of sample 24.15% 20.43% 20.13% 19.50% 15.79% 100%

Languages 71 38 30 27 25 20 17 14 13 12 10 8 7 7 6 5 4 3 3 1 1 1 323 languages

Percentage of sample 22.00% 11.80% 9.30% 8.40% 7.80% 6.23% 4.50% 4.40% 4.10% 3.80% 3.12% 2.50% 2.20% 2.20% 1.90% 1.60% 1.26% 0.94% 0.94% 0.33% 0.33% 0.33% 100%

Genetic composition148

Macrophylum Amerindian Indo-European Austronesian Niger-Kordofanian Afro-Asiatic Australian Indo-Pacific Nilo-Saharan Uralo-Yukaghir Altaic Caucasian Elamo-Dravidian Austro-Asiatic Isolates Sino-Tibetan Na-Dene Khoisan Chukcho-Kamchadal Eskimo-Aleut Daic Miao-Yao Pidgins and Creoles 22 macrophyla/groupings

The languages of our sample do not make random use of the types of relators when it comes to encoding Comitative and Instrumental. There are only two possibilities: Either a relator is syncretistic, i.e., it encodes Comitative and

102 The Companion-Metaphor Instrumental, or it is not and thus encodes only one of the two categories to the exclusion of the other. An additional third type requires that neither of the two categories is encoded, cf. (B3). The plus sign indicates that a given relator encodes the relevant category, whereas a minus states the opposite. ƒZ means that the relator has no Comitative/Instrumental function at all. (B3) Relator ƒX/Y ƒX ƒY ƒZ

Relator types Comitative + + ¬ ¬

Instrumental + ¬ + ¬

Type syncretistic distinct distinct other

Languages have different choices as to the relator types they prefer. As a matter of fact, there are languages which use only syncretistic or only distinct primary markers, whereas others combine the two relator types, cf. (B4).149 The outcome is an inventory of three language types – with the possibility of subdividing one of the types further according to which of the categories has a distinct marker of its own.150 The two constellations for the C-type are exemplified in (B5–6). (B4)

Relator types and language types

Relator 1 Comitative Comitative + Instrumental Comitative Comitative + Instrumental

(B5) (B5.1)

(B5.2)

(B5.3)

Relator 2 Instrumental

Type A-type B-type Comitative + Instrumental C-type Instrumental

C-type/Comitative: Hungarian (Lee 1990: 128–129) sociative család-o-stul el-költöz-ö-m away-move-LIG-1SG family-LIG-X ‘I move with my family.’ ACCOMPANIMENT: comitative-instrumental család-dal költöz-ö-m el away family-X/Y move-LIG-1SG ‘I move with my family.’ TOOL: comitative-instrumental ceruzá-val ír-t-unk pen-X/Y write-PAST-1PL ‘We wrote with a pen.’ ACCOMPANIMENT:

Comitative Instrumental

Global statistics 103

In Hungarian, in addition to the syncretistic comitative-instrumental in -val/-vel, there is also the so-called sociative151 in -stul/-stül, which encodes various relations subsumed under comitativity. It can never be used for the expression of Instrumentality. (B6) (B6.1)

(B6.2)

(B6.3)

C-type/Instrumental: Saramaccan preposition (Byrne 1987: 160) Samo ku di mujée tsà wàta kò a di dàgu Samo X/Y DET woman carry water come to DET dog ‘Samo and/together with the woman brought water to the dog.’ TOOL: preposition (Byrne 1987: 62) a kòti di kumàlu ku di fàkà X/Y DET knife he cut DET fish ‘He cut the (large) fish with the knife.’ TOOL: serial verb (Byrne 1987: 170) mi téì fàkà kòti di fìsi I Y knife cut DET fish ‘I cut the fish with a knife.’ ACCOMPANIMENT:

In the English-based Creole Saramacca, spoken in the hinterland of Suriname, the preposition ku, derived from Portuguese com ‘with’, is used for the full range of Comitativity and Instrumentality. There is, however, a competing construction when it comes to marking TOOLS: In these cases, the serial verb téì from English take introduces the NP representing the 152 TOOL. Téì can never be used in more Comitative-like contexts153 and is also excluded from encoding MEANS OF TRANSPORTATION and HUMAN/ ANIMATE INSTRUMENT. Relators of the kind predicted by the Companion Metaphor occur in the B-type and in both varieties of the C-type. One therefore expects to find many tokens of Comitative-Instrumental syncretism as well. However, the reverse is true. In our sample, the vast majority of languages opt for keeping Comitative and Instrumental strictly apart by formal means. With 209 languages, the A-type clearly dominates, whereas the B-type is represented by just 79 languages. The smallest group of languages belongs to the Ctype, namely 35 cases worldwide. Table (B7) translates these absolute numbers into percentages.

104 The Companion-Metaphor

(B7) Distribution of languages over language types 100,00% 90,00% 80,00% 70,00%

A-type 64,70%

60,00% 50,00% 40,00% 30,00% 20,00%

B-type 24,46% C-type 10,84%

10,00% 0,00%

Two thirds of the sample languages do not follow the Lakoff and Johnson universal. Only a quarter of them are in line with what the universal predicts. The remaining 10% simultaneously diverge from and conform to the universal, as they employ at least one syncretistic marker alongside a nonsyncretistic one. Even if we add the C-type languages to the number of Btype languages for the simple reason that C-type languages also provide evidence for Comitative-Instrumental syncretism, this new combination of types still lags far behind the A-type at a ratio of approximately 1:2. Hardly more is needed to demonstrate that the supposed universal is not a universal at all. It is nevertheless enlightening to see how the various types are distributed in areal and genetic terms. 8.1.1. Areas Each continent is represented by at least 50 languages in our sample, the largest being 78 languages of the Americas. This margin of 28 languages is of course a problem for our comparative task, as the statistics may be distorted because of the different share of the sample the various areas have. Table (B8) contains the absolute numbers of languages according to type and area.

Global statistics 105

(B8)

Distribution of language types over areas

Continent Americas Asia Africa Oceania Europe Total

A-type 54 47 38 54 16 209

B-type 16 12 20 6 25 79

C-type 8 7 7 3 10 35

Total 78 66 65 63 51 323

In spite of the above mentioned problems, three things strike the eye immediately: – with the notable exception of Europe, the B-type is a minority solution everywhere; – the C-type always forms the smallest group of languages; – outside of Europe, and the A-type is the uncontested majority solution. The relations between the two minor types also follow a certain logic. The C-type usually amounts to more or less half of the size the B-type can claim. The dominance of the A-type is especially pronounced in Oceania, whereas it is less significant in Africa. Table (B9) indicates the percentage of each language type within each area. The mirror image is presented in Table (B10): Here we look at the percentage each continent has within a given type and specify the statistical weight of the languages of the various areas in our sample. (B9)

Language type per area

Continent Americas Asia Africa Oceania Europe

A-type 69.23% 71.21% 58.46% 85.71% 31.37%

B-type 20.51% 18.18% 30.76% 9.52% 49.01%

C-type 10.25% 10.60% 10.76% 4.76% 19.60%

Except for Europe and Oceania, the C-type displays almost the same percentages as for the entire sample, namely, slightly more than 10%. For the remaining two types, the percentages are more variegated. In Oceania, the percentage of the B-type drops below 10%. In Africa, and especially in Europe, the B-type has a strong position, whereas in Asia and the Americas, despite its being firmly established, it is clearly ousted by the A-type. In turn, the A-type goes far beyond the two thirds mark in Asia, the Americas and Oceania. In Africa, its dominance is less pronounced and in Europe, the A-

106 The Companion-Metaphor type is only second best after the B-type. The values are especially indicative not only if they diverge from the percentages calculated on the basis of the entire sample but also if they do not correspond to the share that an areallydefined group of languages has among the sample languages. (B10)

Area per language type

Continent Americas Asia Africa Oceania Europe

A-type 25.83% 22.48% 18.18% 25.83% 7.65%

B-type 20.51% 15.38% 25.64% 7.69% 32.05%

C-type 22.85% 20% 20% 8.57% 28.57%

Share in sample 24.14% 20.43% 20.12% 19.50% 15.78%

As Table (B10) shows, the A-type is overrepresented in the Americas (only slightly), Asia, and, with highly significant values, also in Oceania. In Africa and Europe, the tables are turned because the A-type is underrepresented – even dramatically so in Europe, where the actual percentage of A-type languages amounts to only half of the estimated one. In contradistinction, the B-type is overrepresented only in Africa and Europe. In the latter area, the actual percentage is twice as high as the estimate. In all other areas, the percentages of the B-type do not come near their estimated equivalents. For Oceania, they do not reach much more than a third of the expected percentage. As to the C-type, the actual percentages correspond neatly to the estimate in Asia and Africa, as well as in the Americas too, albeit less so. In Europe, we encounter a disproportionately high percentage, whereas the C-type fails to meet the expectations in Oceania to more or less the same extent as the B-type. Thus, the distribution in geographical terms is anything but even, Europe is the odd one out, as it favours B-type languages and is very tolerant as to C-type languages, to the detriment of the A-type languages, which always rank first outside the European continent. One might get the idea that counting languages is perhaps not the best method of investigating the issue at hand. Owing to the fact that we are discussing a facet syncretism, it may be more adequate to look at relators instead of languages. This change of perspective has the advantage of reducing the number of types from three to two because there is no such thing as a mixed marker corresponding to C-type languages. There are altogether 645 relators, the vast majority (= 523) of which are non-syncretistic; only 122 reflect the syncretism predicted by the universal. The percentages are given in Diagram (B11).

Global statistics 107

(B11) Relator types attested in the sample 100,00%

nonsyncretistic 81,08%

90,00% 80,00% 70,00% 60,00% 50,00% 40,00% 30,00%

syncretistic 18,92%

20,00% 10,00% 0,00%

The importance of the Companion Metaphor and the erroneous universal is thus further reduced because the share of non-syncretistic markers rises above 80%. This is of course explicable from the fact that the dominant Atype languages each make use of (at least) one distinct Comitative marker and one distinct Instrumental marker, cf. (B12). (B12)

Syncretistic and non-syncretistic markers

Category Comitative Instrumental Comitative-Instrumental Total

Attestations 272 251 122 645

Percentage 42.17% 38.91% 18.92% 100%

The relators reveal another trend: As soon as we take relators as basic units instead of languages, the picture changes also in those areas where B-type languages are relatively prominent. Syncretistic relators can neither compete with the non-syncretistic ones in absolute numbers nor in percentages. With the sole exception of Europe, this observation holds also for the relation between non-syncretistic Comitatives and syncretistic ComitativesInstrumentals as well as for the one between non-syncretistic Instrumentals

108 The Companion-Metaphor and syncretistic Comitatives-Instrumentals. The latter ones are vastly outnumbered in any case, cf. (B13). (B13)

Frequency of relator types per continent

Continent Americas Oceania Asia Africa Europe Total

Syncretistic 25 10 21 31 35 122

Non-syncretistic 141 125 111 97 49 523

Total 166 135 132 128 84 645

Nowhere does Comitative-Instrumental syncretism affect the majority of relators attested in a given area. It is always a minor pattern whose share oscillates between about 40% in Europe down to slightly above 7% in Oceania. We calculate for the entire sample a share of slightly less than 19% for syncretistic markers and 81% for non-syncretistic ones, cf. (B14). The Americas and Asia host languages where the ratio valid for the entire sample is closely replicated. In all other continents, the actual percentages diverge from the ones of the sample. The most striking cases are again Oceania and Europe: In Oceania non-syncretistic markers have a share of significantly more than 90%, whereas Europe is the only area where syncretism is characteristic of far more than 25%.

(B14) Percentage of relator types per continent 100,00% 84,93%

80,00%

92,59% 84,10%

81,08%

60,00%

58,33% 41,67%

40,00% 24,22%

20,00% 15,07%

0,00%

syncretistic

75,78%

Americas

As ia

18,92%

15,90% 7,41% Africa

Oceania

Euro p e

non-syncretistic average syncretistic average nonsyncretistic

Global statistics 109

Non-syncretistic encoding of Comitative and Instrumental prevails in a global perspective. There are of course areal differences as to the degree of predominance. Syncretism is important only if we look at languages as representatives of types. If the relators themselves are investigated, syncretism becomes even more marginal in the extra-European world. Thus, we feel entitled to suggest two major revisions to the universal put forward by Lakoff and Johnson (1980): – REVISION I With much more than chance frequency, languages will make use of at least two distinct words or grammatical devices, one of which is specialised to indicate ACCOMPANIMENT, with the other indicating INSTRUMENTALITY. – REVISION II With a probability of 4-to-1, the words or grammatical devices used in a given language to indicate ACCOMPANIMENT and INSTRUMENTALITY, respectively, are not syncretistic with each other. These revisions strip the putative universal of the original content. Moreover, the former constraint is turned into the exact opposite: What previously was the only possible choice is now the exception. Languages disfavour Comitative-Instrumental syncretism and prefer distinct markers for each of the categories. Europe stands out from the rest of the continents because of its susceptibility to Comitative-Instrumental syncretism. The universal is in fact a Euroversal – an areal trait of European languages which sets them apart from the bulk of the world’s languages.154 This peculiarity of European languages induces us to reserve Part C to a detailed investigation of Comitatives/Instrumentals in Europe. Before we employ the linguistic looking glass in this way, there remain long lists of topics which have to be addressed in the remainder of Part B. 8.1.2. Phyla Having looked at the areal aspects of the failed universal, it makes sense to check whether or not the universal is genetically biased too. There is of course the intuition that the preference of European languages for Comitative-Instrumental translates into an Indo-European predilection for this pattern. For practical reasons, we revert to counting languages in this section, cf. (B15).

110 The Companion-Metaphor (B15)

Languages per type and macrophylum

Macrophylum Amerindian Indo-European Austronesian Niger-Kordofanian Afro-Asiatic Australian Indo-Pacific Nilo-Saharan Uralo-Yukaghir Altaic Caucasian Elamo-Dravidian Austro-Asiatic Isolates Sino-Tibetan Na-Dene Khoisan Chukcho-Kamchadal Eskimo-Aleutian Daic Miao-Yao Pidgins and Creoles 22 macrophyla/groupings

A-type 50 7 21 16 15 18 16 10 3 9 9 6 5 6 6 4 1 3 2 1 1 0 209

B-type 16 25 7 8 6 0 1 3 5 3 0 0 1 1 0 0 3 0 0 0 0 0 79

C-type 5 6 2 3 4 2 0 1 5 0 1 2 1 0 0 1 0 0 1 0 0 1 35

Total 71 38 30 27 25 20 17 14 13 12 10 8 7 7 6 5 4 3 3 1 1 1 323

Of a total of 22 macrophyla, there are only four where the A-type is not the numerically strongest type. We may count out one of these four because our Pidgins and Creoles category contains exactly one exemplar which is hardly representative155 (and thus, Daic and Miao-Yao which are also represented by just one single language, each share the same fate). For the other three cases, the evidence is more tangible. In the Khoisan macrophylum, B-type languages outnumber A-type languages by 3-to-1. For the Uralo-Yukaghiran languages, the ratio is less impressive but becomes similar to Khoisan ratio if we add C-type languages: The A-type, then, accounts for just a third of the altogether 13 languages. The most remarkable reversal of the otherwise dominant ratios is to be found in the Indo-European macro-phylum, where the B-type is more then three times as strong as the A-type. Even if we add C-type languages (this time) to the A-type, the B-type is still stronger by far than its two competitors taken together.

Global statistics 111

The marked behaviour of the macrophyla which disfavour the A-type becomes clearer when we look at diagram (B16), where the macrophyla appear in the top-down order of decreasing percentages of the A-type. Between the lowest but still dominant percentage for the A-type (= NigerKordofanian, with slightly less than 60%) and the highest but no longer dominant percentage for the same type (= Khoisan, with exactly 25%), there is a huge gap of 35 points on the percentage scale. This remarkable fact turns the divergent macrophyla into real eccentrics among our sample languages. (B16)

Decreasing percentages of the A-type

100,00% 90,00% 80,00% 70,00% 60,00% 50,00% 40,00% 30,00% 20,00% 10,00% 0,00%

B+C-type

A-type

Table (B17) surveys the percentages of the various macrophyla within the three language types. Indo-European languages occupy the first rank (with significantly higher percentages than the second best) for the B-type and the C-type but are down to rank 8 within the A-type. Uralo-Yukaghir ranks at position 2 for the C-type. Within the B-type and the A-type the distance between the percentages of Indo-European languages and those of UraloYukaghir covers several ranks. The relative aversion of Indo-European towards the A-type and the parallel preference for B-type and C-type can be seen clearly from (B17).

112 The Companion-Metaphor (B17) Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Percentage of macrophyla per type B-type Phylum IE AM NK AN AF UY AL, KH, NS IP, IS

% 31.64 20.25 10.12 8.86 7.59 6.33 3.8 1.27

C-type Phylum IE AM, UY AF NK AN, AU, EA AA, CA, ND, NS, PC

% 17.4 14.5 11.7 8.9 6 3

A-type Phylum AM AN AU IP, NK AA NS AL, CA IE EA, IS, ST AS ND CK, UY EA DA, KH, MY

% 23.02 10.04 8.61 7.65 7.17 4.79 4.3 3.34 2.88 2.4 1.92 1.43 0.96 0.48

It is interesting to note that the A-type-phobia of Indo-European languages is not simply genetically determined. Indo-European is one of seven macrophyla/groupings whose branches spread over more than one continent. As (B18) shows, Indo-European languages behave differently in correlation to their whereabouts: those members of the macrophylum which are spoken in Europe indisputably favour the B-type to such an extent that the few languages opting for the A-type amount to less than the tenth part of Indo-European B-type languages of Europe. However, Indo-European languages spoken in Asia have a predilection of over 60% for the A-type. In this way, both subgroups of the Indo-European macrophylum conform to the preferences which are typical for the area where they are spoken. The evidence from Altaic and Uralo-Yukaghir is less telling. (B18) Phylum IE

AN

Multi-areal macrophyla per type Type A B C A B C

Europe 2 21 6

Africa 0 1 0

Asia 5 3 0 1 2 1

Americas

Oceania

20 5 1

Global statistics 113 Phylum AA

UY

AL

IS

EA

Type A B C A B C A B C A B C A B C

Europe 1 0 0 2 3 3 0 1 0 1

1 0 0

Africa 11 5 3

Asia 3 1 1 1 2 2 9 2 0 5 1 0 1 0 0

Americas

Oceania

0 0 1

With a view to support the idea that there is an areal motivation behind the differential distribution of Indo-European languages over types, consider examples (B19–20) from two modern Indo-Aryan languages, namely Kalderash – a Romani variety spoken in Europe – and the largest language of India, Hindi (Stolz and Stroh 2001: 397–403). (B19) Kalderash (B19.1) ACCOMPANIMENT (Paramiþà 12) kana areslo o Jurko po abáv and wedding in when be:3SG.PAST DET.M Jurko for o Danmark e kumpanja le Frinkuloski DET.M Denmark DET.F clan DET.PL Frinkulo:GEN nas le lovar-en-ca NEG:be.PAST DET.PL Lovari-PL-X/Y ‘When Jurko was in Denmark for the wedding, the Frinkulo clan was not together with the Lovaris.’ (B19.2) TOOL (Paramiþà 70) mudarelas le baĜ-en-ca ragatka-sa them stone-PL-X/Y sling-X/Y kill:3SG.PAST ‘He killed them with stones and a sling.’ (B20) Hindi (B20.1) ACCOMPANIMENT (Shukla 2001: 166)156 laDŽkƯ laDŽkƝ kƝ-sƗth Ɨƭ girl boy X come:PAST ‘The girl came along with the boy.’

114 Syncretism (B20.2) TOOL (Sharma and Vermeer 1972: 23) cƗqnj se pensil banƗo knife Y pencil sharpen:IMP ‘Sharpen the pencil with a knife!’ Kalderash behaves like the vast majority of other Romani varieties on European soil, as it employs a syncretistic relator for both Comitative and Instrumental (Boretzky 1994: 108). On the other hand, Hindi, one of its distant relatives, makes use of two distinct expressions, in accordance with the majority of other Neo-Indic languages and their non-Indo-European neighbours in South Asia. In sum, Kalderash is in line with the areal preferences of Europe, and Hindi follows the patterns that dominate in Asia. It is not only the genetic affiliation of a given language that determines which type the language belongs to but areal aspects as well, which also have an important say in this matter. Being spoken in Europe is a much stronger factor than belonging to the Indo-European macrophylum, but the latter is also a powerful prognostic, as the revisions III–IV emphasise. – REVISION III With a 50% probability, a language spoken in Europe will employ Comitative-Instrumental syncretism. – REVISION IV With a 3-to-1 probability, an Indo-European language will employ Comitative-Instrumental syncretism. Lakoff and Johnson (1980) have promoted a typical trait of Indo-European languages of Europe to the status of a universal. In Sections 8–8.2, this universal has been disproved. It is now time to have a look at the competitors of Comitative-Instrumental syncretism. 9. Syncretism In this chapter we look at the synchrony of syncretistic patterns. The dynamic aspects of syncretism will be focused upon separately in Part D, where we discuss grammaticalisation. Owing to the fact that Comitative and Instrumental are by no means formally identical in every instance, the question arises as to whether or not there other categories with which the two – either together or as singles – combine in syncretistic patterns. The answer to this question is a positive one. In Tables (B21–24) we reproduce the statistical results of the check for syncretistic patterns conducted by Stolz (1997a: 164–

Syncretism 115

165) on the basis of the 323-languages sample for the sake of convenience. In what follows, we use the label Agentive for Passive Agent. (B21) Frequency of syncretistic patterns: strictly binary157 Rank 1 2 3 4 5 6 7 8 9 10 11 Sum

Pattern Comitative + Instrumental Instrumental + Locative Comitative + AND Instrumental + Ergative Comitative + Possession Comitative + Locative Instrumental + Possession Instrumental + Agentive Instrumental + AND Instrumental + Benefactive Comitative + Benefactive 11 patterns

Cases 122 91 56 31 30 11 10 8 7 5 1 372

% 32.80% 24.46% 15.05% 8.33% 8.06% 2.95% 2.7% 2.15% 1.9% 1.34% 0.26% 100%

There are 11 strictly binary patterns with 372 tokens in the sample. Comitative-Instrumental syncretism may fail as a universal, but it is still number one among the syncretistic patterns attested in our largest sample. It accounts for almost a third of all strictly binary combinations. Between ranks 5 and 6, there is a sizeable gap of more than 5 points on the percentage-based scale – meaning that the last six patterns are only of marginal importance. Syncretistic Comitative-Instrumental markers, in turn, can enter more extended syncretistic patterns. Almost two thirds (more precisely, 77 out of 122 or 63.11%) of all Comitative-Instrumental markers participate in syncretistic patterns which go beyond the combination of these two categories. Table (B22) lists those cases which contain exactly one additional category. (B22) Frequency of syncretistic patterns involving Comitative-Instrumental (loosely binary) Rank 1 2 3 4 5 6 Sum

Pattern Comitative-Instrumental + AND Comitative-Instrumental + Locative Comitative-Instrumental + Possession Comitative-Instrumental + Causative Comitative-Instrumental + Agentive Comitative-Instrumental + Benefactive 6 patterns

Cases 35 17 11 6 5 3 77

% 45.45% 22.07% 14.28% 7.79% 6.49% 3.89% 100%

116 Syncretism Surprisingly, the total number of tokens for patterns involving the syncretistic Comitative-Instrumental amounts to twice as many cases as those attested for distinct Comitatives and Instrumentals, cf. (B23). (B23) Ternary and larger patterns158 Rank 1 2 3 4 5 6 Sum

Pattern Comitative + AND + Possession Instrumental + Locative + Agentive Instrumental + Locative + Possession Instrumental + Locative + Ergative Instrumental + Locative + Benefactive Comitative + Locative + AND Comitative + Locative + Possession Instrumental + Locative + AND Instrumental + AND + Possession Instrumental + Locative + AND + Possession 9 patterns

Cases 10 6 5 5 3 3 2 2 1 1 38

% 26.31% 15.78% 13.15% 13.15% 7.89% 7.89% 5.26% 5.26% 2.63% 2.63% 100%

This unexpected disproportion becomes even more pronounced when we add the more extended patterns involving syncretistic markers of Comitative-Instrumental to the list in (B22). Table (B24) contains 13 additional cases, and thus the total of tokens for syncretistic patterns which Comitative-Instrumentals partake in rises to 99, i.e., almost three times as many attestations as the ones given in (B23). (B24)

Larger patterns involving Comitative-Instrumental

Rank 1

Pattern Comitative-Instrumental + Locative + Possession Comitative-Instrumental + AND + Possession Comitative-Instrumental + AND + Causative Comitative-Instrumental + Locative + AND + Possession + Agentive + Causative + Benefactive Comitative-Instrumental + Locative + Agentive Comitative-Instrumental + Possession + Causative Comitative-Instrumental + Locative + AND 7 patterns

2

Sum

Cases 3 3 3 1

% 23.07% 23.07% 23.07% 7.69%

1 1 1 13

7.69% 7.69% 7.69% 100%

Under the premise that the above data reflect a trend among the world’s languages, we may speculate that Comitative-Instrumental relators are more prone to further syncretism than distinct Comitative and Instrumental relators.159 This speculation immediately gives rise to the question as to why this should be so. A possible answer is that Comitative-Instrumentals

Syncretism 117

allow for a combination of syncretistic chains, which are normally only possible with either the Comitative or the Instrumental. We claim that Comitatives and Instrumentals have different preferences as to the partners they choose when it comes to syncretism. Upon superficial inspection, this claim is at odds with the observation that it is exactly the A-type which has the highest share of additional syncretistic patterns, i.e., those languages which keep Comitative and Instrumental formally distinct show a certain predilection for syncretisms involving other categories, cf. (B25). (B25) Typological foci of additional syncretistic categories Category Ergative Locative Possession Benefactive AND Agentive Causative

A-type 96.7% 80.5% 74.5% 66.6% 59.5% 53.95% 0

B-type 0 9.4% 13.8% 33.3% 27.0% 30.7% 66.6%

C-type 3.3% 10.1% 11.7% 0 13.5% 15.35% 33.3%

Of all cases of an Ergative being syncretistically encoded with either Comitative or Instrumental, 96.7% are attested in languages of the A-type. This preference for A-type languages comprises Locative, Possession, Benefactive, Coordination and Agentive too. The only category going along with the B-type is the Causative, which happens to have zero frequency with Atype languages. In addition to the typological bias in favour of the A-type, there are also significant areal differences, cf. (B26). (B26)

Areal foci of additional syncretistic categories160

Category Ergative Agentive Causative Benefactive AND Possession Locative

Europe 16.12% 15.38% 16.66% 0 7.3% 13.72% 11.9%

Africa 0 23.07% 16.66% 22.22% 43.75% 37.25% 30.5%

Americas 3.22% 7.70% 50% 33.33% 26.05% 15.68% 23.7%

Asia 22.6% 53.85% 0 0 9.38% 3.93% 21.1%

Oceania 58.06% 0 16.66% 44.44% 13.52% 29.42% 12.8%

The status of the zeroes in (B26) partly depends on the absence of certain language types from an area (such as ergative languages, which are not reported for Africa). Europe is the only continent where no areal focus (other than the one for Comitative-Instrumental syncretism) is located. Moreover,

118 Syncretism the percentages for European languages are rather low. In contradistinction to Europe, the other continents qualify each as hotbeds for at least one syncretistic category: 58.06% of all cases of syncretism of Ergative and Comitative and/or Instrumental are attested in Oceania. Still impressive are the percentages for Agentives in Asia and Causatives in Africa. For the remaining categories, the highest percentages drop below the 50%-mark. Nowhere do we find a close competition for the rank of areal hotbed. Some categories are absent from the inventory of syncretistic categories in some areas, with the Americas being the only continent where all categories are attested. All this adds to the exceptional status of Europe as a linguistic area where Comitatives and Instrumentals behave differently from the rest of the world. Tables (B27–28) give the absolute numbers of tokens for the combination of categories in ternary and more extended syncretistic patterns (Stolz 1997a: 171). According to the figures in (B27), distinct Comitatives and Instrumentals indeed tend to go separate ways when it comes to combining with additional syncretistic partners. The Comitative is excluded from a variety of combinations where the Instrumental is admissible, but not vice versa. In other cases, the token frequencies of combinations of Comitative + Z and Instrumental + Z differ significantly. Where Comitative-Instrumental syncretism applies, further syncretistic combinations are legitimate even in those cases which the distinct Comitative and the distinct Instrumental are banned from. The only exception is the Ergative, which neither distinct Comitative nor syncretistic Comitative-Instrumental combine with. (B27)

Cooccurrence of additional categories in syncretistic patterns I

Category Com Ins Com-Ins Sum

Loc 5 22 6 33

Pos 12 7 8 27

AND 13 4 6 23

Ag 0 6 2 8

Cau 0 0 5 5

Ben 0 3 1 4

Erg 0 5 0 5

When we look at the compatibility of the additional categories among each other (i.e. without taking Comitatives and Instrumentals into account), a similar pattern emerges, cf. (B28). The Ergative is excluded from all combinations except the ones with the Locative, and thus the five instances of Instrumental-Ergative syncretism reported in (B27) must include the Locative as third member of the syncretistic pattern as well. Likewise, all cases of the Benefactive being syncretistic with Comitative and/or Instrumental include the Locative as well. The same holds for the Agentive, etc.

Syncretism 119

(B28)

Cooccurrence of additional categories in syncretistic patterns II

Category Loc Pos AND Ag Cau Ben Erg Sum

Loc – 12 9 8 1 4 5 39

Pos 12 – 16 1 2 1 0 32

AND 9 16 – 1 4 1 0 31

Ag 8 1 1 – 1 1 0 12

Cau 1 2 4 1 – 1 0 9

Ben 4 1 1 1 1 – 0 8

Erg 5 0 0 0 0 0 – 5

What these correlations reveal is that syncretism does not operate on an arbitrary or random basis. There is something to each category that partakes in a syncretistic pattern that qualifies it as a suitable syncretistic partner of the other members of the pattern. The suitability of Comitatives, Instrumentals and Comitative-Instrumentals is far from being identical for all possible combinations. In some cases, it seems to be a matter of degree only, whereas elsewhere the differences are of a categorical nature. Puzzling as it may be, Comitatives and Instrumentals clearly have affinities to one another, as the number of syncretistic Comitative-Instrumentals suggests, but at the same time both categories have individual preferences for other partners in syncretism which are dispreferred by the other. This differential behaviour of the categories under scrutiny calls for closer inspection. Accordingly, a selection of the most prominent among the syncretistic preferences is reviewed in Sections 9.1–9.5. Even before scrutinising the preferences, we are in a position to add further revisions to the above list of comments on the failed universal and its implications: – REVISION V Comitative and Instrumental are prime attractors of one another in syncretism but do not have a monopoly. Other potential partners are likewise attractive – though differently for Comitative and Instrumental. – REVISION VI As syncretisms other than the Comitative-Instrumental pattern are highly frequent in the languages of the world, the Companion Metaphor cannot be the only explanation for the structural options of human languages. Either there are many more metaphors at work, or the non-metaphorical foundations of linguistic structure are much stronger than expected.

120 Syncretism 9.1. Agentivity Conceptual affinity of categories is reflected by the syncretism of their expressions. This axiom is shared by various approaches. Croft (1991), Van Valin (1993) and others put forward different constraints on the compatibility of categories in syncretistic patterns. The validity of these constraints has repeatedly been questioned, and suggestions for revisions have been made (Luraghi 2001a–b). The major issue which concerns us here is the relation that holds or fails to hold between AGENTS and INSTRUMENTALS and/or COMITATIVES. Schlesinger (1989) depicts the “cognitive space” between AGENT and INSTRUMENTAL as a semantic continuum.161 Unsurprisingly, markers of the one category may come to be employed for the other, as there are only gradual differences between the two. The question remains to be answered as to whether this close neighbourhood in “cognitive space” also applies to the COMITATIVE. If we follow Croft (1991: 185) and Van Valin (1993), the answer must be yes. However, this is not corroborated by the facts. In contradistinction to the Instrumental, the Comitative does not freely combine with syncretistic partners which bear the feature [+agentivity]. Evidence ex negativo stems from the list of syncretistic patterns which are unattested in the sample Stolz (1997a: 168–169) used, cf. (B29).162 (B29)

Missing syncretisms Comitative + Causative Comitative + Agentive Comitative + Ergative Comitative-Instrumental + Ergative

What is missing are combinations with highly agentive categories (= Agentive, Causative and Ergative). From this short list of absent combinations, two more general patterns of incompatibility result, cf. (B30). (B30)

Incompatibilities B-type Œ Ergative Comitative/¬B-type Œ [+agentivity]

According to (B30), B-type and C-type languages do not allow for Ergatives to combine with their syncretistic Comitative-Instrumentals (though distinct Instrumentals of the latter type may be syncretistic with Ergatives). More generally, the distinct Comitative, be it in A-type or C-type languages, avoids combinations with categories which are characterised by a high degree of agentivity. This is astonishing, as both Croft (1991) and Van

Agentivity 121

Valin (1993), in comparison to the INSTRUMENTAL, place the COMITATIVE nearer to prototypical AGENT categories, probably because they assume the feature [+human] of the prototypical COMPANION to match the same feature of prototypical AGENTS. Is it perhaps this semantic closeness that prevents Comitatives to syncretise with AGENTS or related categories in order to avoid a hypertrophic accumulation of similar features? Or are there reasons? With a view to solving this riddle, Stolz (2001b) has conducted a special investigation focused on the issue of the agentivity problem using a sample of 122 languages.163 This sample additionally contains many languages which have not been considered for the larger 323 languages sample. The major selective criterion for the inclusion of languages in the special sample was the existence of Ergatives and/or personal passives as fullyestablished grammatical categories. To complement the results of Stolz (2001b), we glance sideways to Causatives as well in order to cover all of the three categories marked [+agentive] above. The statistics stem from Stolz (2001b: 167–169) and thus do not cover Causatives. Comitative, Instrumental and a category Z featured [+agentive] yield five logically possible patterns. This number rises to a theoretical maximum of 15 patterns if we subdivide [+agentive] into the three categories of Causative, Passive AGENT and Ergative, cf. (B31). (B31) No. I II III IV V

Logically possible patterns

Syncretism Comitative zinstrumental z [+agentivity] Comitative = instrumental z [+agentivity] Comitative z instrumental = [+agentivity] Comitative = instrumental = [+agentivity] Comitative = [+agentivity] z instrumental

Causative Causative Causative Causative Causative

[+agentivity] Passive agent Passive agent Passive agent Passive agent Passive agent

Ergative Ergative Ergative Ergative Ergative

In continuation of the statistics of Section 9.1, we start our evaluation by checking the frequency of the above patterns. These patterns are unevenly distributed over the 122 languages of the special agentivity sample. As (B32) suggests, pattern V is an absolute rarity, whereas pattern IV is clearly a minority solution with patterns I-III being indisputably the preferred options with similarly high percentages (each amounting to about a third of the sample languages).

122 Syncretism (B32)

Frequency of patterns worldwide

I 31 25.4%

II 40 32.78%

Pattern III 39 31.96%

Sum IV 11 9.01%

V 1 0.08%

122 100%

The sample is heavily biased for European languages which account for 40.98% of the entire sample used by Stolz (2001b). The absolute figures given in (B33) are therefore skewed in favour of Europe. (B33)

Absolute frequency of patterns according to areas

Continent I 4 7 4 6 10 31

Africa Asia Oceania Americas Europe Sum

II 6 2 0 3 29 40

Pattern III 3 9 17 2 8 39

Sum IV 6 2 0 0 3 11

V 1 0 0 0 0 1

20 20 21 11 50 122

However, (B34) identifies via grey shading those percentages which represent a significant increase of the values calculated in the estimate. Africa and Asia rate higher than their estimates for three patterns each, whereas the remaining three continents diverge upwards from the estimate only once each: 72.5% of languages displaying pattern II are spoken in Europe; this value counts 30 points more than the estimate on the percentage-based scale. None of the other percentages comes close to this European maximum except pattern V in Africa. (B34) Relative frequency of patterns according to areas Continent Africa Asia Oceania Americas Europe

I 19.35% 22.58% 12.9% 19.35% 32.25%

II 15% 5% 0 7.5% 72.5%

Pattern III 7.69% 23.07% 43.58% 5.12% 20.51%

% of sample IV 54.54% 18.18% 0 0 27.27%

V 100% 0 0 0 0

16.39% 16.39% 17.21% 9.01% 40.98%

The mirror image of (B34) is provided by (B35), which focuses on the statistical weight a given type has within an area. Discounting the exceptional unique attestation of pattern V in Africa, every continent displays higher percentages than expected at least once. Again, the various patterns are thus

Agentivity 123

characterised by hotbed areas, on the one hand, and areas where they become ever scarcer or even disappear completely. In (B35), grey shading marks again those percentages which are higher than expected. This time, the most striking figures are the ones for pattern III in Oceania, which accounts for more than 80% of all languages of that region represented in the sample. Note that the percentage is 50 points higher than the estimate! (B35) Percentage of patterns within areas Continent Africa Asia Oceania Americas Europe Sample

Pattern I 20% 35% 19.04% 54.54% 20% 25.4%

II 30% 10% 0 27.27% 58% 32.78%

III 15% 45% 80.95% 18.18% 16% 31.96%

IV 30% 10% 0 0 6% 9.01%

V 0.5% 0 0 0 0 0.08%

We may now unify the results of (B34–35) in order to identify areal preferences irrespective of the different sizes of the various areally defined components of the entire sample. Not only is pattern II disproportionately represented among languages from Oceania, but Oceania also provides the largest group of languages employing pattern III. We may therefore conclude that syncretism of Instrumentals with a [+agentivity] category to the exclusion of the Comitative is a typical trait of languages spoken in Oceania. Likewise, the European evidence shows a preference for a different pattern: It is typical for languages spoken in Europe to combine Comitative and Instrumental to the exclusion of [+agentive] categories. Languages displaying this pattern are overwhelmingly spoken in Europe. It is perhaps due to the overrepresentation of European languages in this special sample that pattern II also has the largest share of all patterns in the sample. For the remaining areas and patterns, the distribution of preferences is less pronounced. Patterns IV–V are obviously disfavoured on global scale. The above statistics can be cross-checked against genetic and typological parameters. In (B36), we contrast the Indo-European sample languages with a genetically-unspecified group other. Two thirds of the Indo-European languages favour pattern II and their share of all languages employing this pattern rises to 70%.164 All other patterns are merely minority options for members of the Indo-European macrophylum.

124 Syncretism (B36)

Genetic statistics

Phylum IE Other Sum

I 4 27 31

Pattern III 7 32 39

II 28 12 40

Sum IV 3 8 11

V 0 1 1

42 80 122

In (B37), we highlight the differences between Ergative languages and other (mostly accusative) languages. Most Ergative languages (almost 70%) opt for pattern III. With the same proportion, Ergative languages also constitute the largest group of those languages which make use of pattern III, namely, the one which combines Instrumental and [+agentivity] but excludes Comitative.165 The only other pattern which seems to be attractive to languages with Ergative alignment is pattern I, namely, the maximally distinct one: Slightly less than a third of the languages employing pattern I belong to the Ergative type. With the exception of pattern III, every pattern is more frequently attested for non-Ergative languages. (B37)

Typological statistics

Type Ergative Other Sum

I 10 21 31

II 2 38 40

Pattern III 27 12 39

Sum IV 0 11 11

V 0 1 1

39 80 122

Certain types and certain macrophyla stand out from the rest of the sample languages, as they display very clear preferences for certain patterns to the detriment of others. Given that ergativity is not a prominent feature of the Indo-European languages selected for the sample used by Stolz (2001b), it is unsurprising to see that the favourite pattern in Indo-European languages is different from the one typically chosen by ergative languages. This IndoEuropean preference for pattern II is also shared by a sizeable group of other non-ergative languages of different genetic affiliation (47.5% of all non-ergative languages employ this pattern). Genetic, typological and areal parameters interact when it comes to determining which of the patterns fits in best with a given language. In the following paragraphs, we present examples for the various patterns in (B31).166 The maximally distinct pattern I is logically excluded from B-type languages. Its stronghold is indeed the A-type, with only a few additional attestations in the languages of the C-type. The Polynesian language Tuvaluan is a good example of an A-type language that also has a

Agentivity 125

distinct relator for [+agentive] – in this case the preposition nee for the Ergative. The Comitative is encoded by the preposition mo, and the Instrumental by the preposition ki, cf. (B38). (B38) Tuvaluan (B38.1) ACCOMPANIMENT (Besnier 2000: 317) koo nofo mo tino koloa e maasani INCEP stay X person those NONPAST used_to a ia i ei CONTR.ABS he at ANAPHOR ‘She is staying with people whom she knows.’ (B38.2) TOOL (Besnier 2000: 315) maatou e kini saale nee ia ki NONPAST strike often ERG he Y we:PL:EXCL te pate kilikiti DET bat cricket ‘He’d commonly beat us up with a cricket bat.’ (B38.3) ERGATIVE (Besnier 2000: 280) toko uke tino ne ave nee vaka kaisoa NUM many person PAST send ERG ship steal ‘Many people were stolen by slave ships.’ Modern Farsi is a typical Indo-European B-type language. It employs the preposition ba for both Comitative and Instrumental. None of these can be used as a marker of categories with the feature [+agentive]. For the passive 167 AGENT , various complex prepositions are used instead, cf. (B39). (B39) Farsi (B39.1) ACCOMPANIMENT (Mahootian 1997: 148) ba mæryæm raft-æm tehran X/Y Maryam go:PAST-1SG Teheran ‘I went to Teheran with Maryam.’ (B39.2) TOOL (Mahootian 1997: 148) pænjere-ro ba ajor šikæst window-OBJ X/Y brick break:PAST ‘She broke the window with a brick.’ (B39.3) PASSIVE AGENT (Mahootian 1997: 143) musiqi-ye æsil be-tævæssote aqa-ye traditional to-intermediary Mister-EZAFE music-EZAFE mehran ænjam mi-šævæd PRES-become Mehran accomplish ‘Traditional music will be performed by Mr. Mehran.’

126 Syncretism Pattern III is exemplified by the Hindi sentences in (B40). In contradistinction to the distantly related Farsi, Hindi is an Indo-European A-type language, cf. also (B20) above. Typologically, Hindi is a split-ergative language. However, there is no Ergative-Instrumental syncretism. Rather, we encounter the Instrumental marker se as marker of the causee in causative constructions and of the AGENT in passive constructions, whereas the postposition ne marks Ergative. (B40) Hindi (B40.1) Causative (Sharma and Vermeer 1972: 25) naukar se makƗn sƗf karƗo servant Y house clean make:IMP ‘Let a servant clean the house!’ (B40.2) Passive (Sharma and Vermeer 1972: 26) mujh se peDŽ nahƭ kƗĠƗ gayƗ tree NEG fall go:PAST I:OBL Y ‘The tree was not felled by me.’ (B40.3) Ergative (Sharma and Vermeer 1972: 20) rƗm ne phal khƗye eat:PAST Ram ERG fruit ‘Ram ate fruit.’ The Amerindian language Quechua (variety of Ayacucho) represents the Ctype as it employs a syncretistic Comitative-Instrumental relator -wan alongside the only partially synonymous relator -nti-, which is specialised on Comitativity. In causative constructions for transitive verbs, -wan is used to mark the causee and thus exemplifies pattern IV, cf. (B41).168 However, -ntican never be used for Instrumentality nor as a marker of the causee. (B41) Quechua (Ayacuchano) (B41.1) ACCOMPANIMENT (Hartmann 1985: 64) Paulina-wan ri-chka-n Paulina-X/Y go-DUR-3SG ‘He is going with Paulina.’ (B41.2) TOOL (Hartmann 1985: 64) chakitaklla-wan llamka-nku spade-X/Y work-3PL ‘They work with spades.’ (B41.3) COMBINATION (Hartmann 1985: 130) puka pikanti-ta kanka-nti-ta miku-chka-n roast-X-ACC eat-DUR-3SG Puka Pikanti-ACC ‘He is eating Puka Pikanti together with roast-beef.’

Agentivity 127

(B41.4) CAUSATIVE (Hartmann 1985: 177) parqo-chi-ni chakra-ta Pedro-wan field-ACC Pedro-X/Y water-CAUS-1SG ‘I make Pedro water the field.’ Swahili provides the isolated example for pattern V. Swahili is a member of the Niger-Kordofanian macrophylum and belongs to the A-type. It employs the relator kwa for the Instrumental. The preposition na is used in ACCOMPANIMENT relations and also as marker of the AGENT in passive constructions, cf. (B42). (B42) Swahili (B42.1) ACCOMPANIMENT (Loogman 1965: 269) a-li-ondoka na baba yake X father his 3SG-PAST-go_away ‘He went away with his father.’ (B42.2) TOOL (Loogman 1965: 287) John a-li-andika barua hii kwa kalamu mpya pen new John 3SG-PAST-write letter this Y ‘John wrote this letter with a new pen.’ (B42.3) Passive AGENT (Loogman 1965: 269) mwindaji a-ka-ua-w-a na ndovu elephant hunter 3SG-PERF-kill-PASS-INDIC X ‘The hunter was killed by an elephant.’ Loogman (1965: 286) notes that the instrumental marker kwa may still be omitted and thus zero-marking applies for the Instrumental NP. Moreover, if the passive construction is agentless, i.e., if no NP introduced by na occurs, kwa may be replaced by na. This latter substitution process seems however to be restricted to stylistically motivated cases of personification of the inanimate TOOL.169 As the above examples additionally emphasise, Instrumentals, as opposed to Comitatives, are prone to syncretise with categories which bear the feature [+agentive]. Let us now look at this preference from the other side. How important is the Instrumental for the AGENT categories? Siewierska (1984: 41) observes that “the agent [in passive constructions] typically takes the instrumental”. Keenan (1985: 261–264) mentions a handful of alternatives but also ascribes the highest importance to the Instrumental. In his brilliant and definitive study of the origin of AGENT markers, Palancar (2002: 22) demonstrates first of all that the vast majority of AGENT markers in his sample partake in syncretistic patterns, namely, 87.5%. The percent-

128 Syncretism age rises to 92.4% with passive AGENTS and drops to 80% with Ergatives. There is thus a clear leaning of agent makers to syncretise. Discounting spatial cases, the Instrumental is among the favourite partners in syncretism of both [+agentive] categories (Palancar 2002: 41). Instrumental-passive AGENT syncretism accounts for almost a quarter of all attested cases (= 23.8%) and is second best after the Cause-passive AGENT with exactly a third of the syncretistic patterns. As to Ergatives, the Instrumental is the uncontested first choice, as it occurs in slightly more than half of the cases (= 51.9%). Even if spatial cases are included, there would be no change of rank positions: The Instrumental remains in second and first position for syncretisms with passive AGENT and Ergative, respectively: Only the percentages are lower. Thus, Instrumental-Agent syncretism is not only a favourite constellation for the Instrumental but also a prime option for Agent categories. Comitatives are mentioned in Palancar’s monograph only in passing because the following implication holds: Whenever a Comitative is present in Agent syncretisms, Instrument is also present […]. [I]t is the presence of the Instrument and not of the Comitative what is important in accounting for the emergence of Agent syncretisms (Palancar 2002: 39).

Palancar’s generalisation should be complemented by adding that Instruments do often syncretise with Agents without the Comitative being part of the syncretistic pattern, whereas Comitatives never syncretise with Agents without the Instrumental going along. Palancar (2002: 104) conceives of the Instrumental as a so-called “Medial-Entity” which shares certain properties with the “Initial-Energizer” and thus is a “secondary causal force of the event”. Comitatives lack this feature. This can be illustrated by the following examples from the Australian language Djamindjung, cf. (B43). (B43) Djamindjung (B43.1) Ergative (Schultze-Berndt 2000: 56) jalig janyungbari-ni gujugu-ni yurl gani-ma-m wuju child other-Y big-Y chase 3SG:3SG-hit-PRES small ‘The other big child chases the little one.’ (B43.2) Instrumental (Schultze-Berndt 2000: 56) burrurrug-biyang gana gurunyung-ni head-Y scatter-now 3SG:3SG:chop.PAST ‘He hit it with his head, scattering it.’

Agentivity 129

(B43.3) Comitative: ACCOMPANIMENT (Schultze-Berndt 2000: 569) dunggulba jawaguny-biya ganurr-ujja-ny knock other_group-now 3SG:3PL-poke-PAST dawuu-mayan waladbari mululurru-mij RED:old_man RED:old_woman-X shoot-CONT ‘He shot the others, he shot the old man together with the old woman.’ (B43.4) Comitative: TOOL/COMBINATION (Schultze-Berndt 2000: 61) thunuj gan-arra-m mununggu-mij carry_under_arm 3SG:3SG-put-PRES string-X ‘She is carrying it in a bundle with a string.’ Djamindjung, as is the case with numerous other Australian languages, displays Instrumental-Ergative syncretism, cf. (B43.1–2). In addition, the language makes use of a comitative case whose functional domain comprises Accompaniment situations, as in (B40.3). However, as Schultze-Berndt (2000: 60–61) points out: The comitative marks a noun phrase whose referent […] is construed as involved in an event together with another participant […]. This concomitant participant will sometimes be interpreted as an ‘instrument’ […].

Schultze-Berndt (2000: 61) goes on to explain that if a potential instrument is not construed as an Effector (Van Valin and Wilkins 1996), it can be marked for comitative as in (B44). (B44)

Djamindjung (Schultze-Berndt 2000: 567) mangurn ga-jga-ny burrag 3PL.OBL whitefellow 3SG-go-PAST ‘A white man went for them with a rifle.’

guyug-mij fire-X

In our reading, guyugmij is an example of a confective (the rifle is carried along by the white man – independent of the white man’s intentions as to what to do with the rifle later on). The choice of the appropriate case marker depends on the role of the participant: If it is required to carry out the action, the instrumental is called for, but if it just happens to be there without forming part of a causal relation, the comitative is the correct solution. The problems the topological models of Van Valin (1993) and Croft (1991) have with the locus of the COMITATIVE and INSTRUMENTAL in relation to [+agentive] categories boil down to the following misinterpretations:

130 Syncretism – the comitative in the above models is generally considered to be a coagent170, although Comitativity cannot be reduced to Co-Agentivity; – co-agents are of course featured [+agentive], but comitatives do not necessarily display this feature; what all varieties of Comitatives have in common, though, is their power to mark co-presence of a more or less egalitarian status. Thus, the supposed closeness of the COMITATIVE to the AGENT and similar categories turns out to be only an incidental combination of two features. The more important one of these is [+co-presence], which happens to combine with [+agentive] if and only if co-presence as to an AGENT is actually the case. Moreover, [+agentive] can be assigned to a NP via the comitative if and only if it agrees with the AGENT NP on the animacy parameter. If it fails in this respect, [+agentive] is ruled out and the resulting constructions are instances of CONFECTIVES, ORNATIVES etc. where only [+co-presence] remains in vigour. The INSTRUMENTAL, however, receives the feature [+agentive] because it is associated with the function of a secondary cause in the chain of events. A secondary cause presupposes a primary cause, i.e., there is a hierarchical order on which the INSTRUMENTAL ranks lower than the AGENT, but the TOOL with which a PATIENT is affected in an event is the entity which enters the scene immediately before the result of the action is achieved. It can therefore be interpreted as the ultimate cause. COMITATIVES are not necessarily associated with any causal effects, whereas INSTRUMENTALS usually are. In terms of agentivity, COMITATIVES and INSTRUMENTALS are rather different categories. One may wonder how they can associate in syncretistic patterns with each other nevertheless. As a matter of fact, COMITATIVES need INSTRUMENTALS to syncretise with [+agentive] categories – but even Comitative-Instrumentals do not fit in well with Ergatives, cf. above. 9.2. Coordination The [+co-presence] feature postulated for the Comitative co-occurs very often with a specification [Įegalitarian]. If the Į can be spelled out as a plus, the Comitative develops traits which render it similar to the coordinating conjunction AND and the like. Stassen (2000) examines a sample of 260 languages which distribute unevenly over two distinct language types, namely, WITH-languages vs. AND-languages. AND-languages are twice

Coordination 131

as many as WITH-languages171 and there are many cases where features of both types occur in a kind of mixture (Stassen 2000: 25). In these two language types, two different encoding strategies are employed, appropriately called Coordinate Strategy vs. Comitative Strategy. We quote Stassen’s definitions directly because there are some aspects which appear to be difficult to accommodate with what the conclusions of the previous section suggest. The Coordinate Strategy is described as encod[ing] the two participants in the construction by way of NPs with equal structural rank. Thus, the two NPs involved are not differentiated as to syntactic function; they have the same thematic role, and in languages in which such NPs receive case marking they will both have the same case. Typically, though not necessarily, the two NPs in such constructions can be seen to form a constituent, viz. a coordinate (plural or dual) NP (Stassen 2000: 7).

The comitative strategy, on the other hand, is the exact opposite of the coordinate strategy as under the Comitative Strategy the two participants in the event are morphosyntactically encoded as NPs of unequal structural rank. While one of the NPs can take any case role, the other NP is invariably encoded as the head of an oblique NP. A prototypical characteristic of comitative structures is that the two NPs involved are not part of the same constituent (Stassen 2000: 18).

Stassen emphasises the unequal structural rank of the NPs combined by the Comitative Strategy. We have argued for a certain kind of egalitarianism which characterises the participants whose NPs are combined by Comitatives. This egalitarianism is not absolute. It may apply only to a selection of features the participants have in common in a given situation. The higher the degree of egalitarianism, the more similar to AND the Comitative becomes.172 The Comitative is better suited for conveying egalitarian ideas, in a manner of speaking, than the Instrumental, which favours a strictly hierarchical interpretation. Nevertheless, there are some cases of InstrumentalAND syncretisms. These, however, show a peculiar distribution (statistics based on Stolz 1997a)173, cf. (B45). (B45)

Comitative and/or Instrumental syncretise with AND

Com + AND 49 55.68%

Com-Ins + AND 35 39.77%

Com + AND & Ins + AND 4 4.54%

Languages 88 100%

132 Syncretism More than half of the sample languages allow only for Comitative-AND syncretism – with most of them belonging to the A-type. InstrumentalAND syncretism is attested in less than 5% – likewise mostly A-type languages. Instrumental-AND syncretism always implies the existence of Comitative-AND syncretism in the same language, cf. (B46). (B46)

Implication Instrumental = AND Š Comitative = AND

The implication (B46) covers two varieties of constellations. There are languages which display the familiar syncretism of Comitative and Instrumental and add AND to the pattern. Besides this relatively frequent solution, there are also some languages which have two distinct expressions for AND, one of which is identical with the comitative marker, while the other syncretises with the Instrumental. Examples for both varieties can be found in Africa. Fon is a member of the Kwa phylum in West Africa. It is a C-type language, as it has a prepositional relator xá dedicated to Comitativity and a discontinuous circumpositional relator kpó (dó) … kpó (kpán) used for Comitatives and Instrumentals alike. Only the latter one can also be used as an NP-coordinator, cf. (B47). (B47) Fon (B47.1) ACCOMPANIMENT (Höftmann 1993: 138) lE! é yì sìn dùn gbè xá gbE~ t O! he go water fetch just X comrade PL ‘He has just gone with his comrades to fetch water.’ (B47.2) ACCOMPANIMENT (Höftmann 1993: 138) é yì sìn dùn gbè kpó (do) gbE~ t O! l´ 8 kpó (kpán) he go water fetch just X/Y comrade PL X/Y ‘He has just gone with his comrades to fetch water.’ (B47.3) TOOL (Höftmann 1993: 138) O! xò ví tO~ n kpó (dó) kpò kpó (kpán) tO! stick X/Y father DET hit child his X/Y ‘The father hit his child with a stick.’ (B47.4) Coordination (Höftmann 1993: 114) kpó nO~ dO~ n é kpó (dó) àsí tO~n he X/Y wife his X/Y live there ‘He and his wife are living there.’ Fon is an example of the pattern Com-Ins + AND in Table (B45). Not too far away from the region where Fon is spoken (Benin) is the home of the

Coordination 133

Chadic language Kwami (North-East Nigeria). There are two prepositional relators, namely, kán for the Comitative and káa for the Instrumental. Both relators can be used in contexts where they are interpreted as coordinating conjunctions, cf. (B48). (B48) Kwami (B48.1) ACCOMPANIMENT (Leger 1994: 266) Ëálí kán yìËì tì Kwáamì nì him to Kwami I go.PERF X ‘I went with him to Kwami.’ (B48.2) TOOL (Leger 1994: 267) káa wóojì yì shìÎÎùgó Y knife he poke:PERF ‘He cut with a knife.’ (B48.3) Coordination I (Leger 1994: 270) Ëíishè shàkká kúrmá kán púgúm deaf_mute X blind make beg:VN ‘The deaf-mute and the blind usually beg.’ (B48.4) Coordination II (Leger 1994: 145) Ëáffé Tátò káa Táatà yìnì kúÎí RECIP Tato Y Tata they reject:NARR ‘The Tato and Tata hated each other.’ It is not absolutely clear whether Kwami belongs to the A-type or the Ctype because káa at least occasionally comes to be used with NPs that are [+human].174 These occurrences outside the realm of Instrumentality fit in well with the following fact: Both prepositions are employed as ANDconjunctions. However, kán can be used for any kind of NP independent of animacy, whereas káa is restricted to [+human] NPs. According to Leger (1994: 270), coordinating káa is admissible only with pronominals. However, sentence (B48.4) speaks in favour of a more extended domain for the use of káa. What adds to this interpretation is the fact that káa may form a correlative conjunction káa … káa (equivalent to English ‘not only … but also’ or ‘both … and’), where full lexical NPs are common, cf. (B49). (B49)

Kwami (Leger 1994: 271) Ëìb-ù-dù példèm káa Kwáamì káa Fíikà Y Kwami Y Fika seize-VN-VEN path tì Yámàl from Yemen ‘Both the Kwami and the Fika moved over from Yemen.’

134 Syncretism Kwami probably represents a transition stage. The relator káa is presently undergoing functional changes. It may even be the case that all examples for the coexistence of Comitative-AND syncretism and Instrumental-AND syncretism in one and the same language reflect an ongoing restructuring of the system. Given this, the minority solution of (B45) may turn out to constitute no distinct type at all, i.e., the association of Instrumentality and coordination would become even weaker.175 As the above examples suggest, it is not always easy to tell Comitative (Instrumental) and coordinating AND apart. Why should they? There has been a long tradition of philosophically-minded discussions about the ANDComitative relationship. The main question asked by authors such as Hetzron (1973), Dyla (1988), Aissen (1989), Landmann (1989), Lasersohn (1990) and McNally (1993) is whether the various encoding strategies evidence a conceptual distinction of two types of groups of entities, namely, one with individuals on a par with each other and another for which inequality applies. In Stassen’s definitions of the two encoding strategies, there are three criteria (Stassen 2000: 21) which are helpful when it comes to deciding whether we are dealing with a group of the one or the other type: – same vs. different structural rank, – same constituent vs. different constituents, – non-singular agreement vs. singular agreement. Owing to the fact that Stassen (2000) provides some 160 examples from over 100 languages in his seminal article, we refrain from giving examples for the straightforward or so-called “pure” cases, which are amply documented by Stassen (2000).176 A case in point is Quechua.177 As the examples in (B41) above demonstrate, Quechua is a C-type language with a syncretistic ComitativeInstrumental marker -wan and a distinct Comitative -ntin- (-ndin- in other varieties, the final -n is often dropped). Both relators may also be used as coordinating devices. How do we tell their different functions apart? On the one hand, there is number agreement on the verb: As long as -wan marks the Comitative and/or Instrumental, the verb remains in the singular, provided the subject is singular as well. If coordination is intended, then plural agreement applies. However, this is not the whole story. In contradistinction to -wan, -ntin- alone often goes along with plural marking on the verb.178 Thus, differentiation of Comitative and AND via differential number agreement is not decisive. There is a solution to the problem: Under coordination, the relator is attached to each of the coordinated constituents

Coordination 135

(except the last one, in longer enumerations). This multiple marking applies not only to -ntin- but also to -wan, cf. (B50). In this way, Quechua does not have to resort to a distinct marker for coordination. (B50) Quechua (B50.1) Ayacuchano: coordinating -wan (Hartmann 1985: 64) Marcelina-wan Felipe-wan Birnacha Marcelina-X/Y Felipe-X/Y Bernabé papa-ta alla-chka-nku harvest-DUR-3PL potato-ACC ‘Marcelina, Felipe and Bernabé are harvesting potatoes.’ (B50.2) Equadoriano: coordinating -ndin (Catta Quelen 1987: 204) shuc cari huarmi-ndi huahua-cuna-ndi one man wife-X child-PL-X Villa-manta shamu-rca return-PAST.3SG Villa-ABL ‘A man, his wife and his children returned from Ibarra.’ Another Amerindian language, Classical Nahuatl, resorts to a rather elegant procedure, as the distinction of Comitative reading vs. conjunction reading of ƯhuƗn is simply a matter of number agreement on the verb. Only subject and object NPs ranking high on the animacy hierarchy trigger number agreement on the verb. With [+human] nouns, number agreement is the rule, though, perhaps not 100% compulsory. If two singular NPs are connected to each other by ƯhuƗn a finite verb marked for the singular is indicative of the Comitative (cf. [A8.1] above), whereas plural marking on the verb suggests a coordinated subject, cf. (B51). (B51)

Classical Nahuatl (Launey 1981: 125) mǀztla huƗl-la-z-quê in DET tomorrow hither-go-FUT-PL Ư-huƗn in Malin-tzin POSS.3SG-X DET Maria-DIM ‘Tomorrow Pedro and Maria will come.’

Pedro Pedro

In Swahili (cf. [B42] above), word order comes into play. If the NPs interconnected by the relator na are immediate syntactic neighbours in pre-verbal or post-verbal position (and if they are on similar ranks on the animacy hierarchy), then coordination is likely to apply, and thus plural agreement on the verb – for either subject or object – is triggered. If the NPs wind up on different sides relative to the verb, other readings are called for. In (B52.1), the two NPs are positioned post-verbally, which is the canonical position of non-

136 Syncretism subjects in Swahili. They form a coordinated object NP that is co-indexed on the verb by the appropriate object prefix -zi- for class 5 nouns in the plural. In (B52.2), the coordination comprises two NPs to the left of the verb, which is the usual subject position. Accordingly, the NPs trigger plural subject marking by the prefix wa- on both verb forms.179 (B52) Swahili (B52.1) Object (Möhlig and Heine 1999: 115) hapo mwanzo Mungu a-li-zi-umba right in_beginning God 3SG-PAST-3PL-create mbingu na nchi sky X earth ‘In the very beginning, God created heaven and earth.’ (B52.2) Subject (Möhlig and Heine 1999: 102) upepo na jua wa-li-kuwa wa-ki-gombana wind X sun 3PL-PAST-be 3PL-CONV-fight ‘The wind and the sun were having a fight.’ On top of that, the distinction of Comitative-Instrumental and AND in Welsh – a C-type language180 – is a matter of word order alone. In the official orthography, the preposition â and the conjunction a are distinguished by diacritics – but this is just a convention of the written register (Stolz 1998a). The unmarked word order in Welsh sentences is VSO with adverbial phrases winding up in the rightmost position, cf. (B53). (B53) Welsh (Jones and Thomas 1977: 35) (B53.1) ACCOMPANIMENT mae John yn dadlau am y be.3SG John in discuss about DET ‘John is discussing the book with Mary.’ (B53.2) Coordination mae John a Mair yn dadlau be.3SG John X/Y Mary in discuss ‘John and Mary are discussing the book.’

llyfr â Mair book X/Y Mary am y llyfr about DET book

Iconically, the two NPs John and Mair are adjacent when functioning as a coordinated subject. If Accompaniment is the intended meaning, then the Comitative phrase is at the far end of the sentence, i.e., distanced from the subject NP proper. What distinguishes the Welsh case from the Swahili one is the simple fact that for lexical subjects there is no number agreement on the verb, i.e., the finite verb is always in the singular, provided the subject is not the pronoun nhw ‘they’ (or any other plural pronoun).

Coordination 137

Maltese and Latvian are two bona fide AND-languages whose structures will be in focus in Part B. For the time being, we make do with giving one example for each of the two in order to show how closely Comitative and coordination are related. Maltese belongs to the A-type. Its comitative relator is the preposition ma, and NP and sentence coordination is the task of the conjunction u. Two NPs combined by u trigger plural agreement on the lexical verb either as subjects or as objects. NPs as parts of PPs headed by ma are of course oblique and thus cannot normally trigger agreement. However, there are exceptions, cf. (B54). (B54)

Maltese (Rajt III 241) fl-Italja kien-u marr-u l-Kumissjarju in:DET-Italy be.PERF-3PL go.PERF-3PL DET-commissioner Axisa li kien Gƫawdxi mas-Supretendent Calleja Axisa REL be.PERF Gozitan X:DET-superintendent Calleja ‘The Commissioner Axisa, who was a Gozitan, had gone to Italy with Superintendent Calleja.’

(B54) is awkward, as it shows plural forms of the finite verbs kienu marru, although the only legitimate candidate for the status of subject is a singular NP, namely, (i)l-Kummissjarju Axisa. The plural is motivated ad sensum by the co-presence of a second NP which indicates the companion of the commissioner, namely, (i)s-Supretendent Calleja. However, the companion NP is part of a PP intended as an adverbial of ACCOMPANIMENT. The only possible form of the finite verbs is the (masculine) singular kien marr ‘he had gone’. The actually attested plurals would require the use of u instead of ma’. As the proper place for u coincides with the one occupied by ma’ in example (B54), the two grammemes are functionally and syntactically very similar to each other.181 Additional factors facilitating the choice of the preposition in lieu of the conjunction are the postverbal position of both the subject NP and the adverbial PP, as well as the intercalation of the relative clause attribute to the subject NP, which expands the distance between the PP and the predicate nucleus, i.e., the second participant is mentioned so late in the sentence that the original plural agreement on the verb was already forgotten. As already mentioned in Part A the coordinating conjunction of Latvian is un. Two NPs conjoined by un trigger plural agreement when in subject function. The preposition ar, the syncretistic Comitative-Instrumental marker, triggers the accusative in the singular and the dative in the plural. Complement NPs of ar are of course in a case which cannot trigger agreement – but cf. (B55), taken from a story published on the internet.

138 Syncretism (B55)

Latvian (www.zz.lv/raktsi.php) beidzot sieva ar vƯr-u bija man-ACC be:PAST.3 finally woman:F X/Y atrad-uši kompromisu find-PTCPL.PL.M compromise ‘Finally, the woman and the man found a compromise.’

As mentioned in connection with example (A15) above, the use of the coordinating conjunction un is absolutely disfavoured in contexts of high(est) animacy. (B55) thus reflects common idiomatic Latvian. What is surprising from a purely grammatical point of view is the fact that again, morphosyntax is overruled by semantics because the agreement markers on the participle have nothing to agree with formally. Sieva, a feminine singular noun, is the legitimate subject NP of the sentence and would require the adequate participle form atradusi. The masculine plural is commonly used in Latvian to refer to groups whose members belong to different genders. The PP ar vƯru, together with the NP sieva, can be interpreted as forming such a mixed-gender group. However, only one NP, sieva, is in the appropriate case determining agreement, namely, the nominative. In contradistinction to the Maltese example in (B53), the remarkable agreement behaviour in (B55) is nothing out of the ordinary for Latvian native speakers. Quite the contrary, agreement ad sensum is the rule, provided the two NPs are [+human]/ [+animate]. Avoiding plural agreement is stylistically highly marked (and considered unLatvian). If animacy is low however plural agreement becomes less acceptable and native speakers tend to use the conjunction un more frequently.182 Obviously, there is no need for distinct primary markers for coordination as long as the morphosyntax of the Comitative allows for some flexibility.183 Coordination is in many languages a secondary distinction belonging to the functional domain of their Comitatives. Due to their prototypical semantics, Comitatives are far better suited for the purpose of group formation than Instrumentals. Thus, Comitatives have a predilection for syncretising with categories that usually fail to appear in the inventories of thematic roles, semantic cases, etc. Both Comitatives and AND function as relators for co-participant relations, i.e., not only for co-agents but for any kind of co-role. In contradistinction to Comitatives, distinct AND is more strictly bound to the requirement of feature egalitarianism of the coordinated NPs. Thus, good exemplars of AND-languages avoid combinations of coordinated NPs which do not share a sufficient number of features, be it animacy-based ones or other. In WITH-languages and generally for Comitatives and Instrumentals, the restrictions are much more lax and thus allow

Coordination 139

for an array of subcategories in the realms of Comitativity and Instrumentality (Stolz 1998a: 111–112). To complement our tour d’horizon of the relationships between ANDconjunctions and the relators of Comitativity and/or Instrumentality, we present data from the Amerindian B-type language Aymara. In Aymara, there is a marker -mpi which fulfils the functions of a syncretistic Comitative-Instrumental marker. At the same time, -mpi is also employed as coordinator, not only of NPs, but also of clauses and sentences.184 Furthermore, this grammeme also functions as the case form for the causee in causative constructions185, cf. (B56). (B56) Aymara (B56.1) Accompaniment (Hardman, Vásquez, and Yapita 1988: 208) k”iti-mpi-s sara-ñani go-1PL:FUT who-X/Y-INTERR ‘Who will we go with?’ (B56.2) Instrumental (Hardman, Vásquez, and Yapita 1988: 224) k”iti-mpi-s lapisa-mpi-x qillqa-y-ta write-CAUS-2SG.3SG who-X/Y-INTERR pencil-X/Y-TOP ‘Who have you made write with a pencil? (B56.3) Coordination (Hardman, Vásquez, and Yapita 1988: 255) Arjintina-mpi-r Wauliwya-mpi-ru-w sar-i go-3SG.PERF Argentina-X/Y-ALL Bolivia-X/Y-ALL-AFF ‘He has gone to Argentina and Bolivia.’ (B56.4) Causative (Hardman, Vásquez, and Yapita 1988: 108) nayaw jupa-r jupa-mp t’ant’ chura-y-ta 3SG-X/Y water give-CAUS-1SG I 3SG-ALL ‘I make him give water to her.’ As examples (B56.2)–(B56.3) clearly show, speakers of Aymara have no difficulty with multiple occurrences of -mpi in one and the same sentence, no matter whether they have different functions (Causee vs. Instrument) or identical ones (coordination, cf. also Quechua [B50] and Kwami [B49]). In this way, Aymara, in contradistinction to Swahili, tolerates a high degree of homonymy, just like Dehu (cf. footnote 167). Nevertheless, the formal identity of expressions for [+agentive] and coordination, including Comitative-Instrumental syncretism, is far from being a frequent pattern. A possible functional explanation for its crosslinguistic rarity may be the danger of too much blurring of the differences between hierarchical and egalitarian descriptions of states of affairs.

140 Syncretism 9.3. Spatial relations In the above, we have repeatedly alluded to the concept of same space. The Comitative is sometimes understood as being based on a concept which comprises two entities which are co-present in the same space (including metaphorical readings of space) (Stolz 1993: 43).186 Given this spatial component of the conceptual basis, it is surprising that Comitative-Locative syncretism is not more prominent. According to the statistics in (B21)– (B24), Comitatives and Locatives associate rather seldom, whereas Instrumentals and Locatives attract one another. The pattern InstrumentalLocative is attested eight times as often as Comitative-Locative. Locatives are second best in the hierarchy of syncretistic partners of the Instrumental, i.e., Locatives compete with Comitatives for the status of syncretistic partner of the Instrumental. The label Locative, however, is a cover term for a variety of spatial concepts, i.e., there is some heterogeneity involved: The Locative stands for static and dynamic spatial concepts as well. Therefore, we encounter Locatives (Inessive, Adessive, Superessive, etc.) proper alongside Latives187 (Ablative, Allative, Elative, etc.), which freely syncretise with the Instrumental. This multitude of possible partners is not a disadvantage: As with [+agentive] categories discussed above in Section 9.1, we are dealing with a certain prominent semantic feature – not a concrete grammatical category – that blocks or attracts associations between categories. The appropriate form to represent this feature could be [+locative]. In the Indo-Aryan language Oriya (an A-type AND-language), there is an inflectional locative -re with a variety of functions, among which we find spatial ones like Locative and Allative on the one side and Instrumental on the other. Comitative functions are encoded by a variety of mostly denominal postpositions, of which s1 hit1 is one of the more prominent, cf. (B57) [cf. also (A77) above]. (B57) Oriya (B57.1) Locative (Neukom and Patnaik 2003: 63) °thik ehi s1 m 1 y 1 -re mo-r1 swami just this time-Y 1SG.GEN husband as-i-p1 h1 nc-il-e come-CONV-arrive-PAST-3 ‘Just at this moment, my husband arrived.’

Spatial relations 141

(B57.2) Instrumental (Neukom and Patnaik 2003: 64) nali k1 l 1 m 1-re lekh-uch-i se go° ti-e write-PROG-3SG she CL-INDEF red pen-Y ‘She is writing with a red pen.’ (B57.3) Comitative (Neukom and Patnaik 2003. 326) s1 hit1 k1 l 1 h 1 sang1 sathi-man1-nk1 friend-PL-OBL X quarrel k1 r-1 tyag1 giving_up do-2PL.IMP ‘Give up quarrelling with your friends.’ In French Guyana (Cayenne), in the country’s hinterland near the border with Brazil there live the Wayampi, whose language forms part of the TupiGuarani phylum. Wayampi is an A-type language.188 The postposition l8 (w8) is used for the purpose of encoding the Comitative, whereas (pu)p8 serves the expression of both the Instrumental and the Locative, cf. (B58). (B58) Wayampi (B58.1) Accompaniment (Grenand 1980: 62) 1 lo-1-ta k1 kGtG yãyã l8 w 8 1DU-go-FUT pyre towards sister X ‘I go with my older sister to the pyre.’ (B58.2) Instrumental (Grenand 1980: 62) 8 l-apama malia pup8 a-piãpi 1SG-smooth my-bow knife Y ‘I am smoothing my bow with a knife.’ (B58.3) Locative (Grenand 1980: 63) 1 - i-m1 - 8 na Wayapuku pup8 Wayãpi Wayãpi 3-REFL-make-home Oyapock Y ‘The Wayãpi live on the Oyapock.’ The same preference for Instrumental-Locative syncretism can be observed in a wide variety of other A-type languages, some of which have already been mentioned above, namely, Guaraní, Hindi, Swahili, Samoan and Finnish. For brevity’s sake, we do not repeat the examples of their Comitatives and Instrumentals but refer back to (A44), (B20), (B42), (A68), and (A57) instead. In Guaraní (an A-type AND-language), the relator -pe is not only typically employed for Instrumentality but also – and probably more basically – for [+locative] categories, cf. (B59).

142 Syncretism (B59)

Guaraní: Instrumental = Locative (Gauto Bejarano 1990: 110) aƭ che kotý-pe be:1SG my room-Y ‘I am in my room.’ In Hindi (likewise an A-type AND-language) (B60), the Instrumental is syncretistic with the Ablative, as both are encoded by the postposition se.189

(B60)

Hindi: Ablative = Instrumental (Sharma and Vermeer 1972: 24) vah chat se gir paDŽƗ he roof Y fall from ‘He fell down from the roof.’

In Swahili (an A-type WITH-language), the relator kwa does not only encode instrumentality but also serves as a Locative [±dynamic] marker, cf. (B61). (B61)

Swahili: Locative = Instrumental (Loogman 1965: 286) ni-li-kuwa kwa Juma Y Juma 1SG-PERF-be ‘I was in the house of Juma.’ In Samoan (equally an A-type WITH-language), the preposition i is multifunctional. Besides the Instrumental, i also encodes various spatial concepts including locative proper and directionals, cf. (B62). (B62)

Samoan: Locative = Instrumental (Mosel and Hovdhaugen 1992: 144) sƗ nonofo i Papa PAST live:PL Y Papa ‘They lived in Papa.’

Last but not least in this short survey is Finnish (an A-type AND-language). The Instrumental belongs to the functional domain of the multipurpose inflectional adessive. The spatial readings of the adessive include the superessive190, cf. (B63). (B63)

Finnish: Adessive = Instrumental (Karlsson 1978: 124) kupit ovat pöydä-llä cup:PL be:3PL table-Y ‘The cups are on the table.’

B-type languages are less often represented among those languages which display the syncretistic patterns under scrutiny. Austronesian Malagasy (Btype AND-language) is a case in point. The preposition àmin’- serves various purposes, the most important of which are Comitative-Instrumental and

Spatial relations 143

Locative, cf. (B64). The accents in (B64) are not part of the standard orthography. (B64) Malagasy (B64.1) Accompaniment (Rasoloson 1997: 119) hiràka amin-areo izahày rahampìtso X/Y-2PL we tomorrow go_along:FUT ‘We will go with you tomorrow.’ (B64.2) Instrumental (Rasoloson 1997: 118) manòrata àmin’-ny pensìly hàzo ny mpiànatra X/Y-DET pencil CL.tree DET pupil:PL write:PL ‘The pupils write with a pencil.’ (B64.3) Locative (Rasoloson 1997: 117) hamàngy òlona àny àmin’-ny hôpitàly àho somebody there X/Y-DET hospital I visit:FUT ‘I will visit someone in the hospital.’ Much more rarely do we encounter cases of Comitative-Locative syncretism in A-type languages, i.e., without the participation of the Instrumental in the same pattern. There are, however, several such exceptions among Caucasian languages, as in Bezhta, a North-East Caucasian A-type ANDlanguage.191 Bezhta has an inflectional instrumental ending in -d and a comitative in -Ƨor/-Ƨör which also encodes spatial proximity192, cf. (B65). (B65) Bezhta (B65.1) Instrumental (Kibrik and Testelets 2004: 235) do õgi-d hudo büþþä I axe-Y wood chop ‘I chop wood with an axe.’ (B65.2) Accompaniment (Kibrik and Testelets 2004: 238) do eá’eþ di:la halmaƧli-Ƨor friend-X I go I:GEN ‘I am walking with my friend.’ (B65.3) Locative (Kibrik and Testelets 2004: 238) do märä-Ƨör eá’eþ I mountain-X go ‘I am going into the mountains.’ Owing to Palancar’s (2002: 205–208) demonstration of the many spatial origins Agent markers display, it is necessary to look from a broader perspective at the preference of Locatives to syncretise with Instrumentals and not with Comitatives. As a matter of fact, Agent markers and Locatives have the

144 Syncretism same taste when it comes to syncretising, as both exhibit a striking leaning towards the Instrumental. Thus it is hardly surprising that we find many instances of syncretistic patterns which comprise the three categories at the same time. Australian languages provide us with ample evidence for Instrumental-Locative-[+agentive] syncretism.193 Blake (1977: 63–67) surveys 115 Australian languages of all known phyla. According to his description, the following picture emerges, cf. (B66). Note that for all the patterns listed in (B66), the abbreviation LOC stands for the static Locative (which in turn may or may not be syncretistic with one of the dynamic Latives).194 (B66) Syncretism in Australian languages Ergative alignment No. Pattern 1 ERG zINS z LOC ERG = INS z LOC 2 ERG z INS = LOC 3 ERG = INS = LOC 4 ERG = LOC z INS 5 Sub-total ERG Accusative alignment 6 7 Sub-total ACC Ins = LOC Ins = ERG

INS z LOC INS = LOC

Languages 12 45 9 13 1 80

35 0 35 22 58

% 15% 56,25% 11.25% 16.25% 1.25% 100%

100% 0% 100% 19.13% 50.43%

In Australian languages, Instrumental-Ergative syncretism is doubtlessly the preferred option. If we include those cases in which also the Locative partakes, Instrumental-Ergative syncretism occurs in 72.5% of all syncretistic patterns attested in languages with Ergative alignment (Blake 1977: 44– 45). (B67) contains three sentences from Ngaliwuru – a relative of Djamindjung. The suffix -ni is polysemous, encoding not only the Ergative but also the Instrumental and the Locative. (B67) Ngaliwuru (B67.1) Ergative (Hoddinott and Kofod 1976: 398) djalig bibi-ni nuwina-ni mun ga-ni-n awu father-Y his-Y look 3SG-3SG.OBJ-see.PRES baby ‘His father looked at the baby.’

Spatial relations 145

(B67.2) Instrumental (Hoddinott and Kofod 1976: 399) gandi-ni djumbul na-manu stick-Y man 1SG-hit ‘I hit the man with a stick.’ (B67.3) Locative (Hoddinott and Kofod 1976: 400) n -wu-njawu yagbali-ni ya-yadj wundju NEG.FUT-be.3SG if 1SG-FUT-see camp-Y ‘If he is in camp I would like to see him.’ Discounting the patterns that the Locative participates in, InstrumentalErgative syncretism still accounts for a robust majority of more than 56%. Even if we widen the scope to include accusative languages as well, the pattern remains the most important with slightly more than half of the tokens. What strikes the eye most, however, is the fact that, with one notable exception, the Locative syncretises with the Ergative only if the Instrumental joins the pattern too.196 On top of that, not only may the Instrumental syncretise with the Ergative without taking the Locative along, but also, Instrumental-Locative syncretism is unattested in accusative languages. Similarly, Instrumental-Locative syncretism excluding the Ergative in ergative languages is only a minority option, amounting to only a fifth of Instrumental-Ergative syncretism in the same language type. Similar patterns can also be observed frequently in languages spoken elsewhere, be they ergative or accusative languages. We have already mentioned Hindi – a split ergative language which displays Instrumentalpassive Agent-Locative syncretism because the Instrumental relator is identical not only with the Agent marker in passive constructions but also with the one used for the Ablative, cf. (B60) above. The Dravidian language Kannada (A-type AND-language, accusative alignment) reflects the same distinctions as Hindi – a fact which is probably the product of long term contacts between Indo-Aryan and Dravidian languages on the Indian subcontinent. There are several postpositions which can be employed to mark the Comitative: samgada, oºane, omdige, knjºa (Jensen 1969: 40–41), whereas the relator of the Instrumental -imda has affix status. This case exponent has a functional domain which comprises the Instrumental proper: kattiy-imda hode ‘to strike with the sword’, the Ablative: mamdirad-imda ‘from/out of the temple’ and the passive Agent: sarpagaƺ-imda nƗsavƗdaru ‘they were killed by the snakes’ (Jensen 1969: 134–136). The nowadays extinct Tungusic language Manchu likewise employed the pattern Locative [±dynamic]-Instrumental-passive Agent: The clitic-like postposition -de fulfilled all these functions while the Comitative was mainly

146 Syncretism encoded by adverbials such as, e.g., emgi or sasa (Haenisch 1986: 42 and 49). However, this pattern is by no means restricted to languages of the Asian continent. It also recurs elsewhere – and sometimes with slight modifications. The unclassified Warao of Venezuela, an A-type AND-languages again, shows a peculiar realisation of the pattern, cf. (B68). (B68) Warao (B68.1) Accompaniment (Vaquero 1965: 103) Teté a rani yata naruae Teté his mother X go_away:PAST ‘Teté went away with his mother.’ (B68.2) Locative (Vaquero 1965: 103) Mérida yata naukuna Mérida X go:POT ‘He would go to Merida’ (B68.3) Instrument (Vaquero 1965: 102) jatabu isía nae arrow Y kill:PAST ‘He killed it with an arrow.’ (B68.4) Passive Agent (Vaquero 1965: 102) ka ina kokotuka Dioso aisía nonaja tanae our world entire God Y create PAST ‘Our entire world has been created by God.’ (B68.5) Locative (Vaquero 1965: 102) jana sanuka isía yurukí canyon little Y enter:ADHORT ‘Let us enter by the small canyon’ In Warao, Comitative and Instrumental receive distinct encodings197, namely, the postpositions (y)ata and (a)isia, respectively.198 As expected, the Instrumental syncretises with the passive Agent and also with one of the many spatial categories. What is surprising, though, is the fact that the Comitative also syncretises with a spatial concept. In the case of the Instrumental, the Perlative is the [+locative] partner while the Comitative combines with a spatial component which covers Allative, Locative and Superessive. Thus, in spite of all similarities, history is repeating itself: Comitative and Instrumental symptomatically go different ways. No matter whether the patterns of syncretism include the Comitative or not, they all violate the principles of Role and Reference Grammar, according to which only micro-roles of the same macro-role plus some neigh-

Possession 147

bourhood relationships across the boundaries allow for syncretistic encoding of categories. (A33) suggests that both Comitative and Instrumental are Actor roles which cannot syncretise with spatial categories which typically belong to the Undergoer macro-role. The same holds for Croft’s (1991) distinction of antecedent and subsequent roles (cf. [A79]), if we accept Luraghi’s (2001a: 37) reading of Croft’s argumentation: According to her, spatial concepts are associated with the status of subsequent roles. Therefore, syncretism between Comitative and Instrumental on the one hand and Locative, Direction Goal is disallowed, on the other, as the respective patterns would automatically cross the borderline between the two major groups of roles. However, violations of both models of syncretism are so frequent that we cannot discard them as unimportant exceptions. The high frequency at which spatial categories syncretise, especially with the Instrumental, are indicative of a very basic shortcoming of the extant models: Both Van Valin (1993) and Croft (1991) rely heavily on two-dimensional linearizations of categories which automatically block associations between categories which happen to wind up at different extremes of the diagram. Instrumental-Locative syncretism, however, calls for a multidimensional approach, cf. Section 9.6. 9.4. Possession Possession199 is another area that attests to the differential behaviour of Comitatives and Instrumentals. Moreover, possession is also a functional subdomain which is hardly ever is addressed in studies of semantic cases or thematic roles – a notable exception is Lehmann, Shin, and Verhoeven (2000). For practical purposes, we focus our discussion of pertinent data on cases of predicative possession (= possession that is predicated via a VP-like structure) (Heine 1997: 183). We acknowledge that adnominal/attributive possession is also an important aspect of Comitatives and Instrumentals (Heine, Claudi, and Hünnemeyer 1991: 166). Accordingly, it is given due consideration in Part C.200 Suffice it here that the vast majority of Australian languages make use of a proprietive case which is distinct from the expressions of Instrumental and Comitative (cf. the examples from Arrarnta above [A38.2]).201 The proprietive, among other things, covers the function of encoding adnominal possession. The presence of distinct Proprietives often excludes the use of Comitatives/Instrumentals inside the NP. Outside of Australia, Proprietives are also attested, though perhaps not in an equally dense geolinguistic distribution.

148 Syncretism Expressions of predicative Possession are syncretistic partners of Comitative or Instrumental, mostly in A-type languages, which cover almost 75% of all the cases reported by Stolz (1997a) (cf. [B25]).202 Furthermore, the Comitative outnumbers the Instrumental as a syncretistic partner of Possession by a ratio of 3-to-1 (cf. [B21]). This predominance of the Comitative is in line with the importance Heine (1997: 53–57) ascribes to the so-called Companion (or Accompaniment) Schema. Comitative-Possession syncretism is a pattern whose geolinguistic distribution displays a very strong African component (cf. [B26]), although it is also remarkably frequent in Oceania. Comitative-Possession syncretism occurs elsewhere as well (Heine 1997: 55), with the exception of Asia, where it is clearly underrepresented. For reasons of data accessibility, we choose African languages for our documentation (followed by some cursory remarks on the European situation). We already know that Swahili is an A-type WITH-language whose Comitative relator is, somewhat unexpected, also employed as a marker of the passive Agent. However, this is not the whole story, because na also functions as the primary marker for predicative possession, cf. (B69). (B69) Swahili (B69.1) Present (Loogman 1965: 285) ni-na kisu knife 1SG-X ‘I have a knife.’ (B69.2) Perfect (Loogman 1965: 418) a-li-kuwa na mali nyingi X many possession 3SG-PERF-be ‘He was very rich (lit. he had/was with many possessions).’ In possessive constructions of the predicative kind, na behaves like a verb in the present tense. In (B69.1), na is the host of the pronominal subject prefix ni- (identical to the one used on inflected verbs). Swahili does indeed have a copula verb in the present tense (affirmative ni, negative si), which, however, may be omitted “where the meaning is not affected by the omission” (Loogman 1965: 230). As (B69.2) demonstrates, in tenses other than the present, the verb kuwa ‘to be’ is compulsory, and the resulting construction with na can be read as ‘to be with’. Thus, possessive na has not yet developed into a full blown verb, although the first stages of reanalysis have been reached in the present tense. It is therefore still legitimate to consider the uses na is put to in (B69) to be synchronically associated with

Possession 149

those discussed in the previous chapters. Moreover, the Instrumental relator kwa cannot replace na in predicative Possession constructions. Kanuri, a Nilo-Saharan language spoken around Lake Chad, belongs to our A-type. There is evidence that Kanuri used to be a WITH-language in former times.203 Synchronically, we classify it as an AND-language. The Instrumental relator -(la)n also encodes various [+locative] categories, while the associative -a is the expression used for Comitativity and inalienable Possession, cf. (B70). Note that the associative affix can be used also for adnominal possession. (B70) Kanuri (B70.1) Instrumental (Cyffer 1991: 109) dâ jána-lan kámgada meat knife-Y cut:3PL ‘They cut the meat with a knife.’ (B70.2) Locative (Cyffer 1991: 109) fáto-lan nápsána house-Y sit:3PL ‘They sit in the house.’ (B70.3) Accompaniment (Cyffer 1991: 97) íshin Áli kámunz6-a Ali wife-X comes ‘Ali is coming with his wife.’ (B70.4) Inalienable Possession (Cyffer 1991: 97) sémo kurúwu-a k6 ri ád6 dog this ear long-X ‘This dog has long ears.’ While Swahili na covers the whole range of possession, Kanuri makes a formal distinction between inalienable and alienable possession. Only the former belongs to the functional domain of the associative -a. Alienable possession is expressed by constructions involving a spatial component, cf. (B71). (B71)

Kanuri: alienable Possession (Cyffer 1991: 39) nâ-nyî-n rédiyo mbéjí radio EXI house-POSS.1SG-Y ‘I have a radio (lit. there is a radio at my place).’

The relator which serves the purposes of Comitativity is also employed to denote the more intimate kind of possessive relation, whereas [+locative] comes into play only where possession is of a much looser nature. Thus, Kanuri is not fully in line with the hypothesis204, that the Companion Schema is

150 Syncretism more likely to express physical and temporary or, more generally, alienable possession rather than inalienable possession (Heine 1997: 92–93).

In the neighbourhood of Kanuri, the Chadic language Hausa is widely spoken. Hausa is a B-type WITH-language. The relator dà translates to English ‘with’, ‘by’, ‘and’; it also has some temporal and spatial functions. With verbs in the so-called extended form of the Efferential, dà also introduces the Patient (Wolff 1993: 374–375). Furthermore, dà is also employed to encode possession, cf. (B72). (B72) Hausa (B72.1) Accompaniment (Wolff 1993: 424) sun zoo dà wânsù they come X/Y brother ‘They have come with their brother.’ (B72.2) Coordination (Wolff 1993: 458)205 goomà dà biyu shaà biyu nèe ten X/Y two twelve be ‘10 + 2 = 12’ (B72.3) Instrument (Robinson 1959: 55) ya sare shi da takobi he cut him X/Y sword ‘He cut him with a sword.’ (B72.4) Possession (Wolff 1993: 440) munàa dà ku4 ìi we X/Y money ‘We have money.’ Hausa does not distinguish alienable from inalienable possession in predicative constructions. It is tempting to ascribe this neutralisation to the fact that the grammeme dà is syncretistic as to Comitative and Instrumental. Kanuri -a is specialised for inalienable possession – and this specialisation fits in with its Comitative functions. With a view to proving that Comitative and Instrumental components determine the functional potential of a given relator in the realm of possession, we have to find an example of an Instrumental marker being specialised for the expression of alienable possession. As a matter of fact, such examples are not easily detected. In the nearby Mande phylum, we encounter a languages which, when adequately reanalysed, may yield something coming close to what we are looking for. Bamanankan, spoken in Mali and the surrounding countries, is C-type WITH-language (but cf. below).206 There are two candidates for the

Possession 151

status of Comitative relator: On the one hand, (a)ni, which can be translated into English either as ‘and’ or as ‘with’, and on the other, the discontinuous formation ni…ye.207 For Kastenholz (1989: 161 note 26), simple (a)ni is clearly a conjunctional coordinator, but cf. below. The ComitativeInstrumental relator, the postposition fè, has a wide range of functions from Locative to Comitative and Instrumental and passive Agent to alienable Possession, cf. (B73).208 (B73) Bamanankan (B73.1) Coordination (Brauner 1974: 25) cè ni muso do man X woman be ‘This is a man and a woman.’ (B73.2) Accompaniment I (Kastenholz 1989: 161) Bákari bé tága kùngo lá Músa fè go jungle in Musa X/Y Bakari AFF ‘Bakari is going into the jungle with Musa.’ (B73.3) Accompaniment II (Kastenholz 1989: 161) Bákari bé tága kùngo lá ní Músa go jungle in X Musa Bakari AFF ‘Bakari is going into the jungle with Musa.’ (B73.4) Passive Agent (Brauner 1974: 89) liwru kalanna an fè Y book read:PTCPL we ‘The book was read by us.’ (B73.5) Locative (Brauner 1974: 45) a bè a fa fè he be his father Y ‘He is at his father’s place.’ (B73.6) Alienable Possession (Brauner 1974: 46) so bè ne fè house be I Y ‘I have a house (lit. there is a house at me).’

yé X

The difference between the two types of Accompaniment is explained by Kastenholz (1989: 161) in terms of the autonomy of the Companion. Example (B73.3) would thus imply that the Companion Musa is taken along by Bakari, who in turn is the more active participant. The logical consequence of Kastenholz’s argument would be that example (B73.2) depicts the situation as one of egalitarian status of the participants. However, Kastenholz (1989:

152 Syncretism 161 note 26) hastens to add that it is absolutely preferable to use a construction of the following type (B74) instead of (B73.2) if the intended meaning implies autonomy of the two participants. (B74)

Bamanankan (Kastenholz 1989: 161 note 26) Bákari àni Músa bé tága kùngo AFF go jungle Bakari X Musa ‘Bakari and Musa go into the jungle.’

lá in

He goes on to say that an equally good translation of (B74) requires mit ‘with’ instead of und ‘and’. Moreover, it is said that even typical Instruments (Tools, Means of Transportation, etc.) could be encoded by (a)ni, provided they are understood as relatively autonomous entities. The discontinuous relator ní…yé, which is also used for confectives and combinations, does not allow for a reading that implies autonomy. If we accept Kastenholz’s line of reasoning, we are left with the problem what to make of (B73.2). We base our solution on Kastenholz’s footnote 26: The use of the postposition fè does not normally invite a Comitative reading. We assume that it can be employed with such a reading only under very special circumstances and thus may be counted out as a primary marker of the Comitative. To our mind, fè is the primary marker of Instrumentality.209 As example (B73.6) demonstrates, fè is also used to encode alienable Possession. For its inalienable counterpart, the locative postposition na ‘in’ is used, cf. (B75). (B75)

Bamanankan: inalienable Possession (Brauner 1974: 46) bana bè ne na sickness be I in ‘I am ill (lit. there is a sickness in me).’

As inalienably possessed items, we mainly find physico-mental states, etc. Concrete objects fall indiscriminately into the class of alienably possessed items. In this way, the distinction of alienable vs. inalienable yields two rather different systems in Kanuri and Bamanankan. However, the exact location of the borderline between what counts as alienable and what is inalienable in an individual language is not important to our present purpose. What is decisive here is the fact that fè is used in combination with possessees which are under the control of the possessor and therefore pass as alienable.210 Being under the control of a higher-level participant applies of course also to the prototypical instrument! The control criterion notwithstanding, clear cases of a stronger association of Instrumentality and Alienability are scarce. Of the few instances of Posses-

Possession 153

sion mentioned in the previous chapters, one attests to an association of Comitativity and Possession in general (cf. Mongolian [A45.2–3]). The Kalmykian case (cf. [A46.4–5]) is more complicated, because there are two distinct haveconstructions, one requiring the inflectional dative, the other the so-called associative (Benzing 1985: 56). The functional motivation for the presence of two distinct constructions is still unclear, although the opposition of permanent (= associative) vs. temporary/actual (= dative) possession seems to be involved to some extent. Irrespective of whether this analysis is actually true, the Instrumental itself is excluded from predicative Possession in Kalmyk. Elsewhere in Europe, we encounter a number of languages with different genetic backgrounds which employ Comitatives and/or Instrumentals in predicative Possession. In this section, therefore, we make do with a fragmentary and simplifying outlook on the European situation (Stolz 2001b) – languages not mentioned in this section will be discussed in due order in Part C. While it is true that Finnish (A58.3) employs the same relator for Instrumental and predicative Possession, Sulkala and Karjalainen (1992: 178–179) claim that there is no compelling evidence for a fully established alienability distinction in Finnish predicative Possession. However, the authors admit that frequency may give a hint as to which possessees have a leaning towards which construction. For inalienable possessed items, they identify genitive marking on the possessor as the slightly more frequent pattern, whereas adessive marking on the possessor is typical of alienable possessed items. These rules hold as long as the possessor is [+animate]; otherwise the inessive is used on the possessor. The inessive may also be used to mark the possessor (= experiencer/location) of physical and mental states, though this is somewhat rare. As there are only few other A-type languages in Europe, it is not easy to verify this covert tendency suggested by the Finnish data. In the Afro-Asiatic A-type AND-language Maltese (cf. Part C), the preposition bi is used both for Instrumentality and – only optionally – also for actual/temporary possession (including mental and physical states), whereas the Comitative marker ma’ is never associated with possession. Bi forms a paradigm with other expressions of predicative possession, namely, the spatial preposition gƫand ‘at one’s place’, diachronic univerbations of the copula and the benefactive preposition li ‘for’ in non-present contexts. These competitors are general markers of predicative possession which may neutralise any other low-level distinction. Note also that bi is also employed in the realm of Comitativity if the Companion is devoid of control. In C-type languages such as Welsh and Icelandic (cf. Part C), the picture is variegated: The C-type WITH-language

154 Syncretism Welsh (Stolz 1998a) uses the Comitative relator gyda(g) in predicative constructions (alongside the preposition gan) but the syncretistic ComitativeInstrumental a in adnominal constructions. Physical and mental states normally fall within the functional domain of spatial prepositions such as, e.g., ar ‘on’. In the C-type AND-language Icelandic, the preposition með ‘with’ governs two different inflectional cases: – The dative, if an Instrumental reading is intended or a Companion with a high degree of autonomy is involved; – the accusative, if it serves the expression of either temporary possession (including physical and mental states), abstract and body part possession or Accompaniment by a companion with a low degree of control. B-type languages are likewise heterogeneous. In Irish, an AND-language and a distant relative of Welsh, the syncretistic Comitative-Instrumental relator le is employed for alienable Possession/legal ownership of the BELONG-type (Heine 1997: 29–33), whereas inalienable Possession requires markers of spatial concepts – mainly the prepositions ag ‘at, on’ and ar ‘on’. Ag is also the general marker of predicative possession of the HAVE-type. The Romance AND-language Portuguese uses its syncretistic Comitative-Instrumental preposition com also in constructions of predicative possession. To what extent its use is still optional remains to be investigated. Suffice it to say that together with the copula estar, which encodes temporary states, com combines with possessees which cannot be controlled by the possessor. Estar com covers many instances of temporary/actual possession, including mental and physical states, whereas Portuguese ter, the direct translation equivalent of English to have, is used as a neutralising general verb of HAVING. All this yields an interesting picture. First of all, we learn that predicative possessive constructions based on Comitatives and/or Instrumentals are never the sole expressions for predicative possession in European languages. Where they are employed they tend to be associated with temporary/actual possession, especially of uncontrollable mental and physical states – a pattern which is familiar from some of the African languages discussed above. However, this tendency does not account for the full range of variation, cf. (B76).

Possession 155

(B76) Type

Comitative/Instrumental-based predicative possession constructions Case

Language Other

A C B

INS INS COM COM COM-INS COM-INS

Finnish Maltese Icelandic Welsh Irish Portuguese

Permanent Body parts X X

Temporary X X X

X X X

As the grey shading in Table (B76) suggests, temporary possession is indeed an important domain. However, temporary possession allows for different possessees, namely, those which are under the control of the possessor (concrete objects) and those which are not (mental and physical states). The latter ones have something in common with inalienable possessees such as, e.g., body parts, namely the fact that they are not subject to control on the part of the possessor. This explains the extension of the grey shading into the rubric of permanent possession in Maltese and Icelandic. One may, of course, claim that the use of Comitatives and Instrumentals in predicative possession is motivated by the fact that the temporarily possessed items (and body parts, for that matter) are physically close to the possessor. This is certainly true of many of the above cases. Yet Welsh and Irish show that this apparent rule cannot be generalised to all instances. Moreover, no pattern emerges from (B76) which would allow us to ascribe any determining power to type-membership of a given language or to functional specialisations of a given relator. In contradistinction to the preceding three syncretistic partners, the differences between Comitatives and Instrumentals are less pronounced with predicative possession. To be precise: Comitatives are much more closely associated with possession than Instrumentals. Only where there is evidence for the latter to be used with possessive functions is it correct to say that their behaviour is not much different from the one observed for Comitatives. If Instrumentals occur in predicative possession at all, they may be used for the same categories as Comitatives. However, there is a weakly developed preference of possessive Instrumentals to be used in constructions of temporary/actual possession especially physical and mental states.

156 Syncretism 9.5. Semantic map Disregarding other attested patterns211 for the time being, we are now in a position to make use of the above evidence for the purpose of constructing a (nuclear version of a) semantic map for Comitatives and Instrumentals. We use the term “semantic map” here only for the sake of convenience, since it is already well established in linguistic discourse. Otherwise we share Haspelmath’s (2003: 219) reservations against the term itself and his scepticism as to the appropriateness of alternatives such as “cognitive map” or “mental map”. We are not entirely convinced that “function(al) map” would be a good replacement either. The internal organisation of semantic maps and other technical and methodological aspects are discussed at length by van der Auwera and Plungian (1998) and Haspelmath (2003). Simplifying, semantic maps represent an attempt at a topographically-inspired visualisation of relative similarity or closeness of concepts/categories/functions. The greater the distance between two labels on the map, the greater their conceptual/semantic/functional distance and vice versa. Maps of this kind may reflect the situation characteristic of an individual language or aspire at universal validity. Whether it makes any difference to distinguish terminologically between universal and non-universal varieties is a methodological issue (Haspelmath 2003: 220) which cannot be dealt with in this book. Owing to the fact that semantic maps are still predominantly produced on conventional paper, their major shortcoming consists in the restriction to two-dimensional representation. We are firmly convinced that operating with just two dimensions is a stage that should be overcome as soon as possible. Schlesinger (1979, 1989), Heine, Claudi, and Hünnemeyer (1991: 163–166) and others address this largely technical problem under the title of multidimensional spaces and discuss the pros and cons of these spaces in comparison to so-called conceptual networks (Haspelmath 2003: 219 note 7). Multidimensionality is indeed called for, as Comitatives and Instrumentals suggest that equidistance and transitivity of relations among categories etc. are not necessarily straightforwardly linear in terms of map topography. What strikes the eye most is the fact that Comitatives have a certain predisposition for syncretistic patterns with partners that normally do not show up in the usual inventories of semantic cases, thematic relations and the like. While it is true that Instrumentals are still the favourite partners of Comitatives, AND-conjunction and possession are almost equally prominent, although they do not belong to the same functional level. This fact supports the idea that Comitatives are indeed borderliners, in the sense that they combine

Semantic map 157

traits of, on the one hand, semantic cases proper and, on the other, domains which (more or less) clearly fall outside the framework of semantic cases, etc. However, it cannot be denied that coordination and possession themselves are relations with semantic content. It is likewise telling that the Instrumental preferably associates with categories which clearly belong to what is usually considered semantic cases, etc. Given these different preferences, it is surprising to see that Comitative and Instrumental are nevertheless strong attractors for each other. In the maps below, we provide an only slightly modified conventional type of semantic map, meaning that we use the two-dimensional technique of representation. For obvious reasons, the map has to rely on a somewhat simplified interpretation of the data and their statistics, because otherwise a highly complicated plotting procedure would have been required in order to adequately represent the differences in terms of distance. For the calculation of the distances on the map, we have employed a simple rule of thumb: For each attested case of syncretism, we have added 0.1 point to the strength of the association lines in those maps which reflect the intensity of the relationship between the categories. To some extent, the map also reflects the relationship that holds between those categories which syncretise with Comitative and/or Instrumental. However, this is not the same as determining the semantic/conceptual/cognitive closeness or distances between these categories independent of their relation to Comitatives/Instrumentals. As a matter of fact, the map is Comitative-centred and all information implied by the map is to be understood in association with the Comitative – be it directly or indirectly. A semantic map for, say, predicative possession itself (i.e. a map centred upon predicative possession) would most probably look different. (B77)

Semantic map I BENEFACTIVE

[+Agentive] COMITATIVE AND POSSESSIVE

INSTRUMENTAL

[+Locative]

158 Syncretism In (B77), the two main association areas of Comitatives and Instrumentals are identified. Grey shading marks the area of the foremost choices of the Comitative, which, as we said, is situated mainly outside the realm of the classic thematic roles. Comitative, coordinator AND and Possession cover one sector of the map, whereas Instrumental, Benefactive, [+agentive] and [+locative] occupy the other sector. The Instrumental displays an evident preference for categories which count among the classic thematic roles. Both areas overlap, as Instrumental is the most important partner of Comitative and vice versa. This is, of course, only part of the whole story. From the above, we know that besides the overlap in the core region of Comitative and Instrumental, there are at least occasional connections between categories of the two different areas. Accordingly, (B78) highlights the intensity of relations between categories: The strength of association lines is meant to be suggestive of the degree of associability. (B78) Semantic map II BENEFACTIVE

[+Agentive] COMITATIVE

INSTRUMENTAL

[+Locative]

AND

POSSESSIVE

The major associations of the additional categories are always with Comitative and/or Instrumental. This is, of course, a consequence of our focus on syncretism of Comitatives and Instrumentals and thus cannot be taken as an adequate representation of the general associability of categories such as Benefactive and Possession or the like. No matter how rare they actually are in terms of token frequency, there are a variety of association lines cutting across the two distinct areas established in (B77). These cases cannot simply be discarded as exceptions. As a matter of fact, there is no need to resort to this measure at all, because upon closer inspection, it turns out that these crosscuts obey certain rules.

Semantic map 159

There are altogether 11 patterns which involve association lines crossing the borders of the areas identified in (B77). These 11 ‘deviating’ patterns are listed in (B79). The table identifies which of the three possible constellations (= distinct Comitative, syncretistic Comitative-Instrumental, distinct Instrumental) a given pattern co-occurs with and how often in Stolz (1997a). In addition, the ratio of the various co-occurrences is given and the preferred option is identified. The outcome is uncontroversial. (B79)

Deviations

Pattern

Additional Syncretism COM-INS INS 11 30 0 19 1 9 3 2 3 0 4 0 0 1 0 45 7 2 4 6 0 2 0 0 1 0 21 10 17 138 23

Ratio

Preference

1:3:0 0:19:0 1:9:0 1:1:1 0:4:0 0:1:0 0:8:1 1:2:3 0:2:0 0:1:0 0:2:1 1:8:1.3

COM-INS COM-INS COM-INS

COM

COM-LOC COM-AG COM-BEN AND-LOC AND-AG AND-BEN AND-INS POSS-LOC POSS-AG POSS-BEN POSS-INS

Total

– COM-INS COM-INS COM-INS (INS) COM-INS COM-INS COM-INS COM-INS

Except for the “illicit” combination of Possession and [+locative], where there is a slight preference for an additional syncretism with the distinct Instrumental and the absolutely even ratio for AND-[+locative], all other patterns show a striking preference for syncretism with the syncretistic Comitative-Instrumental.212 In five cases, there is not even an alternative to this solution. The syncretistic Comitative-Instrumental is on average preferred by a ration of 8-to-1 over the distinct Comitative and the distinct Instrumental. The higher the number of tokens, the more pronounced the trend becomes. Therefore, we assume that the associability of categories which are situated in different areas according to (B77) is facilitated by the combination of the syncretistic potentials of Comitative and Instrumental. Put differently, the syncretistic Comitative-Instrumental is less restrictive as to the choice of additional partners in syncretism, whereas both the distinct Comitative and the distinct Instrumental do not lend themselves easily to associating with categories which lie outside their own area on the map. The Comitative needs the Instrumental in order to combine with [+agentive], cf. Section 9.1 above. It also combines more easily with

160 Syncretism [+locative] or BENEFACTIVE if the Instrumental is also involved in the syncretistic pattern. This function of the Instrumental of allowing the Comitative to enter into syncretistic patterns it is normally banned from exemplifies the bridge-function introduced in Stolz (2001a: 331). The Comitative, in turn, provides the necessary bridge for the Instrumental to syncretise with Possession and AND. Comitative-Instrumental syncretism is also helpful for the coordinator and Possession, on the one hand, and [+agentive], [+locative] and BENEFACTIVE, on the other, to combine in syncretistic patterns. In Comitative-Instrumental syncretism, the syncretistic combinability of both categories is added up and unified. The bridge-function of the other categories in our above inventory is less strong.213 It is noteworthy, however, that combinations with [+locative] also seem to facilitate syncretism of the Instrumental with Possession and/or AND. Schematically, the bridge-function of categories can be best represented as in (B80), where we choose arbitrary variables as substitutes for otherwise semantically defined features. Grey shading marks those cells which have identical fillers. The picture which emerges is the well-known Wittgensteinian pattern of family resemblance, which has been so important in grammaticalisation research and in Lakoff’s (1987: 83–84) model of socalled radial categories (Luraghi 2001a: 38–39). The principle behind family resemblance goes like this: Immediate neighbours share at least one feature, whereas more distant members of the chain may display less similarity – theoretically even none at all. What safeguards their allocation to the same class, however, is the fact that the less similar members are interconnected via another member that they each share more features with than with each other. In (B80), AND does not have any features in common with the Instrumental. However, both overlap with the Comitative in a number of traits. Notwithstanding the fact that the two sets of features they share with the Comitative do not overlap at all (otherwise they would automatically have at least one common feature even without the intermediary Comitative), they may now combine in syncretistic patterns because their feature matrix coalesces with the one of the Comitative and thus acquires those traits which the Comitative per se has in common with other categories. This holds for all the distant relationships covered by (B80) and beyond. Note also that we assume some feature(s) for each of the categories which are never shared with any of the members of the chain. These features guarantee that a category may either attract other categories outside this chain to syncretise or that the category remains unaffected by attraction, as it has something special which makes it an individual.

Semantic map 161

(B80) POSS

B C

Family resemblance AND A B

COM

A B C D E F

INS

LOC

AG

BEN

D E

G H I J K L M N O

K L

L M

M N

P Q R

Our approach emphasises the importance of individual features as opposed to holistic categories (Stolz 1997a: 180, 2001b: 171–172). Categories as wholes are bundles of features. These features must of necessity “fit one another” semantically. Still, the features remain semi-autonomous in the sense that every single one of them may be highlighted separately when it comes to establishing a connection between two categories, i.e. feature bundles. Trivial as it is, the more individual features of two (or more) feature bundles are identical or similar enough, the higher the chances that syncretism applies. However, this is only a condition which facilitates syncretism – it is by no means necessary that many features correspond to each other to trigger syncretism; one may be enough in a given constellation. The heavy focus on features distinguishes our approach from other competitors which operate on the basis of holistic units. These models formulate ideal guidelines for syncretism but cannot account for many unexpected deviations. As we have shown in Chapter 8, the metaphor model has a hard time explaining why the majority of languages do not comply with the supposedly universal schema. Metaphors start from holistic conceptual gestalts and translate them into complex higher units (“propositions”) (Heine, Claudi, and Hünnemeyer 1991: 78–98). We do not deny that metaphors are indeed a powerful tool in the process of understanding and creat-

162 Syncretism ing language structures. However, we also recognise conceptual gestalts for what they are, namely more or less fixed agglomerations of identifiable properties. Some cognitive processes may operate on a metaphorical basis, but this does not preclude the possibility that others (or similar processes on different occasions) are feature-based, etc. Association as a mental activity cannot be sweepingly reduced to the propositional version of metaphor. Feature association may also be seen as a networklike, essentially connectionist activity. Similarly, approaches working with concepts as their basic unit can be shown to make wrong predictions as to legitimate and illegitimate syncretistic patterns. We have already mentioned Croft’s (1991) distinction of various classes of case roles, cf. (A79). For Croft (1991: 187), the syncretistic behaviour of the various case roles is determined by their classmembership: Antecedent roles may syncretise among each other but never with subsequent roles, which, in turn, only allow for syncretism with roles of the same class. As a matter of fact, it is the roles qua holistic entities which allow or disallow syncretism. According to Croft’s model, Comitative and Instrumental may syncretise freely, as both count among the antecedent roles, whereas syncretism of either of the two with the Benefactive is ruled out because the Benefactive belongs to the class of subsequent roles. Benefactive-(Comitativ-)Instrumental syncretism is by no means unheard of, no matter how seldom it is actually attested, cf. above. Disregarding the ways and means with which potential counter-evidence is dealt with in the concept-based model, we only touch upon the third class of thematic roles distinguished by Croft (1991: 206–212), namely, the socalled neutral roles. This class is formed by spatial concepts and possession. Luraghi (2001a: 37), in her critical assessment of Croft’s theory, takes the liberty to subsume the neutral roles under the subsequent ones because the relevant passages in Syntactic categories and grammatical relations may indeed be (wilfully) read as an invitation to lump the two classes together. Yet we do not subscribe to Luraghi’s interpretation as is, for Croft does not explicitly make this claim anywhere in the book. Instead, we point to the fact that our data strongly suggest that Possession and [+locative] go separate ways when it comes to syncretising with Comitative and/or Instrumental (and other categories not under scrutiny here). If two members of one and the same class radically diverge from one another as to their favourite combinations with other categories, it becomes rather doubtful to assume common class-membership in the first place unless the classes themselves are organised according to the prototype model. Moreover, the

Semantic map 163

same applies to Comitative and Instrumental as members of the class of antecedent roles: Their syncretism profiles are anything but similar, except for the fact that they attract one another. Within the class of antecedent roles, this is especially striking with the Comitative’s failure to syncretise with [+agentive]. All this behavioural dissimilarity speaks against the three classes of roles as the prime determining factor for syncretism. Having similar functions in the causal chain of an event is probably not the only – and perhaps not even the strongest criterion for syncretism to apply or not. This judgement can be extended to include the proposals made within the framework of Role and Reference Grammar, cf. (A33). In this case, the problems are even greater than for Croft’s model because Van Valin (1993) still employs notions which are identical or similar to the conventional semantic cases of case-grammar, whereas Croft goes a long way to shape the system anew.214 In Part A, Section 4.2, we have already discussed at length the problems which arise when we try to account for attested syncretisms which do not respect the boundaries imposed by the distinction of macroroles. Again, cases are treated as wholes rather than as feature bundles. If a category in its entirety sometimes complies with the rules formulated by the theory and sometimes disobeys them, this is a serious challenge to the approach itself because the basic units behave in an erratic way. The major troublemakers in (A33) are again Comitatives, Instrumentals and the spatial cases which display a syncretistic behaviour which runs counter to the assumed restrictions: It does not seem to be especially relevant to the categories whether or not they and their syncretistic partners belong to the macrorole ACTOR, INDIRECT or UNDERGOER. It is still an open question as to whether anything goes, but if it turns out that there is no limit to ACTORUNDERGOER syncretism, the approach loses all of its assumed relevance for the study of syncretism. In a way, at least some of the assumptions of the holistically-minded approaches seem better suited for a syncretistic Comitative-Instrumental than for two distinct categories. We claim that the problems of the category-based approaches could be remedied by the adoption of an additional feature-based component. The many instances of “illicit” syncretisms turn into legal ones, if features are ceded a certain degree of autonomy within a given feature bundle. In addition, features are also a more promising basis if we want to account for syncretism outside the narrow field of thematic relations proper.215 Yet what exactly are these features? And how are they activated? These are crucial questions that are not easy to answer. On the one hand, we assume the features to be semantic “primitives”, i.e. minimal units of meaning.

164 Syncretism Hitherto, we have taken for granted that these units largely correspond to what is familiar from compositional semantics: The presence or absence of certain properties such as, e.g., animacy216, volition, agentivity, control, dynamicity, duration, etc. is specified. However, this open list does not fully capture what is relevant here. In the following list, we only mention potential features which are relevant for typical syncretisms of the Comitative. Those features which have a say in the syncretism of the Instrumental are amply discussed in Palancar (2001). – Comitative = AND Both categories imply that the participants which the relators connect to one another have enough properties in common to be considered to be two of a kind. If a given relator is employed, it follows the rule “copy all properties X to make the participants equal in status”. We assume therefore a feature SAME AS. The employment of the SAME AS feature also yields a group of almost identical exemplars, i.e., a group of closely associated and hierarchically largely egalitarian entities. Accordingly, we postulate two more features, namely, GROUP and EGALITARIAN. – Comitative = Possession In contradistinction to the first case, SAME AS and EGALITARIAN are less likely candidates for a syncretistic connection between the two categories. Still, GROUP is an appropriate choice. As EGALITARIAN is ruled out, the resulting group comprises exemplars of different status, i.e., it is hierarchically organised and thus CONTROL comes into play. However, CONTROL by the Possessor is only asserted. – Comitative = Instrumental The standard pattern of syncretism likewise involves CONTROL but this time by an Agent with the full array of features ranging from volitionality to activity. Thus, CONTROL is not only asserted but also integrated into a more complex scenario which also includes the feature FORCE, which is absent from Possession. Note that FORCE is not a feature which belongs to the Comitative itself! FORCE is the element that links the Instrumental to [+agentive]. The activation of the features follows the same principles as are presupposed for the workings of analogy (Anttila 1989: 88–108): Linguistic phenomena evoke each other in the mind of the language user because there are functional/semantic, formal or distributional characteristics (including factors of frequency) which are shared by the items. None of the types of

Semantic map 165

analogy seems to take precedence over the others. Which of them gets the upper hand depends largely upon the context in which analogy is triggered. Again, these contexts can be of various kinds, situational or constructionbased. We come back to this issue in Part D when we discuss grammaticalisation processes. Our feature-based approach fails to account for those cases where crosscuts of the types mentioned in (B79) are direct, i.e., without any intermediary bridge-category. There are, of course, several ways to speculate about the reasons for this in a more or less plausible fashion. However, we do not yet have tangible proof of the validity of such assumptions: – Either diachronic developments have lead to the loss of an erstwhile established function which represented the necessary bridge, – or the feature bundles of individual languages deviate from the more prototype-like patterns described here; if so, chances are that a given feature bundle comprises features which create a more direct link to the syncretistic partner without the detour via a bridge-category. In the absence of uncontroversial evidence, we let these hypotheses stand without any further comment. Follow-up studies may take up these issues to investigate their explanatory power in detail. Recapitulation: Comitative and Instrumental associate with different partners which belong to different areas of grammar. The Instrumental is closer to semantic cases/thematic roles, whereas the Comitative has clear leanings towards categories outside this domain. The picture changes as soon as Comitative and Instrumental syncretise with each other. In this frequent combination, syncretism is allowed to go in any direction. The partners of Comitative and Instrumental, respectively, also comply to this rule: They do not easily combine with categories outside their own domain unless they also syncretise with a Comitative-Instrumental. Comitative and Instrumental serve as bridges for each other when it comes to syncretism beyond the boundaries of their reserved areas. The relationship of Comitative and Instrumental is thus ambivalent: On the one hand, Comitative and Instrumental have markedly distinct preferences for syncretism. This gives rise to the idea that their sets of semantic features are very different. This assumption is supported by the fact that their respective partners in syncretism do not normally syncretise among each other. On the other hand, however, Comitative and Instrumental are favourite syncretistic partners of each other. This implies that they have things in common which function as attractors. This hypothesis receives support from the fact that syncretistic

166 Systematicity of interaction Comitative-Instrumentals attract all sorts of additional partners which are normally excluded from combinations with the distinct categories. It cannot be denied that Comitative and Instrumental are neighbours on the map – and very close ones, too. Nevertheless, they often behave like entities which are situated in two different universes. With a view to a better understanding of their relationship, it is necessary to have a closer look at other factors which speak for or against their close proximity. The most important criterion for this close up investigation is markedness – a concept which is the topic of Chapter 10. 10. Systematicity of interaction Do Comitative and Instrumental belong together only when syncretism between the two applies? The previous chapters have amply demonstrated that Comitatives and Instrumentals each have too many individual traits to simply be considered one. At the same time, one has to admit that they syncretise much too often to pass as distinct and absolutely incompatible units. Comitative-Instrumental syncretism is admittedly especially frequent under certain conditions and in certain areas, cf. Part C. However, one must not oversimplify matters by focusing on formal syncretism alone. There are, of course, other phenomena which may be indicative of a closer relationship between two categories. Even though the relators of two categories are distinct, the two categories may nevertheless interact in a systematic way. The systematicity of their interaction, then, is suggestive of a paradigmatic relationship, which, in turn, supports the hypothesis that Comitative and Instrumental are not just two categories but two categories under a common roof. This chapter is devoted to describing the essentials of the interaction of Comitatives and Instrumentals. For all issues of markedness addressed in the following sections, we adopt the same procedures as Stolz (1996), which are largely based on the common practice within the framework of Natural Morphology (Dressler 1987). Similarly, in Sections 10.1 through 10.6, any allusions to aspects of grammaticalisation are based on Lehmann’s parameters (Lehmann 1995: 123 and 164). An integrative evaluation of our findings is reserved for Section 10.7.

WITHOUT 167

10.1. WITHOUT There are various domains in which Comitatives and Instrumentals can be said to interact indirectly. With a view to demonstrating that, despite their many idiosyncrasies, the two categories are nevertheless in a close relationship to one another, a look at the shape they take under negation is helpful. In Stolz (1996: 23–48; 1997c), the translation equivalents of English without in a small crosslinguistic sample are compared to the Comitatives and Instrumentals of the same sample languages.217 As a working hypothesis, it can be expected that two categories which are definitely distinct also have each distinct negative forms of their own. Conversely, a common negative form for two or more categories points at a close connection of these categories. As a matter of fact, many of the differences observed for Comitatives and Instrumentals are neutralised under negation218, i.e., the two share a negative form common to both of them. In Table (B81), we state which kinds of combinations of syncretism we have encountered: There are three types of combinations, namely, languages which have only one relator for negative Comitative and negative Instrumental besides a syncretistic (positive) Comitative-Instrumental, languages which keep Comitative and Instrumental distinct in both polarities, and languages which have a common negative expression for Comitative and Instrumental, despite the two categories being formally distinct outside negation. Among the four logically possible combinations, only one is not attested: We have not found evidence for a language which distinguishes Comitative and Instrumental only under negation. (B81)

Tetrachoric chart for WITH(OUT)

Meaning WITH

Patterns COM = INS COM  INS

WITHOUT COM = INS + +

COM  INS

¬ +

Furthermore, the three attested types do not occur with the same frequency because distinct forms of Comitative and Instrumental under negation are generally rather rare.219 There is a clear tendency towards a conflation of both categories when negated. The Amerindian A-type language Hixkaryána is one of the rare representatives of those languages which keep Comitative and Instrumental distinct in both polarities, cf. (B82).

168 Systematicity of interaction (B82) Hixkaryána (Derbyshire 1979: 95–96) (B82.1) Comitative: positive G-to-no Waraka y-akoro Waraka POSS.3SG-X 1SG-go-PAST ‘I went with Waraka.’ (B82.2) Comitative: negative ro-hetx y-akoro-hra k-omok-no POSS.1SG-wife POSS.3SG-X-NEG 1SG-come-PAST ‘I came without my wife.’ (B82.3) Instrumental: positive watma ke me-taha-no club Y 2SG-strike-PAST ‘Did you hit him with a club?’ (B82.4) Instrumental: negative we-taha-no watma-hnG ke mekmetxho ke haxa Y CONTR 1SG-strike-PAST club-NEG Y hammer ‘It was not with a club but with a hammer that I hit him.’ The expressions differ a lot structurally. In positive contexts, the Comitative is expressed by the obligatorily possessed postposition -akoro, whereas the Instrumental is encoded by the postposition ke, which never bears a possessor prefix. Under negation, the Comitative marker itself is the morphological host of the negator -hra. This is different for the Instrumental. Here, the negation -hnG is suffixed to the noun denoting the instrument.220 It is worth noting that two different negators in two different slots are used to negate the two categories, i.e., Comitative and Instrumental diverge maximally in terms of their distribution behaviour. Much more common are A-type languages, which neutralise the distinction of Comitative and Instrumental under negation. The genetic isolate Basque, an A-type AND-language, is a case in point, cf. (B83). (B83) Basque (Saltarelli 1989: 158–159) (B83.1) Comitative: positive ne-re senargai-arekin joan fiancé-X go 1SG-GEN nintzen AUX:1SG.ABS:PAST:INTRANS:PAST ‘I travelled to Italy with my fiancé.’

Italia-ra Italy-ALL

WITHOUT 169

(B83.2) Comitative: negative lagun-ik gabe etorri naiz friend-PART X/Y.NEG come AUX:1SG.ABS:PRES:INTRANS ‘I have come without friends.’ (B83.3) Instrumental: positive221 esku-z zati-tzen dut ogia hand-Y divide-HAB AUX:3SG.ABS:TRANS:1SG.ERG bread:ABS.DEF ‘I used to divide the bread with my hands.’ (B83.4) Instrumental: negative haizkora-rik gabe ez-in X/Y.NEG NEG-be_able axe-PART duzu egurr-ik ebaki AUX:3SG.ABS:TRANS:2SG.ERG wood-PART cut ‘Without an axe you won’t be able to cut wood.’ Comitative and Instrumental have distinct inflectional encodings, whereas their negative forms are identical: The postposition gabe ‘without’ (governing the partitive case) serves to negate them both. This neutralisation pattern is also valid for C-type languages such as Hungarian. In Hungarian (cf. [A7]), there is a syncretistic Comitative-Instrumental -val/-vel alongside a less frequently employed inflectional associative -stul/-stül, which is specialised for typical instances of Comitativity222. Both these cases share a common negative expression, namely the postposition nélkül ‘without’, cf. (B84). (B84) Hungarian (Kenesei, Vago, and Fenyvesi 1998: 211–213) (B84.1) Comitative (Confective): negative a rendör fegyver nélkül van DET policeman gun X/Y.NEG be ‘The policeman is without a gun.’ (B84.2) Instrumental: negative kalapács nélkül tör-t-em break-PAST-1SG.DEF hammer X/Y.NEG le a kilincs-et handle-ACC down DET ‘I broke off the door handle without a hammer.’ For B-type languages, German may serve as the paradigm case. (A17–18) can be negated by substituting ohne for mit and changing dativegovernment to accusative-government, cf. (B85).

170 Systematicity of interaction (B85)

German das

Kind DET.ACC child Erich fotographiert Erich takes_photos

die DET

Mutter mother

ohne X/Y.NEG

die DET.ACC

Kamera camera

‘Erich takes photos of the mother without the child/camera.’ The resulting sentences are semantically ambiguous. The PP ohne das Kind refers either to the absence of a Companion of the ACTOR (= no co-agent) or to the absence of a Companion of the UNDERGOER (= no co-patient). The PP ohne die Kamera has three readings: (a) (b) (c)

It identifies a certain mother whose distinctive feature is ‘not having a camera’ (whereas other mothers present in the universe of discourse may have one); it depicts the mother in a new state: Erich asked her to lay aside the camera which is not supposed to be in the picture; Erich is using advanced technology which enables him to take pictures without using a conventional camera.

In (a)–(b), a confective is negated, whereas it is only in (c) that the absence of a certain tool is stated. No matter how the relationship between Comitatives and Instrumentals, on the one hand, and their negative forms, on the other, is spelled out in detail223, there is a very strong tendency for the categories to merge under negation. This merger is tantamount to formal identity and is thus an indicator of a very close association of the categories. In a way, neutralisation under negation can also be viewed as a kind of part-time or conditioned syncretism of Comitative and Instrumental. This evidence calls for a crosscheck using other criteria. 10.2. Formal likeness In a variety of A-type (and C-type) languages, the relators of Comitatives and Instrumentals resemble each other, although they are distinct. This synchronic resemblance is often indicative of a diachronic relation between the two (cf. Part D, Chapters 13–14): The expression of the one category is derived from the other. Disregarding the historical motivation behind this resemblance for the time being, we take the facts at face value.

Formal likeness 171

An interesting pattern emerges: While it is not too difficult to find Comitative relators which are segmentally reminiscent of their Instrumental counterparts in such a way that the expression of the Comitative is more complex than the one used for the Instrumental, we have not encountered any uncontroversial evidence for a reversal of this relationship. (B86) contains a selection of 19 characteristic examples based on Stolz (1997a) and Stolz, Stroh, and Urdze (2005). Boldface identifies those segments which the Comitative relator shares with the Instrumental relator. In all these cases, the Comitative (or the relator which is also responsible for the Comitative) receives a more complex encoding that is partly identical with the less complex encoding of the Instrumental. It is tempting to interpret this as evidence of the Comitative being derived from the Instrumental, but cf. Part D, where we discuss the diachronic developments which lead to at least some of the constellations in (B86). (B86) Type A A A A A A A A A A A A A A A A A C C

Comitatives resemble Instrumentals Language Alyawarr Blackfoot Chemehuevi Chukot Comanche Itelmen Koasati Kot Krongo Lithuanian Mixe Palauan Qawasqar Russian Boro Timucua Warao Ngalakan Welsh

Comitative -ilalinga -wi:xpokim-waËi g8-…-t8 -maËaix k’-…-ƺ stóklon -oš yásu + N[INS] mà • obĊngkel nowaq s(o) + N[INS] -ntoon6 uquale aisiko baĠa-…-yi’ gyda(g)

Instrumental -ila -Mwi:xt-w(a) -t8 -max -ƺ st-ô áN[INS] -m oba -q N[INS] -n6 quale aisía -yi’ a(g)

The formal correspondence on the expression side is, of course, suggestive of a close connection on the content side. Synonymy is ruled out, as the expressions are not absolutely identical. Nevertheless, there must be something that Comitatives and Instrumentals have in common besides those

172 Systematicity of interaction parts of their relators which happen to consist of the same segments. In terms of iconicity, it seems that the Comitative in the above examples is semantically more complex because the morphemes used to encode the Comitative are regularly more complex than the ones representing the Instrumental. Put differently, the Comitative is the Instrumental plus some additional ingredient. This analysis especially fits those cases where the expression of the Comitative consists of two identifiable morphs. The discontinuous morphs, i.e. the circumfixes in the Paleo-Siberian languages and Ngalakan, are rather convincing examples, as are the ones which accord to the Russian pattern, which has an additional preposition for the Comitative governing the Instrumental. Owing to the fact that formal similarity of relators in A-type and C-type languages is not particularly frequent (about 8% of the A-type and C-type languages in Stolz’s [1997a] sample)224, one might ask whether the patterns of (B86) are just exceptional cases to be ignored when it comes to drawing conclusions. Indeed, taken on their own, they do not suffice to serve as a basis for any hypothesis. However, (B86) documents a pattern that has a much wider distribution than formal similarity: In the above table, formal similarity goes along with greater complexity of the Comitative relator as opposed to the Instrumental one. We can reformulate this observation as a conditional implication: – Similarity and Complexity: If, in a given language, Comitative and Instrumental are kept apart formally, and if they display expressions which share significant parts of their segmental chains, then one expression is always more complex than the other, and it is always the Comitative which receives the more complex expression. This higher degree of complexity is by no means restricted to cases where the morphemes of Comitative and Instrumental are look-alikes. More complex Comitatives are commonplace among A-type and C-type languages (Stolz 1998b: 89–93), cf. Section 10.3. 10.3. Complexity For a start, segmental complexity of a relator can be calculated in two ways: Either we rely on the absolute number of segments which together yield the phonological shape of the relator, or we count the syllables225 that

Complexity 173

the relator is made up of. For the present purpose, we treat the two criteria in conjunction, i.e., the expression of a category is said to be more complex than the one of its counterpart if it is at least longer by one unit on one parameter and at the same time no less than co-extensive on the other parameter. In (B87), we survey the logically possible patterns.226 Note that we do not have convincing evidence for contradictory values for syllables and segments: If one of the two receives a plus the other cannot have a minus and vice versa.227 (B87)

Logically possible patterns

Relation X>Y X>Y X>Y X=Y X INS

patterns syllabic ratio

100,00% 90,00% 80,00% 70,00% 2:1 44,10%

60,00% 50,00% 40,00%

X:0 17,70%

30,00%

1:1 25,70%

20,00% 10,00%

~1,5 : 3,70%

3:1 8,80%

0,00%

(B93)

COM > INS

patterns segmental ratio

100,00% 90,00% 80,00% 70,00% 60,00% 50,00%

~1,5 : 1 2 : 1 30,20% 30,90%

40,00% 30,00% 20,00% 10,00%

X:0 1:1 3,70% 2,90%

3:1 >3:1 ~2,5 : 1 13,20% 10,30% 8,80%

0,00%

Typical examples for the various patterns are given in (B94).230 Equal complexity on one parameter does not prevent the other criterion from being fulfilled. It is worth noting that co-extension in terms of segments is more rarely attested than identical number of syllables (a fact that meets the ex-

Boundedness 177

pectations, as variation on lower levels [= segmental chain] may be much stronger than on higher levels, where the smaller units of the lower level(s) are integrated as constituents into larger units [= syllables]). Generally, the variation for syllabic asymmetry ranges from a minimum of 1-to-1 to a maximum of 5-to-3 (discounting the cases of asyllabic Instrumental relators), whereas, on the segmental level, the variation spans from 1-to-1 to 8to-5 (again without zero-marked Instrumentals). The largest gaps between Comitative and Instrumental relators are 4-to-0 for syllables and 7-to-1 for segments. (B94)

Examples for patterns of complexity

Pattern COM > INS

COM = INS COM < INS

Language Akateko Alamblak Marshallese Gola Trumaí Burushaski

COM

-etox -pnë ibba yaa -tam -ina

INS

-u -e kin wee -letsi -yate

Syllables 2:1 1:1 2:1 1:1 1:2 1:1

Ratio Segments 4:1 3:1 1:1 1:1 1:1.6 1:1.3

Independent of the exact numerical values for the various ratios, there is the remarkable fact that the expressions of Comitatives and Instrumentals interact systematically, even though their relators do not display partial identity, as the ones listed in (B86). Admittedly, not all cases are exactly the same. However, even without a 100% homogeneity, the two categories have a strong preference to conform to the pattern consisting of a more complex Comitative and a less complex Instrumental. The tendency towards more complexity of the Comitative can be understood as evidence for a higher degree of markedness of the Comitative as opposed to the Instrumental. In the subsequent sections, we review a selection of other criteria which are commonly used when it comes to assigning markedness values to members of a pair of categories. A general evaluation of the observed facts will be attempted in Section 10.7. 10.4. Boundedness The extension of a relator in terms of segments and syllables is already a reliable indicator of the special status the Comitative has within the pair of categories under review. The assumed higher degree of markedness of the Comitative receives support from the realm of morphological boundedness.

178 Systematicity of interaction Both the Comitative and the Instrumental may be encoded either by free morphemes or by bound morphemes. Logically, there are four possible combinations, namely, with both categories encoded by bound morphemes, both categories encoded by free morphemes, or with a bound relator while the other one uses a free morpheme. Again, these patterns occur with strikingly different frequencies, cf. (B95). (B95)

Tetrachoric chart for boundedness

Comitative

Status F B

Instrumental F + +

B + ¬

Only three patterns are attested by uncontroversial examples in our sample: – B-B = Both Comitative and Instrumental are bound morphemes. Trumaí is an A-type WITH-language spoken in the Amazonian region of Brazil. It makes use of two distinct relators for Comitative (-tam) and Instrumental (-itse). The exact morphological status of these case markers remains to be determined, as they occur both in bound form and sometimes as free postnominal morphemes. They are therefore best classified as enclitics. In the grammatical description at hand, they are treated as bound elements, cf. (B96). (B96) Trumaí (Monod-Becquelin 1975: 178) (B96.1) Accompaniment/Confective uruts-tam ua-kaË šy-n tortoise-X away-go-3SG ‘He went away with the tortoise.’ (B96.2) Means of transportation avião-letsi peš pata-n plane-Y arrive-3SG ‘He arrived by plane.’ – F-F = Both Comitative and Instrumental are free morphemes. The genetically isolated Ainu is an A-type AND-language. It employs the postpositions tura for the Comitative and the postposition ari for Instrumental. Both postpositions are relatively free and do not seem to cliticise to their neighbours in the syntactic chain, cf. (B97).

Boundedness 179

(B97) Ainu (B97.1) Accompaniment (Refsing 1986: 169) ponnispa tura ren an ne onne young_gentleman X three we be be_old huci tura ren an ne wa old_woman X three we be and ukoitak an ruwe ne ASS be chat we ‘With the young gentleman we are three, with this old woman we are three, and we chat together.’ (B97.2) Body part instrument (Refsing 1986: 170) hempak suy ka cikiri ari ku Y 1SG how_many time INDEF foot nisapi kik kane an INCOMPL shin kick ‘He was kicking at my shin several times with his foot.’ – F-B = The Comitative is a free morpheme whereas the Instrumental is a bound one. In the A-type WITH-language So, spoken in Uganda, variously classified as Cushitoid or Nilo-Saharan, etc. (Carlin 1993: 4), the Comitative is encoded by a free morpheme – the preposition ka which also functions as marker of predicative possession and coordinating conjunction. In contradistinction to ka231, the Instrumental relator -Ák is a bound suffixed morpheme with a variety of related functions, cf. (B98), (B98) So (B98.1) Accompaniment (Carlin 1993: 178) ones ica ka yog-ic n1-ga-u NARR-go-IT then he X people-POSS.3SG tar eo umk-1 t to village in_law-PL ‘Then he goes with his people to the village.’ (B98.2) Tool (Carlin 1993: 129)232 in-ko-gas-ak l1 8 bl-Ák NEG-FUT-bleed-IMPS cattle knife-Y ko-gas-ak t8 8m-Ák FUT-bleed-IMPS arrow-Y ‘They won’t bleed the cattle with a knife but with an arrow.’ whereas evidence for the pattern

180 Systematicity of interaction – B-F = the other way around: the Comitative is bound and the Instrumental free is wanting or at least controversial.233 For the general picture234, we only look at A-type and C-type languages of the sample used in Stolz, Stroh, and Urdze (2005), reduced by two languages for which the extant description did not allow a decision as to the degree of boundedness of the markers.235 Tables (B99–100) indicate how often Comitative and Instrumental relators occur as bound or free morphemes in combination: (B99) Bounded vs. free I

COM > INS COM = INS COM < INS

Absolute Percentage Unification Percentage

B-B 68 30 12 110 50.69%

F-F 35 24 10 69 31.79% 179 82.48%

Patterns F-B 35 3 0 38 17.51%

Sum B-F 0 0 0 0 0% 38 17.51%

148 57 22 217 100% 217 100%

(B100) Bound vs. free I 100,00% 80,00% 60,00% 40,00%

B-B 50.69%

F-F 31.79%

F-B 17.51%

20,00% 0,00%

What immediately emerges from the values in (B99–100) is the fact that the relators of the two categories overwhelmingly behave identically. If one of them is bound, the other is more often than not bound too. If one of them is free, the other is more often than not free too. The vast majority of our sample languages of type A and C follow these rules. These cases of identical morpheme status, however, are less interesting than the minority solution where Comitative and Instrumental have different morpheme status. Differences of this kind occur only in 17.51% of our sample languages.

Zero-marking 181

Notwithstanding this relative infrequency, these cases are much more interesting for our present purpose, as they again attest to the systematic inequality of Comitative and Instrumental. In combinations with asymmetrical morpheme status, it is generally the Comitative which is the free morpheme, whereas the Instrumental is encoded by a bound morpheme. This pattern accounts for 100% of the cases of morpheme-status asymmetry. On this statistical basis, it is possible to make another rule-like statement about the relationship of Comitatives and Instrumentals: – Asymmetry of morpheme status: If, in a given language, Comitative and Instrumental are kept apart formally and if their expressions differ as to their degree of boundedness, then it is always the Comitative which is encoded by a free morpheme while the Instrumental relator is a bound morpheme. In terms of Bybee, Pagliuca, and Perkins (1991: 26), Comitatives thus often look like more recent GRAMS or less grammaticalised categories as opposed to Instrumentals, cf. Part D, Chapters 13–14. This interpretation does not lend itself to an easy translation into markedness-theoretical ideas. Tentatively, one could argue that their preference for free morphemic status also in languages where a higher degree of boundedness is actually an option for grammemes marks Comitatives as untypical grammatical categories – and thus as relatively marked ones. This relative markedness as opposed to the Instrumental is also reflected by the restrictions on zero-marking discussed in the next section. 10.5. Zero-marking Zero-marking and the related issue of zero-morphemes have been hotly debated issues in linguistic structuralism. Opinions range from an outright refusal to accept that content can be represented by the absence of form, to the opposite view, which assumes that the maximal number of possible overt distinctions determines the number of slots, be they phonological, morphological or syntactic, which require fillers, including phonologically empty ones (Mel’þuk 2001). We lean towards the former idea, because the zero would automatically become a highly polysemous element which cannot be accounted for adequately in any model of form-function relations. However, we concede that it is possible to identify the absence of special phonological (including non-segmental) means to encode something. Say-

182 Systematicity of interaction ing that a given function does not require a morphological marker is different from claiming that there is a phonologically unrealised marker for this function anyway. Fortunately, we do not have to settle this issue here, as it suffices for our present purpose to consider the presence and absence of overt markers independent of the problem of whether zero-morphemes can possibly exist or not. In our sample (Stolz, Stroh, and Urdze 2005), there are exactly five cases (= Kobon, Mataco, Mangaryi, Salt-Yui and Triqui) in which the absence of an overt marker can be stated with certainty. All these cases occur in A-type languages, and it is always the Instrumental which lacks the overt marker, never the Comitative. Consider the examples from Triqui – an Amerindian A-type AND-language of Mexico – and Kobon – a Papuan Atype AND-language in (B101–102).236 (B101) Triqui (Hollenbach and Hollenbach 1975: 125–126) (B101.1) Accompaniment zo’3 ga4 gwa32 âh ga’na’3 X Juan DEC come:PERF he ‘He has come with Juan.’ (B101.2) Tool zo’5 ne34 mã3 ne’3 âh ga’ne’5 you knife OBJ rope DEC cut:FUT ‘Cut the rope with the knife!’ (B102) Kobon (B102.1) Accompaniment (Davies 1989: 73)237 ne aip ar-nab-in you X go-FUT-1SG ‘I will go with you.’ (B102.2) Tool (Davies 1989: 15) nipe kaj ru rGb-öp he pig axe cut-PERF.3SG ‘He cut the pig with an axe.’ In both languages, the Comitative is encoded by free morphemes: the preposition ga4 in Triqui and the postposition aip in Kobon. At the same time, neither of the two languages encodes the Instrumental by the usual morphological means. As a matter of fact, the NP representing the tool in (B101.2) and (B102.2) is simply an additional argument of the verb and is thus morphologically treated on a par with other arguments such as, e.g., subject and object NPs. Note that this not an idiosyncrasy of the individual verbs in our

Zero-marking 183

examples. This is the normal (and in fact the only conventionalised) way of expressing Instrumentality in these languages. One could argue that Seri (Amerindian A-type AND-language) is a language which comes rather close to not encoding the Comitative and the Instrumental. Not only do Instrumentals lack any morphological marking, but Comitatives also appear to be without relators of their own, cf. (B103). (B103) Seri (Moser 1996: 154–155) (B103.1) Tool

Ëæ´ Ë æ

kom

plant

DEF.LYING

kix

Ë amkanoiin bowl

ko -k-m-ma« š

DEF.SITTING

IO.3SG-IMP-NEG-hit ‘Don’t hit the bowl with the stick!’ (B103.2) Accompaniment Ë i-Ë amiigo kË i-t-aaJ POSS.1SG-friend DEF.NTR OBJ-REAL-accompany k-aYp-Ë a AG-arrive-ASS ‘He came with my friend.’ There is obviously no morpheme characterising the situation as one that involves an instrument. The NP indicating the tool is again one of the arguments of the verb. This is similar to the Accompaniment situation in (B102.2). A lexical verb meaning ‘to accompany’ is regularly used in contexts where motion/movement is involved. This verb behaves morphosyntactically like a normal lexical verb. There is no compelling evidence speaking in favour of an advanced stage of auxiliarisation or grammaticalisation. Is Seri a language without grammatical Comitatives and Instrumentals? A language with a zero-marked Comitative? Surprisingly, the answer is no. Consider (B104): (B104)

Seri (Moser 1996: 155) i-aË a« k-atikpan -Ë a Juan «aË Pedro «aË pti Juan and Pedro and together POSS.3-X AG-work-ASS ‘Juan and Pedro are working together.’

In this sentence, there is an abundance of morphological means indicating that the two participants are acting together. Besides the double coordinating conjunction «aË…«aË and the adverbial pti , there is also the possessormarked postposition -aË a« which is the normal translation equivalent of English with. At the moment, we can only speculate whether -aË a«is re-

184 Systematicity of interaction stricted to situations whose descriptions do not involve motion verbs. What is clear, however, is that it cannot be used to encode Instrumentality. Chances are that it is a relator which only occurs in combination with participants which rank highest on the animacy hierarchy. Actually, “hidden” or covert Comitatives (and Instrumentals) are not rare. The reason for this, again, is syncretism: If a category is always coencoded with another category which appears to be functionally more remote, or if there is a very high degree of polysemy of relators with a general neutralising effect, it may become difficult to recognise the morphological means as also being representative of functions belonging to the domains of Comitativity and Instrumentality. Otomí and Mapudungun, two Amerindian A-type AND-languages, the former of the Otomangue phylum spoken in Mexico, the latter a member of the Andean phylum spoken in Chile, display an interesting pattern of syncretism for the Comitative by way of connecting it to number distinctions. Similar observations can be made with respect to Austronesian Tiri, a C-type AND-language spoken on New Caledonia.238 For a start, consider (B105), which contains two examples from Mapudungun. (B105) Mapudungun (B105.1) Accompaniment (Salas 1992: 99) iñche amuan Temuko ñi chao iñchu 1DU/X I go:FUT Temuco my father ‘I shall go to Temuco with my father.’ (B105.2) Tool (Salas 1992: 104) kafna-pu-ke-fi-ngün toki meo Y form-there-HAB-OBJ.3-PL axe ‘There they give shape to it with an axe.’ In Mapudungun, there is a polysemous postposition meo, which encodes a variety of spatial relations and serves as the primary marker of the Instrumental. Upon superficial inspection, there is no Comitative marker at all, as no postposition is used in Accompaniment situations (Salas 1992: 103– 105). However, Mapudungun has a special set of non-singular pronouns called group-formation pronouns (“grupalizadores” [Salas 1992: 99–100]) which occur in clause-final position as a kind of sum of the participants involved in the situation described, cf. (B106). Of these, the forms of the 1st person are identical with the personal pronouns of the 1st person dual and plural, whereas there are no full-fledged personal pronouns of the 3rd person outside the paradigm of group-formation pronouns.

Zero-marking 185

(B106) Mapudungun: group-formation pronouns Dual Plural

1st person iñch(i)u iñchiñ

2nd person emu emün

3rd person engu engün

These pronouns are by no means restricted to animate participants. As their label suggests, they also occur in other occasions where combinations of entities are stated as, e.g., bread and tea in (B107). (B107)

Mapudungun (Salas 1992: 168) akutulu ñi füta wentru mia-uma REL-go.previously returning POSS.3SG husband man lelfün meo elukefi mate kofke engu field Y give:to_him tea bread 3DU/X ‘When her husband comes back from the fields, she serves him tea and bread.’

In Otomí, autochthonous prepositions are almost nonexistent.239 There is one such word, nangue, which may be classified as a general relator whose functional domain is such that descriptive grammarians consider it to be devoid of any semantics (Voigtlander and Echegoyen 1985: 301–302). Its frequency is said to be marginal and only very rarely is it employed to encode the Instrumental. However, nangue does sound less natural to native speakers than the much more frequent construction with the so-called circumstantial mood, cf. (B108). (B108) Otomí (Voigtlander and Echegoyen 1985: 302). (B108.1) Tool: autochthonous preposition bi gäc ra dąpo nangue ra juai DET machete 3SG.PAST clear DET jungle Y ‘He cleared the jungle with the machete.’ (B108.2) Tool: Circumstantial mood í gäc ra dąpo ra juai machete 3SG.PAST.CIRC/Y clear DET jungle DET ‘He cleared the jungle with the machete.’ Nangue clearly qualifies as a secondary marker – the primary one being the use of the appropriate proclitic of the circumstantial mood. Neither nangue nor the circumstantial mood are specialised to exclusively encode the Instrumental. Rather, they both serve the purpose of indicating that there is an additional NP whose exact semantic relation to the nuclear predication depends entirely on the contextual information. Yet neither of the two en-

186 Systematicity of interaction coding strategies can be used for the Comitative. As is the case in Mapudungun, one has difficulties at first identifying Comitatives in Otomí. Again, one has to look for pronominal forms to find them, cf. (B109). (B109) Otomí (Voigtlander and Echegoyen 1985: 150) ra Xuua ne ra Lipe dí xÀ-he 1.PRES chop_wood-1PL.EXCL/X DET Juan and DET Felipe ‘I cut wood with Juan and Felipe.’ In (B109), the verb bears the suffixed pronominal morpheme for the 1st person plural exclusive, whereas the obligatory proclitic indicates a 1st person irrespective of number distinctions. The addition of the suffixed pronominal element is considered a way of expressing Accompaniment. According to Voigtlander and Echegoyen (1985: 149), the difference between genuine number marking and Comitative readings is especially clear with forms of the 3rd person because the use of the usual suffixed pronominal morphemes to indicate dual or plural subjects is blocked for the 3rd person (which requires a set of demonstratives to fulfil this function). However, the suffixed pronominal morphemes are nevertheless used on forms of the 3rd person when it comes to indicating Accompaniment: Es posible agregar al radical modificado por el proclítico de tercera persona los pronombres complementarios de número dual o plural inclusivo -ui, -hÀ, pero éstos dan significado de acompañamiento, no de sujetos simultáneos.

The pronominal technique can also be used with inanimate participants (Voigtlander and Echegoyen 1985: 307) but is never employed with an Instrumental reading. Likewise, pronouns in Tiri may be used to encode the Comitative but never the Instrumental. Osumi (1995: 40–41) observes that the pronouns behave morphosyntactically like prepositions when used for Comitative function (they trigger the non-subject forms of the pronoun to their immediate right), cf. (B110). (B110) Tiri (B110.1) Accompaniment (Osumi 1995: 40) komu nrî u fi pwere numea Noumea 1DU.EXCL/X 3SG.OBJ 1SG.SUBJ go to ‘I go with him to Noumea.’ (B110.2) Tool (Osumi 1995: 82) nrâ ta traiki nrî ù dog X/Y stick 3SG.SUBJ hit ‘He hit the dog with a stick.’

Distribution 187

On our present basis of knowledge, it cannot be decided conclusively whether the pronominal technique is the preferred way to express Accompaniment relations. As a matter of fact, Tiri makes use of at least eight genuine prepositions, each of which counts the Comitative among its catalogue of functions, whereas nrî is the only preposition with an Instrumental meaning component (Osumi 1995: 89). For the topic of this section, however, it is enough to see that pronouns can be employed in such a way that they give rise to a Comitative-like reading, but fail to invite an Instrumental one. There are thus two more generalisations to be made: – If zero-marking applies in a language, then this language never belongs to the B-type, and it is always the Instrumental which receives zero-marking, whereas the Comitative always requires a phonologically realised marker. – If pronouns are employed as relators in a language, then these pronouns are non-singular ones, and it is always the Comitative and never the Instrumental which is encoded by this technique. The latter observation especially has a bearing on our semantic map, as it introduces an additional link for the Comitative connecting it to the domain of number240 (which could best be located somewhere beyond the area occupied by AND far removed from the Instrumental). Zero-marking, on the other hand, is a privilege of the Instrumental, which is suggestive of unmarked status and a higher degree of grammaticalisation, cf. Section 10.7 below. 10.6. Distribution In the previous chapters of Part A and B, we have encountered various examples of Comitatives and Instrumentals which belong to different distribution classes. Table (B111) provides a bird’s eye view of typical examples of these differences. Initially, divergent distribution seems to speak against a paradigmatic relationship between Comitatives and Instrumentals. With a view of clarifying this point, we discuss three patterns which hitherto have not been in focus. (B111) Different distribution patterns Ex. A56 B108–109 A47–48

Language Chukot Otomí Armenian

Comitative circumfix suffix postposition

Instrumental suffix prefix suffix

Comitative-Instrumental

188 Systematicity of interaction Ex. B98 B73–75 Fn. 241 B6 B105–107 Fn. 108 B114

Language So Bamanankan Senoufo241 Saramacca Mapudungun Aleut Yidiny

Comitative preposition

pronoun preposition derivation

Instrumental suffix postposition serial verb serial verb preposition

Comitative-Instrumental circumposition circumposition preposition verbal postbase

inflection

In the Colombian language Achagua (A-type, AND-language), the Instrumental marker -ju is a regular case suffix directly attached to the noun. Yet this direct affixation is blocked for the Comitative marker -aht‰a. Notwithstanding the fact that -aht‰a too is a bound morpheme, it needs a kind of dummy base as host. This dummy base is either a possessor prefix or the semantically empty ja- (which also functions as a dummy base for a number of other spatial relators [Meléndez Lozano 1989: 47–49]). Attached to its pronominal or dummy base, -aht‰a is placed in postnominal position, cf. (B112). (B112) Achagua (B112.1) Accompaniment (Meléndez Lozano 1989: 48) tana inu Pedlo ja-aht‰ a-i INTERR come Pedro DUMMY-X-EMPH ‘Who is coming there with Pedro?’ (B112.2) Body part instrument (Meléndez Lozano 1989: 49) lija jai-ri nu-li li-kahi-ju 1SG-DAT POSS.3SG-hand-Y he give-M ‘He hit me with his hands.’ The Instrumental is more tightly integrated into the morphological paradigm of the noun, and the degree of boundedness is higher for the Instrumental than for the Comitative, as the former fuses with the noun which is in its scope. The Comitative, on the other hand, needs a kind of empty linker morpheme to interact syntactically with the noun. Thus, the Instrumental seems to be more grammaticalised than the Comitative. In the Eskimo-Aleutan A-type AND-language Yupik, there are two inflectional cases which fulfil functions within the domain of Instrumentality242, namely, the so-called ablative-modalis ending in -mek and the vialis marked by -kun. These are regular morphological case forms of Yupik nouns. The Comitative, however, allows for a wide variety of coding strategies, all of which are verb-based. The most prominent one involves the use of the verbal postbase -rtuumar-, which could be translated as ‘to act together with’, cf. (B113).

Distribution 189

(B113) Yupik (B113.1) Accompaniment (Jacobson 1984: 577) tekite-llru-uq nulia-rtuumar-mi wife-X-3SG.CONNECT arrive-have_done-3SG.INTRANS ‘He arrived with his wife.’ (B113.2) Tool I: ablative-modalis (Jacobson 1984: 519) ilavkug-mek nanilugur-mek qillrutaa rope-Y1 short_piece-Y1 tie:3>3 ‘He tied it with a short piece of rope.’ (B113.3) Tool II: vialis (Jacobson 1984: 551) urluve-rrlainarteg-gun pissula-llru-ut bow-only-Y2 hunt-have_done-3PL ‘They hunted only with bows.’ What strikes the eye first is the fact that Instrumental and Comitative relators choose different word classes as their potential host categories. Thus, they simply cannot belong to the same morphological paradigm. Furthermore, the Instrumental functions are covered by two members of the regular case inventory of Yupik nouns, i.e., -mek and -kun are paradigmatically well integrated and thus display evident signs of a high degree of grammaticalisation. For the Comitative, however, the paradigm is less restricted, if not an open class. -rtuumar- may be replaced by a wide range of other postbases whose meaning includes the component [accompaniment] plus a variety of other specifications such as, e.g., kind of activity, movement, etc. Generally, postbases have meanings of their own which are very close to that of lexical morphemes, whereas the case-suffixes do not. What the various morphemes nevertheless have in common is the fact that they are all bound morphemes: Postbases need a base which they can be attached to. In sum, the Comitative relator is relatively less grammaticalised than the Instrumental relators. It is common practice among specialists of Australian languages to classify the Comitative as a derivational category while the Instrumental normally counts as an inflectional one (Dixon 1980: 324). As the examples from Yidiny (an A-type language with Instrumental-Locative syncretism and a distinct ergative) suggest, this distinction is justified, because syntactic words marked for Comitative may additionally be marked for further case relations, cf. (B114).

190 Systematicity of interaction (B114) Yidiny (B114.1) Ornative (Dixon 1980: 325) dagul-ji-n gu waguja-n gu wagal-ji-n gu wife-X-ERG three-X-ERG man-ERG gudaaga wawa-l dog look_at-PRES ‘The man with three wives is looking at the dog.’ (B114.2) Instrumental (Dixon 1980: 299) wagal bunja-n bangaal-da waguja-n gu wife hit-PRES axe-Y man-ERG ‘The man hit his wife with an axe.’ The Comitative – when used attributively as in (B114.1) – serves as the base for the agreement morphemes which copy the case of the head noun. In this way, Comitative and Instrumental occupy different slots in the morphotactic chain of the morphological word. This is evidence against the hypothesis according to which the two categories belong to one and the same paradigm. All these differences in their distribution properties seem to create a possible objection to studying the relationship of Comitatives and Instrumentals because there are so many discrepancies as to their morpheme class membership. Admittedly, the two categories are often at a distance from one another if we look at their allocation in terms of word classes and morpheme classes. However, this kind of divergent behaviour is by no means the rule, though the many examples of Comitatives and Instrumentals belonging to the same class of morphemes suggest it, cf. above. Moreover, even where the two categories disagree on the level of distribution over classes, this disagreement reflects a certain pattern. In many of the cases highlighted in this section, the divergences of Comitative and Instrumental are such that it is the Comitative which tends to be encoded by means of the less grammaticalised strategy, while the Instrumental receives a relatively more grammaticalised expression. The pattern of adpositions for the Comitative vs. affixes for the Instrumental represents the paradigm case. Where the degree of grammaticalisation is not at issue, Comitatives require the more complex or crosslinguistically more marked strategy. The use of discontinuous relators (circumfixes, circumpositions) is more often attested for Comitatives than for Instrumentals. Owing to the fact that the two categories also enter into relations with other categories, the degree of complexity and grammaticalisation of their expressions is also determined by the network of additional relations, meaning that there might be no other way to map the content-based relation of Comitative and Instrumental unto mor-

Markedness and grammaticalisation 191

phological form except to resort either to highly complex expressions or to strategies which do not belong to the same distribution class and/or degree of grammaticalisation. A lesser degree of grammaticalisation may be reinterpreted as a kind of higher markedness (though not vice versa) when looked at from the point of view of grammar understood as a morphosyntactic component. 10.7. Markedness and grammaticalisation Wälchli (2003: 31–37) emphasises that there are many different kinds of markedness, i.e., the determination of markedness values cannot be reduced to a single criterion. With a view to checking the relative markedness of Comitative and Instrumental, we briefly review his catalogue of kinds of markedness while we indicate which of the two categories under scrutiny qualifies for the status of marked or unmarked under a certain kind of markedness. – Formal markedness In terms of the presence vs. absence of morphosyntactic markers, Comitatives and Instrumentals behave moderately differently from a crosslinguistic perspective, as Comitatives are always encoded overtly, whereas Instrumentals also allow for the absence of morphosyntactic markers. Ergo: The Comitative is more marked. – Distinctive markedness As to formal distinctiveness of the means of expression used for the two categories, no clear picture emerges, because both the Comitative and the Instrumental are frequently involved in syncretism – and syncretism implies reduced formal distinctiveness. However, in Atype and C-type languages, Comitatives are attested to partake in 113 syncretistic patterns while the Instrumental boasts of a total of 175 cases, i.e., the probability for the Instrumental to be involved in syncretism is 1.55 times higher than for the Comitative (cf. Chapter 9). Ergo: The Comitative is more marked. – Structural markedness This kind of markedness presupposes that there is a binary opposition comprising Comitative and Instrumental – a constellation Chapter 10 was devoted to prove. The evidence for an opposition to exist stems from the systematicity of interaction. What is still wanting, though, is the identification of the feature whose presence or absence is necessary

192 Systematicity of interaction to determine which of the two members of the opposition is marked or unmarked. We have yet to run into convincing candidates for the status of category value. This is perhaps evidence of a less tightly knit relationship between the two categories. Ergo: (as yet) indecisive. – Local markedness The relative frequency of an expression in a specific domain requires a corpus-based approach. This is one of the topics of Part C below. Ergo: (as yet) indecisive. – Typological markedness With regard to the crosslinguistic frequency of a feature (ideally in an unbiased sample), we have to distinguish various constellations: First of all, Comitatives as well as Instrumentals are attested in all of our sample languages (a fact which is due to our function-based definition of the two categories). If internal categorical differentiation counts here, Comitatives come in larger numbers than Instrumentals (cf. Part A, Section 4.2.2) because they allow for more subcategories. Ergo: The Instrumental is more marked, but cf. below (B116). – Textual markedness The appropriate choice of an expression in a given context requires a text-based approach. This is again one of the topics of Part C below. Ergo: (as yet) indecisive. Wälchli (2003: 31) himself points out that there are other kinds of markedness, namely pragmatic and distributional markedness, whose importance he acknowledges but which he considers irrelevant for the topic he has in mind. What this selective handling of kinds of markedness suggests is that it largely depends upon the linguist’s research interests which kind(s) of markedness will be taken into consideration for the analysis of data. Thus, it is by no means necessary to determine the markedness values on all of the above parameters and others – an appropriate subset of the catalogue may suffice. Furthermore, other (more traditional) kinds of markedness are missing from the list. Token frequency à la Greenberg (1966), semiotic principles of system-independent naturalness (Mayerthaler 1987) and system-congruity (Wurzel 1987), with their relatively elaborate apparatus of identification procedures243, are still helpful tools when it comes to deciding whether two or more categories are in an asymmetrical, i.e., hierarchically organised relationship. Owing to the fact that at least some of the criteria can best be applied to a representative text-based corpus, the actual situation is not entirely uncontroversial and thus needs to be looked at more closely in Part C. If – for the time being – we add the insights we have won

Markedness and grammaticalisation 193

by determining the parameter values of Comitatives and Instrumentals for complexity of expression, boundedness, zero-marking, distribution and morphological similarity, it nevertheless becomes more plausible to assign the status of marked category to the Comitative, which leaves the Instrumental as the unmarked member of an opposition, the essence of which, however, has so far escaped detection. The higher degree of markedness of the Comitative is perfectly in line with what the ubiquitous case hierarchies and the like suggest. The Comitative is almost always considered a peripheral category, whereas the Instrumental tends to be located much nearer to the centre or, if the Instrumental happens to show up among the peripheral categories, the Comitative is not made mention of at all or is put into brackets (Lehmann 1983; Givón 1984; Seiler 1988). The Comitative is depicted as having closer ties with full lexemes while the Instrumental is much closer to pure grammemes. The Comitative often seems to have some kind of meaning, whereas the Instrumental only has functions. Again, the fact that Comitative and Instrumental are often located at different places on the lexicon-grammar continuum neatly fits in with the ideas of grammaticalisation research: In this framework, it is commonplace to postulate the Comitative as the less or least grammaticalised (= still largely lexical) source category for other more grammaticalised categories which develop on the basis of the Comitative – among them most prominently the Instrumental (Heine, Claudi, and Hünnemeyer 1991: 158; Lehmann 1995: 112). As we have not yet tackled diachronic issues, this grammaticalisation channel and its general theoretical implications are discussed in Part D. At this point, it suffices to state that, synchronically, Instrumentals often bear evident signs of a higher degree of grammaticalisation, as they fulfil many of the conditions imposed by Lehmann’s (1995: 164) famous parameters of grammaticalisation, cf. (B115). In this table, we indicate the overall tendency of Comitatives and Instrumentals to conform to the criteria for weak or strong degrees of grammaticalisation. Brackets are indicative of less compelling evidence. As is evident from (B115), weak grammaticalisation is exclusively associated with Comitatives, whereas strong grammaticalisation is typically attested for Instrumentals. Note that this does not preclude individual Comitatives from being very strongly grammaticalised and individual Instrumentals from being rather weakly grammaticalised. What the table suggests is that if the two categories in a given language happen to be formally distinct, then it is very likely that the Instrumental is more grammaticalised than the Comitative. Table (B115), however, cannot account for the cases of syncretistic Comitatives-Instrumentals.

194 Systematicity of interaction (B115) Degree of grammaticalisation Parameter integrity

Weak

Weak vs. strong grammaticalisation Strong many vs. few semantic features INS many vs. few segments paradigmaticity COM large vs. small paradigm INS paradigmatic variability COM free vs. constrained variation INS structural scope (COM) constituent-based vs. word/stem-based INS bondedness COM free vs. bound INS syntagmatic variability (COM) movable vs. fixed position INS COM

Disregarding a number of problematic issues, we are now in a position to postulate an integrated model in which the degree of grammaticalisation and the degree of markedness of the two categories correlate. (B116) is an attempt to capture this correlation in a schematic way. (B116)

Two sides of a coin

Å---- increasing markedness decreasing -----Æ COMITATIVE

INSTRUMENTAL Å---- decreasing grammaticalisation increasing --Æ

The above discussion has shown that Comitative and Instrumental behave ambiguously: On the one hand, they do not necessarily go together in syncretistic patterns, yet on the other hand, there is evidence for a systematic interaction of the two categories. Add to this the fact that ComitativeInstrumental syncretism is at least a non-negligible pattern. It then becomes almost impossible to decide whether or not it is legitimate to consider the two categories as belonging to the same (still undefined) supercategory. The best we can say at the moment is that Comitative and Instrumental are close to one another, but not close enough to be simply lumped together independent of their formal properties. In Part C, we therefore inspect more closely an area where Comitative-Instrumental syncretism is notoriously frequent, namely, Europe. At the same time, our inspection of the European aspects of the matter at hand allow us to address a variety of issues which

Markedness and grammaticalisation 195

so far have not been dealt with adequately because the data-base did not lend itself to a corpus-linguistic analysis. This methodological shortcoming is remedied in the subsequent chapters of Part C.

Part C Europe: A continent where many things appear to be the same but turn out to be different under the looking-glass

11. A corpus-based analysis In Chapter 8, the worldwide survey of Comitatives and Instrumentals has identified Europe244 as the only area where the supposedly universal syncretism of Comitative and Instrumental occurs with a substantially high frequency. Everywhere else on the globe, the formal identity of the two categories is a dispreferred minority solution and thus Comitative-Instrumental syncretism smacks of an areal trait, a Europeanism. Two questions arise: – Firstly, does the predominance of Comitative-Instrumental syncretism have any repercussion on the linguistic map of the continent, meaning: Do all European languages behave similarly as to their Comitatives and Instrumentals, independent of their genetic affiliation, typological class membership and micro-areal location? – Secondly, what exactly do we know about the degree of similarity/dissimilarity of languages which are identified as A-type, B-type or C-type languages, meaning: If we know that two or more languages keep Comitatives and Instrumentals formally apart (or not), does this piece of knowledge allow us to formulate further-reaching claims as to their degree of similarity? These questions cannot be answered on the basis of our large global samples on which we drew on extensively in the previous chapters. Moreover, questions of this kind escape any attempt at an answer if we only go by isolated examples provided by descriptive grammars and other such linguistic sources. A different kind of empirical basis is called for, namely one which allows the observer to look at the details of actual usage of an expression in a natural context. In Part C, it is our main goal to determine how similar the languages of Europe are, going beyond the simple fact that many of them belong to the

A corpus-based analysis 197

same type, namely, the B-type. It is not sufficient to state that two languages each use one primary marker in a context-free translation of English with and thus qualify as members of the B-type. On this rather superficial level of observation, it is still absolutely unclear whether the two languages put their relators to similar or different uses outside the somewhat artificial context of dictionary translation. Going solely by the information that can be gathered while one peruses grammars and dictionaries, one might get illconceived ideas about how similar or dissimilar the languages of a sample actually are. Relators may be translation equivalents of each other when it comes to compiling bilingual dictionaries, but this does not imply that they also remain mutual translation equivalents in discourse or in written texts. Two relators which are considered to be mutual translation equivalents may diverge in their domains of use in such a way that their token frequency is radically different – even in texts which are translations from one (or both) of the two languages. Similarity on one level or in one context does not necessarily require similarity on the next level or in the next context. Theoretically, at least, there is also the possibility that languages whose primary markers behave differently in the dictionary context converge as to the use they are put to in coherent texts. With a view to determining how similar the languages of Europe are among each other and whether or not Europe constitutes a homogeneous linguistic area with respect to Comitatives and Instrumentals, we start from a statistical analysis of a sample of 64 languages and non-standard varieties currently spoken in Europe for which translations of one and the same literary text exist. This is thus a separate corpus-based comparative study. Our quantitative findings are occasionally complemented by data from other languages not included in the sample. We then proceed to three case studies – one for each of the three types postulated in Part B. The functional profile of the relevant markers is described separately for three individual languages representing the A-type, B-type or C-type. These case studies are based on (mostly) original text material and are thus absolutely independent of the text employed for the corpus-based analysis. In the final chapter of Part C, an attempt is made at integrating the results from the two separate lines of investigation in order to come to grips with the European variety of the relationship between the categories under scrutiny. This in turn serves as the basis for a short comparison with the insights which the analysis of our global samples have yielded.

198 A corpus-based analysis 11.1. The Little Prince For various reasons, we have chosen the translations (and the French original) of Le petit prince by Antoine de Saint-Exupéry (originally published in 1944) for our corpus. A small number of examples used in the previous sections were already drawn from this corpus. The ready accessibility245of the various translations together with the impressive amount of translations in numerous standard and non-standard varieties of languages (not only of those spoken in Europe) have proved to be a major argument in favour of The Little Prince. Our sample comprises 64 units, cf. (C1). Boldface marks those languages which also appear in the sample lists of Stolz (1997a) and Stolz, Stroh, and Urdze (2003 and 2005). (C1) Groups IE 55

Make-up of the corpus-based sample (Macro-)Phyla Germanic 14

Romance 20

Slavonic 11

Various 10

Non-IE 9 Uralic 4 Various 5

Standard Danish, Dutch, English, Faroese, German, Icelandic, Luxembourgeois, Norwegian (Bokmål), Swedish Catalan, French, Italian, Portuguese, Romanian, Spanish

Regional/Non-standard Alsatian, Frisian (West), Limburgisch (North), Limburgisch (South), Yiddish Aragonese, Asturian, Badiotto, Corsican, Friulian, Galician, Gascon, Ladin, Languedocien, Moldavian, Provençal, Sardinian, Sursilvan, Vallader

Bulgarian, Belarusan, Croatian, Czech, Macedonian, Polish, Russian, Serbian, Slovak, Slovenian, Ukrainian Albanian (Gheg), Romani Albanian (Tosk), Armenian (East), Breton, (Lovari) Greek, Kurdish (Kurmanþi), Latvian, Lithuanian, Welsh Estonian, Finnish, Hungarian, Saami (North) Azerbaijani, Basque, Georgian, Maltese, Turkish

The Little Prince 199

The sample is biased for Indo-European languages with a strong GermanoRomance component: 86% of the sample is Indo-European, of which 25.5% and 31.25% are Germanic and Romance languages, respectively. Germanic and Romance are also the two phyla with the largest amount of regional and non-standard varieties.246 We acknowledge that the composition of this sample does not conform to the ideals discussed in Part A, Chapter 6. However, we attempt to determine how similar languages with identical starting points (= the above mentioned dictionary translations) remain when we look at the actual usage of markers. Thus, close genetic relationship is not a disadvantage – quite the opposite: It allows us to test whether or not sister languages or sister varieties always behave similarly. Translations also reduce the necessity of determining whether two utterances are semantically equivalent, because one already knows beforehand that the texts deal with the same subject and that the text-internal chronology and organisation is at least grossly identical. A sentence in a given place in the text is very likely to contain much the same information as a sentence in the same place in a different translation. In addition, the text contains about 1,650 sentences247 and thus has medium length, meaning that the data can still be controlled even for a larger sample of languages. An undeniably positive trait of Le petit prince is the mixture of narrative text and direct speech. One encounters rather complex syntax with embedding and subordination alongside holophrastic utterances and the entire spectre of intermediate varieties of complexity. Notwithstanding its being a piece of belles-lettres, the French text itself is stylistically not particularly artificial. As the author at least pretends to have written the text for an audience of children, the text lacks the mannerisms associated with more experimentally-minded writing. Eventually, these and other arguments in favour outweigh the many methodological objections one may put forward. Of course, we are well aware of the many pitfalls there are when linguists work with translations as their empirical basis. It is this awareness of the dangers which impelled us to complement the corpus-based analysis with the textually independent case studies. As for The Little Prince, the translators may inadvertently let themselves be guided too much by the grammar, wording and style of the original, and thus the outcome is a less than natural version of the language which the text has (supposedly) been translated into. On the other hand, translators may consciously diverge from the patterns of either the original or other languages which the text has already been translated into. This happens to be the case with the Galician version, whose translator evidently made an effort to use a style (and thus indirectly also a grammar) which is

200 A corpus-based analysis equidistant from the two powerful and thus somewhat threatening Romance neighbours Spanish and Portuguese (Luna Alonso 2000). In a number of other cases, a close reading of the translations reveals that, despite the explicit claims to the contrary on the inside cover, the translation was not always made directly from the French original, at least not solely. There is evidence that sometimes extensive use was made of already available translations into languages which the translator – perhaps – was more familiar with than French. These facts guarantee that the dependence of the translations on the original is less rigid than one might assume. 11.1.1. General characteristics Before we look more closely at the statistics, it is necessary to familiarise ourselves with the general properties of the sample languages and their relators. As a matter of fact, all three language types are attested in our European subsample, albeit in rather disproportionate amounts, cf. (C2). (C2)

Phyla and types

Unit

Phylum

IE

Non-IE Total

Romance Germanic Slavic Various Uralic Various

A 0 0 8 1 1 2 12

Type B 20 13 3 7 2 2 47

Total C 0 1 0 2 1 1 5

20 14 11 10 4 5 64

As expected, the B-type by far outnumbers the other two types. With 73.43%, the B-type accounts for almost three quarters of the languages in the corpus. Within the global sample, the respective percentage for Europe was 49%. This enormous increase is probably at least partially brought about by the strong Germano-Romance bias. Indeed, Romance languages alone cover 42.55% of all B-type languages – and together with the Germanic ones the percentage clearly exceeds the 70% mark. Outside these two phyla, the B-type is still strong but by no means the predominant option (12 A-type languages and 4 C-type languages compete with 14 B-type languages). Note also that the B-type is a minority solution in the Slavic phylum, where the A-type dominates. In the non-Indo-European chapter of the sample, the three types have relatively similar frequencies: 3 A-type

The Little Prince 201

and 2 C-type languages as opposed to 4 B-type languages. With 43 (= 91%) out of 47 languages, the B-type is a typical Indo-European property. In the subsequent sections, we will also elaborate on the geographical aspects of the distribution of types in Europe. In other respects, the European subsample conforms nicely to the general picture which has emerged from the discussion in Part B. Consider the inventory of relators in Table (C3). (C3)

Relators

Language Albanian (Gheg) Albanian (Tosk) Aragonese

ComitativeInstrumental me

Instrumental

Comitative

me

WITHOUT WITHOUT I II pa pa

con

sin

Armenian (East) Asturian

-ov

het

con

ensin

Azerbaijani

ilä; -lä/-la

Badiotto

cun

-siz/-suz/sız/-süz sënza -(e)kin

aĚanc‘

Basque

-(e)z

Belarusan

Instrumental ç(ca) + Instrumental áåç

gabe

Breton Bulgarian

gant ñ /ñúñ

hep áåç

Catalan

amb

sense

Corsican

(in)cù

Croatian

senza Instrumental s/sa + Instrumental bez

Czech

Instrumental s/se + Instrumental bez

Danish

med

uden

Dutch

met

zonder

English

with

without

Estonian

-ga

ilma

Faroese

við

uttan

Finnish

-lla/-llä

French

avec

Galician

co

Georgian

-it

-ine-; kanssa

ilman sans sen

-dan ertad

gareshe

-ta -tta /-ttä

202 A corpus-based analysis Language German Ladin

ComitativeInstrumental mit cun

Instrumental

Comitative

Greek

ì‘ / ìå

Hungarian

-val/-vel

-stul/-stül

nélkül

Icelandic

með + DAT

með + ACC

án

Italian Latvian

con ar

÷ùñßò

senza bez

Luxembourmat geois Limburgisch I mèt

ouni zónger

Limburgisch II mèt Lithuanian

su + Instrumental

Maltese

WITHOUT WITHOUT I II ohne sënza

zónger Instrumental bi

be ma‘

bla

Macedonian

co

áåç

Moldavian

cu

fără

Norwegian

med

Polish

uten Instrumental z/ze + Instrumental bez

Portuguese

com

sem

Romani

-sa/-ca

bi

Romanian

cu

fără

Russian Saami (Northern) Swedish

mingƫajr

Instrumental s/so + Instrumental bez -(gu)in

-kheahtta

med

utan

Serbian

Instrumental c/ca + Instrumental áåç

Slovak Slovenian

Instrumental s + Instrumental

Spanish

s/z (ž) + Instrumental con

Sursilvan

cun

Turkish

ile; -le/-la

Ukrainian Vallader

cun

Welsh

â

bez brez sin senza

-siz/-suz/sız/-süz Instrumental ç (³ç) + Instrumental áåç sainza gyda

heb

-heapmi-

The Little Prince 203

In Europe, like elsewhere on the globe, Comitatives and Instrumentals, if distinct, share a common negative form. Where there are two translation equivalents of English without, these are never specialised for the negation of only one of the two categories.248 In Europe, two encoding strategies exhaust the attested range of possibilities, namely, adpositional marking and/or affixation – the latter being possible only in languages with inflectional case marking (many of which belong to the Slavic phylum). There, the two strategies often occur in combination. If they do, it is almost always the Comitative which is encoded in this way, whereas the Instrumental requires only a case suffix, cf. (C4).249 (C4) Frequency of encoding strategies Comitative Instrumental COM-INS Total

Adposition 2 1 35 38

Case 3 13 4 20

Combination 11 1 2 14

Total 16 15 41 72

The combination of adposition plus noun inflected for case accounts for 68% of all Comitatives. With 78%, Comitatives are by far the largest group of all attested combinations. Similarly, Instrumentals encoded solely by affixation cover 65% of all categories encoded by case affixes. At the same time, we encounter affixal case marking in 86% of all attested Instrumentals. For syncretistic markers, the relations are even more pronounced: Adpositional relators are the favourite means of encoding with a percentage of 92%. Comitatives-Instrumentals are the most frequent out of all adpositionally encoded categories, with 85%. Thus, there are clear preferences of each of the categories for different options on the expression side. Comitatives are prone to receive the more complex encoding, as their expressions often involve two components (an adposition plus a case affix). As a matter of fact, the expressions of Comitatives and Instrumentals in our European subsample behave in a predictable way, i.e., according to the patterns established on the basis of our global samples, cf. Chapter 10: If there is a difference in complexity on the expression side, it is the Comitative which tends to be expressed by the more complex relator. Of the 17 languages which do not belong to the B-type, 15 display Comitatives (including one Comitative-Instrumental) which are more complex than the respective Instrumentals (including one Comitative-Instrumental) either on the syllabic level and/or on the segmental level. There is not a single instance of the inverse constellation: The two remaining cases attest to equal segmental complex-

204 A corpus-based analysis ity. As to the parameters of bondedness and complexity, European Comitatives and Instrumentals are absolutely in line with the global tendencies. Eight out of nine A-type languages and four out of five C-type languages are case languages, i.e., they employ inflectional case morphology. There is a unilateral tendency towards a correlation of A/C-type and case language status. However, many case languages belong to the B-type (Slovenian, Saami [Northern], German [and its non-standard varieties], Faroese, Estonian, Turkish, Azerbaijani). Adpositional languages belong overwhelmingly, but not exclusively, to the B-type (Maltese is an A-type language, Welsh a C-type language). The scarcity of case languages in the Western part of Europe goes along with a minimum of A-type languages in the same region (Basque is the exception). Exactly the opposite holds for the European East: There, A-type languages abound, as do case languages. Now that we know that European relators as such are not significantly different from their equivalents outside Europe, we are in a position to formulate three working hypotheses, the validity of which will be checked against the statistical evidence in the subsequent sections: – Genealogy wins out: If two languages are genetically related to each other, their relators behave similarly in such a way that they do not converge to a higher degree with the distribution profile of relators of other unrelated or less closely related languages. – Typology wins out: If two languages belong to the same type, their relators behave similarly in such a way that they do not converge to a higher degree with the distribution profile of relators of languages allocated in a different type. – Areality wins out: If two languages are geographical neighbours, their relators behave similarly in such a way that they do not converge to a higher degree with the distribution profile of relators of languages situated in a different region. 11.1.2. Relators and their frequencies What strikes the eyes immediately is the enormous range of relator frequency in the text.250 In Table (C5), the discrepancy is most pronounced between the language with the highest token frequency and the one ranking lowest on the list, namely, the Gheg variety of Albanian and French. The Gheg relator – the

The Little Prince 205

syncretistic me – occurs almost 11 times as often as the likewise syncretistic French avec. Since all other sample languages display a significantly higher token frequency (with ratios ranging from 1.4-to-1 to 10.9-to-1 additionally indicated in Table [C5]) for their relators than the one attested for French avec, it is unlikely that the French original influenced the use of Comitatives and/or Instrumentals in the translations to any noticeable degree. As a matter of fact, the French relator is even deviant and could easily be considered to be an exceptional case. In addition, the simple fact that there are 49 different values for the frequencies is indicative of the relative diversity of the distribution profiles of the various relators. (C5) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Frequency of relators Language Albanian (Gheg) Basque Kurdish Belarusan Maltese Albanian (Tosk) Polish Russian Romanian Ukrainian Moldavian Faroese Armenian (East) Vallader Finnish Welsh Hungarian Greek Swedish Limburgisch (South) Lithuanian Azerbaijani Danish Yiddish Luxembourgeois Norwegian (Bokmål) Limburgisch (North) Portuguese

Tokens 403 360 341 225 224 219 213 201 198 192 177 166 165 157 152 145 138 134 133 129

Ratio 10.9 9.7 9.2 6.1 6.0 5.9 5.7 5.4 5.3 5.2 4.8 4.5 4.4 4.2 4.1 3.9 3.7 3.6 3.5 3.4

125

3.3

124 121 120 118

3.3 3.2 3.2 3.1

28 29 30 31 32 33 34 35

36 37 38 39 40 41 42 43 44 45 46

Language Breton Bulgarian Dutch Ladin Serbian Badiotto Sursilvan Estonian Slovenian Czech English Galician German Turkish Friulian Latvian Aragonese Croatian Macedonian Slovak Icelandic Catalan Alsatian Spanish Sardinian Italian Saami (North) Provençal

Tokens 108 106

Ratio 2.9 2.8

101

2.7

99

2.7

96 95 94 91

2.6 2.5 2.5 2.4

89

2.4

88 84

2.3 2.2

80 79 78 77 73 65 60

2.1 2.1 2.1 2.0 1.9 1.7 1.6

55

1.5

206 A corpus-based analysis

26 27

Language Asturian Frisian Georgian Romani (Lovari)

Tokens 113 111

Ratio 3.1 3.0

47 48 49

Language Corsican Gascon Languedocien French Total Average

Tokens 53 52

Ratio 1.4 1.4

37 8,122 126.9

1.0

Languages with the same genetic background do not necessarily have the same frequency values – nor do identical frequency values imply that the languages in question are close relatives of each other. Languages like the two insular North-Germanic languages Faroese and Icelandic, which are bona fide sister languages, are distanced by 28 rank positions (or 87 tokens). On the other hand, there are only 11 frequencies which are the same for more than one language. In these cases, genetic relationship is hardly ever crucial – boldface marks those cases where languages of the same phylum share identical frequency values: – Token frequency 129 shared by Germanic Limburgisch (South) and Baltic Lithuanian, – token frequency 118 shared by Germanic Limburgisch (North) and Romance Portuguese, – token frequency 111 shared by Germanic Frisian, Indo-Aryan Romani and South Caucasian Georgian, – token frequency 106 shared by Germanic Dutch and Slavic Bulgarian, – token frequency 101 shared by Romance Ladin and Slavic Serbian, – token frequency 99 shared by (Rhaeto-)Romance Sursilvan and Badiotto, – token frequency 91 shared by Germanic English and German, Romance Galician and Altaic Turkish, – token frequency 89 shared by Romance Friulian and Baltic Latvian, – token frequency 84 shared by Slavic Croatian and Macedonian, – token frequency of 60 shared by Romance Italian and Uralic Saami, – token frequency 52 shared by Romance Gascon and Languedocien. However, as languages are individuals (and translators as well251), absolute identity can hardly be expected to apply anywhere in the linguistic phenomenology of two or more languages. Moreover, similarity is not restricted to identity. Accordingly, we observe that languages with a common genetic background often accumulate around a certain value or oscillate between two not too distant values. We take up this issue below.

The Little Prince 207

In (C6), we scrutinise the ratio of relators dedicated to the Comitative and (Comitative)-Instrumental, respectively, where they are distinct, i.e., in A-type and C-type languages. (C6)

Ratio in A-type/C-type languages

Language Total Hungarian Georgian Basque Finnish Welsh Maltese Armenian (East) Polish Belarusan Russian Ukrainian Croatian Serbian Lithuanian Icelandic Czech Slovak Total

138 111 360 152 145 224 165 213 225 201 192 84 101 129 79 94 80 2,693

Tokens (Comitative-) Instrumental 137 109 332 140 129 194 137 158 162 144 134 54 63 81 44 48 30 2,096

Comitative

Ratio INS : COM

1 2 28 12 16 30 28 55 63 57 58 30 38 48 35 46 50 597

137.0 54.5 11.8 11.6 8.0 6.4 4.9 2.9 2.6 2.5 2.3 1.8 1.6 1.6 1.2 1.1 0.6 3.5

With the notable exception of Slovak, the Comitative is always the junior partner of the Instrumental. On the average, the Instrumental is 3.5 times more frequent than the Comitative when it comes to counting tokens. In Czech, the closest relative of Slovak, and Icelandic, however, the token frequencies of both categories are almost identical. Everywhere else, the Comitative cannot compete with the frequencies attested for the Instrumental. This clear tendency towards a higher frequency of the Instrumental helps us answer one of the open questions of Section 10.7 above: – In terms of Greenbergian token frequency in texts, the Comitative is clearly the less frequent and therefore marked category, whereas the Instrumental is the more frequent and thus unmarked category. Furthermore, the values in (C.5–6) clearly suggest that identity or even a higher degree of similarity of distribution profiles is not the usual case. The range of variation for the Comitative reaches from the minimal frequency of one isolated example in Hungarian to the maximum of 63 attestations in Belarusan, i.e., there is a large gap of 63-to-1 between the two languages, although both belong to the A-type. Similarly, the relatively low frequency for the Instrumen-

208 A corpus-based analysis tal in Slovak, where only 30 tokens are reported, contrasts with Basque with 332 cases, i.e., there is a gap of 11-to-1 between the two values. Under such circumstances, it must be assumed that one Comitative is not automatically like the other, nor is one Instrumental just a clone of the other Instrumental. Because we are comparing A-type, B-type and C-type languages, the values for Comitatives and Instrumentals have to be reunified as the category with the least degree of differentiation (= the syncretistic ComitativeInstrumental) imposes the limits. There are several A-type/C-type languages whose frequency values for the Comitative are still significantly higher than the ones for the syncretistic French avec and various other low frequency relators of (C5). Even the bare Instrumental values of A-type and C-type languages normally exceed the frequencies of the low ranking relators in Table (C5) – the three exceptions being Icelandic, Czech and Slovak. At the same time, Georgian and Hungarian are characterised by extremely low frequencies of their Comitatives. The striking differences in the total frequency values thus cannot be explained by the fact that some languages employ distinct relators for the two categories whereas others do not. This is not to say, however, that the distribution of A-type languages of frequency values is absolutely random. (C7)

Map of Europe

The Little Prince 209

With a view to answering the question as to which factor is decisive for the differences in relator frequencies reported above, it is necessary to get a grip on the interaction of genealogy, typology252 and areality. For this purpose, we have divided the linguistic map of Europe into four quadrants according to the North-South and West-East divide, cf. (C7).253 The quadrants reflect different degrees of genetic diversity. In the North West, only two languages do not belong to the Germanic phylum, namely, Welsh and Saami. Similarly, there are only two non-Romance languages in the South West, namely, Breton and Basque. Moreover, the typological variation within these quadrants (i.e. within the Western half of Europe) is minimal. It is tempting to correlate genetic homogeneity and typological homogeneity. However, this correlation is not necessarily a very strict one. The Eastern half of Europe suggests that genetic heterogeneity does not preclude typological homogeneity. In the northeast quadrant, Slavic, Baltic, Uralic (and Germanic Yiddish) are represented by several languages each. Nevertheless, the A-type clearly dominates. In the South East, the genetic diversity is strongest, as this quadrant hosts various Indo-European phyla and isolates alongside Afro-Asiatic, Caucasian and Altaic languages. Admittedly, there are representatives of all three types in this quadrant, but the B-type clearly dominates with twelve out of 20 languages. With the exception of Slovenian, the B-type languages cover a solid block reaching from the Adriatic Sea to East Anatolia and the Caspian Sea. This area comprises the members of the Balkan Sprachbund (both Albanian varieties, Greek, Macedonian, Bulgarian, Romanian, Moldavian), Romani, Kurdish, and the Altaic languages Turkish and Azerbaijani, and thus this group is genetically highly mixed. In what follows, we will see whether the genetic composition of the quadrants allows any predictions as to the frequency values of the relators of the individual languages. Table (C8) combines the information about absolute relator frequencies and type-membership, genetic affiliation and location of a given language. The different types (bold, italic, small caps, normal) and their combinations identify subcategories of frequency values (Stolz, Stroh, and Urdze 2003: 75 note 6): Boldface Normal Italics Bold italics SMALL CAPS BOLD SMALL CAPS

Highest frequency High frequency Higher medium frequency Lower medium frequency Low frequency Lowest frequency

n > 300 300 > n > 134 134 > n > 105 105 > n > 80 80 > n > 50 50 > n

210 A corpus-based analysis The table is additionally subdivided into eight parts according to the increase/decrease of the relator frequency in steps of 37 – the basic unit represented by the minimal frequency of French avec. For practical reasons, we subsume the languages with the three highest frequency rates under one group. (C8)

Frequencies over type, phylum and area I

(C8.1)

n • 9X

Language

Tokens

Albanian (Gheg) Basque Kurdish

403 360 341

(C8.2)

S S S

NIE B

V

E W E

225 224 219

A A

SL

N NIE

B

E E E

S S

V

213 201 198 192

A A B

SL SL

N N

SL

N

R

A

E E E E

S

5X > n > 4X

Moldavian Faroese Armenian Vallader Finnish

(C8.5)

V

A

6X > n > 5X

Polish Russian Romanian Ukrainian

(C8.4)

B

Area N S W E

7X > n > 6X

Belarusan Maltese Albanian (Tosk)

(C8.3)

Type Phylum A B C G R SL V NIE

177 166 165 157 152

B B

R

S

G

N

A

V B

S S

R

A

NIE

E W E W

N

E

4X > n > 3X

Welsh Hungarian Greek Swedish Limburgisch (South) Lithuanian

145 138 134 133 129 129

C C B B B

V

N NIE

V G G C

V

W S S

N N N

E E W W E

The Little Prince 211 Language

Tokens

Azerbaijani Danish Yiddish Luxembourgeois Norwegian Limburgisch (North) Portuguese Asturian

125 125 124 121 120 118 118 113

(C8.6)

Type Phylum A B C G R SL V NIE B B B B B B B B

Area N S W E

NIE G G G G G

S N N N N N

R R

E W E

S S

W W W W W

3X > n > 2X

Frisian Georgian Romani Breton Bulgarian Dutch Ladin Serbian Badiotto Sursilvan Estonian Slovenian Czech English German Galician Turkish Friulian Latvian Aragonese Croatian Macedonian Slovak ICELANDIC CATALAN ALSATIAN

111 111 111 108 106 106 101 101 99 99 96 95 94 91 91 91 91 89 89 88 84 84 80 79 78 77

B

G

N

C B B B B B

NIE V V SL G

N R

A

S S S S

SL B B B B

R R NIE

N N R R V

N R

G

E W W E E E W W W E W E

S S S S

SL SL SL C G

E W W

N

R

A B B

S S S

NIE

B

W

S S

G G

A

E E

N

SL SL

A B B B B B B B

W S S S S

S N

W E E E W W W

212 A corpus-based analysis (C8.7)

2>n>X

Language

Tokens

SPANISH SARDINIAN ITALIAN SAAMI (NORTH) PROVENÇAL CORSICAN GASCON LANGUEDOCIEN

73 65 60 60 55 53 52 52

B B B B B B B B

R R R R R R R

S S S S

W W W W W W W W

37

B

R

S

W

(C8.8)

Type Phylum A B C G R SL V NIE

Area N S W E S S S

NIE

N

n=X

FRENCH

Superficially, there seems to be no discernible pattern. However, this is a mistake. It is indeed possible to detect correlations between frequency values, type membership, genetic affiliation and areal location. To this end, we have to operate with the frequency-based groups instead of individual languages. In (C9), we specify how many languages of what type, phylum and region associate with what frequency-based group. (C9)

C8.1 C8.2 C8.3 C8.4 C8.5 C8.6 C8.7 C8.8 Sum

Frequencies over type, phylum and area II Type B 2 1 1 3 11 4 20 8 1 12 47 A 1 2 3 2

C G

1 3 7 2 6

5 14

R

1 2 2 7 7 1 20

Phylum SL V 2 1 1 3 1 3 7 3

11

10

NIE 1 1

N

1 2 3 1

1 3 2 9 8 1

9

24

Area S W 3 1 2 1 3 2 5 9 18 14 7 8 1 1 40 35

Sum E 2 3 4 3 5 12

29

3 3 4 5 14 26 8 1 64

With altogether 40 languages, the frequency-based groups (C8.5) and (C8.6) have the largest share of our sample. Their frequency ranges from 74 to 145 and thus covers the categories of lower medium and higher medium frequency. All other groups comprise significantly less languages. The largest single group (= [C8.6]) is also the only one including languages of all types, phyla and areas. Every other group is characterised by certain

The Little Prince 213

gaps on at least one parameter. Changing the vantage point, we notice also that the B-type is the only type which is represented in all frequency-based groups. None of the phyla has such an omnipresence – and of the four quadrants, it is only the South which recurs throughout the list. It is thus legitimate to suspect that there are preferred patterns of combination in lieu of a random distribution. For a start, consider the A-type first: Two thirds of all A-type languages belong to the groups (C8.1) through (C8.4) and are thus generally associated with high and highest frequency values. The remaining four A-type languages, however, occur in group (C8.6), where they have lower medium frequency only. All of these latter A-type languages belong to the Slavic phylum – more precisely to the Southern branch (Croatian, Serbian) and the Western branch (Czech, Slovak). If we now check the frequency values of the other Slavic languages, the following distribution emerges: All other South Slavic languages (Slovenian, Macedonian, Bulgarian) are in the same frequency-based group as the ‘deviant’ A-type languages just mentioned, whereas the members of the Eastern branch of the Slavic phylum (Belarusan, Russian, Ukrainian) occur in (C8.2) and (C8.3), i.e., high and highest frequency values apply. Polish, the remaining West Slavic language, sides with its neighbouring sister languages in the East and thus belongs to (C8.3). We interpret this constellation as the result of interaction between typological and genetic parameters. A-type languages have a tendency to have high frequency values. This is corroborated by the fact that the majority of A-type languages occur, independent of other parameters, in groups (C8.1) through (C8.4). However, this tendency is countered by genetic factors which are evidently irresistible in the case of the Southern branch of the Slavic phylum and still strong for the Western branch. Put differently, there is a typologically determined preference for a correlation of A-type and high frequency which may be overruled by genetic parameters. Geography may also play a role because the deviant A-type languages are all spoken in the southeast quadrant, whereas the Slavic languages with high frequent relators are situated in the northeast. C-type languages are restricted to the two largest frequency-based groups, namely, (C8.5) and (C8.6). They are therefore typical medium frequency languages. Admittedly, the distance within the C-type between the minimum, represented by Icelandic with 79 tokens, and the maximum, represented by Welsh with 145 tokens, is considerable. Owing to the small size of the C-type, there is no absolutely clear motivation which suggests itself. The type is genetically heterogeneous (no two languages belong to the

214 A corpus-based analysis same phylum). Geographically, only the southwest quadrant is free of C-type languages. Still, Georgian, Icelandic and Welsh are situated rather near to fringes of the European continent (but cf. Hungarian and Lithuanian). Similar to the C-type, B-type languages cumulate in (C8.5) and (C8.6), where they amount to 31 cases (= 66%). In contradistinction to the C-type, however, Btype languages occur all over the place. Notwithstanding their ubiquity, Btype languages do not go together well with high frequency values: In the groups (C8.1) to (C8.4), there are just seven B-type languages (= 15% of all B-type languages), i.e., one language less than the A-type. For the latter, eight languages make up 66.6%. Since A-type languages account for more the 50% of the members of the high frequency groups, it is overrepresented there, whereas B-type languages clearly fail to reach their predicted random share (the fictitious number of 21 languages for the first four frequency-based groups as opposed to the equally fictitious six languages for the A-type). Of the seven B-type languages in the high frequency groups, the five with the highest frequency values are all located in the southeast quadrant. Again, the three top-ranking B-type languages are members of minor Endo-European phyla. We may thus state that, on the one hand, location in the East and, on the other, not being a member of one of the major phyla are factors which facilitate high relator frequency of a B-type language. The picture is much clearer with the low frequency groups (C8.7) and (C8.8) because there the B-type is without competitors. 100% of these languages belong to the B-type (amounting to slightly less than 20% of all Btype languages) – and with seven out of nine languages, the Romance phylum predominates by far (77.8%). There is also a 100% preference for location in the southwest quadrant for low frequency languages, which is not only an epiphenomenon of the Romance predominance. The Romance languages (including French) with the lowest frequencies are all spoken around the Western part of the Mediterranean, whereas those members of the Romance whose territory does not have a Mediterranean coastline (Portuguese, Galician, Asturian in the West, Rhaeto-Romance in the centre and the Balkan Romance languages in the East) display a tendency towards relatively higher frequencies. Looking beyond low frequency groups, languages from the southwest, with one notable exception, belong to the Romance phylum. They cluster in groups (C8.5) through (C8.8), i.e., they cover medium down to minimal frequencies. However, there are two exceptions, namely, Rhaeto-Romance Vallader in group (C8.4) and the only non-Romance language of the South West, Basque, which occupies rank two of the list. Note that Basque is the only language with inflectional case

The Little Prince 215

in the South West – all other languages there are prepositional. Vallader, a member of the Western branch of the Romance phylum, behaves similarly to the two representatives of the Eastern branch in the Balkans, Romanian and Moldavian. In contradistinction to the bulk of the Romance phylum, the Balkan Romance languages display rather high frequency values and accordingly belong to group (C8.4) and (C8.3), respectively. Their next-ofkin, the East Romance language Italian, however, is a typical representative of the South West with a very low frequency (grossly a third of the Balkan Romance values). Romanian and Moldavian neither behave like the majority of their sister languages nor like their closest relative. Their deviant values thus can hardly be explained as being determined by purely genetic factors. However, the two Balkan Romance languages behave very much like their Slavic neighbours to the east, Ukrainian and Russian. More generally, Moldavian and Romanian have the properties typical of languages located in an eastern quadrant: Of the 15 languages of groups (C8.1)– (C8.4), twelve (= 80%) are located in the East. High frequency and location in the East go often hand in hand. Of altogether 30 Eastern languages, 40% display high frequencies, while there is not a single case in the two groups with the lowest frequency rates. Just like the A-type, location in the East correlates (moderately) with high frequency. To a lesser degree, similar correlations can be observed ex negativo for the North and the West: The latter is only weakly represented among high frequency groups (three languages, i.e., 20% of high frequency languages and only 8.5% of all languages allocated in the West), whereas the former is especially infrequent among the low frequency groups (one isolated example, i.e., 11% of low frequency languages and 4% of all languages spoken in the North). All this however cannot, however, explain satisfactorily why Vallader displays the unexpectedly high frequency values. Its closest relatives within the RhaetoRomance phylum (Sursilvan, Badiotto, Ladin and less so also Friulian) at least have frequency values which also go clearly beyond the ones attested for French, Italian, Spanish and Catalan. The Germanic phylum does not pose many problems because its members cluster around higher and lower medium frequencies in groups (C8.5) and (C8.6). However, Faroese, with 166 tokens, is exceptional in many respects. First of all, it exceeds the second highest Germanic value (Swedish with 133 tokens) by 33 units. It therefore deviates from the vast majority of the Germanic phylum. It also diverges from the bulk of B-type languages, as its relator is highly frequent, which is not typical of the B-type. Moreover, Faroese is located in the West, whereas it is normal for a high-

216 A corpus-based analysis frequency language to be located in the East. At the moment, we do not have sufficient proof for the hypothesis that the isolated position of Faroese among the Germanic phylum can be accounted for in terms of language contact with Goidelic and Brythonic Celtic.254 A major obstacle for this hypothesis is the fact that Icelandic does not go along with Faroese. Even more remarkable is the position of Alsatian, the German dialect spoken on the French side of the German-French border. It is likely that the relatively low frequency value of Alsatian is an effect of long term contact with Gallo-Romance, be it standard French or one of its regional varieties (but cf. Luxembourgeois, whose exposure to French influence has also been relatively strong – without affecting the relator frequency, however). Except for the outlier insular languages Faroese and Icelandic, together with the probably contact-induced frequency of Alsatian, the Germanic phylum is relatively homogeneous. In a way, the Germanic phylum reflects the prototype of our sample languages as to frequency. The minor phyla of the Indo-European macrophylum and the non-IndoEuropean languages show almost identical distribution patterns. Both are underrepresented among the low frequency groups (no attestation for the minor Indo-European phyla, only one attestation for a single non-IndoEuropean language). A third or more of both occurs in the high frequency groups. Unsurprisingly, their largest share is again with (C8.5)–(C8.6). Members of minor Indo-European phyla which have higher frequency rates are all located in the East. The Baltic and the Celtic phyla are represented by two languages each. In both cases, one language has a higher medium frequency (Welsh [145 tokens] and Lithuanian [129]), whereas the other one has a considerably lower medium frequency (Breton [108 tokens] and Latvian [89 tokens]). For both these divergences, language contact is a highly probable explanation: Breton has experienced long term contact with prestigious French, the language with the lowest frequency value, and Romanicization of Breton is obvious in many other areas of grammar and beyond (Press 1986: 3). For almost a millennium, Latvian was exposed to heavy influence on the part of Germanic languages, especially Swedish, Low German and German. The same Germanicization has also affected the neighbouring Uralic language Estonian, which differs from its sister language Finnish in much the same way that Latvian differs from Lithuanian. We come back to the diachronic aspects of this issue in Part D. Many things speak in favour of strong areality. This includes the high frequency values of Kurdish and Armenian – although some patterns escape an areally-minded analysis, and among these difficult cases we encounter the differences between Azerbai-

The Little Prince 217

jani and Turkish. Such difficulties notwithstanding, it has become clear that the distribution of languages over frequencies is not entirely random. Quite to the contrary, there are several competing motivations – typology, genealogy and areality. The hierarchy of these three parameters favours genealogy and areality to the detriment of typology – a fact which we will come back to in Section 11.2. The following section, which treats a rather surprising observation, demonstrates how strong areality can actually be. 11.1.3. Positive vs. negative relators According to Table (C3) above, all the sample languages have a common negative form for their Comitatives and Instrumentals, independent of the type the languages belong to. As a matter of fact, all the translations as well as the original of The Little Prince contain instances of the respective negative form. As with the differential frequency values observed for the positive relators, their negative counterparts are statistically rather heterogeneous: The frequency of translation equivalents of English without ranges from the maximum of 56 tokens in Kurdish down to the two attestations in Slovak. The frequencies of the two relators are never identical for any of our sample languages because the negative is always less frequent than the positive relator. There are altogether 1,092 tokens, which yields an average of slightly more than 17 tokens per language (this figure is attested only in five languages: Breton, French, Frisian, Portuguese and Romanian). Languages with low frequency rates for the negative relator are typically located in the East, as the last 14 ranks are all occupied by languages (of diverse genetic background) of the northeast and southeast quadrants (of these 14, the last twelve are all case-inflecting A-type and C-type languages). On the opposite end of the list, the situation is less clear. Languages from the East as well as the West display strikingly high frequencies. The first six ranks are occupied by languages from the South, among the 24 highest ranking languages, we encounter only two Northern ones (Welsh and Saami). The vast majority of these high-frequency languages are prepositional and belong to the B-type (as opposed to 3 A-type and Ctype languages). No Romance language has less than 14 tokens, while no Slavic language exceeds the frequency of 11. The Germanic languages oscillate between the upper limit of 18 tokens and the lower limit of 9 tokens. Needless to say, a picture emerges which resembles the one observed

218 A corpus-based analysis for the interaction of typology, genealogy and geography with positive relators. If we correlate the token frequencies of positive and negative relators for each language, a somewhat surprising pattern is discernible (Stolz and Gugeler 2000: 59–60), cf. (C10), where the languages appear in the order of decreasing positive-to-negative ratios. (C10)

Positive-negative ratio I

Language Ukrainian Belarusan Polish Slovak Georgian Czech Croatian Armenian Hungarian Albanian (Gheg) Russian Latvian Lithuanian Faroese Serbian Yiddish Basque Luxembourgeois Moldavian Romanian Bulgarian Slovenian Finnish Dutch Estonian Alsatian Limburgisch (North) Greek Danish German Limburgisch (South)

Tokens 3 5 5 2 3 3 3 7 7 22 11 5 8 10 7 9 29 10 15 17 10 9 16 11 11 9 14 16 15 11 16

Ratio 64 45 42.6 40 37 31.3 28 23.5 19.7 18.3 18.2 18 17.8 16.6 14.4 13.7 12.4 12.1 11.8 11.6 10.6 10.5 10.1 9.6 9.5 8.5 8.4 8.4 8.3 8.2 8.1

N N N N

S

Location W

S S S S S S S N N N N

W S

E E

N S N

W W

S S S S N N N N N

E E E E E W E W W

S N N N

E E E E E E E E E E E E E E

E W W W

The Little Prince 219 Language

Tokens

Ratio N

Romani (Lovari) Icelandic Macedonian Norwegian Swedish Maltese Portuguese Frisian English Breton Kurdish Albanian (Tosk) Vallader Ladin Welsh Galician Badiotto Sursilvan Provençal Sardinian Friulian Catalan Italian Azerbaijani Turkish Asturian Aragonese Languedocien Corsican Gascon French Saami (North) Spanish

14 10 11 16 18 32 17 17 14 17 56 37 27 20 29 19 22 23 14 18 25 22 18 38 28 36 29 18 20 22 17 31 38

7.9 7.9 7.6 7.5 7.4 7 6.9 6.5 6.5 6.3 6.1 5.9 5.8 5.1 5 4.8 4.5 4.3 3.9 3.6 3.5 3.5 3.3 3.3 3.2 3.1 3 2.9 2.6 2.3 2.1 1.9 1.9

S S

N

Location W

E E

W S

N N

E W W

S S N N S S S S S N S S S S S S S S S S S S S S S S N S

E W W W W E E W W W W W W W W W W W E E W W W W W W W W

Irrespective of the absolute figures for the token frequencies of positive and negative relators, the ratio between the two values clearly follows an areallydetermined rule. Languages spoken in the East may have rather different absolute token frequencies which diverge considerably, like the ones reported for Albanian (Gheg), Russian and Latvian, with twenty-two, eleven and five tokens, respectively. Nevertheless, their positive-negative ratios are relatively

220 A corpus-based analysis similar to each other, as they oscillate between 18.3 and 18.0. Likewise, Asturian, Aragonese and Languedocien all have relatively high but nevertheless significantly different token frequency values, namely 36, 29 and 18. These differences notwithstanding, their positive-negative ratios converge, as they all approximate 3.0. As a matter of fact, languages of different quadrants behave differently, cf. (C11). (C11)

Positive-negative ratio II

Frequency n > 10 10 > n > 6 6>n>1 Sum

N 10 12 2 24

S 13 7 20 40

W 3 13 19 35

E 20 6 3 29

Sum 23 19 22 64

Of the 23 languages with a ratio higher than 10, 20 (= 86% of this frequency group) are located in the East. These 20 languages account for 69% of all Eastern languages. Languages from the West are a clear minority in the higher ratio group, although they outnumber Eastern languages in the two other ratio groups: Location in the West occurs more than twice as often for medium ratio languages and more than three times as often with low ratio languages. In the latter group, the relation of the high ratio group is exactly reversed: 86% Western languages as opposed to 14% Eastern languages. Of all Western languages, only 8.5% display high ratios. North differs from South in much the same way. With high ratio languages, there is somewhat of an equilibrium between the two areas. This stalemate is turned to a clear majority of Northern languages in the medium ratio group (63%) and an overwhelming majority of Southern languages in the low ratio group (90%). Only 8% of all Northern languages go along with low ratios, whereas 50% of all Southern languages are compatible with low ratios. If we combine the values for the locations according to quadrants, the results are telling, cf. (C12). (C12)

Average ratio per quadrant

Quadrants N S Average

W 8.16 4.29 6.2

E 26.5 15.3 20.9

Hemisphere Average 17.3 9.8 Hemisphere

There is again a cline from the South West to the North East and from the North West to the South East. The ratios are generally much higher in the

The Little Prince 221

East as opposed to the West, and much higher in the North than in the South. This is remarkable, because we find languages with extremely high frequencies of the negative relator alongside neighbouring languages with rather low frequency in the South East, such as Kurdish with 56 tokens and Georgian with three tokens, and in the North West as well, such as Welsh with 29 tokens and Faroese with ten tokens. In the South West, Basque stands out again with 29 tokens, as opposed to Provençal with 14 tokens. And the North East is the home of Ukrainian with three and Finnish with 16 tokens. This picture largely corresponds to the distribution of the average frequencies of the positive relators over quadrants, cf. (C13). (C13)

Average frequency (positive relator) per quadrant

Quadrants N S Average

W 115.1 97.4 106.25

E 148.1 154.5 151.1

Hemisphere Average 131.6 126 Hemisphere

Irrespective of the fact that this time, the South East has the highest average of all quadrants, there is still a pronounced difference between the West and the East. In addition, the cline from the South West to the North East and from the North West to the South East is an undeniable truth. Therefore, we may legitimately conclude that it makes a difference structurally where in Europe a language is situated. These differences between the West and the East and the North and the South of Europe fit in well with similar observations made by students of the European linguistic area, according to which (certain) subareas behave differently on various parameters in a significant way. The evidence thus speaks in favour of a distinction between SAE languages and peripheral members of the European area situated in the East (Haspelmath 2001). What makes our data more interesting is the fact that they reveal a continuum-like character of similarity/dissimilarity relations in the European area. The quadrants are characterised by markedly different average frequency values and ratios. However, these differences are not of a categorical nature. As a matter of fact, there is also an internal variation within the quadrants – and this internal variation at times corresponds to a gradual increase/decrease of values and ratios mapped onto the centre-periphery dichotomy. This means that those languages within a quadrant which are located nearer to the centre of the continent tend to have traits in common

222 A corpus-based analysis with their neighbours in adjacent quadrants, while those languages of the same quadrant which are more remote from the centre of the continent tend towards the extreme values and ratios. The most striking examples are perhaps the leaning of North Western Alsatian towards its neighbours in the South West quadrant, of North Western Swedish towards its neighbour in the North East, Finnish, and the relative convergence of Czech, Slovak and Slovenian (all located in the South East quadrant) with neighbouring German in the North West. The Westernmost Ibero-Romance languages (especially Portuguese, Asturian) depart again from the low frequency values of their sister languages next door – just like Faroese, as opposed to Icelandic and Norwegian, and Welsh, in comparison to English or Breton. This could be seen as evidence of a periphery effect. There are, of course, languages whose values and ratios do not conform to this streamline pattern. The fact that there is a similarity of ratios even though the absolute frequency values do not necessarily converge, calls for an extension of our viewpoint: It is not sufficient to operate only with relator frequencies and their ratios if one wants to determine to what degree two or more languages can be said to differ from each other or in what sense they resemble each other. If we want to understand the reasons for the divergence or convergence of certain languages in terms of their frequencies and ratios, we have to review the functional domain of their relators – and this can only be executed on the basis of a comparison of the usage the various relators are put to in the sample text. 11.2. On being (dis)similar in Europe 11.2.1. The similarity coefficient Relator frequencies alone do not allow us to make conclusive statements about the degree of similarity or dissimilarity of two or more languages. Identical or similar quantities may not imply identical or similar qualities. The fact that one language may use a relator with much the same frequency as another language uses its translation equivalent does not necessarily mean that the two relators are employed for the same purposes, i.e., their functional domains may be rather different. With a view to determining how similar our sample languages are in actual fact, we compare the employment of Comitatives and Instrumentals in individual contexts for the entire sample. For this purpose, we follow the lead of Altmann and

On being (dis)similar in Europe 223

Lehfeldt (1973), without, however, claiming any more than just a layman’s competence in the realm of quantitative linguistics. We are confident, however, that our simplified approach is sufficient to prove the major points. The procedure we adopt runs as follows: – – – – –

If in a sentence of The Little Prince a given language makes use of its primary marker of Comitative/Instrumental, we check every other sample language for the same sentence.255 If other languages also employ their primary markers in the same sentence, we determine whether or not the relator is used in the same context.256 If the relators of two languages in one and the same sentence are indeed used in the same context, we register this fact as one instance of similarity under both languages. If several instances of similarity for the same pair of languages can be identified, they are added up for each language to yield the total number of cases of convergence of the two languages compared. When all sentences of the sample text are checked, the similarity coefficient for every possible combination of two languages of the sample is calculated by the following formula (Stolz, Stroh, and Urdze 2003: 77 note 7): X(a) X(b) X(a + b)  +  =  x 100 = coefficient Y(a) Y(b) Y(a + b) The variables stand for: Y = total of tokens in a given language: X = number of tokens identical to the other language: and (a) and (b) identify the two languages to be compared. In a fictitious example, L(a) has 100 tokens altogether. Of these 100 tokens, ten are functionally identical to the ones of L(b). L(b), however, has a total of altogether 200 tokens. Thus, the coefficient is 6.666: 10  100

+

10  200

x 100

= 6.666

The coefficient allows us to disregard the otherwise very interesting effect that similarity is an asymmetrical concept. As the above formula suggests, the share of convergent tokens is much higher with L(a) than with L(b) – and accordingly, L(a) is more similar to L(b) than L(b) is similar to L(a). This

224 A corpus-based analysis differential relation is one of the reasons why it is technically hardly possible to plot the similarities onto a two-dimensional map.257 For practical reasons, we only operate with the coefficient in the remainder of this section. Reviewing the coefficient for all 64²–64 = 4,032 language pairs, we notice immediately that there is no coefficient of 100, nor is there any evidence for the theoretically possible minimum of 0, i.e., no two languages are the same as to their usage of Comitatives and Instrumentals, and none employ the relators in absolutely incompatible ways. All languages of our sample resemble each other at least to some extent – however small it may be. At the same time they all retain more or less pronounced traits of individuality. The coefficient oscillates between the maximum of 80.16 reported for the two varieties of Limburgisch and the minimum of 4.25 calculated for the pair Basque-Georgian. Thus, there is still a wide margin of over 75 points. For convenience, in our exemplification, we focus on the values on the two extremes of the scale: In the subsequent paragraphs, we discuss the five highest ranking and the five lowest ranking positions of the coefficient for each of our sample languages. We start with a survey of each genetically defined group of the sample, which will subsequently be evaluated by way of comparing the values. Then we proceed to formulate generalisations on this basis. Without denying that some of the data are suspiciously messy, for the sake of the argument we focus especially on those sets of data which lend themselves to a positive interpretation. The less conclusive part of our statistics will be discussed in some detail in Stolz, Stroh, and Urdze (forthcoming). Table258 (C14) contains the coefficient for each member of the Germanic phylum. If we take up the working hypotheses formulated above (Section 11.1.1), then one possible prediction is that, among the languages with a comparatively high coefficient, we will find only or predominantly close relatives – in this case, members of the Germanic phylum. On the other hand, it is highly unlikely that Germanic languages will show up among those languages characterised by a low coefficient. (C14)

Coefficient: Germanic phylum 1

2

3

4

5

59

60

61

62

63

Als 35.22

Fris 51.06

Dut 50.27

Spa 49.33

Ger 48.81

Letz 47.47

Geo 18.09

Alb-G 16.67

Kur 16.27

Arm 15.7

Bas 11.9

Dan 34.1

Nor 62.04

Swe 55.81

Far 51.55

Ger 47.22

Dut 45.02

Arm 18.62

Aze 18.4

Geo 17.8

Kur 17.6

Bas 9.07

Dut 36.27

Fris 64.52

Ger 54.82

Spa 50.28

Als 50.27

Mac 48.42

Geo 19.35

Alb-G 19.25

Arm 18.45

Kur 17.45

Bas 11.59

On being (dis)similar in Europe 225 1

2

3

4

5

59

60

61

62

63

Eng 32.2

Spa 48.78

Ita 47.68

Gal 43.96

Ger 43.96

Sur 43.16

Arm 17.19

Fin 16.46

Geo 15.84

Kur 15.28

Bas 8.43

Far 24.76

Dan 51.55

Nor 42.66

Swe 37.46

Lim-S 31.86

Sur 30.94

Arm 13.29

Fin 13.21

Aze 12.37

Geo 11.55

Bas 7.6

Fris 36.27

Dut 64.52

Lim-S 55.83

Ger 55.45

Lim-N Letz 54.15 51.72

Alb-G 20.62

Geo 18.92

Arm 18.12

Kur 17.7

Bas 11.46

Ger 41.38

Letz 64.15

Mac 61.71

Spa 58.54

Lim-S 58.18

Gher 57.29

Alb-G 21.46

Arm 21.09

Geo 20.79

Kur 16.2

Bas 12.42

Ice 26.71

Als 41.03

Spa 40.79

Ger 40

Ara 38.32

Sur 35.96

Alb-G 14.11

Fin 13.85

Kur 11.9

Arm 8.2

Bas 7.74

Letz 35.93

Ger 64.15

Mac 53.66

Fris 51.72

Lim-S 51.2

Gher 47.75

Alb-G 19.08

Arm 18.18

Geo 16.38

Kur 15.58

Bas 12.47

Lim-N Lim-S 35.56 80.16

Fris 54.15

Ger 53.59

Spa 50.26

Dut 47.7

Geo 20.09

Kur 18.74

Alb-G 18.43

Arm 17.67

Bas 12.55

Lim-S 36.53

Lim-N Ger 80.16 58.18

Fris 55.83

Letz 51.2

Dut 47.66

Geo 21.67

Alb-G 20.68

Arm 19.73

Kur 19.57

Bas 12.68

Nor 31.2

Dan 62.04

Swe 49.8

Ger 43.6

Cze 42.99

Far 42.66

Geo 18.18

Alb-G 16.83

Fin 16.37

Kur 15.62

Bas 7.08

Swe 34.02

Dan 55.81

Ger 53.57

Nor 49.8

Letz 46.46

Sur 44.83

Alb-G 19.4

Aze 19.38

Arm 18.12

Geo 17.21

Bas 8.52

Yid 36.64

Spa 51.78

Mac 50.96

Ger 49.3

Ara 47.17

Bul 46.09

Arm 20.76

Alb-G 20.49

Geo 19.57

Kur 19.35

Bas 14.46

The hypotheses are partly borne out by the values because there are indeed no Germanic languages from ranks 59 to 63. Moreover, three of the languages (Danish, Frisian and Limburgisch [South]) allow only Germanic sister-languages on the five highest ranks. The close relationship among the North Germanic languages (with the exception of Icelandic) is reflected by the fact that Danish, Norwegian, Swedish and Faroese are favourite partners of each other. Likewise, the pairings (of mutual preference) GermanLuxembourgeois, Dutch-Frisian and Limburgisch (North)-Limburgisch (South) corroborate the importance of genealogy as a determining factor of similarity. However, English and Yiddish do not have a particularly Germanic leaning. With the sole exception of German, which is the only Germanic partner of both these exceptional languages, English prefers members of the Romance phylum, whereas Romance and (Balkan) Slavic languages are equally strong partners of Yiddish. The English closeness to Romance is shared to a slightly lesser extent by Icelandic. For German, Germanic languages compete with Slavic and Romance ones, too. In the remainder of the Germanic phylum, Germanic sister languages form the

226 A corpus-based analysis numerically strongest group for the five highest ranks. Note that none of the languages – be they Germanic or not - attested on the five first ranks ever appear a single time on the lowest five ranks. With the exception of Czech as one of the partners of Norwegian, only B-type languages appear in the first five ranks for all Germanic languages. Since Icelandic belongs to the C-type, there are altogether six pairings of languages of different typological class membership. In 11 out of 14 cases, the relations cross the boundaries of our quadrants. As all members of the Germanic phylum are located in the North West, these border crossings trivially involve association with members of other phyla. English prefers the South West (like Icelandic), whereas Yiddish divides its preferences over the South West, the South East and the North West. Only Danish, Frisian and Limburgisch (South) remain faithful to their quadrant. With the same genetically oriented hypotheses in mind, we now extend our view to include the Romance phylum. Owing to the fact that some Germanic languages show a certain predilection for converging with certain Romance languages, the idea suggests itself that we will also encounter these Germanic languages among the preferred partners of some members of the Romance phylum. (C15) Ara 39.63 Ast 30.97 Bad 33.16 Cat 37.95 Cor 29.84 Fre 31.08 Friu 34.88 Gal 36.53 Gas 34.95

Coefficient: Romance phylum 1 Spa 78.26 Spa 48.39 Gher 50 Ara 71.08 Ita 51.33 Gas 71.91 Ita 48.32 Spa 63.41 Fre 71.91

2 Cat 71.08 Ara 46.77 Ger 48.42 Spa 62.25 Spa 50.79 Ita 59.79 Sar 48.05 Ara 58.1 Ita 60.71

3 Gal 58.1 Gal 43.14 Ita 46.54 Gal 55.62 Lan 47.62 Spa 56.36 Spa 46.91 Cat 55.62 Spa 56

4 Ita 54.05 Mac 42.64 Spa 46.51 Gas 53.85 Fre 46.67 Lan 56.18 Sur 45.74 Ita 52.98 Cat 53.85

5 Ger 51.4 Als 40 Ara 44.92 Ita 50.72 Friu 45.07 Pro 52.17 Gas 45.39 Fre 50 Lan 53.85

59 Rus 20.76 Alb-G 17.44 Rus 20 Rus 20.07 Rus 14.96 Geo 13.51 Arm 19.69 Geo 18.81 Arm 15.67

60 Geo 20.1 Fin 16.6 Kur 18.64 Kur 19.57 Fin 14.63 Alb-G 12.73 Fin 19.09 Arm 17.97 Kur 15.27

61 Kur 19.58 Kur 14.54 Geo 16.19 Alb-G 19.13 Kur 13.71 Kur 12.7 Rus 17.93 Bie-R 17.72 Rus 15.02

62 Arm 18.97 Geo 12.5 Alb-G 15.94 Arm 18.93 Alb-G 12.28 Rus 11.76 Alb-G 17.89 Rus 16.44 Alb-G 14.07

63 Bas 14.73 Bas 11.42 Bas 12.2 Bas 11.87 Bas 7.26 Bas 6.05 Bas 10.69 Bas 11.53 Bas 7.28

On being (dis)similar in Europe 227

Gher 35.41 Ita 35.16 Lan 32.97 Mol 30.66 Por 28.59 Pro 24.94 Rum 30.67 Sar 32.45 Spa 41.6 Sur 37.23 Val 32.96

1 Ger 57.29 Spa 61.65 Fre 56.18 Rum 51.2 Spa 43.98 Fre 52.17 Mol 51.2 Gal 50 Ara 78.26 Ger 56.84 Sur 48.44

2 Bad 50 Gas 60.71 Gas 53.85 Ara 40.75 Gal 40.19 Gas 44.86 Gre 41.57 Cat 48.95 Gal 63.41 Spa 53.49 Gher 44.96

3 Sur 50 Fre 59.79 Ita 50 Letz 39.6 Ara 39.81 Lan 41.12 Gal 39.45 Friu 48.05 Cat 62.25 Gher 50 Ara 44.9

4 Cat 48.04 Ara 54.05 Spa 49.6 Gal 39.55 Mac 39.6 Ita 38.26 Ara 38.46 Ita 48 Ita 61.65 Ara 49.2 Spa 43.48

5 Letz 47.75 Gal 52.98 Cat 49.23 Gher 38.85 Cat 38.78 Friu 37.5 Val 38.31 Gas 47.86 Ger 58.54 Cat 48.59 Lim-S 43.36

59 Rus 19.87 Arm 16.89 Fin 18.63 Alb-G 20 Rus 16.3 Fin 13.53 Rus 20.05 Bie-R 17.24 Alb-G 21.43 Arm 21.21 Fin 19.42

60 Alb-G 19.05 Rus 16.86 Geo 18.4 Arm 17.54 Fin 15.56 Kur 12.12 Arm 19.83 Arm 16.52 Geo 19.57 Kur 20.91 Rus 18.99

61 Geo 18.87 Geo 16.37 Kur 14.25 Fin 17.02 Arm 14.13 Geo 12.05 Fin 17.71 Rus 14.29 Kur 18.84 Fin 20.72 Arm 18.01

62 Arm 18.80 Alb-G 15.55 Alb-G 13.19 Geo 15.97 Geo 13.97 Alb-G 11.35 Geo 14.24 Alb-G 13.68 Arm 18.49 Geo 20 Geo 15.67

63 Bas 13.02 Bas 7.14 Bas 10.68 Bas 12.66 Bas 10.04 Bas 7.23 Bas 12.54 Bas 10.35 Bas 12.47 Bas 12.64 Bas 13.54

There are ten languages which only have close relatives on the five highest ranks. Eight languages still have four Romance sisters in the same positions and only two languages (Asturian and Ladin) make do with just three intraRomance pairings. In stark contrast to the Germanic phylum, however, Romance languages generally prefer partnerships within the same phylum over outward-looking relations. Where non-Romance languages come in, they are predominantly members of the Germanic phylum (particularly German, with five appearances), with the Slavic phylum being involved twice (Macedonian) and the minor Indo-European phyla just once (Greek). We thus may state a higher degree of genetic solidarity of Romance languages. Again, the pattern observed for the Germanic phylum above also holds for the Romance languages: None of the languages which occur on the five highest ranks is ever among the five lowest ranks and vice versa. In terms of geography, Romance languages leap relatively seldom over boundaries: In 9 out of 20 languages, associations involving two different quadrants are attested. The number of boundary-crossings is especially high with the two Balkan Romance varieties Romanian and Moldavian. However, in these cases, there is no automatic parallel association to members

228 A corpus-based analysis of different phyla. Romanian and Moldavian still prefer pairings with other Romance languages which, however, happen to be located in the neighbouring quadrant. Moreover, Romanian and Moldavian are – predictably – each other‘s number one favourite, whereas neither of the two makes it onto the list of the first five partners of any of the other Romance languages. This supports the idea that the easternmost Romance languages form a unit at least moderately distinct from the bulk of the phylum. If we look at the first position alone, we immediately see that the majority of the Romance languages have an inclination towards associating with a close relative that is also spoken in the immediate vicinity. Only four languages deviate from this pattern, namely Italian, which takes Spanish as the preferred partner (whereas, for Spanish, Aragonese is unsurprisingly more important), Sardinian, which opts for Galician (which in turn chooses Spanish), and two Rhaeto-Romance languages, Ladin and Sursilvan, which both give preference to German. This latter pattern may reflect the impact exerted by co-territorial German in South Tyrol and Switzerland on the two minor languages. The Romance phylum displays a predilection for associations with members of the same type class, even more so than the Germanic phylum, because the 100 pairings are all examples of combinations of two B-type languages. The Slavic phylum largely conforms to the above patterns – with a few exceptions, cf. (C16). (C16) Bie-R 23.44 Bul 38.35 Cro 26.69 Cze 35.67 Mac 40.44 Pol 24.78 Rus 22.2 Ser 35.44

Slavic phylum 1 Rus 56.81 Mac 63.16 Ser 52.97 Slok 54.02 Bul 63.16 Cze 33.88 Ukr 62.6 Mac 57.3

2 Ukr 53.24 Lat 56.41 Mac 46.43 Ger 52.97 Ger 61.71 Bul 33.86 Bie-R 56.81 Cro 52.97

3 Ser 34.36 Ger 54.82 Bul 44.21 Mac 52.81 Ser 57.3 Yid 33.83 Lat 32.41 Cze 51.28

4 Pol 33.33 Spa 51.4 Cze 43.82 Ser 51.28 Letz 53.66 Bie-R 33.33 Arm 31.15 Bul 51.21

5 Cze 31.97 Ser 51.21 Slok 41.46 Bul 51 Slok 53.66 Mac 32.32 Ser 31.13 Lat 48.42

59 Alb-G 14.65 Fin 24.03 Arm 17.67 Aze 21 Arm 23.29 Fin 15.89 Fin 13.6 Arm 18.05

60 Fre 14.5 Geo 23.04 Geo 14.36 Geo 19.51 Geo 22.56 Arm 15.87 Fre 11.76 Geo 17.92

61 Pro 14.29 Alb-G 21.61 Kur 14.12 Alb-G 16.9 Alb-G 21.77 Pro 14.93 Alb-G 11.59 Alb-G 17.86

62 Fin 12.73 Kur 17.9 Alb-G 13.14 Kur 15.17 Kur 16 Bas 13.62 Kur 11.07 Kur 16.29

63 Bas 10.94 Bas 14.59 Bas 10.36 Bas 11.01 Bas 12.16 Geo 12.35 Bas 9.63 Bas 10.85

On being (dis)similar in Europe 229

Slok 34.02 Slov 35.57 Ukr 25.4

1 Ger 56.14 Mac 53.63 Rus 62.6

2 Cze 54.02 Cze 49.74 Bie-R 53.24

3 Mac 53.66 Slok 49.14 Ser 38.23

4 Slov 49.14 Ger 47.31 Lat 37.01

5 Bul 46.24 Ser 46.94 Bul 35.57

59 Arm 17.96 Arm 21.54 Kur 15.38

60 Alb-G 16.98 Geo 18.45 Geo 14.52

61 Geo 14.66 Alb-G 18.07 Fin 13.95

62 Kur 12.35 Kur 16.51 Alb-G 13.11

63 Bas 8.18 Bas 11.43 Bas 9.42

Genetically, Croatian and Belarusan have the most homogeneous group of languages on the highest ranks, as they exclusively associate with sister languages. The majority of the Slavic languages allow for one non-Slavic language on the same positions; Russian and Macedonian, with two partners from outside their own phylum, are more tolerant. Bulgarian is the only Slavic language with a non-Slavic majority in the upper regions of the coefficient. Thus, only two out of eleven languages remain purely Slavic when it comes to preferred associations. Among the non-Slavic partners, German is again the most frequent with five appearances, closely followed by Latvian with four attestations. It is worth noting that there is only one Romance language on the highest ranks, namely Spanish as the fourth best among the partners of Bulgarian. Romance languages are, however, attested repeatedly at the opposite end of the scale, where French and Provençal turn up twice each. Interestingly, Russian violates the rule of the other major IndoEuropean phyla, according to which one and the same language may not occur at both ends of the scale. Armenian is one of Russian’s preferred partners, whereas it is among the least attractive partners for six other Slavic languages. In nine cases, the first position is occupied by a close relative spoken close by. While the combinations Belarusan-Russian, BulgarianMacedonian, Croatian-Serbian, Czech-Slovak, etc. meet our expectations (meaning that East Slavic associates with East Slavic, West Slavic with West Slavic and South Slavic with South Slavic), only Slovenian, with its preference for the slightly more distant Macedonian and Slovak leaning towards German, deviate from this pattern. Geographically, there are altogether 17 border-crossings – including seven cases which remain inside the phylum. More complicated is the situation as to the typological solidarity of the Slavic languages. There are many pairings of languages which belong to different types. Only Belarusan is typologically consistent in the sense that it exclusively goes along with members of its own A-type. Everywhere else, B-type mixes with A-type and the other way round. Slavic languages thus obey genetics more than they obey geography or typology. The three major phyla converge in so far as they do not allow for non-IndoEuropean languages on the highest ranks. Minor Indo-European phyla are

230 A corpus-based analysis strong only for the Slavic phylum (Armenian and Latvian) – or generally in the East, as Romanian also associates with Greek. Table (C17) identifies the degree of deviations observed for the three major phyla on the three parameters. (C17)

Degree of deviations

Phylum Germanic Slavic Romance

Genealogy 32.8% 23.6% 11%

Geography 34.3% 31% 16%

Typology 8.57% 45.45% 0%

A third of the combinations of Germanic languages do not stay within the phylum or the quadrant. Yet combinations with members of a different type are relatively marginal. The Slavic languages display a slightly higher degree of genetic and regional solidarity than the Germanic languages, but in contradistinction to both Germanic and Romance, they do not care much for typological consistency: Almost half of the combinations involve members of different types. Solidarity among Romance languages is very high – and this applies also to geography. Yet what is even more pronounced is the 100% watertight typological solidarity: There is not a single case of association with a member of a different type class. In sum, Romance languages show the highest degree of homogeneity on all three parameters. Before we look at the genetically more heterogeneous groups, it is in order that we check the quality of the favourite language pairings in the genetically more homogeneous groups. We distinguish bilateral (or mutual/reciprocal) preferences from unilateral ones. Bilateral preferences imply that the two languages are the top ranking candidates on the list of coefficient values for each of the pair. Unilateral preference applies to those cases where B is the preferred partner of A but not vice versa. In such cases, we extend our view to the preferred partner of B and indicate the rank position (among the five highest ranks) this language C occupies on the list for A and vice versa, cf. (C18). (C18)

Bilateral vs. unilateral preferences

(C18.1) Germanic phylum Bilateral AlB DanlNor DutlFris GerlLetz

AoB AlsoFris EngoSpa FaroDan

Unilateral BoC FrislDut SpalAra DanlNor

AoC 2 0 2

Rank CoA 4 0 5

On being (dis)similar in Europe 231 Bilateral AlB

Unilateral BoC

AoB

Lim NlLim S

IceoAls SweoDan YidoSpa

AlsoFris DanlNor SpalAra

AoC 0 3 4

Rank CoA 0 2 0

AoC 2 2 2 2 3 2 4 3 2 0 0 0

Rank CoA 0 0 3 0 0 3 4 0 0 0 0 0

(C18.2) Romance phylum Bilateral AlB

Unilateral BoC

AoB

SpalAra FrelGas MollRum

AstoSpa BadoGher CatoAra CoroIta FriuoIta GaloSpa ItaoSpa PoroSpa ProoFre SaroGal SuroGer ValoSur

SpalAra GheroGer AralSpa ItaoSpa ItaoSpa SpalAra SpalAra SpalAra FrelGas GaloSpa GeroLetz SuroGer

(C18.3) Slavic phylum Bilateral AlB RuslUkr BullMac

Unilateral AoB Bie-RoRus CrooSer CzeoSlok PoloCze SeroMac SlokoGer SlovoMac

BoC RuslUkr SeroMac SlokoGer CzeoSlok MaclBul GerlLetz MaclBul

AoC 2 2 2 0 5 0 0

Rank CoA 2 0 0 0 5 0 0

In the Germanic phylum, bilateral preferences are more common than among the two other phyla. Nevertheless, their importance for the whole network of relations connecting languages to each other is relatively small. In all three phyla, the unilateral preferences outnumber the bilateral ones. This fact is, of course, detrimental to any attempt at predicting similarity, because asymmetry is involved. However, a certain degree of probability is still guaranteed, because in the majority of the unilateral cases (= 17 cases),

232 A corpus-based analysis the third language C (= the preferred partner of B) ranks high on the coefficient-based list of A, too. Very interestingly, the reverse is not true: A appears much more rarely on the first five positions of C, namely, only seven times as opposed to 17 for C as one of A’s favourites. It is possible, however, to formulate the following probabilistic implication: –

If B is the preferred partner of A and if C is the preferred partner of B, then there is a high probability that C will also be among the high ranking partners of A.

It remains to be seen whether this implication also holds with languages which form part of genetically heterogeneous groups. The remaining two groups are indeed genetically heterogeneous. We may thus expect a lesser degree of genetic solidarity, if at all. Among the minor Indo-European languages, there are three combinations of two closely related languages: the two co-territorial Albanian varieties, Latvian and Lithuanian (Baltic), which are also next-door neighbours of each other, Breton and Welsh (Brythonnic Celtic), and Kurdish and Romani, which are (very) remotely related via the Indo-Iranian background. Greek and Armenian are isolates. Does this heterogeneity have an influence on the distribution of high coefficient values over languages? (C19) helps us to find an answer to this question. (C19)

Minor Indo-European phyla 1

2

3

4

5

59

60

61

62

63

Alb-G 16.98

Alb-T 24.44

Sur 23.51

Mac 21.77

Bul 21.61

Gre 21.6

Saa 11.23

Arm 10.92

Geo 10.51

Fin 10.45

Bas 8.91

Alb-T 24.47

Gre 35.69

Bul 35.69

Mac 33

Rum 32.13

Fris 30.3

Saa 15.05

Pro 14.6

Fin 14.56

Wel 14.29

Bas 10.71

Arm 18.6

Aze 33.79

Rus 31.15

Ukr 26.89

Bie-R 26.67

Bul 24.35

Fin 12.62

Kur 11.46

Alb-G 10.92

Bas 8.76

Ice 8.2

Bre 32.78

Ara 47.95

Lan 46.25

Gal 44.22

Spa 44.2

Cat 44.09

Geo 19.18

Kur 19.15

Arm 19.05

Alb-G 17.22

Bas 10.68

Gre 30.98

Mac 50.46

Bul 45

Ger 43.56

Yid 41.86

Rum 41.57

Bie-R 18.38

Fin 18.18

Arm 18.06

Rus 16.12

Bas 13.36

Kur 16.52

Rum 24.12

Friu 23.72

Mal 23.01

Sur 20.91

Val 20.88

Fin 11.76

Bas 11.7

Arm 11.46

Wel 11.11

Rus 11.07

Lat 34.4

Bul 56.41

Ger 52.22

Mac 50.87

Ser 48.42

Cze 46.99

Fin 21.58

Geo 18

Alb-G 16.26

Kur 13.95

Bas 12.03

Lit

Ser

Lat

Mac

Slok

Cze

Pro

Kur

Geo

Alb-G

Bas

29.38

47.83

42.2

40.38

39.23

38.57

17.39

16.6

15.83

15.04

11.86

On being (dis)similar in Europe 233 1

2

3

4

5

59

60

61

62

63

Rom

Hun

Spa

Est

Ger

Slov

Arm

Alb-G

Geo

Kur

Bas

27.55

53.82

35.87

35.75

35.64

34.95

17.39

15.18

13.51

13.27

12.31

Wel

Ger

Sur

Est

Mac

Dan

Aze

Arm

Kur

Geo

Bas

22.88

33.05

31.97

30.71

30.57

30.37

14.07

12.9

11.11

7.81

7.13

Judging by their degree of genetic solidarity, the minor Indo-European phyla are something of a disappointment because only the Gheg variety of Albanian and Lithuanian behave more or less in the expected way: For Gheg, its Tosk sister variety is indeed the preferred partner, whereas Latvian is at least second best for Lithuanian. This is already the extent of genetically fitting pairings. There is no place for the Gheg variety among the first five ranks of Tosk Albanian, nor do we find Lithuanian on the top positions of Latvian. The two Celtic languages behave like foreigners of one another, too, as do Kurdish and Romani. The most striking fact is that the preference of Gheg for Tosk is unilateral. However, the preference lists of both varieties overlap: Greek, Bulgarian and Macedonian are favourites of Gheg as well as of Tosk. In addition, Macedonian and Bulgarian occupy the first two positions on the list for Greek. Tosk and Greek also share Romanian as one among the five favourites. Owing to the fact that Greek is second best on the list of Romanian, one thing comes immediately to mind: All these languages are members of the Balkan Sprachbund. In this case, areality wins out against genealogy (although this still does not explain the discrepancies among the two Albanian varieties). As for the Baltic languages, both Latvian and Lithuanian have a strong inclination to associate with members of the Slavic phylum. Serbian, Czech and Macedonian occur on both lists, Bulgarian is number 1 for Latvian, and Slovak comes in fourth place for Lithuanian. The latter accepts favourite partners only from the East – with the four Slavic languages being located in the Southeast, i.e., in a different quadrant. Similarly, all Slavic partners of Latvian stem from the southeast quadrant – only German is from the Northwest. For both Baltic languages, associations are with A-type and B-type languages alike. Thus, they escape any explanation based on our three parameters: Genealogy is important for Lithuanian, but not for Latvian; geographic vicinity is largely unimportant (it applies only in the case of Latvian as second best on the Lithuanian list and thus is conflated with the genetic factor), as is typology. In Part D, we discuss the possibility of German influence in the past especially on Latvian. What still surprises us is the fact that Latvian and Lithuanian display largely the same unexpected preferences. The two Celtic languages do not even show a trace of parallel behaviour. Welsh is characterised by a rather heterogeneous mixture of Germanic, Ro-

234 A corpus-based analysis mance, Slavic and non-Indo-European languages with German in front position. This heterogeneity contrasts sharply with the very homogeneous picture one observes for Breton. Breton sides preferably with Romance languages (although French, the most convincing guess, is missing from the list of high coefficient values). All five positions are occupied by Romance languages – the majority from the Ibero-Romance branch. While Breton appears to be a cloaked associate member of the Romance phylum, Welsh does not belong anywhere in particular. Breton, like Welsh, prefers associations with B-type languages. For both languages, geographic vicinity is of no importance when it comes to choosing favourites. Less surprisingly, Romani and Kurdish go separate ways. Kurdish has four Romance languages on the first five positions and, in addition, Maltese. The Romance languages belong to the Btype, just as Kurdish itself – but Maltese is an A-type language. Immediate neighbourhood is not decisive. Put differently, the Kurdish favourites do not form an arbitrary list. Romani allows for two Uralic languages and one Slavic, Germanic and Romance language each. The first position of Hungarian is almost self-explanatory: The variety of Romani used for our sample is the Lovari spoken in Hungary. Chances are that the high coefficient value reflects a high degree of Magyarisation of the local variety of Romani. The pairing Romani-Hungarian is a case of bilateral preference! With the exception of Slovenian, which is spoken in the vicinity, the other partners of Romani are less easy to account for. At the very least, they are all representatives of the B-type and thus belong to the same class as Romani, whereas Hungarian is a C-type language. The preference lists of Kurdish and Romani fail to overlap. The two isolates Greek and Armenian cannot be expected to converge much. As mentioned above, Greek associates strongly with other members of the Balkan Sprachbund. Nevertheless, Bulgarian is on both lists. Armenian fits in well with Slavic languages – especially with those from the Eastern branch. With Russian, Ukrainian and Belarusan, Armenian shares the property of being an A-type language. Russian is also spoken close by (and was the superstrate for a long period of time). Yet Armenian, together with Romani, is one of two Indo-European languages which opt for a non-IndoEuropean language on the highest rank. For Armenian, the preferred partner is Azerbaijani – its closest neighbour in the Trans-Caucasian region. Note that Azerbaijani belongs to the B-type and thus differs typologically from Armenian. The relation Armenian-Azerbaijani is a bilateral preference! Armenian, Romani, Hungarian and Azerbaijani disregard genetics and typology. For these pairings, geography is the most important factor.

On being (dis)similar in Europe 235

How does this fact fit into the picture painted by the remaining nonIndo-European languages? In accordance with out leading working hypothesis, we would expect to find a higher degree of similarity among the Uralic phylum and the two Altaic languages, cf. Table (C20). (C20)

Non-Indo-European phyla 1

2

3

4

5

59

60

61

62

63

Aze 23.08

Arm 33.79

Gher 32.74

Tur 32.41

Rus 30.67

Ukr 30.28

Ice 14.71

Fin 14.44

Wel 14.07

Far 12.37

Bas 10.72

Bas 10.7

Ara 14.73

Bul 14.59

Yid 14.46

Hun 14.46

Pol 13.61

Wel 7.13

Nor 7.08

Fre 6.05

Fin 5.47

Geo 4.25

Est 32.65

Ger 49.2

Bul 44.55

Ara 43.48

Mac 43.33

Spa 42.6

Arm 18.39

Geo 16.43

Kur 15.56

Alb-G 15.23

Bas 12.28

Fin 18.52

Saa 27.36

Mac 27.12

Fris 24.33

Bul 24.03

Dut 24.03

Arm 12.62

Geo 12.17

Kur 11.76

Alb-G 10.45

Bas 5.47

Geo 16.89

Cat 23.28

Bul 23.04

Mac 22.56

Arm 22.46

Friu 22

Kur 11.95

Far 11.55

Alb-G 10.51

Wel 7.81

Bas 4.25

Hun 31.66

Rom 53.82

Ger 43.67

Mac 42.34

Bul 41.8

Ser 40.17

Fin 19.31

Alb-G 17.74

Kur 17.12

Geo 14.46

Bas 14.46

Mal 25.31

Gre 36.87

Ara 36.54

Spa 36.36

Yid 36.21

Gal 35.56

Rus 15.53

Fin 15.43

Arm 14.4

Geo 14.33

Bas 13.36

Saa 28.22

Lat 42.95

Als 39.42

Spa 39.1

Ara 37.84

Mac 37.5

Arm 16.89

Alb-T 15.05

Kur 12.97

Alb-G 11.23

Bas 10

Tur 29.32

Mac 42.29

Cat 40.24

Bul 39.59

Ger 39.56

Spa 39.02

Ice 17.65

Fin 17.28

Far 17.12

Alb-G 17

Bas 10.64

Starting with the Altaic phylum, we again face the surprising fact that two very closely related languages display strikingly divergent preferences. Turkish is at least among the five highest ranking languages for Azerbaijani. However, this is not reciprocal, as Azerbaijani is wanting from the list of Turkish favourites. Moreover, these lists do not overlap at all. Azerbaijani shares two of the East Slavic favourites of Armenian, with the surprising second position for Ladin. Turkish, on the other hand, only associates with Indo-European languages from three different phyla. It is perhaps most notable that the two Balkan Slavic languages count among the favourites of Turkish. The Uralic phylum holds something special in store for us, namely the unique case that a member of the same phylum occurs among the five lowest positions. For Hungarian, Finnish is one of the least attractive partners. Finnish is also the only Uralic language which allows a member of the same phylum to occupy the highest rank: Saami. This is exceptional because none of the other Uralic languages even allows for a sister language to occupy one

236 A corpus-based analysis of the first five positions. Besides a certain leaning towards German, Hungarian shows sympathies with Slavic languages from the wider neighbourhood of the Balkans. Of these Slavic languages, Macedonian recurs with every Uralic language: Bulgarian is also present on the lists of Finnish and Estonian. Hungarian associates with partners from the A-type as well as from the B-type. All other Uralic languages accept only B-type languages on the first five ranks. Germanic partners are prominent not only for Hungarian but also for Saami (Alsatian!), Estonian (German) and Finnish (Dutch and Frisian), although one would have expected Swedish to be more important for Finnish. The strong position of German is probably the result of intensive historical contacts in the Baltic region and the Danube monarchy, cf. Part D. Estonian and Saami also share the Ibero-Romance languages Aragonese and Spanish as important partners. It is somewhat surprising to encounter Latvian as highest ranking favourite of Saami while it is absent from the list of Estonian, its closest neighbour to the North, with which it has shared much of a common linguistic past. The three remaining languages are situated in different locations on the European map. Basque is in the distant Southwest, Maltese is on the boundary between Southwest and Southeast and Georgian is at the far end of the Southeast quadrant. Basque and Maltese belong to the Atype, whereas Georgian is a C-type language. Maltese prefers combinations with B-type languages, which is also the prevalent pattern for the other two languages. However, Basque also has one C-type language (Hungarian) and one A-type language (Polish) among its five favourites. Georgian counts Armenian, an A-type language, among those languages which are characterised by high coefficient values. This position of Armenian as one of the first five partners of Georgian may also be interpreted as a reflection of the geographic neighbourhood relation that has been strong for both languages over several millennia. All three languages have a certain leaning towards Romance, especially Ibero-Romance languages. The fact that Aragonese is the preferred partner of Basque is most probably another example of areality, as both languages are next door neighbours on Spanish territory. This geography-based reasoning, however, does not apply to Georgian and Maltese. The Iberian orientation is especially strong with Maltese with Aragonese, Spanish and Galician at positions 2, 3 and 5. It cannot be denied that Maltese has experienced a long lasting Romanicisation, although the most direct sources of influence have always been Sicilian and Italian.259 Likewise, the first rank occupied by Greek becomes plausible if viewed from the vantage point of areality and maritime language contact between (Byzantine) Greek and Maltese (Brincat 2000). There is no handy explanation for the other high ranking

On being (dis)similar in Europe 237

partners of the three languages under scrutiny. The double occurrence of Yiddish on the lists of Basque and Maltese escapes any attempt at a motivated explanation. This difficulty of finding arguments other than pure chance is also reflected by the distribution of bilateral and unilateral preferences in the two genetically heterogeneous groups, cf. (C21). (C21) Genetically mixed groups (C21.1) Minor Indo-European phyla Bilateral AlB ArmlAze RomlHun

Unilateral AoB Alb-GoAlb-T Alb-ToGre BreoAra GreoMac KuroRum LatoBul LitoSer WeloGer

BoC Alb-ToGre GreoMac AralSpa MaclBul RumlMol BullMac SeroMac GerlLetz

AoC 5 3 4 2 0 3 3 0

Rank CoA 0 0 0 0 0 0 0 0

(C21.2) Non-Indo-European phyla Bilateral AlB AzelArm HunlRom

AoB BasoAra EstoGer FinoSaa GeooCat MaloGre SaaoLat TuroMac

BoC

Unilateral

AralSpa GerlLetz SaaoLat CatoAra GreoMac LatoGer MaclBul

AoC 0 0 0 0 0 0 3

Rank CoA 0 0 0 0 0 0 0

The minor Indo-European phyla attest to patterns which resemble the ones of the major phyla, i.e., unilateral preferences constitute the unmarked case. At the same time, language C, in the majority of cases, is also among the favourites of A. This is also in line with the tendency formulated above. Yet there is a slight difference in comparison to what we learned from the Germanic, Romance and Slavic phyla: In (C21.1), language C for the most part occupies a lower rank position than in the major phyla. On top of that, non-Indo-European languages diverge radically from this pattern because, with the exception of Turkish, none of them allow for language C being among A’s favourites nor vice versa. A ready explanation for this deviation is the fact that non-Indo-European languages usually display an average

238 A corpus-based analysis coefficient value which is significantly lower than the overall average of the whole sample – and, of course, much lower than the average for each single genetically defined subgroup, cf. (C22). (C22)

The average of the average

Phylum Germanic Romance Slavic Minor Indo-European Non-Indo-European Sum

Added averages 476.79 671.62 342 254.54 216.35 1,961.3

Average per phylum 34.05 33.58 31.09 25.46 24.03 30.64

Not only do the various phyla behave differently as to their average values but the differences are even more pronounced on the level of individual languages. In Tables (C23–24), we list languages of our sample according to the number of occurrences among the first five and list five partners, respectively. The percentages refer to the potential of 63 occurrences (= 100%). There are altogether 13 languages which never make to the higher ranks and 47 which never occur at the lower end of the list of coefficient values. Ten of these languages occur both on the highest and on the lowest ranks. Six languages (Asturian, Breton, Corsican, English, Lithuanian and Portuguese), all of them members of the Indo-European phylum, are always located somewhere in between the two extremes. (C23)

Prime candidates for ranks 1 through 5

Language German Spanish Macedonian Bulgarian Aragonese Italian Catalan; Galician Serbian Sursilvan Czech Luxembourgeois Frisian; Gascon; Ladin Dutch; French; Latvian; Limburgisch (South); Friulian; Languedocien

Times 29 28 26 19 18 13 12 11 10 9 8 7 6 5

% of 63 46% 44% 41% 30% 28.5% 20.5% 19% 17.5% 16% 14% 13% 11% 9.5% 8%

On being (dis)similar in Europe 239 Language Alsatian; Belarusan; Danish; Greek; Romanian; Russian; Slovak; Yiddish Armenian; Norwegian; Swedish; Ukrainian Estonian; Faroese; Hungarian; Limburgisch (North); Polish; Slovenian; Vallader Albanian (Tosk); Azerbaijani; Badiotto; Croatian; Maltese; Moldavian; Provençal; Romani; Saami; Sardinian; Turkish

Times 4

% of 63 6%

3 2

5% 3%

1

1.5%

The majority of our sample languages are represented at least once on the higher ranks. Languages which recur significantly often in the upper regions of the statistics are, however, not that numerous. If we arbitrarily choose 25% as the borderline separating significantly high values from less significant ones, we find five languages which exceed this percentage: German, Spanish, Macedonian, Bulgarian and Aragonese – all of them Indo-European languages (from the three major phyla) belonging to the Btype. In areal terms, the South prevails, as only German is located in one of the northern quadrants. More generally, among the first 22 languages listed in (C23), there is only one – incidentally, French – which also occurs on the lowest positions. Czech and Serbian are the only representatives of the Atype, side by side with 20 B-type languages. More than half (= eleven) of the Romance languages of our sample occur among the 22 first positions (together with six Germanic, four Slavic and one Baltic language). These languages cover a continuous area in the continental West and Centre of Europe with outliers in the Baltic and the Balkans, i.e., those languages which occur most often as partners of other languages are also for the most geographical neighbours of each other. Note also that non-Indo-European languages are squarely underrepresented: Only Estonian and Hungarian occur more than once (exactly twice each), while all other non-IndoEuropean are attested only once or even not at all. (C24)

Notoriously last in line

Language Basque Georgian Albanian (Gheg) Armenian; Kurdish Finnish Russian Azerbaijani; Welsh Provençal

Times 63 48 45 43 30 17 5 4

% of 63 100% 76% 71% 68% 48% 27% 8% 6%

240 A corpus-based analysis Language Belarusan; Faroese; Icelandic; French; Saami Albanian (Tosk); Norwegian

Times 3 1

% of 63 5% 1.5%

The distribution of languages over lower ranks differs strikingly from that observed for the higher ranks. The most striking fact, in any case, is the high number of occurrences of Basque, Georgian, Albanian (Gheg), Armenian and Kurdish from positions 59 to 63. For all of these languages, more than 50% of their partnerships are located in the nether regions of the list. Moreover, Basque occurs there exclusively – except for three cases, where Basque winds up one or two ranks higher, it appears 60 times at the very bottom of the list, i.e., hardly any other language ever turns up there (exceptions being Russian, Icelandic and Georgian with one last rank position each). Thus the behaviour of Basque is maximally predictable. This can hardly be due to chance. Georgian, Albanian (Gheg) and Kurdish are largely confined to the lower ranks, although there also is the occasional example of a coefficient value on an intermediate rank. Armenian deviates from this pattern, as it is also attested among the favourites of three languages. Finnish, on the other hand, occurs with slightly less than half of its pairings in the lower regions – but we never encounter it in the upper ones. These six languages alone account for 85% of all coefficient values between ranks 59 and 63. With Albanian (Gheg) and Kurdish, the B-type is clearly underrepresented in comparison to three A-type languages and one C-type language. Similarly, non-Indo-European languages are as strong as Indo-European languages in these positions, i.e., the distribution is clearly skewed in favour of non-Indo-European phyla. With the exception of Basque, all these languages are located in the East. Moreover, Finnish is the only language from the North in this group of six. There is thus a clear prevalence of the southeast quadrant. In addition, five out of the six languages are case languages, four are postpositional, three are V-final and two have some ergative alignment – all of which are features which set these languages apart from the SAE type (Haspelmath 2001). Thus, these languages tend to diverge from other languages of the continent, not only in the realm of Comitatives and Instrumentals, but also in many other respects of grammar and structure. A glance at the number of relators which are isolates in a given translation of the sample is revealing. We count a relator as an isolated instance if and only if it has no equivalent in the same context of the same sentence in any of the other languages of our sample. There are three languages for which no isolates are attested, namely Macedonian, Languedocien and Ma-

On being (dis)similar in Europe 241

cedonian. At the opposite end of the scale, we encounter exactly those languages which are also prominent in (C24) – with Welsh as the only addition to the list, cf. (C25). (C25)

Isolated tokens n > 33.33%

Language Basque Albanian (Gheg) Finnish Kurdish Welsh Armenian Georgian

Relators 360 403 152 341 145 165 111

Isolates 252 255 92 178 65 64 43

% Isolates/Relators 70.00% 63.28% 60.53% 52.20% 44.83% 38.79% 38.74%

In Basque, Albanian (Gheg), Finnish and Kurdish, more than half of the tokens have no equivalents in the other languages of the sample. All of these but Kurdish even reach values oscillating at about the two thirds mark! What this means is that it is actually an exception that a Comitative and/or Instrumental of these languages corresponds to a Comitative/Instrumental in other languages. For Welsh, Armenian and Georgian, the percentages are less impressive but still imply that a considerable amount of tokens do not have a translation equivalent that is also an instance of a Comitative and/or Instrumental. These figures are in stark contrast to the minimal number of isolates found in Spanish, Italian, Corsican and French, where only one example in each language fails to have a Comitative/Instrumental as a translation equivalent. Add to these languages those three mentioned above, where none of the relators is without an equivalent translation, and the high numbers in (C25) become even more remarkable. The average for isolates for the entire sample is 28.09, corresponding to a percentage of 15.71% – two values which the seven languages in (C25) exceed by far. Languages with few or no isolates at all are without exception members of the Indo-European phylum – with the vast majority B-type languages. Estonian and Saami are the two non-Indo-European languages with the lowest values (11.46% and 11.65%), with 33 Indo-European languages on the ranks below following them. What all this amounts to is the gradual distinction of what makes a language more or less typically European. Some of the languages converge more freely with others, whereas some languages, in a manner of speaking, keep more to themselves, i.e., display a higher degree of individualism. In Table (C26), we compare the average values of the similarity coefficient for the languages of our sample. For convenience, the values have been

242 A corpus-based analysis rounded off and up, respectively. To give meaning to the table, we, rather pointedly, present it as a scale of decreasing Europeanness of the languages of our sample. (C26) From more to less European Group I II

Average 41% 40% 38% 37% 36%

III

35% 34% 33% 32% 31% 30% 29% 28% 27% 25% 24% 23% 22% 19% 17% 11%

IV

V

VI VII

Languages German; Spanish Aragonese; Macedonian Bulgarian; Catalan Galician; Limburgisch (South); Sursilvan; Yiddish Czech; Dutch; Frisian; Limburgisch (North); Luxembourgeois; Slovenian Alsatian; Friulian; Gascon; Ladin; Italian; Serbian Danish; Latvian; Slovak; Swedish Badiotto; Breton; Estonian; Languedocien; Vallader English; Hungarian; Sardinian Asturian; French; Greek; Moldavian; Romanian; Norwegian Corsican Lithuanian; Portuguese; Turkish Romani; Saami Croatian; Icelandic Faroese; Maltese; Polish; Provençal; Ukrainian Albanian (Tosk) Azerbaijani; Belarusan; Welsh Russian Armenian; Finnish Albanian (Gheg); Georgian; Kurdish Basque

Going by steps of 5%, we divide the scale into seven degrees. At level I, we find German and Spanish – the two languages which behave the most European, as their average similarity coefficient exceeds the 40% mark. They are closely followed by Aragonese and Macedonian, which display very similar averages. Those languages which are characterised by averages of over 35% form a subset of the 22 highest ranking languages in (C23) and are thus located in an area on the European mainland reaching from the Iberian Peninsula over Central Europe down to the Balkans. The numerically strongest group is at level III, where 24 languages are assembled. These languages fill the gaps to cover the whole area of SAE in Europe. From level IV down to VII, we mainly encounter languages which are situated at the outskirts of the SAE area; none of the major Germanic

On being (dis)similar in Europe 243

and Romance languages, except Portuguese, occur at these levels, whereas seven out of the nine non-Indo-European languages are to be found there. More generally, non-Indo-European languages do not make it upwards beyond level III. Similarly, there are seven languages from the group of minor Indo-European phyla distributed over levels IV through VI. The highest a minor Indo-European language can get is, again, level III (Latvian). Moreover, the entire East Slavic phylum appears at level V. The two insular North-Germanic languages Faroese and Icelandic are also allocated on the lower levels. From levels IV through VII, nine of the 24 languages belong to the A-type and four represent the C-type, i.e., there is a disproportionately low number of languages from the B-type which in turn dominate on the higher levels (of the 16 languages at levels I-II, Czech is the only A-type language). Geographically, the lower levels preponderantly host languages from the eastern quadrants and from the periphery of the continent, i.e., mostly languages which are structurally not fully integrated into the SAE type, if at all. We know now that two large groups of languages can be identified in our sample, namely, languages with a higher degree of convergence and languages with a higher degree of individuality. The Janusian nature of the concepts of linguistic convergence and individuality are discussed by Bechert (1988) regarding several phenomena and their distribution on the European map. As Bechert demonstrates between the lines, it is not sufficient to look at the statistics alone. Rather, one needs to see the qualitative side of the convergence and individuality of the languages, i.e., we have to take account of the actual data too. 11.2.2. Dancing with girls, seeing with one’s heart and other shared contexts In the foregoing section, we emphasised the differences that exist between the languages of our sample. Irrespective of this at times striking variation that separates more SAE-like languages from less SAE-like ones, we have also seen that there is always at least a small amount of similarity even among languages which are located at the opposite extremes of the scale of Europeanness. Thus, it would be incorrect to state sweepingly that languages on the fringes of the continent are absolutely incompatible with ones spoken in the interior. We therefore address the issue of identifying the bonds which tie together all of our sample languages. For this purpose,

244 A corpus-based analysis it is necessary to look at the contexts in which convergence is most pronounced.260 Incidentally, there is only one single sentence in our sample text which contains a Comitative/Instrumental in each and every translation: Puis il s’épongea le front avec un mouchoir à carreaux rouges, ‘Then he mopped his forehead with a handkerchief decorated with red squares.’ Except for this singularity, the best we can get are 63 languages for one and the same sentence. Hits with more than 59 languages are attested only five times. These and many other gaps in the most extended isoglosses261 are often (but by no means always) due to chance, i.e., they are not systematically motivated by some special structural trait of the missing language(s).262 The two longest chains of functionally identical usage instantiate two kinds of Instrumentality, namely, TOOL (64 languages) and BODY PART INSTRUMENT (63 languages). First, we give the full list of examples of the TOOL-example according to genetically-defined groups. The sentence XIV.29 describes a situation in which a temporary acquaintance of the Little Prince, the lamplighter (AGENS/USER) wipes his forehead (PATIENT) with a checkered handkerchief (TOOL). What is important for us is the question how the NP representing the TOOL is marked, if at all. Discounting some variation especially in the semantics of the lexical verbs (all of which, however, designate more or less similar activities), the basic structure of the situation is the same for all of the languages, cf. (C27). (C27)

Ingredients of XIV.29 Participants

NucPred

Relator

N

V[+transitive]

Preposition, Postposition, inflectional case (and combinations thereof)

Material object handkerchief, cloth

to wipe, to sweep, to mop, to clean, to dry

AGENT/USER

PATIENT

TOOL

Pro[3SG.(M)] and/or person inflection on the finite verb

N

refers anaphorically to the lamplighter

Body part front, forehead, temple, face, head Subject/Ergative Object/Absolutive

Adverbial

The intra-participant relation between the handkerchief and the coloured pattern that appears on its surface is expressed in various ways in the sample languages. Since Comitatives/Instrumentals also play a role in a significant number of languages, we come back to this issue below.

On being (dis)similar in Europe 245

We start with the list of sentences from the Germanic phylum. For simplicity, we refrain from giving transmorphemisations for every single one, but we mark the relator by boldface and identify the NP representing the TOOL by indexed square brackets. Wherever there is a second instance of a Comitative and/or Instrumental which marks the attribute of the head noun, we put it in square brackets and give it the index ORNATIVE. What comes to the fore immediately is the fact that there is no variation as to the construction type used for the TOOL-NP and its relator. The only exception to this rule is Icelandic: In all Germanic languages, the relator and the TOOL-NP together form a PP. This is also the case for Icelandic, where, however, the case governed by the preposition is also crucial; in the present context, the noun is in the dative (cf. Section 12.3). (C28)

Germanic phylum XIV.29

D’rno het’r sini Stirn mit [me rotkàrrierte Nàstüech]TOOL àbg’wischt Så tørrede han sig i panden med [et rødternet lommetørklæde]TOOL Toen veegde hij zich het voorhoofd met [een roodgeruite zakdoek]TOOL Then he mopped his forehead with [a handkerchief decorated with [red squares]ORNATIVE]TOOL Far Hann turkaði sveittan av enninum við [einum reyðpuntutum lummaklúti]TOOL Fris Doe switfage er syn foarholle mei [in rearútsjese bûsdoek]TOOL Ger Dann trocknete er sich die Stirn mit [einem rotkarierten Taschentuch]TOOL Ice Síðan þerraði hann sér um ennið með [rauðtiglóttum vasaklút]TOOL Letz an sech duerno d’Stir mat [engem routkaréierten Duch]TOOL ofgebotzt Lim-N Doe vaegdje hae ziene kop aaf mèt [eine roeëje, geroete tesseplak]TOOL Lim-S Doew vreef heë zich d’r kop drueg mit [inne roewe gerüdde sjnoefplak]TOOL Nor Etterpå tørket han pannen med [et rødrutet lommetørkle]TOOL Swe Sedan torkade han svetten ur pannan med [en rödrutig näsduk]TOOL Yid Nokh dem hot er zikh opgevisht dem shtern mit [a royt-kvadratn tikhl]TOOL

Als Dan Dut Eng

In all Germanic languages, the relator is the one which is syncretistic as to the distinction of Comitative and Instrumental. This is largely trivial, in so far as only Icelandic is not a B-type language. English is the only language to use the relator twice – with the second occurrence serving the purpose of

246 A corpus-based analysis connecting the attribute to the nominal head of the TOOL-NP – and thus this constitutes an ORNATIVE. This is an option that potentially every Germanic language has. Nevertheless, it does not seem to be the first stylistic choice for the majority of the languages of this phylum. However, the pattern recurs elsewhere as well, cf. below. (C29) contains the same sentence in the 20 Romance languages of our sample. The picture is by and large the same as in the Germanic phylum, because the TOOL-NP and the relator are constituents of a PP. The relator itself is – trivially again – syncretistic, as the entire phylum belongs to the Btype. The different word order attested in Badiotto is caused by focus. Four languages display double occurrences of the relator, namely, Moldavian, Romanian, Sursilvan and Vallader. Except for Moldavian, these patterns of double employment of the relator are slightly different from that reported for English above, as the PPs are not dependent on a participle. (C29) Ara Ast Bad Cat Cor Fre Friu Gal Gas Gher Ita Lan Mol Por Pro Rum Sar Spa Sur Val

Romance phylum XIV.29 Dimpués s’ixugó a fren con [un moquero de cuadros royos]TOOL Llueu llimpióse la frente con [un pañuelu pintu]TOOL Y cun [n fazorel da cadrí cöci]TOOL s'âl spo assuié ía la frunt Després s’eixuga el front amb [un mocador de quadres vermells]TOOL Dopu s’asciuvò u fronte incù [un mandigliulu quadrittatu rossu]TOOL Puis il s’épongea le front avec [un mouchoir à carreaux rouges]TOOL Po al sujà il cernêli cun t[un fassolet a cuadris ros] TOOL Despois enxugou a fronte c[un pano de cadros vermellos]TOOL Puish que’s boishè lo temp dab [un mocader de quarrèus rotges]TOOL L s’ova pò suiá jú l fruent cun [n fazulët da chedri cueceni]TOOL Poi si asciugò la fronte con [un fazzoletto a quadri rossi]TOOL Puèi se freguèt lo front amb [un mocador dels carrèus roges]TOOL Apoi îúi úterse fruntea cu [o batistă cadrilată cu [roúu]ORNATIVE]TOOL Depois enxugou a testa com [um lenço aos quadrados vermelhos]TOOL Pièi s’espounguè lou front em’[un moucadou di carrèu rouge]TOOL Apoi îúi úterse fruntea cu [o batistă cu [pătrăĠele roúii]ORNATIVE]TOOL Tando s’at assuttadu su sudore de cara chin d'[unu muccadore a quadros rujos]TOOL Luego se enjugó la frente con [un pañuelo a cuadros rojos]TOOL Lu schigenta el siu frunt cun [in fazalet cun [quadrels cotschens] ORNATIVE TOOL ] Lura ha’l süantà seis frunt cun [ün fazöl cun [quaders cotschens] ORNATIVE TOOL ]

On being (dis)similar in Europe 247

This apparent homogeneity across genetic borders is gradually reduced by the examples of the other groups. In the Slavic phylum (C30), there are at least two different cases. (C30) Bie-R Bul Cro Cze Mac Pol Rus Ser Slok Slov Ukr

Slavic phylum XIV.29 Potym vytser uspatsely lob [þyrvonaj kljatþastaj nasowkaj]TOOL i skazaw Setne izbărsa þelo s [edna kărpa na þerveni kvadrati]TOOL Zatim obrise celo [rupþiüem s [crvenim kvadratima]ORNATIVE]TOOL Potom si otĜel þelo [þervenƟ kostkovaným kapesníkem]TOOL Potoa go izbrišal þeloto so [edno karirano tsrveno ša miþe]TOOL NastĊpnie otará sobie czoáo [chustką w [czerwoną kratĊ]ORNATIVE]TOOL Potom [krasnym kletþatym platkom]TOOL utër pot so lba i skazal Zatim obrisa þelo [tsrvenom kariranom maramitsom]TOOL Potom si utrel þelo [vreckovkou s [þervenými kockami]ORNATIVE]TOOL Nato si je z [rdeþe kockastim robcem]TOOL otrl þelo Potim [kartatoju þervonoju xustynkoju]TOOL vyter z litsja pit i skazav

In nine out of the eleven languages, the relator is either exclusively or additionally marked by bound morphology on the component parts of the TOOLNP. The relator may occur more than once per NP, since the attributes of the head noun agree with it in case, number and gender. Whenever the inflectional case forms of the instrumental vary within one and the same sentence, this is mostly due to word class dependent paradigms and different declension classes (cf. also Lithuanian below). In three languages, a preposition is used either alongside the nominal case morphology, or, in its absence, as the sole indicator of the relation. This is an interesting fact, because only Bulgarian and Macedonian do not employ case morphology at all, and thus the preposition is the only way to encode the relation via an overt grammeme. In this way, these two Slavic languages behave like the Romance languages and many members of the Germanic phylum. This does not apply to Slovenian – a language whose inflectional system is still operative on nouns. However, in Slovenian, the inflectional instrumental can no longer appear on its own. It always co-occurs with the preposition z. In Slovak and Croatian, the attribute of the nominal head of the TOOL-NP is introduced by the preposition s, which again triggers the inflectional instrumental on its dependents (cf. English and the four Romance languages above). Note that this preposition is also employed for the encoding of Comitativity, cf. below. Polish, however, uses a different preposition, w ‘in’, in combination with the inflectional accusative. With the exception of

248 A corpus-based analysis the three B-type languages Bulgarian, Macedonian and Slovenian, all the other Slavic languages (which belong to the A-type) employ the relator that is reserved for Instrumentality. The minor Indo-European phyla hold examples of both the A-type and the B-type in store. Lithuanian and Armenian follow the pattern familiar from the Slavic phylum, as they employ their inflectional instrumental case alone to encode the relation. In all other cases, the typical B-type pattern applies (also in the C-type language Welsh): A preposition, a case affix or a combination of both is used to encode the relation. The expression employed is always syncretistic as to Comitative and Instrumental. Four languages (the Tosk variety of Albanian, Greek, Welsh and Kurdish) display double occurrences of their relator and thus behave similarly to those languages with double occurrences identified in the major Indo-European phyla. (C31)

Minor Indo-European phyla XIV.29

Alb-G Mandej fshiu ballin me [nji faculetë të kuqe, kutija-kutija]TOOL Alb-T Pastaj fshiu ballin me [një shami me [kutia të kuqe]ORNATIVE]TOOL Arm Heto [karmir vandakavor taškinakov]TOOL þakati k’rtnink’6 srbec’ w asac‘ Bre Hag e sec’has e dal gant [ur frilien karrezennoù ruz]TOOL Gre Épeita skoúpise to métǀpó tou m‘ [éna mantƝli me [kókkina karrǀ]ORNATIVE]TOOL Kur Paúê wî xwêdana aniya xwe bi [destmaleke bi [damikên sor]ORNATIVE]TOOL zu ha kir Lat PƝc tam viƼš noslaucƯja no piere sviedrus ar [sƗrti rnjtotu kabatas lakatiƼu]TOOL Lit Paskui [raudona languota nosine]TOOL nusišluostơ kaktą Rom Palakodi [jekha posotyake kotoresa]TOOL khoslas pesko chikat Wel Yna sychodd ei dalcen â [chadach a [sgwarau cochion arno]ORNATIVE]TOOL Among the non-Indo-European languages in (C32), there is only one which has a double occurrence of the relator. Maltese uses the preposition bi twice, first to introduce the TOOL-NP and then to add the attribute to the nominal head of this NP. What appears to be another instance of double occurrence in Finnish, on closer inspection, turns out to be a case of NP-internal agreement: The adjectival attribute agrees in case and number with its head noun. Nowhere else in this group does this kind of agreement apply.

On being (dis)similar in Europe 249

(C32) Non-Indo-European phyla XIV.29 Aze Bas Est Fin Geo Hun Mal Saa Tur

Sonra [qırmızı dama-dama däsmalla]TOOL alnının tärini silib dedi [Sudur-zapi gorri-koadratu batez]TOOL, bekokiko izerdia txukatu zuen Siis ta pühkis [punaseruudulise taskurätikuga]TOOL oma otsaesist Sitten hän pyyhkäisi hien otsaltaan [punaruutuisella nenäliinalla]TOOL shemdeg [c’iteludjredebiani cxvirsaxocit]TOOL shublze opli sheimshrala da tkva Aztán [egy piros kockás zsebkendĘvel]TOOL törölgetni kezdte a homlokát Imbagƫad mesaƫ moƫƫu b’[maktur bi[l-kaxxi ƫomor]ORNATIVE]TOOL Son sihkui bivastaga [gállus ruksesruvttot njunneliinniin]TOOL Sonra [kırmızı kareli bir mendille]TOOL alnını sildi

In all non-Indo-European languages, the relator which comes to be employed is the one which encodes Instrumentality, be it exclusively or syncretistically. Thus, the A-type languages Basque, Finnish and Maltese use their Instrumental in lieu of their Comitative, and the C-type languages Georgian and Hungarian use their Comitative-Instrumental in place of their Comitative. It is therefore beyond doubt that the languages of our European sample are prone to convergence as to the use of their Instrumentals in descriptions of situations which involve a typical TOOL. Does this also include less typical cases of Instrumentality? Sentence XXV.34 exemplifies such a less typical case because the Instrument is a body part, the heart (which in this special case is meant in a rather metaphorical way). The Little Prince criticises the world as it is, including himself, and appeals to people’s feeling. One’s heart (= the metaphorical centre of feelings and emotions), he says, should show the way to a better understanding of others and the world in general. The rendering of this philosophical statement is not homogeneous in the translations. The French original contains an impersonal construction. Accordingly, in the translations, there are many cases of impersonal pronouns or the like as AGENTS. We also encounter personal pronouns proper that may be understood as generalising expressions and thus as equivalents of impersonal ones. The epistemic modality of the statement is encoded by a modal verb (or an equivalent complex construction). (C33) gives a schematic outline of the structural appearance of the sentences.

250 A corpus-based analysis (C33)

Ingredients of XXV.34

Participants AGENT/USER Pro[impersonal], Pro[1Pl], Pro[2SG] and/or person inflection on the finite verb refers to humankind Subject/Ergative

Aux

NucPred

Relator

N

ModV/Particle[+obligation]

V[±transitive]

Preposition, Postposition, inflectional case (and combinations thereof)

Body part heart Adverbial

must, should, it is necessary,

to search, to look,

TOOL

The exemplification follows the same order as above. Likewise, the same marking conventions are employed. We start out with the Germanic phylum (C34). As the picture resembles closely the one we encountered above, there is no need for a detailed comment on the data: The same relator as in (C28) is used with the same special remark for Icelandic being valid here too. Wherever the word order appears to be extraordinary (as in Dutch), this divergence is explicable in terms of focus and emphasis. In Luxembourgeois, the preposition and the definite article (in the dative) coalesce to yield the form mam. This fusion is blocked outside combinations with definite articles, at least for the written register. (C34)

Germanic phylum XXV.34

Als Dan Dut Eng Far Fris Ger Ice Letz Lim-N Lim-S Nor Swe Yid

M’r müess mit’[m Harz]BODY PART süeche Det gælder om at søge med [hjertet]BODY PART Met [het hart]BODY PART moet men zoeken One must look with [the heart]BODY PART Tað ræður um at leita við [hjartanum]BODY PART Moatst sykje mei [it hert]BODY PART Man muß mit [dem Herzen]BODY PART suchen Það verður að leita með [hjartanu]BODY PART ǥT muss ee ma[m Häerz]BODY PART sichen Doe mós zeuke mèt [dien hert]BODY PART Me mót mit [‘t hats]BODY PART zeuke Man må lete med [hjertet]BODY PART Man måste söka med [hjärtat]BODY PART Men darf zukhn mit[n harts]BODY PART

On being (dis)similar in Europe 251

Romance languages too follow largely the same pattern as before. If we compare (C29) to (C35), we only detect minor differences in the phonological shape of the relator. (C35)

Romance phylum XXV.34

Ara Ast Bad Cat Cor Fre Friu Gal Gas Gher Ita Lan Mol Por Pro Rum Sar Spa Sur Val

Cal rechirar con [o corazón]BODY PART Hai que buscar co[l corazón]BODY PART An mëss chirí cun [le cör]BODY PART Cal cercar amb [el cor]BODY PART Sogn’à circà cù [u core]BODY PART Il faut chercher avec [le cœur]BODY PART Bisugne cirî cu[l cûr]BODY PART Hai que buscar c[o corazón]BODY PART Que cau cercar dab [lo còr]BODY PART N dë crí cun [l cuer]BODY PART Bisogna cercare co[l cuore]BODY PART Amb [lo còr]BODY PART cal cercar Trebuie să cauĠi cu [inima]BODY PART Deve-se é procurar com [o coração]BODY PART Es emé [lou cor]BODY PART que fau bousca Cu [inima]BODY PART trebuie să cauĠi! Cheret a chircare chin [su coro]BODY PART Es necesario buscar con [el corazón]BODY PART Ins sto encurir cun [il cor]BODY PART Id es da tscherchar cu[l cour]BODY PART

These variations are mostly instances of phonologically-conditioned allomorphy of the sandhi kind: The shape of the relator varies according to the phonological quality of the initial segment of the immediately following word. Sardinian, for instances, uses chin d’ in front of vowels and simply chin when the next word starts with a consonant (cf. also Friulian). In various Romance languages, the preposition and the definite article (sometimes also the indefinite article) fuse to form a new complex word, as in Vallader, where cun ün contrasts with cul. Corsican, however, is special in so far as the alternation of cù and incù is not simply induced by the phonological context. Rather, the opposition of definiteness versus indefiniteness is decisive. The longer allomorph is regularly used in front of indefinite NPs, whereas the short form occurs normally in combination with definite NPs.

252 A corpus-based analysis The monotony continues through the Slavic phylum. (C36) is a faithful remake of (C30). The same expressions as in (C30) are used throughout (C36), i.e., wherever the bare instrumental suffices with proper TOOLS, it also suffices to encode BODY PART INSTRUMENTS, and wherever a preposition is needed either in addition to the inflectional instrumental or without accompanying case morphology, it is the same in (C30) and (C36). (C36)

Slavic phylum XXV.34

Bie-R šukacy treba [sercam]BODY PART Bul þovek trjabva da tărsi săs [sărdeto si]BODY PART Cro Treba tražiti [srcem]BODY PART Cze Musíme hledat [srdcem]BODY PART Mac treba da se bara so [srceto]BODY PART Pol Szukaü naleĪy [sercem]BODY PART Rus Iskat’ nado [serdcem]BODY PART Ser treba tražiti [crcem]BODY PART Slok Treba hĐadaĢ [srdcom]BODY PART Slov Iskati moraš s [srcem]BODY PART Ukr treba šukati [sercem]BODY PART With the exception of Welsh, the above pattern repeats itself in the minor Indo-European phyla: The languages employ the same marker for TOOLS and for BODY PART INSTRUMENTS, cf. (C37). (C37) Alb-G Alb-T Arm Bre Gre Kur Lat Lit Rom Wel

Minor Indo-European phyla XXV.34 Duhet me kërkue me [zemër]BODY PART do kërkuar, me [zemër]BODY PART Harkavor Ɲ [srtov]BODY PART p’ntrel Ret eo klask gant [ar galon]BODY PART prépei na psáxnei kaneís me [ten kardiá]BODY PART Mirov gerek bi [dilê]BODY PART xwe li tiútan bigerin JƗmeklƝ ar [sirdi]BODY PART Reikia ieškoti [širdimi]BODY PART [Amare ilesa]BODY PART trubuj te roden Rhaid chwilio efo[’r galon]BODY PART

The Gheg variety of Albanian stands out in so far as it employs the relator me twice. However, the first occurrence of me in this sentence instantiates its use as introductory particle of the so-called infinitive. The problems posed by this co-occurrence of one and the same morph with widely different functions are discussed in Part D, when we tackle the diachronic side of our topic.

On being (dis)similar in Europe 253

Welsh is exceptional in this group because it employs different markers in (C31) and (C37), namely, â, a and efo. While â and a are best interpreted as merely orthographic variants of one and the same relator (Stolz 1998a), the case of efo is something completely different. Efo is a partial synonym of â, i.e., efo may replace â as long as the intended meaning is the one associated with the Comitative-Instrumental. However, the replacement is blocked for those cases where a functions as coordinating conjunction (Jones and Thomas 1977: 384–387). Efo is likewise ruled out as a substitute for â if the latter is part of the prohibitive construction or the equative construction, cf. (C69)– (C70). In addition, efo has a twinge of a regional trait typical of the Northern variety of Welsh (though it is largely considered an equivalent of Southern Welsh gyda(g)). Since efo occurs just four times in The Little Prince and is thus clearly outnumbered by both the other relators of Welsh, namely â and gyda, it hardly qualifies as a primary marker. Moreover, it does not have a monopoly on any of the functions on the Comitative-Instrumental continuum. There are two examples for Comitative proper and again two for BODY PART INSTRUMENT. In the former case, gyda is the dominant choice, and in the latter, â is more frequent than efo. We are not in a position to decide whether the translator occasionally opted for efo in lieu of â or gyda in order to render the text stylistically more flexible. Be that as it may, efo – like its competitor â – is syncretistic as to Comitative and Instrumental and thus fits in well with the general picture: The minor Indo-European languages and their distant relatives in the major phyla use in these contexts a relator which has an instrumental component. The non-Indo-European languages conform to the general picture established so far: Discounting Basque, which is missing from the isogloss263, the languages in (C38) employ the very same markers as in (C32) above.264 This observation holds for languages of all types: A-type, B-type and Ctype all behave in identical ways. The double occurrence of the body part noun gulit ‘with the heart’ inflected for the instrumental in Georgian is a stylistic device of the translator, who presumably wants to add emphasis and drama to the sentence. (C38) Non-Indo-European phyla XXV.34 Aze Est Fin Geo Hun Mal

[üräklä]BODY PART axtarmaq lazımdır Tuleb otsida [südamega]BODY PART Pitää etsiä [sydämellä]BODY PART [gulit]BODY PART unda edzebo, [gulit]BODY PART! [A szívünkkel]BODY PART kell keresni Jeƫtieƥ li jfittxu b’[qalbhom]BODY PART

254 A corpus-based analysis Saa Tur

Galga ohcat [váimuin]BODY PART ønsan, [yüre÷iyle]BODY PART aramalı

In principle, the exact kind of Instrumental relation is not an obstacle for the languages of our sample to converge massively. We know from the discussion in the previous sections that at least some of the languages (especially French and Finnish) distinguish BODY PART INSTRUMENTS from TOOLS by formal means, cf. (A57), (A58), (A60). However, these relators compete with the respective Instrumental or Comitative-Instrumental, which have thus far had the upper hand in the sentences we have scrutinised. The question arises as to whether the same degree of conformity among the languages also applies in the realm of Comitativity. Sentence XXI.103 yields the most extended isogloss for a Comitative context in our sample text. Of the 64 languages, only three are missing from the list, namely, Icelandic, Languedocien and Georgian.265 The sentence stems from the explanation the fox is giving the Little Prince for typical behaviour of social beings: In the fox’s eyes, the habit of the hunters of dancing with the village girls on Thursdays is a paradigm case of predictability. Disregarding the temporal adverbial, the description of the situation involves an ACCOMPANEE (the hunters) and a COMPANION (the village girls), who are both ACTORS, alongside a nuclear predication represented by an intransitive action verb (to dance). (C39) is the corresponding schematic representation of the common core of the sample sentences below. The languages disagree as to how they handle the information about the village semantico-syntactically: Many languages use the NP representing the village as an attribute of the NP representing the female dancers, while others locate the regular dancing event in the village. This is the reason why the syntactic boundaries of the COMPANION-NP in the examples below do not always coincide. (C39)

Ingredients of XXI.103

Participants

NucPred

Relator

N[+human]: girls, women,

V[intransitive]

Preposition, Postposition, inflectional case (and combinations thereof)

refers to the female dancers Adverbial

to dance, to play

ACCOMPANEE

COMPANION

Pro[3Pl], they N[+human]: hunters, youngsters, and/or person inflection on the finite verb refers to the male dancers Subject

On being (dis)similar in Europe 255

The Germanic phylum attests to the very same pattern as with the two kinds of Instrumentality above: Throughout the paradigm of languages, the syncretistic relator is used, i.e., the expressions for ACCOMPANIMENT, TOOL and BODY PART INSTRUMENT are identical, as is the expected pattern for the B-type, cf. (C40). (C40)

Germanic phylum XXI.103

sie tànze àm Dunnerschtig mit [de Maidle vom Dorf]COMPANION Hver torsdag danser de med [landsbyens unge piger]COMPANION Op donderdag dansen zij met [de meisjes uit het dorp]COMPANION Every Thursday they dance with [the village girls]COMPANION Mínir veiðimenn til dømis hava tað til skikk hvønn hósdag at dansa við [teimum ungu gentunum í bygdini]COMPANION Fris Se dûnsje op tongersdei mei [de doarpsfamkes]COMPANION Ger Sie tanzen am Donnerstag mit [den Mädchen des Dorfes]COMPANION Letz Donneschdes danzen se mat [de Méedercher äus dem Duerf]COMPANION Lim-N Op dónderdjig danse die mèt [de maedjes van ǥt dörp]COMPANION Lim-S Zie danse ‘t donnesjdigs mit [de meëdjes va g’n dörp]COMPANION Nor Hver torsdag danser de med [de unge pikene i landsbyen]COMPANION Swe På torsdagarna brukar de dansa med [flickorna i byn]COMPANION Yid Zey tantsn yedn donershtik mit [di meydlekh fun dorf]COMPANION Als Dan Dut Eng Far

Similarly, the picture presented by the Romance phylum in (C41) meets our expectations of typical representatives of the B-type. (C41)

Romance phylum XXI.103

Ara Ast Bad Cat Cor Fre Friu Gal Gas Gher Ita Mol Por Pro

O chuebes bailan con [as mesachas d’o lugar]COMPANION el xueves valsien co[les mociquines del llugar]COMPANION la jöbia bali düæ canæ cun [les jones dl paîsc]COMPANION El dijous ballen amb [les noies del poble]COMPANION U ghjovi ballanu cù [e femine di u paese]COMPANION Ils dansent le jeudi avec [les filles du village]COMPANION A van a balâ la joibe cu [lis fantatis dal paîs]COMPANION Os xoves bailan co[as mozas da vila]COMPANION Que dànçan lo dijaus dab [las gojatas deu vilatge]COMPANION Ëi bala de juebia cun [la mutans dl paésc]COMPANION Il giovedí ballano con [le ragazze del villaggio]COMPANION în fiecare joi dansează în sat cu [fetele]COMPANION À quinta-feira, vão ao baile com [as raparigas da aldeia]COMPANION Lou dijòu danson emé [li fiho dóu vilage]COMPANION

256 A corpus-based analysis Rum Sard Spa Sur Val

Se duc să joace, joia, cu [fetele din sat]COMPANION Donnia jovia arrumban a ballare in bidda chin [sas pizzinnas]COMPANION El jueves bailan con [las muchachas del pueblo]COMPANION La gievgia saultan els cun [las mattauns dil vitg]COMPANION La gövgia vana a far la trais-cha cu[llas mattas dal cumün]COMPANION

The Slavic data are much more interesting in terms of structural variation, cf. (C42). Since the majority of the Slavic languages belong to the A-type, sentence XXI.103 forms a kind of shibboleth, in a manner of speaking. It is in this context that we expect an expression of the Comitative that is formally different from the one used for the Instrumental. This hypothesis is corroborated by the eight members of the A-type among the Slavic languages. In addition, Russian and Serbian also mark the temporal adverbial by the inflectional instrumental. (C42)

Slavic phylum XXI.103

Bie-R Bul Cro Cze Mac Pol Rus Ser Slok Slov Ukr

u þacver jany tancujuc’ z [mestackovymi dzjauþatami]COMPANION v þetvărtăk te tancuvat săs [selskite momiceta]COMPANION U þetvrtak plešu sa [seoskim djevojkama]COMPANION Tanþí každý þtvrtek s [dƟvþaty z vesnice]COMPANION vo þetvrtok tie igraat v selo so [devojkite]COMPANION W czwartek taĔczą z [wioskowymi dziewczĊtami]COMPANION po þetvergam oni tancujut s [derevenskimi devuškami]COMPANION oni þetvrtkom igraju sa [seoskim devojkama]COMPANION Vo štvrtok tancujú s [dedinskými dievþencami]COMPANION Ob þetrtkih plešejo z [vaškimi dekleti]COMPANION u þetver voni tancjujut’ iz [sil’c’kimi divþatami]COMPANION

In (C42), all the Slavic languages employ prepositions. This is unsurprising for Bulgarian, Macedonian and Slovenian, as we know that the first two lack case inflection and that Slovenian requires the inflectional instrumental to co-occur with the preposition z, independent of the distinction of Comitative vs. Instrumental. However, the other eight Slavic languages constitute the first group of languages which make use of formally distinct expressions for Instrumental and Comitative. While the Instrumental is encoded by the bare instrumental, as (C30) and (C36) demonstrate, the Comitative requires the additional compulsory employment of a preposition. Thus, their Comitatives receive the more complex encoding which, moreover, is also formally based on the simpler one of the Instrumental. As to

On being (dis)similar in Europe 257

markedness, the Comitative conforms to the general pattern described in Part B, Chapter 10. The minor Indo-European phyla behave largely as expected too, cf. (C43). The B-type languages employ the same marker for the Comitative and the Instrumental. Only the two A-type languages, Armenian and Lithuanian, display formal differences for the two categories. Like Serbian and Russian above, Lithuanian inflects the temporal adverbial for the instrumental, too. (C43)

Minor Indo-European phyla XXI.103

Alb-G Alb-T Arm Bre Gre Kur Lat Lit Rom Wel

Ata vallzojnë t’ejten bashkë me [vajzat e katundit]COMPANION Ata vallëzojnë të ënjten me [vashat e fshatit]COMPANION hingšabt‘i ǀrer6 nrank‘ parwm en [gywMac‘i aMj¸ikneri]COMPANION het D’ar yaou e tañsont gant [merc’hed ar geriadenn]COMPANION xoreúoun tƝn PúmptƝ me [ta korítsia tou xǀrioú]COMPANION Ew roja pêncúembê bi [keçikên gund re]COMPANION dîlanê dikin CeturtdienƗs viƼi iet dejot ar [ciema meitenƝm]COMPANION Ketvirtadieniais jie šoka su [kaimo merginomis]COMPANION Haj kade sako zholyine zhantar te khelen anda gav [e shejenca]COMPANION Ar ddydd Iau maen nhw’n dawnsio efo [merched y pentref]COMPANION

The Baltic language Lithuanian follows closely the pattern familiar from Slavic A-type languages because it simply adds a preposition to the NP whose constituents already bear the marker of the inflectional instrumental. Armenian is different in so far as the postposition het, which is the primary marker of the Comitative, does not combine with an NP inflected for the instrumental (the noun is in the dative instead), i.e., the expressions of Comitative and Instrumental are in no way related to each other. In addition, the C-type language Welsh employs again the regional variant efo as in (C37) and neither â nor gyda – the latter being the relator which is functionally specialised to the encoding of the Comitative. As both efo and â are syncretistic relators, their occurrence in this context is nothing extraordinary, cf. above. Among the non-Indo-European phyla, there is a strict separation of Atype languages and the rest. While C-type Hungarian and the B-type languages use their syncretistic relators, Basque, Finnish and Maltese make a clear formal distinction between Instrumentals and Comitatives. Note also that in none of the three latter cases does the expression of the Comitative derive from the one of the Instrumental nor vice versa. The formal differences of the relator used in the Azerbaijani examples (C32), (C38) on the one hand and (C44) on the other have no categorical implications: -la/-lä

258 A corpus-based analysis are only the cliticised allomorphs of the free postposition ilä. The rules regulating their distribution are still largely unclear although rhythmic patterns seem to play a role. (C44)

Non-Indo-European phyla XXI.103

Aze Bas Est Fin Hun

cümä axúamları [känd qızları]COMPANION ilä räqs edirlär ostegunero [herriko neskatxekin]COMPANION dantzan egiten dute Neljapäeviti nad tantsivad [külatüdrukutega]COMPANION He tanssivat joka torstai [kylän tyttöjen]COMPANION kanssa Eszerint minden csütörtökön elmennek táncolni [a falubeli lányokkal]COMPANION huma jiĪfnu ma[t-tfajliet tar-raƫal]COMPANION nhar ta’ ƪamis ahte sii lávejit dánsut [nieiddaiguin]COMPANION duorastaga Örne÷in, perúembe günleri, [köyün kızlarıyla]COMPANION dansa giderler

Mal Saa Tur

Except for those cases where a language is unexpectedly absent from an isogloss, the sample languages behave predictably. The representatives of the A-type and the B-type consistently go separate ways, whereas the Ctype languages go along with the B-type, as they generally employ their syncretistic relators in the above examples and not the functionally more specialised one. Languages belonging to the A-type strictly distinguish between Comitative and Instrumental whereas B-type and C-type languages lump the two categories together. So far, the behaviour of the sample languages suggests that the distinctions and patterns identified in the previous sections are indeed solid and robust for the prototypical contexts. However, what happens in those areas which are located in the transition zone between the two prototypes? For brevity’s sake, we only look at two more contexts, namely, an ORNATIVE and a borderline case which is situated somewhere between a BODY PART and a CONFECTIVE. In sentence XXVI.150, the Little Prince tells his adult companion about what will happen after the Little Prince’s departure from Earth. The Little Prince predicts that he will look at the stars and, in his imagination, they will turn into wells with a rusty pulley (an allusion to a shared experience of the two characters in the desert). In the situation described, there is an inanimate entity, the well, which has a certain property or part, the rusty pulley. The relation between the well and its rusty pulley is marked by a relator which happens to be identical with the Comitative and/or Instrumental marker of 59 of our sample languages. With the exception of Ladin, the gaps in the isogloss are all non-Indo-European languages. Syntactically, the relation is a typical attribute-head noun construction with the at-

On being (dis)similar in Europe 259

tribute coming in the guise of a PP or adverbial. Owing to the fact that the nuclear predication is based mainly on a copular verb (to be, to become, to turn into), we skip it in the schematic representation, cf. (C45). (C45)

Ingredients of XXVI.150

Participants

Relator

REFERENCE POINT

ORNATIVE

N[inanimate]: well and/or person inflection on the finite verb Subject

N[inanimate]: pulley

Preposition, Postposition, inflectional case (and combinations thereof)

Attribute

The possibility of connecting an attribute to its nominal head by means of a Comitative-Instrumental PP is realised by all members of the Germanic phylum, which thus demonstrates that they are faithful members of the Btype, cf. (C46). The one C-type language, Icelandic, uses the same casegovernment as in the Instrumentality and Comitativity contexts above, i.e., með governs the inflectional dative. There is thus no formal distinction between Comitative, Ornative and Instrumental. (C46)

Germanic phylum XXVI.150

Àlli Starne sin Brenne mit [roschtige Rolle]ORNATIVE Og så vil jeg se på stjernerne, og alle stjernerne bliver til brønde med [rustent hejseværk]ORNATIVE Dut en voor mij zal elke ster een put zijn met [een roestige katrol]ORNATIVE Eng All the stars will be wells with [a rusty pulley]ORNATIVE Far Og so fari eg at hyggja at stjørnunum, og allar stjørnurnar verða til brunnar við [rustaðum spæli]ORNATIVE Fris Alle stjerren sille putten wêze mei [in rustige katrol]ORNATIVE Ger Alle Sterne werden Brunnen sein mit [einer verrosteten Winde]ORNATIVE Ice Allar stjörnurnar verða brunnar með [ryðgaðri vindu]ORNATIVE Letz All Stäre sin da wéi Pëtze mat [enger verraschtener Poulie]ORNATIVE Lim-N Alle sjterre zulle pötte zeen mèt [eine verroesjdje zjwingel]ORNATIVE Lim-S Alle sjterre zalle putte zieje mit [inne verrosde zjwungel]ORNATIVE Nor Og så vil jeg se på stjernene, og alle stjernene vil bli til brønner med [rustent heiseverk]ORNATIVE Swe Alla stjärnor blir brunnar med [rostiga block för mig]ORNATIVE Yid Ale shtern veln zayn brunems mit [farzhaverte tritses]ORNATIVE Als Dan

Discounting Ladin266, which happens to be missing from the list, the observation made for the Germanic phylum also holds for the Romance lan-

260 A corpus-based analysis guages, cf. (C47). Each member of this phylum employs ComitativeInstrumental PPs to connect an attribute to its head noun. All of them use, for their Ornatives, the same relator that is also employed in contexts of Instrumentality and Comitativity. (C47)

Romance phylum XXVI.150

Ara

Todas as estrelas estarán pozos con [una carrucha enrobinada]ORNATIVE Toles estrelles serán pozos con [una roldana medio podre]ORNATIVE Dötes les stëres sará n zitl cun [na cirela inrujada]ORNATIVE Tots els estels seran com pous amb [politges rovellades]ORNATIVE Tutte le stelle saranu pozzi cù [a so tagliola rughjinosa]ORNATIVE Toutes les étoiles seront des puits avec [une poulie rouillée]ORNATIVE Dutis lis stelis a saran poþs cun t[une cidule inrusinide]ORNATIVE Tódalas estrelas serán pozos cu[nha polea enferruxada]ORNATIVE Totas las estelas e seran putz dab [ua poleja rolhada]ORNATIVE Tutte le stelle saranno dei pozzi con [una carrucola arrugginita]ORNATIVE Cada estela serà un potz amb [una carrèla rovilhada]ORNATIVE Stelele vor fi toate ca niúte fîntîni cu [scripeĠi ruginiĠi]ORNATIVE E todas as estrelas vão ser poços com [uma roldana enferrujada]ORNATIVE Tóuti lis estello saran de pous em’[uno carrelo rouvihado]ORNATIVE Toate stelele vor fi niúte fîntîni, cu [cîte un scripet ruginit]ORNATIVE Tottu sos isteddos an a esser putos chin d’[una carroccula ruinada]ORNATIVE Todas las estrellas serán pozos con [una roldana enmohecida]ORNATIVE Tuttas las steilas vegnan ad esser fontaunas cun [schirellas de ruina]ORNATIVE Tuot las stailas saran puozs cun [ün tuorn da ruina]ORNATIVE

Ast Bad Cat Cor Fre Friu Gal Gas Ita Lan Mol Por Pro Rum Sar Spa Sur Val

The data from the Slavic phylum in (C48) are much more interesting than the two phyla just discussed. The Slavic Ornatives require the use of the same preposition, as with the Comitatives in (C42), i.e., ORNATIVES behave exactly like Comitatives and are thus formally distinct from Instrumentals – with the exception of the notorious cases of Bulgarian, Macedonian and Slovenian, which belong to the B-type. In the Slavic phylum, the A-type languages display a closer connection of Comitatives and ORNATIVES. The second occur-

On being (dis)similar in Europe 261

rence of the inflectional instrumental in Polish is an instance of the predicative instrumental used to mark the predicate nominal, cf. Section 11.2.3. (C48)

Slavic phylum XXVI.150

Bie-R Bul Cro Cze Mac Pol Rus Ser Slok Slov Ukr

Use zorki buduc’ jak kalodzež z [zaržavelym kalaurotam]ORNATIVE Vsiþki zvezdi šþe bădat kladenci s [răždjasal cekrăk]ORNATIVE Sve üe zvijezde biti zdenci sa [zahrÿalim vitlom]ORNATIVE Všechny budou studny se [zrezavƟlým rumpálem]ORNATIVE Site svezdi gde bidat bunari so [rg’osali þekreci]ORNATIVE Wszystkie gwiazdy bĊdą studniami z [zardzewiaáym blokiem]ORNATIVE I vse zvëzdy budut toþno staryje kolodcy so [skripuþim vorotom]ORNATIVE Sve zvezde dze biti bunari sa [zardzalim þekrkom]ORNATIVE Všetky hviezdy budú studne so [zhrdzaveným hriadeĐom]ORNATIVE Vse bodo vodnjaki s [cvileþim škripcem]ORNATIVE I vsi zori budut’ nibi krinici z [iržavoju korboju]ORNATIVE

In the minor Indo-European phyla (C49), only three languages deserve some comment, because the bulk of the languages of this group behave exactly as the B-type languages of the major phyla, i.e., the same marker is used for Comitative, Ornative and Instrumental. This is different in Armenian, Lithuanian and Welsh. As to Lithuanian, it again follows the pattern characteristic of Slavic A-type languages: The Ornative requires the same encoding strategy as the Comitative and is thus distinct from the Instrumental. Armenian, however, employs the inflectional instrumental for the attribute (cf. [A47]–[A48] above), which means that this language diverges from the bulk of the other Indo-European A-type languages, where Ornative and Comitative are syncretistic. Welsh confronts us once more with the alternation of its relators â and efo. This time, it is the turn of â to be employed. Owing to the fact that â is syncretistic, Welsh can be said to behave largely like Germanic and Romance languages. (C49)

Minor Indo-European phyla XXVI.150

Alb-G Alb-T

Gjithë yjt do të jenë puse me [nji çekërk të ndryshkun]ORNATIVE Dhe do të jenë që të gjitha si ca puse me [një çikrik të ndryshkur]ORNATIVE Ev bolor astMer6 inj hamar klinen onc’ [or þĚþĚac’oM þ’arxov]ORNATIVE hin Ƶrhorner An holl stered a vo puñsoù gant [ur pole raouliet]ORNATIVE óla t’astéria tha ´xoun tƝgádia me [maggáli tou gurídzei]ORNATIVE

Arm Bre Gre

262 A corpus-based analysis Kur Lat Lit Rom Wel

Stêr hemû wê ji min re bibin bîrên bi [çirikên zingargirtî]ORNATIVE, hemû stêr wê ava wexarinê bidine min... Visas zvaigznes bnjs kƗ akas ar [þƯkstošu grieztuvi]ORNATIVE Visos žvaigždơs bus šuliniai su [aprnjdijusiais skridiniais]ORNATIVE Sako jek cherhaj xalying avla, [dulmutane buverosa]ORNATIVE Fe fydd y sêr i gyd yn bydewau â [phwlis rhydlyd]ORNATIVE ganddyn nhw

The absentees from the isogloss accumulate in the non-Indo-European phyla, cf. (C50). Four out of the nine languages do not participate in the isogloss. Of those which use their Comitative/Instrumental relators, Estonian follows the lead of all the other B-type languages mentioned so far, as it employs the same marker for Ornative as for Comitative and Instrumental. The four other languages belong to the A- and C-type, respectively. The two A-type languages Basque and Maltese use their Instrumentals to encode the Ornative, whereas C-type Hungarian employs the syncretistic Comitative-Instrumental relator. In contradistinction to these three languages, Georgian makes use of a strategy that is reminiscent of what we have encountered among Slavic and Baltic A-type languages: In addition to the inflectional instrumental, there is also the mildly cliticised postposition urt, which is normally employed in Accompaniment situations. Thus, Georgian treats the Ornative more like a Comitative than like an Instrumental. (C50)

Non-Indo-European phyla XXVI.150

Bas Est Geo

Izar guztiek edukiko dute [beren pertz herdoilaz]ORNATIVE putzu bat Kõik tähed muutuvad [roostes rattaga]ORNATIVE kaevudeks qvela varsk’vlavi udziro ch’ad c’armogvichndeba, [tavisi moch’riale oc’inarit]ORNATIVE-urt Minden egyes csillag kút lesz, [rozsdás csigával]ORNATIVE Il-kwiekeb kollha jkun fihom il-bjar bi[t-tarjola]ORNATIVE ddur

Hun Mal

The absence of the two representatives of the Altaic macrophylum constitutes the first example of a systematically induced divergence: Azerbaijani and Turkish employ derivational categories for the encoding of Ornatives (and partially also Confective) if the intra-participant relationship is of the attributive kind (cf. [A43] above). The relators ilä ~ -lä/-la and ile ~ -le/-la, respectively, cannot properly express such relationships. Sentence VIII.31 is something of a problem when it comes to classifying it: On the one hand, it involves a body part relationship, as the claws belong to the body of the tigers, yet on the other hand, the claws function as the weapons carried along by the tigers to threaten the rose which is boasting about its courage. For the sake of the argument, we consider the sen-

On being (dis)similar in Europe 263

tence an example of a CONFECTIVE in lieu of another instance of the ORNAThe relevant parts of the situation depicted are schematically identified in (C51). The sentence itself is most often an imperative; the auxiliary verb carrying the grammatical markers of the imperative corresponds to English to let. Since it is of no relevance for our topic, we do not expressly mention this component of the construction in (C51). TIVE.

(C51)

Ingredients of VIII.31

Participants

NucPred

REFERENCE POINT

CONFECTIVE

N[+human]: tigers and person inflection on the finite verb

N[inanimate]: claws

Subject

Adverbial/ Attribute

Relator

V[intransitive] Preposition, Postposito come tion, inflectional case (and combinations thereof)

With one exception, all Germanic languages conform to the same pattern. Icelandic still employs the same preposition með, as in the previous contexts, but this time the preposition governs the inflectional accusative which is crucial. The differential case government serves the purpose of distinguishing different kinds of Comitativity and Instrumentality. Confectives are in a different class than Ornatives. In this respect, Icelandic diverges considerably from the bulk of its sister languages. (C52)

Germanic phylum VIII.31

Als Dan Dut Eng Far Fris Ger Ice Letz Lim-N Lim-S Nor Swe Yid

Sie kenne kumme d’Tiger mit [ehre Kralle]CONFECTIVE For tigrene kan jo komme med [deres skarpe kløer]CONFECTIVE Laat ze maar opkomen, de tijgers met [hun klauwen]CONFECTIVE! Let the tigers come with [their claws]CONFECTIVE! Tí tikararnir kunnu altíð koma við [sínum hvøssu klóm]CONFECTIVE Lit se mar opkomme, dy tigers mei [har klauwen]CONFECTIVE! Sie sollen nur kommen, die Tiger, mit [ihren Krallen]CONFECTIVE! Þau mega koma, tígrisdýrin, með [klærnar sínar]CONFECTIVE! Lo kënnen d’Tigere roueg komme mat [hire Krallen]CONFECTIVE! Lot die tiegere mar kómme mit [hön klauwe]CONFECTIVE! Laot de tiegers mer kómme mèt [häör klauwe]CONFECTIVE! For tigrene kan jo komme med [klørne sine]CONFECTIVE! Låt bara tigrarna komma med [sina klor]CONFECTIVE! Zey kenen kumen di tigers, mit [zeyere negl]CONFECTIVE!

264 A corpus-based analysis The entire Romance phylum knows only one pattern, namely the very same we already encountered above: Without any exception, Comitatives, Ornatives, Confectives and Instrumentals have identical expressions, cf. (C53). The double occurrence of the relator in Moldavian is explicable in stylistic terms because the second occurrence of cu is part of a lexicalised PP with a slightly idiomatised meaning corresponding to English and stuff, and everything. Judging from the French original, cu tot is indeed an addition which is irrelevant to the Confective relation itself. (C53)

Romance phylum VIII.31

Ara Ast Bad Cat Cor Fre Friu Gal Gas Gher Ita Lan Mol Port Pro Rum Sar Spa Sur Val

Ya pueden benir os tigres con [a suyas unglas]CONFECTIVE! Ya puen venir los tigres co[les sos garres]CONFECTIVE! Ares dess mâ gní les tigres cun [sües grifes]CONFECTIVE! Ja poden venir tigres amb [les seves urpes]CONFECTIVE! Chì elli venganu i tigari cù [e so ranfie]CONFECTIVE! Ils peuvent venir, les tigres, avec [leurs griffes]CONFECTIVE! A puédin ben vignî, lis tîgris, cu [lis lôr sgrifis]CONFECTIVE! ¡Xa poden vi-los tigres co[as súas poutas]CONFECTIVE! Que pòden viéner, los tigres, dab [lors charpas]CONFECTIVE! Les dëssa mé uní, la tigres, cun [si sgrinfles]CONFECTIVE! Possono venire le tigri, con [i loro artigli]CONFECTIVE! I se pòdon venir frelhar, los tigres, amb [lors arpas]CONFECTIVE! N-au decît să poftească tigrii cu [ghearele lor]CONFECTIVE cu tot! Se os tigres, com [aquelas garras]CONFECTIVE, pensam que me metem medo. Podon ié veni, li tigre, emé [sis arpo]CONFECTIVE! N-au decît să poftească tigrii, cu [ghearele lor]CONFECTIVE! A ite no benin sas tigres chin [sas ungras issoro] CONFECTIVE! Ya pueden venir los tigres con [sus garras]CONFECTIVE! Els dueian mo vegnir cun [lur greflas]CONFECTIVE, quels tighers! Pür chi vegnan ils tighers cun [lur griflas]CONFECTIVE!

What strikes the eye immediately in (C54) is the fact that three A-type languages (Polish, Belarusan and Russian) which are geographical neighbours in the East are missing from the list. In connection to this phenomenon, it is worth noting that in conservative registers of these Slavic languages, it is still more common that nominal attributes of a head noun are turned into adjectives, and thus PPs are ruled out as a strategy. This preference for adjectivisation is said to be giving way to Comitative-like PPs, especially in colloquial Russian. For the remaining eight languages, the pattern is identical – and it is also identical with the one observed in (C48): All Slavic lan-

On being (dis)similar in Europe 265

guages use a preposition either with or without accompanying instrumental inflection. Thus, Ornatives, Comitatives and Confectives form a unity, whereas Instrumentality is distinct from them. (C54)

Slavic phylum VIII.31

Bul Cro Cze Mac Ser Slok Slov Ukr

Sega tigrite săs [svoite nokti]CONFECTIVE mogat da dojdat! Sad neka doÿu tigrovi, sa [svojim pandžama]CONFECTIVE! Jen at’ si pĜijdou tygĜi se [svými drápy]CONFECTIVE! Tie možat da dojdat, tigrite, so [svoite kanci]CONFECTIVE Mogu dodzi tigrovi sa [svojim kancama]CONFECTIVE! Nech len prídu tie tigre so [svojimi pazúrmi]CONFECTIVE! Naj le pridejo tigri s [svojimi kremplji]CONFECTIVE! Nexaj prixodjat’ xoþ tigri z [pazurami]CONFECTIVE – ne strasno !

Armenian is the only language missing from the isogloss among the minor Indo-European phyla, cf. (C55). The coding strategies meet our expectations. This is hardly worth commenting upon in the case of the B-type languages. Lithuanian conforms once again to the pattern familiar from the Slavic phylum and thus displays a closer connection between Comitative, Ornative and Confective. Welsh too uses the same relator â as with the Ornative, cf. (C49). (C55)

Minor Indo-European phyla VIII.31

Alb-G Alb-T Bre Gre Kur Lat Lit Rom Wel

Le të vijnë edhe tigrat me [kthetrat e tyne]CONFECTIVE! Le të vijnë, po deshën, tigrat me [ato kthetrat e tyre]CONFECTIVE! Gallout a reont dont, an tigred, gant [o skilfoù]CONFECTIVE! Mporeí na kataphtásoun oi tígreis me [ta núxia tous]CONFECTIVE De bila piling bi [navperûškên]CONFECTIVE xwe bên ku ez wan bibînim! Lai nu nƗk tƯƧeri ar [saviem nagiem]CONFECTIVE! Tegul tik jie ateina, tie tigrai su [savo nagais]CONFECTIVE! Akanak aba shaj aven e tigrishura [peske vundyenca]CONFECTIVE! Gadewch i’r teigrod â’[u crafangau]CONFECTIVE ddod

The larger gaps of the isogloss are again non-Indo-European languages. From (C56), three languages are missing (Azerbaijani, Turkish and Basque). Estonian and Saami are two more instances of typical B-type languages and thus employ identical markers for all the relevant categories. The C-type language Hungarian and the A-type language Maltese follow the same pattern as with their Ornatives. For Maltese, it is the Instrumental that is employed. Somewhat surprisingly, Georgian uses the bare inflectional instrumental and thus has two distinct constructions for the Ornative

266 A corpus-based analysis and the Confective. Moreover, Finnish employs the inflectional comitative. The non-Indo-European languages are heterogeneous as to the dividing line which separates Comitativity from Instrumentality. Ornatives and Confectives constitute the terrain where the languages diverge from each other, because some languages subsume these categories en bloc under Comitativity or Instrumentality, whereas others distribute Ornatives and Confectives differently over the two sub-areas of the continuum. (C56)

Non-Indo-European phyla VIII.31

Est Fin Geo

Nüüd võivad nad tulla, need tiigrid [oma küüntega]CONFECTIVE! Tiikerit saavat tulla [kynsineen]CONFECTIVE, minä en pelkää! aba, erti gabedon da movidnen vepxvebi [tavianti basri k'lanch'ebit]CONFECTIVE! Most aztán jöhetnek a tigrisek [a karmaikkal]CONFECTIVE! It-tigri jistgƫu jiƥu hawn, bi[d-dwiefer tagƫhom]CONFECTIVE! Diigerat gal ožžot boahtit [gaccaideasetguin]CONFECTIVE, mun gal in balla!

Hun Mal Saa

All in all, the languages of our sample display sufficiently similar patterns to justify the statement that they resemble each other (though not necessarily to the same extent). The five sentences or contexts we have just had a closer look at point to the fact that the prototypes of Comitatives and Instrumentals are indeed encoded by the primary markers of these categories in the vast majority of the sample languages. The distribution of the encoding strategies over language types is largely uncontroversial with the prototypes: A-type languages have two different strategies, whereas B-type languages employ only one. C-type languages do not behave homogeneously; some follow the B-type and others the A-type. It is worth mentioning that Lithuanian, and less obviously so Georgian too, follow the practice of its Slavic A-type neighbours, whereas Icelandic, Hungarian and Welsh prefer the patterns established for the neighbouring Germanic, i.e., B-type languages. The less prototypical areas are also the ones where more variation can be observed. CONFECTIVES and ORNATIVES do not always share the same fate. In B-type languages, both CONFECTIVES and ORNATIVES take the same relator as the syncretistic Comitative-Instrumental, if at all. In the two other types, the solutions are more variegated. Owing to the large number of Slavic A-type languages, there is a certain trend towards subsuming both ORNATIVE and CONFECTIVE under the Comitative – with Lithuanian and Finnish following in the Slavic footsteps, in a manner of speaking. Georgian splits the intermediate categories in two. Table (C57) summarises our

On being (dis)similar in Europe 267

above findings. Grey shading indicates where normally no Comitative and/or Instrumental relator is used. The ʌ marks those cases where there are at least potential alternatives which, however, are not attested in the sentences we discussed above.267 (C57)

Distribution of relators over contexts according to language groups

Languages Germanic (except Icelandic) Romance Bul, Mac, Slov Alb G/T, Bre, Gre, Kur, Lat, Rom Estonian, Saami Hungarian Turkish, Azerbaijani Welsh Icelandic Slavic Lit Fin Geo Arm Bas Mal

X/Y

Body part X/Y

Tool Type X/Y B

X/Y X/Y X/Y

X/Y X/Y X/Y

X/Y X/Y X/Y

X/Y X/Y X/Y

B B B

X/Y X/Yʌ ʌ X/Y1ʌ X/Y Xʌ X X2 X Y Y Y

X/Y X/Yʌ ʌ X/Y1ʌ X Xʌ X X1 Y Y Y Y

X/Y X/Y X/Y X/Y2 X/Y Y Y Y Y Y Y Y

X/Y X/Y X/Y X/Y1 X/Y Y Y Y Y Y Y Y

B C B C C A A A C A A A

COM

ORN

CON

X/Y

X/Y

X/Y X/Y X/Y X/Y X/Yʌ X/Y X/Y2ʌ X/Y X X X1 X X X X

According to the distribution patterns of the relators, we can identify two foci as distinct with a typological basis. Independent of their genetic background, the B-type languages of our sample share a large part of the general profile of the functional domains of their markers. At the opposite end of the scale, there is the homestead of the A-type, which comes in two varieties: First, there are the A-type languages which subsume both ORNATIVE and CONFECTIVE under the Comitative. All of these languages are located in one of the Eastern quadrants. Together they form a solid geographical block with a very strong Slavic component. Then there are three languages spoken in non-contiguous regions with widely diverse genealogies which nevertheless converge as to their distribution profiles: Armenian, Basque and Maltese draw the borderline between Comitative and Instrumental between the prototypical Comitative and the ORNATIVE and thus subsume ORNATIVE and CONFECTIVE under Instrumentality. Between these two

268 A corpus-based analysis groups is the C-type language Georgian, which treats the ORNATIVE as a kind of Comitative and the CONFECTIVE as a kind of Instrumental. Given that the sample languages largely converge as to the use of Comitatives and Instrumentals in prototypical contexts, one may ask why the similarity coefficient for the sample languages is not much higher. There is a simple answer to this question: The Little Prince does not contain too many sentences which involve appropriate contexts. This is perhaps one of the great disadvantages of our textual basis. As a matter of fact, TOOL and ACCOMPANIMENT in the narrowly defined sense are only a minority in the text. Interestingly, most of these cases are the basis for the most extended isoglosses. In (C58), we enumerate the appropriate cases for 29 isoglosses down to the arbitrarily chosen length of 40 languages, establish the length of the isogloss according to the number of languages participating in it, identify the absentees and determine the context. The labels for the contexts are such that they contain the English preposition with even where this violates the rules of English grammar. The contexts appear in the reversed order of (A30), i.e., the top line starts with prototypical TOOLS whereas the bottom line contains a specimen of Comitativity. Grey shading marks those cases where a group does not allow for absentees. (C58)

Further contexts in The Little Prince

Sentence Ger Als

Rom Ita, Por

I.7a

57

IV.4a

41

Dan, Eng, Far, Ger, Ice, Lim-S, Nor

Ast, Fre, Gas, Gher, Pro, Sur

V.36a

50

Als, Fris, Ice

Fre, Gas, Pro, Sar, Val

XXI.135a 44

Eng, Fris, Bad, Fre, Letz, Nor Friu, Gal, Pro

XV.58a

Eng, Nor

51

Ast, Cor, Fre, Friu, Gher, Ita, Mol, Por, Pro, Sar ,

Absentees Sla Pol, Slok

Context Minor Rom

Bie-R, Lat, Pol, Rus, Lit, Slok, Ukr Rom, Wel

Ser, Slov

NIE Bas Hun

draw with pencil see with telescope

Lat, Wel

Bas, Est, Fin, Saa

Alb-T, Lat, Rom

Aze, Est, shelter with Fin, Geo, screen Hun, Saa

Wel

pierce with roots

write with pencil

On being (dis)similar in Europe 269 Sentence XV.59a

54

Ger Eng

II.21a

42

Far, Ice

Absentees Sla Minor Wel

Rom Ast, Fre, Gal, Gas, Ita, Por, Pro, Sar

Cor, Lan, Bie-R, Por, Pro, Cro, Cze, Rus, Ser, Slok, Slov

XXI.144a 57

Bie-R, Rus, Ukr

Context NIE write with ink

Arm, Bre, Lat, Lit, Rom

Aze, Bas, Geo, Saa

look with eyes

Alb-T, Arm,

Aze, Geo

see with heart

VIII.29a

50

Nor

Fre, Friu, Bie-R, Pro Cro

Alb-G, Arm, Gre

Aze, Bas, Geo, Mal, Saa

torment with vanity

VII.22a

45

Eng, Nor

Por, Pro, Sur

Alb-G, Arm, Kur, Lit, Wel

Aze, Geo, Saa, Tur

threaten with thorns

IV.39a

40

Als, Dan, Ara, Ast, Cro, Pol Eng, Far, Bad, Por, Ice, Lim- Spa N, Swe

Alb-G, Kur, Lit, Wel

Aze, Bas, Fin, Mal, Saa, Tur

bother with questions

XIII.51a

47

Cro; Mac Alb-T, Arm, Gre, Kur

Geo, Tur

do with stars

XIII.48a

44

Cro, Rus, Alb-T, Ukr Arm, Gre, Kur

Aze, Bas, Geo, Tur

do with stars

XIII.82a

40

Cat, Cor, Fre, Friu, Gas, Ita, Lan, Pro, Sar Cat, Cor, Fre, Friu, Gas, Ita, Lan, Pro, Sar Dan, Far, Ara, Bad, Ice Cat, Cor, Fre, Friu, Gas, Gher, Ita, Lan, Pro, Sar

Bul

Bas, Est, Geo, Hun, Tur

do with stars

Bie-R, Cze, Rus, Slok, Slov

Arm, Gre, Kur

270 A corpus-based analysis Sentence XVII.35a 44

Ger Far, Fris, Ice, Nor

XXVI.73a 44

41

NIE Bas, Geo, Mal

Aze, Fin, Geo, Mal, Tur

like with water

Alb-T, Arm, Bre, Kur, Mal

Aze, Fin, Geo, Saa, Tur

like with flower

Arm, Bre, Kur, Wel

Aze, Bas, Est, Fin, Saa, Tur

be with hammer

Gre, Kur, Lit, Rom Cze, Rus, Arm, Ukr Rom, Wel

Bas, Fin, Saa

escape with noise

Aze, Bas, Est, Fin, Hun, Saa, Tur

house with geraniums

Gre, Rom

Aze, Bas, Geo

talk with person

Geo

talk with snake

Fin

laugh with me

XXVI.26a 47

Far, Ice, Nor

Bie-R, Cro, Mac, Slok, Slov, Ukr Bad, Cat, Cro, Cze, Gher, Sar Ser

IV.30a

49

Far

Bad

II.1a

45

Dut, Eng, Cor, Friu, Ice, Lim- Gher, Ita, N Lan, Mol, Pro, Sar, Sur, Val

XXVI.29a 62 XXVI.101 60

Context Minor Arm, Kur, Lat, Lit, Rom Alb-T, Arm, Bre, Gre, Kur

Cor, Fre, Friu, Gal, Gas, Gher, Ita, Lan, Pro, Sar Bad, Cor, Fre, Friu, Gal, Gas, Gher, Ita, Lan, Pro, Sar Cor, Fre, Lan, Por, Pro, Sar, Sur

XXVI.69a 42

VII.35a

Absentees Rom Sla Ara, Cor, Bie-R, Por, Spa, Rus, Ukr Sur

Ice Rum

Slov

Gre

problems with flower

On being (dis)similar in Europe 271 Sentence XXI.9a

49

IV.50a

59

XXI.11a

59

I.27a

Ger Rom Als, Fris, Ice, LimN, Lim-S

Absentees Sla Minor Ser Alb-G, Alb-T, Gre, Kur, Lit

Mol, Rum, Val

Arm, Wel

Ice

Cor

Gre

44

Fris, Ice, Lim-N, Nor

XXIV.49a 51

Ice, Letz

Context NIE Aze, Bas, Geo, Tur

play with me

walk with lamb Fin, Geo

play with you

Bad, Cor, Cro, Cze, Arm, Mol, Pro Rus, Rom Slok, Slov

Aze, Fin, Geo, Saa

mess with adults

Ast, Friu, Mol, Por, Sar

Bas, Geo, Saa

agree with fox

Alb-T, Kur, Lit

The most extended isoglosses in (C58) are on the Comitative side: Talking with snakes, laughing with somebody, playing with somebody and walking with a lamb are contexts which are shared by 59–62 languages. Within the frame of Instrumentality, only drawing with a pencil and seeing with the heart, with 57 languages each, come near these numbers. As one can see from the list of absentees for each context/isogloss, there are also some striking cases of parallel behaviour ex negativo. The Slavic phylum in its entirety forms part of ten isoglosses, and the Germanic phylum appears as a block in eight isoglosses, whereas the Romance phylum (three isoglosses), the minor Indo-European phyla (one isogloss) and the non-Indo-European phyla (three isoglosses) only rarely participate fully in the isoglosses. Starting with the Romance phylum, we immediately see that certain contexts are disfavoured by many Romance languages. A typical example of TOOL, namely, hitting with a hammer, does not involve a Comitative-Instrumental marker in ten out of twenty Romance languages. The same rate (with up to +3 or –1) applies to an example of MATERIAL, namely, writing with pencil, the EQUATIVE constructions, namely, be like with water/flower, the PATIENT-like context, namely, doing (something) with stars, and the RECIPROCITY example of talking with somebody. These are all contexts which do not seem to pose any problems for the other language groups – with the exception of the RECIPROCITY case, which excludes four of the fourteen Germanic languages. Note that talking with snakes involves Comitatives/

272 A corpus-based analysis Instrumentals in all the languages except Icelandic and Georgian! For Germanic languages, the telescope is an object which does not invite the use of the Comitative-Instrumental relator: 50% of the phylum employ a different relator. We encounter the same percentage of Germanic absentees with the context of bothering (somebody) with questions. As to the Slavic languages, there is one context in which the vast majority of the phylum requires an expression different from the Comitative and/or Instrumental, namely, looking with the eyes is disfavoured by eight of the eleven Slavic languages. Over 50% of the Slavic phylum do not use a Comitative and/or Instrumental in the context of be standing with a hammer (in hand). Since the two remaining groups are genetically heterogeneous, the results are less consistent. Nevertheless, with seven out of the nine non-Indo-European languages missing, the ORNATIVE context house with geraniums is evidently one where the use of a Comitative/Instrumental is largely an IndoEuropean solution. This is in line with the fact that be standing with a hammer is also disfavoured in non-Indo-European languages as a context requiring the employment of a Comitative/Instrumental. If we include those contexts which are only represented in isoglosses with less than 40 languages, there are about a dozen situations which contain contexts which may pass as bona fide instances of TOOL, the same amount of contexts where (mostly identical) BODY PART INSTRUMENTS are used. In addition, there are also twelve instances of ACCOMPANIMENT and about another half dozen of ORNATIVES and CONFECTIVES. These add up to 42 cases where our expectations are high as to pan-European convergence. However, this number, on the one hand, goes far beyond many of the similarity coefficient values calculated above and, on the other, it does not come near the attested instances of identical use of relators in numerous pairs of languages (cf. above). Going by the absolute number of identical uses of the relators, we learn that there is not a single pairing of languages involving Basque with n • 42, i.e., all 63 language pairs with Basque fail to reach the calculated amount. On the other extreme of the scale of Europeanness, German participates in 33 combinations which reach or even go beyond 42 identical tokens. This still leaves us with 31 language pairs which fail to fulfil our expectations. A language with a medium similarity coefficient is Breton. For Breton, 17 pairings can be identified for which n • 42 holds, and thus more than twice as many cases are located below this threshold level. This means that the potential common ground of the sample languages is not fully exhausted. Where the languages converge the most, prototypical cases of Instrumentality and Comitativity are involved.

On being (dis)similar in Europe 273

These contexts, however, are outnumbered by others, among them very prominently MANNER – and it is not only for this function that the sample languages diverge widely. Comitatives and Instrumentals fulfil functions in some languages which are excluded in other languages. In Section 11.2.3, we have a cursory look at some of these functions which contribute to the diversity of solutions in our sample. 11.2.3. Where European languages disagree In this section, we look at a selection of cases where the languages of our sample tend to diverge from each other and thus either form smaller groups of languages with identical solutions or turn out to be full-fledged isolates. With a view to keeping the text length within reasonable bounds, we refrain from presenting examples for every singularity. In lieu of a full catalogue of idiosyncrasies, we survey those areas where divergence is especially interesting. – MANNER Many languages of the sample frequently employ PP as MANNER adverbials which correspond to proper adverbs of other languages. As far as we can judge from the evidence of The Little Prince, MANNERencoding PPs are in principle possible in every sample language. However, some of the languages extensively rely on this phrasal strategy, whereas others only occasionally resort to it. The Galician example in (C59) is symptomatic of the predilection for Comitative-Instrumental PPs with MANNER functions in the Romance phylum. (C59)

Galician (LPP Galician III.47) entón o principiño advertiume then DET.M prince:DIM notify:PRET.3SG:me con gravidade X/Y earnestness ‘Then the Little Prince told me earnestly.’

While it is true that English too uses PPs occasionally to encode MANNER and thus converges with Galician in a small number of cases268. Galician is nevertheless much more prone to use PPs introduced by the syncretistic Comitative-Instrumental preposition. This is remarkable because Galician and its sister languages of course have productive means at their disposal to freely form adverbs. In (C60), we list those examples where Galician opts

274 A corpus-based analysis for the PP-strategy in contradistinction to regularly derived adverbs in the English version of The Little Prince. Note that there is only scarce evidence for the reverse relation, i.e., English hardly ever uses a PP where Galician employs an adverb. (C60)

PPs versus adverbs

Sentence II.51 III.25 III.47 IV.48 VIII.33 IX.3 IX.10 X.63 XXVI.33 XXVI.42

Galician con indulxencia con dozura con gravidade con lixeireza con dozura con coidado con regularidade con firmeza con moita tristeza con moita melancholía

English indulgently gently earnestly carelessly sweetly carefully steadily firmly gravely sadly

Thus, the preference of Galician for the PP-strategy for the purpose of encoding MANNER accounts for ten cases of divergence between Galician and English alone. Similar rates apply to many of the other Romance languages – and also to Maltese, both varieties of Albanian, etc. The differences are simply a matter of quantities, not of qualities: Nor is the PPstrategy as such alien to English or other languages, nor is there a morphosyntactic reason for the avoidance of proper adverbs in Galician. The divergent preferences are more a matter of style than a grammatically induced phenomenon. – Spatial In addition to encoding Comitativity and/or Instrumentality, the relators often fulfil other functions among which spatial relations are widespread, although The Little Prince does not contain particularly many examples, as the adequate contexts of a mostly prolative or perlative meaning are rare. (C61) is a typical example from Faroese, whose relator við has a particularly large functional domain (Barnes 1994). (C61)

Faroese (LPP Faroese XXVI.1) við síðuna av brunninum X/Y side:ACC:DET:ACC of well:DAT:DET:DAT stóðu toftirnar stand:PRET:3PL remains:NOM.PL:DET:NOM.PL

On being (dis)similar in Europe 275

av einum gomlum múri of a:DAT old:DAT wall:DAT ‘At the side of the well, there stood the remains of an old wall.’ The examples we encounter in text are mostly from non-Indo-European languages, though not exclusively so. The spatial relation is one of proximity (with or without contact) mostly at the side of the reference object and extended along the reference object. (C62) surveys a handful of examples from a variety of languages. The English glosses are the terms used in the English translation of The Little Prince. (C62)

Spatial examples

Sentence XXIV.31 X.1 I.23 XXVI.171

Examples for COM/INS Georgian gverdit Finnish alueella Maltese mad-dinja kollha Faroese við føtur hansara

English beside in the neighbourhood over all parts of the world close to his ankle

Notwithstanding the fact that English with itself has spatial functions too (not attested in the sample text!), (C61–62) contain examples which normally do not invite the use of English with as a translation equivalent of the Comitatives/Instrumentals of the other languages. Romance and Germanic languages – except Faroese – disfavour the usage of their ComitativeInstrumentals for spatial functions. It comes as no surprise, however, that Finnish employs the inflectional adessive relatively frequently also for purely spatial functions: There are altogether 39 cases in the Finnish translation of The Little Prince which lend themselves to a classification as spatial relations. – Temporal Closely related to the spatial usage of the relators under scrutiny, there are temporal functions fulfilled by them too. Owing to the fact that time adverbials are much more frequent in our sample text, there is no shortage of examples (as opposed to spatial ones). Consider (C63) from Lithuanian. (C63)

Lithuanian (LPP Lithuanian IV.8) šƳ asteroidą tik vieną kartą 1909 this:ACC asteroid:ACC only one:ACC time:ACC 1909 met-ais pastebơjo pro teleskopą telescope:ACC year-X sight:PRET.3 through

276 A corpus-based analysis vienas turkǐ astronomas Turk:GEN.PL astronomer:NOM one:NOM ‘Only once in 1909 did a Turkish astronomer catch sight of this asteroid through a telescope.’ In (C64), various examples from different languages contrast with their English translation equivalents which never involve the preposition with. (C64) Sentence II.8 III.35 II.8 XIV.32 X.19 I.19

Temporal examples Examples for COM/INS Lithuanian apyaušriu Finnish yöllä Maltese mas-sebƫ Russian veþerom Belarusan gadami Ladin cun sies ani

English at sunrise at night at sunrise in the evening (it is) years (since) at the age of six

The employment of Comitatives and Instrumentals for the expression of time adverbials is rather common in the Slavic phylum and in Lithuanian. Outside the Indo-European languages, this strategy is particularly strong in Finnish and Maltese. However, these preferences do not prevent an occasional example from being found in Romance and Germanic languages. – Passive agent Where there are personal passives in European languages, we first of all encounter the general distinction of passives which allow for the expression of the agent and those which do not. Among those which express passive agents overtly, a small number employ their Comitatives/Instrumentals for this purpose (sometimes in opposition to other encoding strategies which have to be used with certain kinds of agents). In (C65), Belarusan encodes the animate passive agent by a PP as an Instrument adverbial. (C65)

Belarusan (LPP Belarusan IV.9) gety asteroid byu zauvažany u teleskop this asteroid was seen through telescope tol’ki adein raz u 1909 godze adn-ym one-Y only one time in 1909 year:LOC tureck-im astranom-am Turkish-Y astronomer-Y ‘This asteroid was seen just once in 1909 through the telescope by a Turkish astronomer.’

On being (dis)similar in Europe 277

Passives, however, are again rather infrequent in The Little Prince and thus do not account for many cases of divergence of our sample languages. In (C66), we only give two more examples of sentence IV.9. (C66) Sentence IV.9

Passive agents Examples for COM/INS Breton gant ur steredoniour turk Russian odnim tureckim astronomom

English Cf. (C66)

The marking of passive agents is one of those areas where the languages of our sample differ categorically from each other because the choice of category and expression are not entirely free. Owing to the small number of full blown passive constructions in the sample text, the structures of the kind exemplified in (C65–66) are only responsible for a small segment of all divergences. – Predicative possession Predicative possession is another area where only a small number of our sample languages employ their Comitatives/Instrumentals. The language which is responsible for the highest number of cases in the sample text is Finnish, cf. (C67). (C67)

Finnish (LPP Finnish IV.25) kuinka monta veljeä how many:PART brother:PART ‘How many brothers has he?’

häne-llä he-Y

on? be

However, for the other languages, which otherwise employ similar strategies, it is hard to come by convincing examples in The Little Prince. Thus, Welsh is missing from (C68) probably because the translator has opted for a style which follows more closely the practice of written Northern Welsh, in which the use of the Comitative relator gyda(g) for predicative possession is unusual, cf. (C43) above. (C68) Sentence II.53 XVII.17 XVII.45

Predicative possession Examples for COM/INS Icelandic hann er með horn Portuguese estava com medo Maltese jiena b’saƫƫti aktar

English it has horns be afraid (= had fear) I am more powerful (= have my power more)

278 A corpus-based analysis As a matter of fact, the examples in (C68) do not strictly require a translation into English which would contain the verb to have. This is explicable as an effect of the more fine-grained differentiation of various kinds of possessive relations in the above languages in comparison to English – a language which neutralises many potential subcategories of possession. In none of the languages mentioned in (C68) is the number of cases even remotely similar to the one attested for Finnish in our sample text: The inflectional adessive is used 60 times for the purpose of marking the possessor in a construction of predicative possession. Thus, this function also outnumbers the instances of the spatial usages of the same inflectional case, cf. above. – Equative One of the idiosyncrasies of Welsh which has a bearing on the degree of similarity of the Comitatives/Instrumentals of our sample languages is the equative, cf. (C69). (C69)

Welsh (LPP Welsh VII.36) rwyt ti-n siarad DEC:be.2SG you-in speak fath â phobl mewn kind X/Y people in ‘You are talking just like the adults!’

yr DET

un one

oed age

The preposition â obligatorily forms part of the equative constructions yr un N â Q ‘(of) the same N as Q’ and mor P â Q ‘as P as Q’. With 28 examples, these constructions occur relatively often in the sample text. These constructions are the most grammaticalised expression of equative relations in Welsh. The use of â is compulsory and thus cannot be explained as a matter of style. – Prohibitive The other major idiosyncrasy of Welsh is the obligatory use of â in the prohibitive construction, cf. (C70). (C70)

Welsh (LPP Welsh II.22) peidiwch ag anghofio fy mod my be stop:2PL X/Y forget filltiroedd o unrhyw fro from some region mile:PL ‘Do not forget that I was thousands of habited place.’

i fil o I thousand from gyfannedd inhabited miles away from any in-

The construction consists of the prohibitive auxiliary peidio ‘to stop’, inflected for person and imperative, plus the verb noun representing the lexi-

On being (dis)similar in Europe 279

cal verb, which is connected to the auxiliary via the relator â. Â can neither be dropped nor can any of its quasi-synonyms replace it in the prohibitive construction. There is no other equally highly grammaticalised construction with a similar function in Welsh. In the Welsh translation of The Little Prince, we have encountered 18 tokens of the prohibitive construction. – Arithmetic Idiosyncrasies are of course not restricted to Welsh. In the Romanian version of The Little Prince, the businessman has a peculiar way of making sums, cf. (C71). (C71)

Romanian (LPP Romanian XIII.5) trei úi cu doi fac three and X/Y two make ‘Three plus two make five.’

cinci five

In (C71), the plus function in the addition is expressed twice, namely once by the usual coordinating conjunction úi ‘and’ and once again by the Comitative-Instrumental relator cu ‘with’. This doubling occurs consistently wherever the businessman is busy counting his goods (altogether seven occasions). However, cu is never used in the Moldavian version, which otherwise closely resembles the Romanian one. In Moldavian, plus alone is sufficient as the verbal equivalent of the plus sign (which is also possible in Romanian proper).269 The motivation behind the use of cu in the Romanian version is perhaps the attempt to avoid misunderstandings, as the complex numerals like douăzeci úi úase ‘twenty-six’ regularly include the coordinating conjunction úi, and thus parsing problems may arise when úi could be interpreted either as part of a complex numeral or as the verbal plus sign. In contradistinction to the two previous examples from Welsh, the Romanian idiosyncrasies are clearly related to deliberate stylistic choices of the translator. They are by no means of a systematic nature. – Predicative instrumental With the notable exception of Bulgarian and Macedonian, the predicative instrumental is a typical trait of Slavic languages (see the numerous contributions to Comrie 2001), especially of the Eastern branch, Polish and Lithuanian as well. A typical Russian example is (C72). (C72)

Russian (LPP Russian X.88) ja naznaþu tebja you.SCC I declare:FUT.1SG ‘I make you a minister.’

ministr-om minister-Y

280 A corpus-based analysis The predicative instrumental is used on the predicate nominal, especially in equative, essive and translative constructions. The predicate nominal is most often an adjective or a noun, but, more rarely, it may also belong to other word classes (numerals), cf. Table (C73), which contains a selection of cases from Belarusan. Single underlying identifies the full predicate nominal; boldface, as usual, indicates the relators. (C73)

Predicative instrumental in Belarusan

Sentence IV.54 VIII.51 X.11

Belarusan ja bajusja stac’ padobnym adþubau sjabe vel’mi njašþasnym moža byc’ nekamu karalem

XXVI.100

zastanešcja maim sjabram

English I may become like the grown-ups It made him very unhappy being […] a king over somebody You will always be my friend

Interestingly, the languages which employ predicative instrumentals do so to different extents. Polish has the highest number of tokens, namely, 65. In the translation of The Little Prince into Belarusan, we count 46 instances of the predicative instrumental, as opposed to 36 tokens in Russian and only ten examples in the Lithuanian version. Notwithstanding these differences in frequency, the predicative instrumental is surely not a mere stylistic device, but a systematic difference of grammars. – Infinitive particle Similarly, the infinitive particle me is a systematic trait which distinguishes the Gheg variety of Albanian from the rest of the sample languages, including Gheg’s closest relative, the Tosk variety of Albanian (which is renowned for the typical Balkanism of infinitive avoidance).270 In Gheg, a subordinate non-finite verb form – here: A participle – is connected to the matrix verb via me, cf. (C74). (C74)

Albanian-Gheg (LPP Albanian-Gheg I.25) u mësova me dallue Kinën REFL learn:AOR:1SG X/Y discriminate:PTCPL China:ACC nga Arizona me nji të vështruem from Arizona X/Y one of look ‘I learned to distinguish China from Arizona at the first look.’

The construction is structurally similar to the English one, in which the infinitive particle to corresponds largely to Gheg me. The frequency of the Gheg infinitive particle is expectedly high: With 142 tokens, it accounts for more than a third of all instances of me attested in the Gheg translation.

On being (dis)similar in Europe 281

Moreover, the infinitive particle is responsible for over 55% of all isolates of Gheg Albanian. Note, however, that even if we discount the infinitive particle, me is still more frequent in Gheg than in Tosk (in Gheg, 261 tokens remain as opposed to the 219 attestations reported for Tosk). Thus the striking differences between the two Albanian varieties cannot be reduced to the absence and presence of an infinitive particle alone, although this is admittedly a very strong factor determining dissimilarity.271 The above areas of divergence can be summarised and ordered according to their relative location on the style-grammar continuum, cf. (C75). The no-man’s land in between the two extremes is characterised by an increase of grammatical obligatoriness from bottom to top and a decrease of grammatical obligatoriness from top to bottom. Spatial and temporal categories are located in this grey area because there are often alternative and equally acceptable ways of expressing the content without the involvement of a Comitative/Instrumental. (C75)

Style-grammar continuum

Grammar

Predicative Prohibitive Equative Passive Agent Infinitive Instrumental Particle predicative possession spatial relations Time MANNER

style

Arithmetic

What is especially remarkable about (C75) is the fact that there is a split: In A-type languages or C-type languages, those functions which are nearer to the top of the table are fulfilled by the Instrumental and never by the corresponding Comitative. The closer we get to the bottom-line of Table (C75), the more likely it becomes that the expression of the Comitative is employed. This observation fits in nicely with the ones we made in Chapter 10 above, where the different degrees of grammaticalisation of the categories under scrutiny were discussed more generally. As is evident, the sample text imposes too many restrictions on the corpus-based analysis to yield absolutely satisfactory results. The Little Prince serves as general guideline and allows us to identify areas which are worth being investigated more thoroughly. Furthermore, the data scrutinised so far already show clear tendencies as to where languages converge and where they diverge. However, many aspects and functions of the relators are not covered by the text and the translations. With a view to determining

282 Case studies whether or not the three different types of languages identified in the previous parts and chapters behave similarly on the functional level, it is necessary to look at the functional domains of the relators in individual languages, independent of the sample text. Therefore, we continue our treatment of Comitatives and Instrumentals with three case studies. In Section 12.1, we look at a representative of the A-type, namely Maltese. The Btype is the topic of Section 12.2, where we describe the state of affairs in Latvian. Icelandic is the C-type language chosen for the discussion in Section 12.3. The results of the case studies and of the corpus-based study are compared and evaluated in Section 12.4. 12. Case studies For the following case studies, we rely on three different text corpora of the languages involved. The texts are listed in the bibliography. They complement the findings we made on the basis of the translations of Le Petit Prince. As this sample text has been shown to be much too restricted to draw definitive conclusions for a wide range of aspects, an analysis of further texts is called for. However, this time the corpus is not subject to a quantitative study. 12.1. A-Type: Maltese 12.1.1. Generalities Maltese is a Type-A language. There are two prepositions which serve as primary translation equivalents of English with, namely ma’ (glossed X) and bi (glossed Y). Their functional domain is described in some detail in Aquilina (1987: 113–114; 1990: 780) and Borg and Azzopardi-Alexander (1997: 139–170). Historically, the two Maltese prepositions correspond to the prepositions ma’a and bi of Classical Arabic, in which they had a very similar catalogue of functions. Modern regional varieties of Arabic – be they spoken in the Maghreb or elsewhere – still employ cognate prepositions, at least for the purpose of marking the instrumental relation.272 There are two negative forms for both bi and ma’: The preposition bla ‘without’ (< Arabic bi + la ‘with not’) and the complex preposition mingƫajr ‘without’ (< minn ‘from’ + gƫajr ‘except’), which neutralise many of the functional distinctions associated with ma’ and bi, respectively.273

A-Type: Maltese 283

In their lexical form, ma’ and bi are of equal segmental complexity reflecting the same syllabic structure. It is therefore not possible to decide on the markedness relation that may hold between ma’ and bi on the basis of the lexical forms. However, in combination with a noun or a full phrase (DP, QP, NP) as their complement, the prepositions are subject to morphonological changes determined by the phonological properties of the immediately adjacent elements to the right. Just like most of the monosyllabic prepositions274 of Maltese, ma’ and bi behave like proclitics in this context, i.e., they coalesce with the definite article which itself is a proclitic (or even a prefix in statu nascendi) of the NP. Ma’ is reduced to the initial consonant if it occurs to the left of a vowel-initial autosegmental word.275 Bi, however, goes further, because it also cliticises to nouns or initial words of NPs which lack the definite article and start with a consonant. In this respect, bi behaves like the spatial preposition fi ‘in’: Both lose their vowel segment and become proclitics to words starting with a vowel or a single consonant, whereas ma’ keeps the phonological shape of the lexical form in front of consonants, cf. (C76). (C76) Maltese – Uninflected prepositions (C76.1) bi (Ƥrajjiet 57) kompla n-tafa’ b-ruƫ-u u REFL-throw Y-soul-3SG and continue ƥism-u fuq ix-xogƫol on DEF-work body-3SG ‘He kept on dedicating himself (with body and soul) to the work.’ (C76.2) ma’ (Ƥrajjiet 98) min j-gƫid li ra-w-h gƫaddej fuq pass:PTCPL on who 3-say REL see-3PL-3SG.M karretun ma’ ƫabib tiegƫ-u Fidiel cart X friend of-3SG.M Fidiel ‘Some say they saw him pass by on a cart with his friend Fidiel.’ Owing to the fact that, in comparison with ma’, bi is more prone to reduction and cliticisation, one may conclude that bi is associated with at least a slightly higher degree of grammaticalisation. This assumption is in line with the frequency with which bi is employed in Maltese texts: Both prepositions are rather frequent items, but bi occurs significantly more often than ma’. A random check of the text frequency on the basis of the first 30 pages of a story (= [L’Imsella]) yields 114 occurrences of bi, as opposed to 52 occurrences of ma’. This pattern is symptomatic of the higher functional load of bi.

284 Case studies Before we scrutinise the functional aspects of the two prepositions, it is in order that we have a look at their pronominal forms. As is the case for the majority of Maltese prepositions, bi and ma’ have special pronominal paradigms when their complement is pronominal, cf. (C77). The full paradigm of pronominal forms of the two prepositions under scrutiny is given in Table (C78). (C77) Maltese – Pronominal prepositions (C77.1) bi (Ƥrajjiet 12) kif wasal-t sib-t l omm-i ti-stennie-ni how arrive-1SG find-1SG to mother-1SG 3SG.F-wait-1SG mara xwejƫa inkwietata bi-ja u b-il-maltemp woman old:F concerned:F Y-1SG and Y-DEF-bad.weather ‘When I arrived, I found my mother waiting for me, an old woman who was concerned because of me and because of the bad weather.’ (C77.2) ma’ (Ƥrajjiet 37) nies li n-iltaqa’ magƫ-ƫom bde-w REL 1SG-meet X-3PL start-3PL people j-ieqf-u j-kellm-u-ni 3-speak-PL-1SG 3-stand-PL ‘People (with) whom I met began talking to me.’ (C78)

Paradigms of pronominal prepositions

Word forms free 1SG 2SG 3SG.M 3SG.F 1PL 2PL 3PL

ma’ ma’ miegƫi miegƫek miegƫu magƫha magƫna magƫkom magƫhom

bi bi bija bik bih biha bina bikom bihom

Translation ‘with’ ‘with me’ ‘with you’ ‘with him’ ‘with her’ ‘with us’ ‘with you’ ‘with them’

Both prepositions have also contributed to the creation of secondary conjunctions: billi ‘by doing, because of’276 and malli ‘as soon as’277 are based on the univerbation of a bipartite construction PREPOSITION + illi ‘that’ (also found with ta’ ‘of’ = talli ‘because of’, minn ‘from’ = milli ‘from that’). Other former PPs with bi as head have been lexicalised, such as, e.g., adverbials like bilfors ‘perforce, obviously’ < bil-fors ‘with the force’, biĪĪejed ‘enough’ < biĪ-Īejed ‘with the increase’, etc. 278 Combinations of ma’ and other elements share the fate of those which involve bi. That is why we find matul ‘during’ < ma’ + tul ‘with length’, madwar ‘around, surroundings’ < ma’ + dwar ‘with about, ma-

A-Type: Maltese 285

ƥenb ‘close to’ < ma’ + ƥenb ‘with side’279, etc. Interestingly, the univerbations with ma’ as the initial component are more grammatical in the sense that the new words function as prepositions, whereas the univerbations with bi as initial component tend more to the lexical side, being mostly adverbial modifiers. We will take up this issue again in Section 12.1.3 below. 12.1.2. The functional domain of bi Irrespective of the fact that ma’ and bi both translate into English as with in the appropriate contexts, the two Maltese prepositions are not synonymous – not even partially. Of the two, bi is the one which is used to indicate instrumental relations proper. It is the primary option as a marker of instrumental relations – with alternative expressions like permezz ta’ ‘through, by means of’ being marginal at best.280 The functional domain of bi covers the full range of instrumentality, as attested by the examples in (C79). (C79) Maltese – Instrumental (C79.1) Tool (Qrempuƛu 55) u minnufih Īewƥ friefet il-lejl qiegƫd-u and immediately two butterfly DEF-night set-3PL l-gƫatu fuq il-basket u qafl-u-h DEF-lid on DEF-basket and close-3PL-3SG.M b-biƛƛ-tejn spag Y-piece-DU string ‘And immediately two bats put the lid on the basket and locked it with two pieces of string.’ (C79.2) Secondary instrument (Delitti 209) Andrea Caruana ta’ 27 sena aƛƛidentalment Andrea Caruana of 27 year accidentally qatel lil omm-u b-revolver. Y-revolver kill to mother-3SG.M ‘27years-old Andrea Caruana accidentally killed his mother with a gun.’ (C79.3) Unusual instrument (Delitti 424) l-arma li bi-ha n-qatel Micallef ma DEF-weapon REL Y-3SG.F PASS-kill Micallef NEG1 kien-et xejn gƫajr pipa NEG2 except pipe be.PAST-3SG.F ‘The weapon with which Micallef was killed was nothing but a pipe.’

286 Case studies (C79.4) Body part instrument (Ƥrajjiet 65) raj-t-u stess b-gƫajn-ej-ja Y-eye-PL-1SG see-1SG-3SG.M self ‘I saw him with my very eyes.’ (C79.5) Animate instrument (Xindi 7) wasl-u s-Saqqajja li sab-u-ha DEF-Saqqajja REL find-3PL-3SG.F arrive-3PL fgat-a bi-n-nies choke:PTCPL-F Y-DEF-people ‘The arrived in Saqqajja which they found to be choking with people.’ It does not matter whether the objects used as instruments in the situations described by the above examples are typical tools, as the pieces of string with which a lid is fastened to a basket in (C79.1), or secondary instruments, like the revolver in (C79.2) from which the lethal bullet emerged that killed the victim. Unusual instruments, like the pipe in (C79.3) used to murder somebody, are likewise marked by bi. The same holds for corporal instruments, like the eyes with which a certain individual was seen in (C79.4), and animate instruments like the people that the streets of Saqqajja were blocked with in (C79.5). The use of bi extends to the description of situations which involve means of transportation as the instrument. In (C80.1), the agent – encoded by the verbal morphology only – drives the character Annetta back home in a coach. In contradistinction to English and many other languages, Maltese does not treat means of transportation as containers or locations. In (C80.2), the regular mail service is ruled out as the channel through which a fake invitation by letter had been delivered to the mailbox of the addressee. (C80) Maltese (C80.1) Means of transportation (Serafin 24) kien j-erƥa’ j-ieƫu lil Annetta lura be.PAST 3SG.M-return 3SG.M-take to Annetta back lejn Dar iƛ-ƚriev bi-l-karozzella towards Dar iƛ-ƚriev Y-DEF-coach ‘He returned to take Annetta back to Dar iƛ-ƚriev in the coach.’ (C80.2) Means of conveyance (L’Imsella 71) kien j-idher ƛar li ma be.PAST 3SG.M-seem clear REL NEG1

A-Type: Maltese 287

kin-et-x t-wassl-et be.PAST-3SG.F-NEG2 PASS-arrive.CAUS-3SG.F bi-l-posta iĪda b-l-id-ejn Y-DEF-hand-PL Y-DEF-mail but ‘It seemed clear that it (= the dubious letter) had not arrived by mail but was delivered personally.’ The instrument does not have to be a concrete object to trigger the use of bi. This preposition also combines freely with less concrete objects and even with abstract notions, cf. (C81). The good words used by the cleric in (C81.1) to convince the members of a split up family to be on good terms are of course still somewhat concrete, in the sense that he must have produced sound to utter them. Marking the use of a language for any act of communication – speaking, reading, or writing – also falls within the functional domain of bi as can be seen from (C81.2). In (C81.3), the extraterrestrial leads a spaceship by his mental force alone to a safe landing ground. (C81) Maltese – Abstract instrument (C81.1) Words (Ƥrajjiet 121) b-il-kelma t-tajba ressaq-t-hom Y-DEF-word:F DEF-good:F approach:CAUS-1SG-3PL ƫafna mil-l-ƥdid lejn xulxin very from-DEF-new towards RECIP ‘Using good words, I very much made them approach each other again.’ (C81.2) Language (L’Imsella 17) David fehem li kien xi be.PAST some David understand REL rumanz bi-t-Taljan novel Y-DEF-Italian ‘David understood that it (= the book) was a novel in Italian.’ (C81.3) Telepathy (Ulied 95) l-aljen kien qiegƫed i-mexxi-h DEF-alien be.PAST PROG 3SG.M-lead-3SG.M bi-l-qawwa ta’ moƫƫ-u of mind-3SG.M Y-DEF-force ‘The alien was steering it with the force of his mind.’ Beyond the realm of instruments proper, bi is used in combination with nouns designating the material of which an entity is built or the parts of

288 Case studies which it is composed.281 Two mice – the protagonists of the story from which (C82.1) is drawn – are surprised to notice that the buildings in the city they just entered are partly made of cheese (which of course is a relatively unusual substance to be used for construction purposes in the real world). In (C82.2), the structure of a certain pillar is discussed and the object is found to be made up of a number of special building blocks instead of being one solid piece of stone. (C82) Maltese (C82.1) Material (Qrempuƛu 62) bd-ew j-osserva-w aƫjar il-bini 3-observe-PL better DEF-building begin-3PL kollu ta’ madwar il-pjazza intebƫ-u DEF-square become.aware-3PL all of around li biƛƛa minn-u kien mibni REL piece from-3SG.M build:PTCPL be.PAST bi-l-ƥobon ta-l-bĪar of-DEF-pepper Y-DEF-cheese ‘They began to scrutinise better all the buildings around the square, they noticed that part of them were built of peppered cheese.’ (C82.2) Material/component parts (L’Imsella 103) da-l-plier Īgur li m-hu-x biƛƛa waƫda REL NEG1-he-NEG2 piece:F one:F this-DEF-pillar sure gƫax mibni bi-l-knaten because build:PTCPL.PASS Y-DEF-square.block:PL ‘This pillar surely isn’t one piece because it is built of quadrangle building stones.’ As a kind of extension of the use of bi in combination with component parts of a whole, the ascription of corporal or mental properties and physical states for instance to human beings also relies on the use of the same preposition. In this function, bi can be used both in attributive constructions, cf. (C83), and in predication (cf. [C86] under POSSESSION). In (C83.1), the trait of being prudent is ascribed to a boy named Joe. The property is represented by a noun gƫaqal ‘prudence’, which is connected to the head noun tifel ‘boy’ by the preposition bi which functions as a kind of attribute marker. The same head noun occurs in (C83.2). However, this time the property ascribed to the boy is not an abstract quality of character but a corporal trait: Being a redhead. The same structure as in the previous example is employed here: The two nouns/NPs are linked by the preposition bi.

A-Type: Maltese 289

(C83) Maltese – Attribution (C83.1) Mental properties/character trait (Serafin 122) lil Joe mil-l-ewwel immarkaj-t-u li hu mark-1SG-3SG.M REL he to Joe from-DEF-first tifel bi-l-gƫaqal boy Y-DEF-prudence ‘Right from the start, I marked Joe as a prudent boy.’ (C83.2) Corporal properties (ornative) (L’Imsella 81) fejn l-istatwa kien hemm dak it-tifel that DEF-boy where DEF-statue be.PAST there b-xagƫr-u aƫmar li suppost kien red REL supposed be.PAST Y-hair-3SG.M maqful fi-r-remissa in-DEF-shed lock.PTCPL.PASS ‘Near the statue, there was that red-haired boy who was supposed to be locked in the shed.’ The ornative function exemplified in (C83.2) leads us directly to another function called confective. In (C84), two consecutive situations are described: First, the advent of Napoleon and his fleet to Malta in (C84.1) and then landfall of Napoleon and his army on Malta in (C84.2). It is important to understand that neither (C84.1) nor (C84.2) has a proper instrumental reading. What is intended with (C84.1) is a statement about the strength of the naval forces Napoleon brought along (and not about the fact that he travelled by ship). Likewise, (C84.2) highlights the manpower of the army that Napoleon disembarked. Nowhere is it said that these 40,000 men were used as an (animate) instrument to achieve a military goal. The sentence only states that Napoleon brought a certain number of soldiers to Malta who acted according to the orders they had received and not on their own initiative (cf. the discussion in connection with [C91–[C92]] below). Sentence (C85) describes a situation less difficult to analyse: a man takes a pail and carries (= walks with) it. (C84) Maltese (C84.1) Confective – Inanimate (Xindi 11) meta ti-ftakar li Napuljun ƥie when 2SG-remember:REFL REL Napoleon come Malta bi 13-il ƥifen, 90 bastiment Malta Y 13 man-o’-war 90 ship

hawn here Īgƫir small

290 Case studies ta-l-gwerra u 400 bastiment ta-t-trasport, and 400 ship of-DEF-transport of-DEF-war minbarra l-ƫafna frejgat-i, DEF-many frigate-PL besides j-aqb-ek biĪa’ kbir fear great 3SG.M-grasp-2SG ‘When you remember that Napoleon came hither to Malta with 13 men-o’-war, 90 small war ships and 400 ships for transport not to mention the many frigates, you will be in great fear.’ (C84.2) Confective – Animate (Xindi 11) meta Napuljun niĪel Malta b-40,000 ruƫ, when Napoleon descend Malta Y-40,000 soul korvett-i ninima kien-x hemm ƫlief be.PAST-NEG2 there except corvette-PL NEG1 ftit xalupp-i, u xebec ta’ l-Ordni DEF-Order few boat-PL and galley of ta-l-Kavallier-i ji-stenne-w-h 3-wait-PL-3SG.M of-DEF-Knight-PL ‘When Napoleon disembarked on Malta with 40,000 men, there were nothing but a few boats, corvettes and the galley of the Order of the Knights waiting for him.’ (C85) Maltese (C85.1) Confective – Inanimate (Ƥrajjiet 120) minn waraj-ha nqala’ bniedem li ƫad-ilha appear man REL take-3SG.F from behind-3SG.F s-satal minn id-ej-ha u mexa bi-h hu DEF-pail from hand-PL-3SG.F and walk Y-3SG.M he ‘From behind her appeared a man who took the pail out of her hands and carried it himself.’ Confective and ornative are relations close to possession because they involve the sometimes only temporary association of an entity with property that may or may not be under the control of that entity. There are constructions of predicative possession in which bi is employed as a marker of the predicate noun. The resulting PP headed by bi is usually combined with the copula (which most often is zero in the present tense).282 In the majority of cases, the construction COPULA + [bi NP]PP is used for non-prototypical forms of possession.283 In (C86.1), the possessee is an abstract entity – ƫsieb ‘thoughts’ – where a translation with the English verbs “to possess”

A-Type: Maltese 291

or “to own” is absolutely ruled out. The same applies to (C86.2), where the possessee is a temporary body part – baffi ‘beard’. (C86) Maltese (C86.1) Abstract possession (Ulied 11) kon-t bi ƫsieb-ni n-ieƫu lil 1SG-take to be.PAST-1SG Y thought-1SG Apollo gƫal dawra Apollo for round ‘I intended to take Apollo for a walk.’ (C86.2) Body part possession (Qrempuƛu 71) il-Baruni Garmiƥo kien bi-l-baff-i DEF-Baron Garmiƥo be.PAST Y-DEF-beard-PL ‘Baron Garmiƥo had a beard.’ In the languages of the world, instrumentals very often acquire the function of expressing cause. This pattern is also valid for Maltese.284 In (C87), bi is used to link a NP as a causal adverbial to a clause. In (C87.1), the pain forces Samuel to bite the handkerchief, whereas it is the heat and the wine which make the heads of people become hotter in (C87.2). In all of the cases, the NP indicating the cause is preceded by the preposition bi. (C87) Maltese – Cause (C87.1) (Ulied 42) Samwel ƫass il-qafla t-aqta’ f-laƫm-u DEF-string:F 3SG.F-cut in-flesh-3SG.M Samuel feel u b-l-uƥigƫ gidem il-maktur bite DEF-handkerchief and Y-DEF-pain issikka-t ma’ ƫalq-u mouth-3SG.M tighten-PTCPL X ‘Samuel felt the string cut into his flesh and because of the pain he bit into the handkerchief tightened over his mouth.’ (C87.2) (Ƥrajjiet 108) l-irjus bde-w ji-sƫn-u DEF-head:PL begin-3PL 3-become.hot-PL bi-s-sƫana u b-l-inbid and Y-DEF-wine Y-DEF-heat ‘The heads started to become hot because of the heat and the wine.’ As a matter of fact, bi is used for a wide variety of adverbials. Among these we find the frequently used positionals for standing and sitting, cf. (C88).

292 Case studies (C88) Maltese – Positionals (C88.1) Standing (Ulied 13) Samwel qam bi-l-wieqfa Samuel rise Y-DEF-standing ‘Samuel stood up.’ (C88.2) Sitting (Ulied 11) in-nannu qagƫad bi-l-qiegƫda ma’ ƥenb Samwel DEF-grandpa sit Y-DEF-sitting X side Samuel b-magg kafè j-agƫli quddiem-u in.front-3SG.M Y-mug coffee 3SG.M-boil ‘Grandpa sat down beside Samwel with a mug of hot coffee in front of him.’ Other adverbial usages abound. Sentence (C89) alone already contains three examples of modal expressions which come in the shape of a PP headed by bi. Bil-mod ‘slowly’ modifies the verb fetaƫ ‘he opened’, bis-salt ‘suddenly’ modifies the verb jaƫtfu ‘he grabs him’, and finally bil-ƫerra ‘brutally’ modifies the verb jigbdu ‘he drags him’. This list of cases of modal bi could easily be continued. Bi is especially productive when it comes to creating new adverbial expressions. For the temporal use of bi, cf. (C99) under temporal usages of ma’. (C89)

Maltese Modal bi (Ulied 41) Samwel fetaƫ it-tieqa bi-l-mod u DEF-window Y-DEF-manner and Samuel open kien kwaĪi nofs-u ƥewwa meta ƫass feel be:PAST almost half-3SG.M inside when xi ƫadd j-aƫtf-u bi-s-salt Y-DEF-jump somebody 3SG.M-grasp-3SG.M mi-ƛ-ƛinta ta-l-qalz-iet u of-DEF-trouser-PL and from-DEF-belt j-iƥbd-u l isfel bi-l-ƫerra to down Y-DEF-roughness 3SG.M-pull-3SG.M ‘Samuel opened the window slowly and he was almost half inside when he felt somebody suddenly grasping him by the belt of his trousers and dragging him down brutally.’

In addition to the semantically still somewhat transparent functions of bi discussed so far, there are also cases of valency bound bi. In these cases, the verb governs what might be called a prepositional object. The depend-

A-Type: Maltese 293

ent PP is often headed by bi. As far as we can judge, bi seems to be especially frequent with verbs of saying, perception and cognition. So you usually hear of, tell about, dream of, become aware of, or notice something which is introduced by bi as in the sentences under (C90). (C90) Maltese – Valency-bound bi (C90.1) To hear (Xindi 76) l-aƫƫar kaĪ li smaj-t bi-h DEF-last case REL hear-1SG Y-3SG.M kien dwar tliet Malt-in u FranƛiĪ be.PAST about three Maltese-PL and Frenchman ‘The last case I heard of was about three Maltese and a Frenchman.’ (C90.2) To notice (Ulied 41) kien ix-xufier li induna bi-t-tfal Y-DEF-child:PL be:PAST DEF-driver REL notice ‘It was the driver who noticed the children.’ (C90.3) To become aware (Ulied 108) il-kap xjenzat intebaƫ bi-l-biĪa f-gƫajn-ej-ha DEF-chief scientist become.aware Y-DEF-fear in-eye-PL-3SG.F ‘The chief scientist noticed the fear in her eyes.’ (C90.4) To know (Ƥrajjiet 13) bi knisja da-Ī-Īagƫzugƫ ma ried-x j-af Y church this-DEF-youngster NEG1 want-NEG2 3SG.M-know ‘This youngster did not want to know about church.’ (C90.5) To dream (Xindi 33) int dejjem t-oƫlom bi-l-ftajjar Y-DEF-tart:PL you always 2SG-dream ‘You always dream of tarts.’ (C90.6) To like (Serafin 8) is-surmast ƫa pjaƛir bi-h wisq DEF-head_master take pleasure Y-3SG.M very.much ‘The head-master was very much pleased with him.’ As the above survey suggests, bi really covers a wide range of functions, of which instrumentality in the strict sense of the term is only a relatively small segment of the entire paradigm of functions of the preposition. Notwithstanding this large functional domain, it does not overlap with that described below for ma’. Those functions of bi which seem to suggest that there is nevertheless some overlap between the domains are discussed in due course in the following sections.

294 Case studies 12.1.3. The functional domain of ma’ For convenience, let us assume that the prototypical instance in which ma’ comes to be used is the description of a situation of accompaniment. Normally, this situation involves two animate participates – preferably both human, cf. (C91). Below we will have to reconsider the issue of the prototype again. (C91)

Maltese Accompaniment (human participants) (L’Imsella 49) fi-r-razzett David gƫalef il-kanal-i DEF-canary-PL in-DEF-farm_house David feed l-ƫomor u mbagƫad mar ma’ DEF-red:PL and then go X Gerry fejn it-tiƥieƥ Gerry where DEF-hen:PL ‘In the farm house, David fed the red canaries and then he went with Gerry to the place where the chicken were kept.’

In (C91), both the accompanee and the companion rank high on the animacy hierarchy, as they are represented by proper nouns (= first names of human beings). The use of ma’ in this context indicates that the two participants are more or less on an equal footing in the situation described. Both David and Gerry act as full blown agents, whereas in (C84.2), the situation of accompaniment is depicted in a different way: Notwithstanding the fact that both Napoleon and his soldiers are humans with the same degree of animacy, the French general is the one who acts out of his own volition. His soldiers’ participation in the act of landing on the island of Malta depends on Napoleon’s decisions and orders. This reduction of agentivity on the part of the soldiers requires bi to be employed.285 If ma’ had been used instead, the reading would have been that Napoleon personally went together with his soldiers to the shore. The fact that both prepositions can take a complement NP provided with the feature [+human] disproves the handy rule given by some of the practical languages courses286, according to which the main criterion that distinguishes ma’ from bi is the restriction on animacy of the NP governed by bi. Things are not as easy as this: Ma’ may also be used in contexts where only the accompanee carries the feature [+human], cf. (C92).

A-Type: Maltese 295

(C92) Maltese (C92.1) Human companion (L’Imsella 78) u jien dejjem wieƫed ƥib-t miegƫ-i hawn here and I always one bring-1SG X-1SG ‘And I always took one (of the twin brothers) along with me here.’ (C92.2) Animate companion (L’Imsella 90) lil Shark ƫad-u-h magƫ-hom ukoll also to Shark take-3PL-3SG.M X-3PL ‘They also took Shark along.’ (C92.3) Inanimate companion (L’Imsella 91) dakinhar ƫad-u l-ktieb magƫ-hom DEF-book X-3PL that.day take-3PL moƫbi fi-l-basket in-DEF-basket hide:PTCPL ‘On that day, they took the book with them hidden in the basket.’ Two of the sentences under (C92) involve the verb ƫa ‘take’ whose reading is one of transportation – just as the reading for the main verb ƥieb ‘bring’ in (C92.1): Something – the book, some animate being, the dog named Shark or one of the twin brothers – is taken along by the agents on their way to some destination. The animacy of the transported object is irrelevant for the choice of the appropriate preposition that marks the pronominal agent as the accompanee. The situation described is one that fits in with the confective relation discussed above (cf. [C84] and [C90]). One of the participants is more agentlike than the other. Nevertheless, ma’ is used as the appropriate preposition because its complement is the accompanee who is prototypically [+human] and not the companion whose animacy may be down to zero. Ma’ cannot be replaced by bi in any of the contexts discussed so far. If we rephrase the sentences in (C92)287 using a motion verb whose subject is coreferential with the complement of the preposition ma’ in (C92.1–3), then ma’ could be used as head of the PP whose dependent NP is the companion, if the companion is [+animate] as well (= C92.1–2). However, bi can replace ma’ in these contexts if the referent of the dependent NP is not acting on his own initiative, i.e., in cases where the companion’s volition is low. If the companion is [¬animate] (= C92.3), bi is the only choice. Thus, in a way, ma’ seems to be sensitive not only to animacy (and to a lesser degree than expected) but also to the egalitarian relation among the participants, in a manner of speaking. On closer inspection, this sensitivity to animacy is less strict than expected. Ma’ appears in a wide range of contexts where animacy is of no

296 Case studies importance. Fits of all, ma’ is frequently employed as a spatial preposition. The situations described by the sentences in (C93) include perlative and prolative relations. (C93) Maltese – Spatial ma’ (C93.1) (Xindi 10) Vinƛenzo u sƫab-u qabd-u take-3PL Vincenzo and comrade:PL-3SG.M j-dur-u ma-t-toroq X-street:PL 3-go.about-PL ‘Vincenzo and his comrades began to walk about the streets.’ (C93.2) (Qrempuƛu 138) biex n-asl-u gƫa-n-naƫa l-oƫra ta-l-bajja in.order.to 1-arrive-PL for-DEF-side DEF-other of-DEF-bay se j-koll-na n-imx-u ma-d-dawra tagƫ-ha koll-ha FUT 3-have-1PL 1-walk-PL X-DEF-turn of-3SG.F all-3SG.F ‘To reach the other side of the bay, we will have to walk fulllength along its turn.’ (C93.3) (Ulied 72) mesah xofft-ej-h mi-r-ragƫwa ta-l-birra of-DEF-beer wipe lip-PL-3SG.M from-DEF-foam ma-l-komma ta-l-qmis of-DEF-shirt X-DEF-sleeve ‘He wiped the foam off his lips with the sleeve of his shirt (lit. he wiped his lips clean of the foam along the sleeve of the shirt).’ (C93.4) (L’Imsella 104) iĪda t-tliet ƫut-u xxabbt-u ma-l-plier DEF-three brother-3SG.M climb-3PL X-DEF-pillar but ‘But his three brothers climbed up the pillar.’ In all of the cases under (C93), the spatial relation involves an extended object along which a movement of some kind takes place. Streets (C93.1), stretches of beach (C93.2), the sleeve of a shirt (C93.3), and the pillar (C93.4) have two properties in common, namely their extension in space and their lack of animacy. In spite of the fact that the nouns are all inanimate, they are nevertheless the complements of ma’. Likewise, spatial ma’ does not obligatorily require the trajector to be animate, cf. (C93.3)288 and (C94).

A-Type: Maltese 297

(C94)

Maltese (L’Imsella 83) wara li rabat-lu wkoll il-maktur REL tie-3SG.M too DEF-handkerchief after ma’ ƫalq-u ƫareƥ barra ma-t-tfal X-DEF-child:PL X mouth-3SG.M go.out outside ‘After he had fastened the handkerchief over his mouth he went outside with the children.’

In (C94), the mouth (= ground) over which the handkerchief (= trajector) is fastened is viewed as having a certain spatial extension corresponding to other stretched out objects along which something can be moved or placed. The connection of the spatial usage of ma’ with the more comitativelike ones is evident in cases like the following. In (C95.1), the government is the institution for which the retired person had been working. The government at the same time provided the place where the work of the secretary had to be carried out. Since the concept of a government is rather multi-faceted, it may also be viewed as including the actual members of a government – animate beings. In this way, the reading of ma’ oscillates between a clearly spatial one (= English at) and a comitative one (= English together with). In (C95.2), the ground (the machinery used on a construction site) relative to which the animate trajector (the workers) is located, lacks any element of animacy. Irrespective of this absence of otherwise crucial properties (cf. above), ma’ is used nevertheless. The reading of the preposition is somewhere in between a spatial one and a comitative one. In (C95.3), a situation is described in which the unexpected find of some ancient letters and a holy cross is mentioned. The letters are said to be found mal-kurƛifiss ‘with the cross’ – meaning they were also there in the same chest that the children had dug out somewhere. Again, the reading of ma’ oscillates between a spatial one and one that resembles a comitative. (C95) Maltese – Spatial ma’ (C95.1) Animate ground (L’Imsella 11) mindu spiƛƛa minn skrivan ma-l-Gvern since end from secretary X-DEF-government kien iktar i-sib Īmien j-aqra more 3SG.M-find time 3SG.M-read be.PAST ‘Since he quit work as a secretary of the government, he found more time to read.’

298 Case studies (C95.2) Animate trajector and inanimate ground (Kappa 53) il-post fejn kien hemm id-dwal DEF-place where be.PAST there DEF-light:PL bi-l-lejl kien mimli b-makkinarju full Y-machinery Y-DEF-night be.PAST ta-t-tqattigƫ u ta-l-bini iĪda and of-DEF-building but of-DEF-digging ma kien-x hemm ƫaddiem-a magƫ-hom NEG1 be.PAST-NEG2 there worker-PL Y-3PL ‘The place where there had been the lights at night was full up with machinery for digging and building, but there were no workers around (lit. with them [= the machines]).’ (C95.3) Inanimate trajector and inanimate ground (Avventura 83) it-tfal mi-n-naƫa tagƫ-hom iktar kien DEF-child:PL from-DEF-side of-3PL more be.PAST moƫƫ-hom f-silta li sab-u miktuba find-3PL write:PTCPL mind-3PL in-paragraph REL f-waƫda minn dawk l-ittr-i from those DEF-letter-PL in-one:F li kien hemm ma-l-kurƛifiss REL be.PAST there X-DEF-cross ‘The children for their part were more concerned with a paragraph they found written in one of those letters that were there together with the cross.’ As a matter of fact, ma’ marks the co-presence of two entities in one and the same segment of space – be it a concrete location or something more abstract like time, etc. However, there are restrictions on the entities involved in a situation described by ma’. Consider (C96) for that matter. (C96)

Maltese Co-presence marked by ma’ (Ulied 10) niĪĪel gidma ƫobĪ ma’ nofs magg ƫalib bread X half mug milk descend.CAUS bite ‘He swallowed a slice of bread and half a mug of milk.’

The two entities combined by ma’ are both bona fide cases of inanimate nouns. The bread and the (half mug of) milk are co-present in a situation where the agent eats his breakfast. Both nutritial components of the breakfast are treated on a par with each other. Neither of the two is dependent on the other; they are both autonomous parts of the breakfast event. However,

A-Type: Maltese 299

if one of the parts of such an event is considered to be subordinate to a higher level part of the same event, the use of ma’ is blocked and bi has to be used instead, cf. (C97). (C97)

Maltese Co-presence marked by bi (Serafin 27) ƫa ƫsieb i-tella’ tazza tè bi-l-lumi take thought 3SG.M-rise.CAUS glass tea Y-DEF-lemon ‘He took care to bring a glass of tea with lemon upstairs.’

In (C97), the lemon is an ingredient to the tea. It is viewed as being a dependent part of the tea that could not be separated from the liquid. If ma’ had been used, native speakers probably would have visualised the situation as involving some lemon fruit – not yet sliced – sitting on the tray somewhere beside a glass of tea. The opposition of ma’ and bi continues in other contexts as well. (C98.1–2) describe situations where different degrees of mixture of entities apply. In (C98.1), the mixture is complete, i.e., the substances used to produce the paste for painting the icon are mixed in such a way that they yield a new material and none of the original ingredients remains unaffected by the process. In contradistinction, (C98.2) describes the slow process of one’s hair becoming grey. This process has not advanced much yet, and thus the grey is only an addition to the original colour of the character’s hair. If an asymmetry of this kind applies, bi is the only appropriate choice.289 (C98) Maltese (C98.1) Mixture with ma’ (Avventura 89) ƫaƥa interessanti hi fi-x-xogƫol ta’ dawn of these thing interesting she in-DEF-work l-ikon-i kien-u j-uĪa-w l-isfar DEF-icon-PL be:PAST-3PL 3-use-PL DEF-yellow ta-l-bajd imƫallat ma-t-trab X-DEF-dust of-DEF-egg mix.PTCPL ta-l-kulur u ilma and water of-DEF-colour ‘An interesting fact is that for painting these icons, they used egg yoke mixed with coloured dust and water.’ (C98.2) Mixture with bi (L’Imsella 6) xagƫr-u kien beda be.PAST begin hair-3SG.M

300 Case studies ji-ĪĪewwaq b-l-abjad Y-DEF-white 3SG.M-mottle:REFL ‘His hair had begun to become mottled with white.’ In addition, both prepositions may be used to mark a temporal adverbial. In (C99), both ma’ and bi occur as head of a PP whose dependent is the noun lejl ‘night’. Notwithstanding this structural coincidence, the resulting adverbial constructions are not synonymous. (C99) Maltese – Temporal adverbials (C99.1) ma’ (L’Imsella 12) in-Nannu Dumink mar j-agƫlef gƫax DEF-grandpa Dominic go 3SG.M-feed because il-fniek i-kun-u j-rid-u xi DEF-rabbit:PL 3-be.FUT-PL 3-want-PL some j-gerrm-u ma-l-lejl X-DEF-night 3-nibble-PL ‘Grandpa Dominic went to feed (the animals) because the rabbits would want something to nibble on during the night.’ (C99.2) bi (Ulied 13) il-bieraƫ bi-l-lejl raj-t xi dwal l hawn DEF-yesterday Y-DEF-night see-1SG some light to here ‘Yesterday at night, I saw some light around here.’ When the PP is headed by ma’, the reading is one of duration – a reading that is in line with the spatial use of ma’ as a prolative marker. The night is viewed as a time span along/during which a certain action (the rabbits’ gnawing on the feed) extends over much of the nighttime period. However, the use of bi as head of the PP indicates that the event described does not cover the whole night but happened at some point of the night. Temporal ma’ is quite frequent. However, the durative reading as in (C99.1) above is not always the preferred one. In (C100), we list a number of typical cases of ma’ employed in temporal functions. (C100) Maltese – Temporal ma’ (C100.1) (Ulied 45) ma-s-sebƫ l-annimal-i fi-l-maqjel bd-ew in-DEF-barn begin-3PL X-DEF-dawn DEF-animal-PL ji-t-ƫarrk-u mi-r-raqda tagƫ-hom from-DEF-sleep of-3PL 3-REFL-move-PL ‘At dawn, the animals in the barn began to get up from their sleep.’

A-Type: Maltese 301

(C100.2) (Qrempuƛu 86) ma’ nĪul ix-xemx Qrempuƛu u Qrempuƛellu DEF-sun Qrempuƛu and Qrempuƛellu X descending telq-u lejn il-palazz ta-l-Gran-Ƥurdien of-DEF-Great-Rat leave-3PL towards DEF-palace ‘At sundown, Qrempuƛu and Qrempuƛellu went to the palace of the Great Rat.’ (C100.3) (Ƥrajjiet 101) xhur wara li siefer mart-u Karmena go.abroad wife-3SG.M Karmena month after REL miet-et ma-t-twelid ta-t-tarbija of-DEF-baby die-3SG.F X-DEF-birth ‘A month after he had gone abroad his wife Karmena died while giving birth to the baby.’ (C100.4) (Ƥrajjiet 107) ma’ nofsinhar marr-et qis-ha mara consider-3SG.F woman X noon go-3SG.F ‘By noon, she went like a grown-up woman.’ (C100.5) (L’Imsella 51) ma’ dawn il-kelm-iet waqa’ skiet kbir fall silence big X these DEF-word-PL ‘With these words, everyone fell silent.’ In (C100.1–3), the temporal adverbial involving ma’ can be interpreted as referring to a time span: Sunrise and sundown are processes which take some time until the sun has actually risen above or sunk beneath the horizon. Likewise, giving birth to a child is not a punctual event, but the labours may be rather drawn out for hours on end. In this way, the use of ma’ in these contexts corresponds neatly to mal-lejl ‘during the night’ in (C99.1). In (C100.4–5) however the time factor differs clearly from the previous examples. Noon time and the uttering of words do not represent time spans of a duration compatible with the one required for the processes described in (C99.1) and (C100.1–3). In fact, for (C99.4–5), it is more appropriate to speak of punctual events. What motivates the use of ma’ here is the fact that two events happen at a certain point in time: The temporal adverbial indicates the one event, namely, the time or action close to which the other event indicated by the main predication occurs. In a rather daring manner of speaking, the events encoded by adverbial and main predication accompany each other; they are co-present.

302 Case studies However, it is not simple co-presence of entities that triggers the use of ma’ – the two co-present entities have to meet one crucial criterion, namely that of relative independence of each other. Consider sentence (C101), from which ma’ is absent, and compare it to (C96). (C101)

Maltese – Coordination (L’Imsella 15) meta reƥgƫ-u lura mil-l-quddies ƫad-u kikkra when return-3PL back from-DEF-Mass take-3PL cup kafè bi-l-ƫalib u kisra ƫobĪ mixwija coffee Y-DEF-milk and slice.F bread roast.PTCPL.F ‘When they came back from church, they had a cup of coffee with milk and a slice of toasted bread.’

The two sentences describe similar situation where the agent eats and drinks something for breakfast. The solid and liquid substances used in these situations are depicted as autonomous components of the breakfast. This is aptly marked by the choice of the coordinating conjunction u ‘and’ in (C101), which corresponds quite closely to ma’ in (C96). As a matter of fact, in many cases u could substitute for ma’ and vice versa, as long as the two participants linked to one another by one of the two morphemes under scrutiny are considered separate and largely independent entities. If the participants fail to meet this criterion, their relationship is one of dependence. Therefore, bi is used to connect the additional ingredient to coffee and tea – i.e. milk and lemon – in (C101) and (C97), respectively. The additional ingredients are almost in a part-whole relation to the coffee or tea. These additions to the basic substance are no longer independent because, once they are mixed, the two liquids/substances cannot be separated (without chemical wizardry, that is). Since the coffee and the tea are considered the basic substance which something else is added to and the basic substance has more weight, physically speaking, the additional ingredients are subordinated to the basic substances. The high degree of autonomy of the two participants of a situation linked by ma’ can be emphasised by adding the adverb flimkien, which translates into English as ‘together’, cf. (C102). (C102) Maltese – Together (C102.1) Animate (Avventura 59) dik il-lejla l-avukat kien ƫareƥ DEF-evening DEF-lawyer be.PAST go.out that mi-d-dar flimkien ma-l-mara tiegƫu X-DEF-wife of:3SG.M from-DEF-house together ‘That evening, the lawyer had left the house together with his wife.’

A-Type: Maltese 303

(C102.2) Inanimate (L’Imsella 39) David reƥa’ kebbeb il-kart-i romblu DEF-paper-PL roll David return roll:CAUS qegƫid-hom mil-l-ƥdid fi-l-qasba flimkien from-DEF-new in-DEF-rod together put-3PL ma’ l-imsella DEF-bodkin X ‘David rolled the papers again and put them back into the rod together with the bodkin.’ As with the simple preposition ma’, the adverbially reinforced PP does not obey any animacy-based restrictions. In (C102.1), the two participants that join in the action described (= leaving the house) are both high on the animacy scale, whereas the objects put back in place in (C102.2) are inanimate. The relatively transparent etymology of the adverb flimkien suggests a strong spatial component in the semantics of the construction: flimkien can be segmented into f-l-imkien [in-DEF-place], which is a univerbated PP meaning ‘in the place’ (Aquilina 1987: 350). Thus, the adverb originally stressed the fact that the two participants are in the same place at the same time. Again, copresence and relative independence of the participants are the most important factors. It almost goes without saying that flimkien cannot combine with bi. For brevity’s sake, we only mention in passing some of the other purposes for which ma’ is employed. Sentence (C103) exemplifies ma’ as a marker of an approximate numerical value. (C103)

Maltese Approximately (Xindi 74) f-ƥurnata waƫda ƥie-li telq-u ma’ one:F come-REL leave-3PL X in-day:F l-400ruƫ u anke 500 DEF-400soul and even 500 ‘It happened that, on one single day, about 400 or even 500 people left.’

There are also instances of valency-bound ma’. Typical examples are the ones listed under (C104). Ma’ is obligatory with verbs whose meaning involves contact – be it concrete or abstract – of two entities: One hitting against the other, one having to do with the other, one reaching the other, etc. (C104) Maltese – Valency-bound ma’ (C104.1) To hit against (Ulied 35) Īrara oƫra ƫabt-et hit-3SG.F stone:F other:F

ma-t-tieqa X-DEF-window

304 Case studies ‘Another pebble hit the window.’ (C104.2) To have to do with (L’Imsella 39) u x-kell-ha x-t-aqsam ma-t-teĪor and what-have.PAST-3SG.F what-3SG.F-part X-DEF-treasure dik l-imsella msadda DEF-bodkin:F rust:PTCPL.F that ‘And what had this rusty bodkin to do with the treasure?’ (C104.3) To cope with (Qrempuƛu 7–8)290 imma Pirpirella ma setgƫ-et-x NEG1 can-3SG.F-NEG2 but Pirpirell t-laƫƫaq ma’ dawk X those 3SG.F-reach:INTENS il-mistoqs-ijiet kollha f-daqqa DEF-question-PL all in-stroke ‘But Pipirella couldn’t cope with all these questions in one go.’ (C104.4) To fancy/like (Xindi 38) jien gƫand-i grazzja magƫ-ha grace X-3SG.F I at-1SG ‘I like her.’ In addition, ma’ is also used as second component of a variety of complex prepositions whose range of meaning is largely confined to spatial relations (Borg and Azzopardi-Alexander 1997: 160–164). One good specimen of such combinations is minn ma’ ‘from (+ LATERAL CONTACT)’, cf. (C105). (C105)

Maltese (Ulied 25) Īul minn move_away.IMP from ‘Get away from the car!’

ma-l-karozza X-DEF-car

It is almost needless to say that bi can replace ma’ in none of these cases (valency-bound ma’ and complex prepositions). Notwithstanding the fact that we have had to skip many details, the above discussion yields a relatively clear picture of the division of labour among the two primary translation equivalents of English with in Maltese. In the subsequent section, we systemise our findings.

A-Type: Maltese 305

12.1.4. Summing up the Maltese situation In Table (C106), we survey the correlation of animacy properties of the participants of a situation with the choice of the appropriate preposition. The table simplifies the complex network a bit by skipping a number of details, which can, however, be easily recovered from the discussion above. Grey shading marks the participant whose NP is dependent on the preposition. Black cells indicate that the use of the preposition is ruled out in the case at hand. (C106) Animacy correlation Participants’ animacy Accompanee Companion + + + +

Relator ma’ bi + + +

+ +

¬ ¬

+ +

¬ ¬ ¬ ¬

+ + ¬ ¬

+ + +

+ + + +

Comments only ma’ is admissible ma’ = autonomous participants bi = Companion with reduced volition only ma’ is admissible ma’ = valency-bound bi = instrumental, ornative, etc. ma’ = inanimate locative only bi is admissible ma’ = inanimate locative ma’ = and bi = part-whole relationship

Animacy alone does not fully determine the choice of the appropriate preposition. This results from the occurrence of ma’ and bi in combination with NPs independent of their position on the animacy hierarchy. However, there is another criterion interacting with animacy that helps to give a structure to the above distribution. This criterion is location of the marking: If the Accompanee is the dependent of the preposition, i.e., if the Accompanee is the marked constituent, then ma’ is the right choice – no matter whether the dependent noun is [+animate] or [¬animate]. In addition, ma’ is banned completely only from one single context, namely where the Companion is part of the PP and [+animate] while the Accompanee is [¬animate]. Here, only bi is admissible. In all other cases, ma’ is at least an option, even if none of the participants involved in the situation bear the feature [+animate]. Therefore, criteria other than animacy are crucial, at least in these cases of doubt. Indeed, there are three criteria which interact. Besides animacy, there are two further parameters that have a say in the distribution of the preposi-

306 Case studies tions. These parameters are autonomy and co-presence of participants. Table (C107) reflects the interaction of the three criteria. (C107) Interaction of parameters Animacy + + + ¬ ¬ ¬ + ¬

Autonomy + + ¬ + ¬ + ¬ ¬

Co-presence + ¬ + + + ¬ ¬ ¬

Preposition ma’ ma’ bi ma’ bi ma’ bi bi

The degrees of animacy, autonomy, and co-presence of the participants have to be mapped onto the marking pattern as described for Table (C106), i.e., the pluses and minuses of Table (C107) have to be checked against the grey shaded cells of (C106) because of the prominent role of the marked Accompanee. Owing to this intricate interaction of parameters it is not possible to assign clear-cut boundaries to the functional domains of bi and ma’. Therefore, Table (C108) is to be taken cum grano salis. Grey shading marks those categories for which bi is normally employed, while unmarked cells belong to the realm of ma’. The function baptised “temporal” occurs twice because both prepositions share this subdomain – in a way that excludes free variation, however! (C108)

Distribution of labour among bi and ma’

instrumental tool secondary instrument body part animate means of transportation means of conveyance abstract material

part-whole mental/physical properties ornative confective inanimate confective animate abstract possession temporary possession cause positional

modal temporal temporal co-presence spatial perlative spatial prolative accompaniment inanimate accompaniment animate coordination/comitative

Irrespective of minor details, bi clearly stands out as the preposition with the larger functional domain. It covers those functions which are bona fide instances of prototypical instrumentals. It also goes far beyond the boundaries of such prototypical instrumentals. Ma’ has a much smaller segment of the range of functions because it combines prototypical aspects of a comitative in the strict sense and those of a locative-like spatial category. In con-

B-Type: Latvian 307

tradistinction, spatial features are almost completely absent from bi. Therefore, we may summarise the functional affinities of the categories involved as follows: INS = ORN = CONF = POSS = MOD z COM = LOC. 12.2. B-Type: Latvian 12.2.1. Generalities While the closest relative of Latvian, Lithuanian, is a representative of the C-type, Latvian follows the majority of our European sample languages, i.e., it belongs to the B-type. Superficially, there does not seem to be much to say about a B-type language, as we would expect it to show ideally no variation at all as to the expressions employed for the encoding of the categories which are of interest to us here. However, there are some interesting facts about the functional domain of ar ‘with’291 which show that the homogeneity has leaks, in a manner of speaking. Ar is a preposition which has split case government – a phenomenon which is the rule for Latvian prepositions (in contrast to the postpositions, which remain faithful to the case they govern independent of number [Stolz 1984]).292 In the singular, ar governs the inflectional accusative, whereas, in the plural, ar joins the other prepositions in so far as they all govern the inflectional dative, cf. (C109). Split case government has no functional implications. (C109) Latvian – Case-government (C109.1) Singular (DoƼuleja 249) turp viƼu bija aizvedis take:PTCPL thither her be:PAST.3 PƝteris ar zirg-u X/Y horse-ACC Peter:NOM ‘Peter had taken her there with the horse.’ (C109.2) Plural (DoƼuleja 22–23) jo ne-bija ne-kƗda prieka NEG-some:GEN fun:GEN because NEG-be:PAST.3 dzƯvot zem sveša jumta un vƝl roof:GEN and still live:INF under foreign:GEN ar vis-iem zirg-iem un rat-iem and cart-DAT.PL X/Y all-DAT.PL horse-DAT.PL ‘Because it was no fun to live under someone else’s roof, especially with all the horses and the cart.’

308 Case studies In (C109.1), the complement noun of the preposition is marked for the accusative as it is in the singular. This word form zirgu contrasts with zirgiem in (C109.2), where more than one horse is involved in the description of the situation. The only difference between the two PPs is in the realm of number. Case government by prepositions is sensitive to number in Latvian. Since ar does not enter into any other construction (i.e., it never combines with any other cases than the two mentioned above), it is safe to state that it is the syncretistic primary marker of the Comitative-Instrumental. This observation is not completely uncontroversial, as there are traditionally minded views in Latvian philology according to which there still is a distinct inflectional instrumental case, either independent of the preposition ar or in combination with it. For PauliƼš, Rozenbergs, and VilƗns (1978: 57), there is a separate paradigmatic slot for an instrumental case in Latvian.293 In contradistinction to all other cases of this paradigm, this putative instrumental has a complex expression made up of the preposition ar and a case suffix which happens to be identical with the accusative in the singular and the dative in the plural. Superficially, this seems to be a fully justified solution, as neither the inflectional accusative nor the inflectional dative alone do the job of the Instrumental. However, there are several problems with the solution favoured by traditional grammar. The most serious one is the potential multiplication of cases, because there is no apparent reason to restrict the arguments of the Latvian grammarians to ar, as all other prepositions display similar number-based splits (PauliƼš, Rozenbergs, and VilƗns 1978: 127–132), cf. (C110). Wherever two different cases governed by the same preposition are mentioned for the singular, these patterns reflect semantic differences. The alpha (Į) indicates that there are several morphologically complex prepositions with the final constituent -pus (< puse ‘side’) which all require the genitive in the singular. (C110) Number-sensitive case government of prepositions Preposition aiz, apakš, bez, kopš, no, pƝc, pie, pirms, priekš, Į-pus, virs, zem lƯdz ap, ar, caur, gar, par, pƗr, pret, starp pa, uz

Singular genitive

Plural dative

dative accusative dative or accusative

Thus, there would be relatively good reason to postulate another 24 (or even more) cases for Latvian, because the pattern of split case government applies everywhere in the system (Lötzsch 1978: 670). This is of course a

B-Type: Latvian 309

highly unattractive solution. The better way to deal with the situation in Latvian is to restrict the case paradigm to the inflectional cases and reserve their interaction with prepositions for a morphosyntactic interpretation. The motivation for postulating a distinct instrumental in many extant descriptions of Latvian is most probably a historical one. There used to be a distinct inflectional instrumental in older stages of the language, the domain of which is meticulously described in GƗters (1993: 162–183). Synchronically, there are only a few, indeed highly marked, remnants of this erstwhile instrumental, in cases when it is used as a stylistic device to archaise the language. For the personal pronouns of the 1st and 2nd person singular and the reflexive pronoun, there exist two alloforms (PauliƼš, Rozenbergs, and VilƗns 1978: 80), cf. (C111). In the 3rd person, no traces of the inflectional instrumentals remain. In the plural of all persons, the preposition ar consistently governs the dative (ar mums ‘with us’, ar jums ‘with you[PLURAL]’, ar viƼiem ‘with them[MASCULINE]’, ar viƼƗm ‘with them[FEMININE]’). The inflectional instrumental is thus a stylistic option at best. (C111) Pronouns Case accusative “instrumental”

new old

1st person singular mani ar mani ar manim

2nd person singular tevi ar tevi ar tevim

In addition to this residual evidence of an erstwhile functional instrumental, there is also another phenomenon which points to a diachronically different status of the categories under scrutiny. The bare noun or NP (adjective plus head noun) inflected for accusative in the singular and for dative in the plural may be used with typical functions of Comitatives and Instrumentals, cf. (C112). (C112)

Latvian – Bare instrumentals (DoƼuleja 78–79) bet to kƗ jƗ-strƗdƗ ar sav-u but this.ACC how DEB-work X/Y POSS-ACC paraug-u rƗdƯja kalpone Alma meita show:PAST.3 maid Alma girl pattern-ACC pƗri trƯsdesmitiem [augst-u pier-i]ORNATIVE over thirty:DAT.PL high-ACC forehead-ACC [plƗn-iem mat-iem]ORNATIVE [plat-iem plec-iem]ORNATIVE plain-DAT.PL hair-DAT.PL wide-DAT.PL shoulder-DAT.PL [sastrƗdƗt-Ɨm rok-Ɨm]ORNATIVE work.PTCPL-DAT.PL.F hand-DAT.PL.F

310 Case studies ‘However, the maid Alma, a girl over thirty with a high forehead, plain hair, wide shoulders and hands worn down by work, showed by her example, how one has to work.’ In (C112), there is only one instance of a PP headed by ar which is a bona fide case of an (abstract) Instrument, namely the one which serves to express the exemplary behaviour of the maid Alma, which provides an example for the protagonist of the story of how on should learn the work of the farmhands. In addition, there are also four prepositionless NPs – all indexed as ORNATIVES – in the same sentence which exemplify the so-called bare instrumental. As one is in the singular and three are in the plural, we also observe the usual pattern of number-sensitive case-changing, familiar from prepositional government. As a matter of fact, the preposition ar may be added to all of these ORNATIVE constructions, as is often done in colloquial Latvian. There are, however, other cases where the use of ar is blocked, cf. below. As Lötzsch (1978: 669–670) convincingly argues, many of these prepositionless constructions are frozen expressions, while others appear to be adverbialisations rather than regular case forms of the nouns in question.294 These are again relics of an earlier, more important category. Nowadays, there is a mild degree of productivity of the bare noun/NP constructions for a closed class of functions which we highlight in the subsequent sections. 12.2.2. The functional domain of ar For obvious reasons, we restrict our comments on the observed facts to a minimum because many instances are self-evident characteristics of an average B-type language. Thus, most of what follows in the first part of this section has more of an enumeration of contexts in which ar commonly occurs. The preposition ar covers the full range of typical functions within the areas of Comitativity and Instrumentality. The examples in (C113) document the ubiquity of ar in the realm of the instrumental. (C113) Latvian – Instrumental relations (C113.1) Primary instrument/tool (DoƼuleja 126) sarkanmatis ar gar-u kok-u red_head:NOM X/Y long:ACC stick:ACC

nocƝla lift_off:PAST.3

B-Type: Latvian 311

no griestiem peleksvitrotu uzvalku grey_striped:ACC suit:ACC from ceiling:DAT.PL ‘The red-haired man lifted a grey-striped suit from the ceiling with a long stick.’ (C113.2) Secondary instrument/machine (DoƼuleja 26) Alise ne-prata rakstƯt ar mašƯn-u write:INF X/Y machine-ACC Alise NEG-be_able:PAST.3 ‘Alise could not write with a typewriter.’ (C113.3) Imaginary instrument (DoƼuleja 318) mežs ar sav-u noslƝpumainƯb-u mysteriousness-ACC wood:NOM X/Y REFL.POSS-ACC noglƗsta man-i ar dziedinoš-u mier-u X/Y salutary-ACC calmness-ACC caress:3 I-ACC ‘The forest with its mysteriousness caresses me with salutary calmness.’ (C113.4) Money/values (DoƼuleja 56) ja jums zirgu uz ceƺa atƼems way:GEN take_away:FUT if you.DAT.PL horse:ACC on jums bnjs ar ko man you.DAT.PL be:FUT X/Y what.ACC me samaksƗt zaudƝjumus? loss:ACC.PL pay:INF ‘If he will take away the horse from you on the journey, will you have something to compensate me with for my losses?’ (C113.5) Body part instrument (DoƼuleja 30) ar vien-u pirkst-u mƝƧinƗja izspƝlƝt finger-ACC try:PAST:3 PREV:play:INF X/Y one-ACC vienkƗršas dziesmiƼas song:DIM:ACC:PL:F simple:ACC.PL:F ‘She tried to play simple tunes with one finger [on the piano].’ (C113.6) Animate instrument (DoƼuleja 465) jnjs strƗdƗjƗt savƗ laikƗ time:LOC you work:PAST:2PL REFL.POSS:LOC ar zirdziƼ-u X/Y horse:DIM-ACC ‘In your time, you worked with a little horse.’ Means of transportation also fall within the domain of ar. The use of ar is independent of the nature of the means of transportation and it is also ir-

312 Case studies relevant for the choice of the preposition whether the person transported is in control of the process or not, cf. (C114). (C114) Latvian – Means of transportation (C114.1) Animal (DoƼuleja 90) man braukt desmit kilometru ar zirg-u kilometre:GEN.PL X/Y horse-ACC me ride:INF ten ‘I have to go on horseback for ten kilometres.’ (C114.2) Bicycle (DoƼuleja 467) pƗrbraucis ar velosipƝd-u mƗjƗs viƼš X/Y bicycle-ACC home he go_home:PTCPL lƗgƗ ne-paƝdis ielikƗs gultƗ at_all NEG-eat_enough:PTCPL lie_down:PAST.3:REFL bed:LOC ‘On returning home by bicycle he went to bed without having eaten enough.’ (C114.3) Train (DoƼuleja 44) jƗ-brauc-ot lƯdz BruƧes pilsƝtai ar vilcien-u DEB-ride-MR till Bruge:GEN town:DAT X/Y train-ACC ‘She would have to go by train to Bruge.’ Similarly, ar is used to encode material of any kind, (C115). (C115) Latvian – Material (C115.1) Material/tool (DoƼuleja 328) es gribƝju vƝlƗk sasiet ar drƗt-i I want:PAST:1SG later tie_together:INF X/Y wire-ACC ‘I wanted to tie it together with a wire later.’ (C115.2) Clothes (DoƼuleja 260) skapis vel bija pilns ar X/Y wardrobe:NOM still be:PAST.3 full:NOM Alises mant-Ɨm dress-DAT.PL Alise:GEN ‘The wardrobe was still full of Alise’s clothes.’ (C115.3) Liquid (DoƼuleja 14) Alise aizgƗja pie pumpja izmazgƗja pump:GEN wash:PAST.3 Alise go_away:PAST.3 to burciƼu piepildƯja to ar X/Y glass:DIM:ACC fill_up:PAST.3 this:ACC tir-u njden-i water-ACC clean-ACC

B-Type: Latvian 313

‘Alise went back to the pump, washed the little glass vase, and filled it up with clean water.’ We also encounter ar when we leave the realm of instrumentality and enter the areas associated with comitativity. (C116) marks the transition from instrumental to ornative: The dancing shoes are most probably not seen as a means by which the character is able to achieve some goal, but rather they are understood as a kind of inappropriate dresscode of the character in an everyday work situation. (C116) Latvian – Ornative (DoƼuleja 79) kas tad ar balles kurp-Ɲm iet uz knjti who then X/Y ball:GEN shoe-DAT.PL go:INF to stable:ACC ‘Now who would walk to the stable in dancing shoes!’ In (C117), ar is used to encode a certain bodily posture of the protagonist. This usage borders on the one of manner adverbial with ar, cf. (C127) below. (C117)

Latvian – Bodily posture (DoƼuleja 150) un ar augst-u pacelt-u galv-u head-ACC and X/Y high-ACC lift:PTCPL-ACC viƼa gƗja uz durvƯm to door-DAT.PL she go:PAST.3 ‘And with her head held high she went to the door.’

Confectives too are usually encoded in the shape of PPs headed by ar. (C118) is a typical example of this usage. (C118)

Latvian – Confective (DoƼuleja 135) beidzot ienƗca Alise ar aboƺu finally enter:PAST.3 Alise X/Y apple:GEN.PL ‘Finally, Alise came in with a plate of apples.’

šƷƯv-i plate-ACC

Moreover, it is also common to find ar heading a PP which includes a nominal attribute to a head noun. These are often instances of part-whole relations, some of which are difficult to distinguish from ornatives proper, cf. (C119). (C119)

Latvian – Ornative/part-whole (DoƼuleja 59–60) mugurƗ vec-ajai dƗmai bija back:LOC old-DEF:DAT.F lady:DAT.F be:PAST melna kleita ar plat-u balt-u apkakl-i white-ACC collar-ACC black dress X/Y wide-ACC ‘The old lady wore a black dress with a wide collar.’

314 Case studies The situation is less clear when it comes to those contexts which are nearest to the prototype of the Comitative. Superficially, ar seems to be the only option for the description of situations where the accompanee and the companion are both active, cf. (C120): it is reasonable to assume that both participants actually talk to each other in (C120.1) – and ballroom dancing requires both members of the couple to be active (though probably with different degrees of control). (C120) Latvian – Comitativity (C120.1) Reciprocal: talking (DoƼuleja 66) ErnestƯne par Alises aizƼemtƯbu runƗja Ernestine about Alise:GEN being_busy:ACC speak:PAST.3 ar Vintera kundz-i wife-ACC X/Y Vinter:GEN ‘Ernestine talked to/with Mrs Vinter about Alise’s being busy.’ (C120.2) Reciprocal: dancing (DoƼuleja 98) Juris dejoja ar Olg-u dance:PAST:3 X/Y Olga-ACC Juris:NOM ‘Juris danced with Olga.’ However, the use of ar in other cases is associated with a certain orientation of the situation. Consider sentences (C121–122). (C121) Latvian – Accompanee-oriented (DoƼuleja 187) lai nƗk ar mani let come X/Y me:ACC ‘Let him come with me!’ (C122) Latvian – Companion-oriented (DoƼuleja 226) omammu arƯ Ƽemsim lƯdz take:FUT:1PL along granny:ACC also ‘Are we going to take granny along, too?’ In (C121), the situation is depicted from the vantage point of the accompanee, who happens to be identical with the person uttering the sentence. He is the one to determine where one has to go. The use of ar is appropriate wherever accompanee-orientation applies,. However, ar is counted out when the situation is reoriented towards the companion – as is the case in (C122). In this example, the NP representing the companion is in the focus position (sentence-initial). This is clear evidence for companion-orientation. The grandmother is allowed to participate in the outing and is not the initiator of the activities. Companion orientation goes along with the use of the relator lƯdz(i) ‘along (with)’, which is a borderline case between adverb and adposition. LƯdz(i) is typically employed

B-Type: Latvian 315

for the description of situations from the point of view of the participant who has less control over the action than the accompanee, if there is one at all. Thus, lƯdzi is appropriate in (C123) where the young horse trots after its master, who leads the way and is thus exerting control. (C123)

Latvian – Companion-orientation (DoƼuleja 326) un zirga bƝrns nƗca viƼam lƯdzi and horse:GEN child:NOM come:PAST.3 he:DAT along uz malkas šƷnjni stable:ACC to wood:GEN ‘And the young of the horse went along with him to the wooden stable.’

However, neither is the use of lƯdzi restricted to situations which involve a motion verb as predicate nucleus nor does it automatically presuppose reduced active involvement of a participant. In (C124), both the accompanee and the companion rank high on the animacy hierarchy and are both depicted as doing the same kind of work with a high degree of physical involvement. (C124)

Latvian – Companion-orientation (DoƼuleja 365) IlmƗrs lƯdzi tƝv-am rƗvƗs X/Y father-DAT pull:PAST.3 Ilmars:NOM vaiga sviedros sweat:LOC.PL cheek:GEN ‘Ilmars toiled alongside his father (lit. pulled [bathed] in sweat of the cheeks).’

This example invites the reading that it is the father who was in command deciding on his own what was to do while his son obeyed his orders. If ar replaces lƯdzi in the above sentence, not only case government has to change from dative to accusative on the dependent noun, but the meaning conveyed is different too: ar tƝvu ‘with the father’ suggests that Ilmars is the foreman and his father is just lending a helping hand. The distinction of accompanee-orientation vs. companion-orientation is relatively strict in Latvian – and it is grammatically relevant as it requires a different relator. Before we have a look at the adverbial functions of PPs headed by ar, we come back shortly to the issue of coordination. As we already mentioned in Section 4.1, (A15), Latvian is an AND-language with interesting traits of a WITH-language. In (C125), both the preposition ar and the coordinating conjunction un ‘and’ co-occur side by side.

316 Case studies (C125)

Latvian – Coordination (DoƼuleja 102) no tƗlienes ne-varƝja lƗgƗ saskatƯt bet at_all look:INF but from far:GEN NEG-can:PAST:3 likƗs ka tie ir Lekužu ƺaudis seem:PAST.3:REFL hat these be.3 Lekuzis:GEN.PL people [Edgars ar Elz-u] un [Juris ar Olg-u] Edgar:NOM X/Y Elza-ACC and Juris:NOM X/Y Olga-ACC ‘From a distance, one could not see at all but it seemed that these were the people from Lekuzis, Edgar with Elsa and Juris with Olga.’

The conjunction interconnects two complex NPs, both of which contain a PP introduced by ar. As a matter of fact, ar is used here as a means to indicate that the accompanee and the companion form a group together. It is likely that the group is also depicted as a hierarchically organised structure. Two such groups are conjoined by the conjunction. If un replaces ar in (C125), this yields a reading according to which all four participants act independent of each other individually. The substitution of un with ar, on the other hand, appears to be less acceptable because of parsing problems. In (C126), the agreement of the cardinal numeral in number, case and gender with the two female proper nouns is suggestive of the coordinating character of ar: As ar governs the accusative, one of the two nouns should not have any bearing on the outcome of agreement. However, semantics beat morphosyntax in this case: Irrespective of the formally subordinating characteristics of the construction, the two NPs are understood as forming a group together, which triggers agreement. (C126)

Latvian – Coordination [DoƼuleja 117] kopš Alises vairs nav Olga ar Elz-u since Alise:GEN still NEG.be Olga:NOM X/Y Elza-ACC istabƗ guƺ div-as vien sleep two-NOM.PL.F alone room:LOC ‘Since Alise was no longer there, Olga and Elza slept alone in the room.’

The agreement pattern is the same if we substitute un for ar in (C126). Very frequently, we encounter PPs of the structure [ar [NP]]PP as manner adverbials of various kinds. In (C127), two NPs depend upon the preposition ar – and both modify the manner in which Alise looked at her mother. One cannot interpret neticƯba ‘disbelief’ and cerƯbas ‘hopes’ as abstract instruments with the help of which some goal is achieved.

B-Type: Latvian 317

(C127)

Latvian – Manner (DoƼuleja 149) Alise paskatƯjƗs mƗtƝ ar ne-ticƯb-u Alise look:PAST:3:REFL mother:LOC X/Y NEG-belief-ACC un tikko jaušam-Ɨm cerƯb-Ɨm and scarcely feeling-DAT.PL hope-DAT.PL ‘Alise looked at her mother unbelievingly and with scarcely felt hopes.’

In connection with manner, there is also the frequently employed idiom for ‘being content’, cf. (C128). (C128)

Latvian – Possession (physical state/state of mind)/manner (DoƼuleja 137) ja jnjs bnjtu ar mier-u if you be.CONDIT X/Y peace-ACC ‘If you were only content!’

One may consider this an example of possession of the kind Heine (1997) classifies as a physico-mental state. However, this construction is isolated in Latvian. Other physico-mental states are not normally expressed by an ar-PP. Ar occurs also in time adverbials, as in (C129), and in causal constructions where ar introduces the cause, cf. (C130). (C129)

Latvian – Time (DoƼuleja 312) bet PƝteris jau ar aprƯƺa sƗkum-u but Peter:NOM already X/Y April:GEN start-ACC gƗja strƗdƗt jauno zemi new:ACC.DEF land:ACC go.PAST.3 work:INF ‘But already at the beginning of April Peter went to work in the new field.’

(C130)

Latvian – Cause (DoƼuleja 402) vecƗkais dƝls jau pirms kara war:GEN old:COMPL.DEF.NOM son:NOM already before miris ar tuberkuloz-i X/Y tuberculosis-ACC die:PTCPL ‘The oldest son was said to have died of tuberculosis even before the war.’

In sum, the domain or ar spans the whole paradigm of contexts with the exception of those constructions which are companion-oriented. Companion-orientation, however, is not the only area where the dominance of ar is restricted.

318 Case studies 12.2.3. Bare instrumental vs. PP EndzelƯns (1951: 580–586) observed that the bare instrumental is a stylistically marked residue of an earlier stage of Latvian language history. Nevertheless, it would be wrong to declare those forms which instantiate the bare instrumental as archaisms across the board. Admittedly, many instances of the bare instrumental are lexicalised phrases such as, e.g., citiem vƗrdiem ‘with other words’. Besides these cases, we encounter the bare instrumental relatively often as a means of encoding body part instruments or something which very much looks like a body part instrument, cf. (C131). (C131) Latvian – Body part instrument (C131.1) Bare feet (DoƼuleja 156) ziemƗ ne-skrietu desmit NEG-run:CONDIT ten winter:LOC verstis bas-Ɨm kƗj-Ɨm leg-DAT.PL verst bare-DAT.PL ‘In the winter, she would not run barefooted for ten versts.’ (C131.2) Dirty feet (DoƼuleja 352) un ne-nƗc dubƺain-Ɨm kƗj-Ɨm istabƗ dirty-DAT.PL leg-DAT.PL room:LOC and NEG-come ‘And don’t come into the room with dirty feet!’ In (C131.2), it is doubtful whether the dirty feet are depicted as a body part instrument or as a temporary property (= ornative) of the addressee. As a matter of fact, the bare instrumental is especially strong in those cases where the body part does not serve as an instrument proper but modifies the action expressed by the lexical verb or characterises some temporary property of one of the participants; cf. (C132) as opposed to (C133), where ar is used to introduce a bona fide body part instrument. (C132)

Latvian – Ornative/manner (DoƼuleja 372) Ženija sƝdƝja kƗ pamirusi ieplƝst-Ɨm Ženija:NOM sit:PAST.3 as die:PTCPL wide_open-DAT.PL ac-Ưm raudzƯdamƗs lielajƗ vƯrietƯ look:PTCPL huge:DEF.LOC man:LOC eye-DAT.PL ‘Ženija sat there like dead and looked with her eyes wide open at the huge man.’

(C133)

Latvian – Body part instrument (DoƼuleja 198) lƯdz pusdienai Lizete notƗƺis ar ac-Ưm until noon Lizete from_far X/Y eye-DAT.PL

B-Type: Latvian 319

pavadƯja tikai vienu pajnjgu accompany:PAST.3 only one:ACC vehicle:ACC ‘Until noon, Lizete observed only one vehicle from a distance with her eyes.’ Bare instrumentals occur very often on body part nouns but mostly in functions which combine properties of an ornative and a manner adverbial, cf. (C134). (C134) Latvian – Ornative/manner (C134.1) Body part (DoƼuleja 351) sakniebt-Ɨm lnjp-Ɨm viƼa press_closed-DAT.PL lip-DAT.PL she lika galdƗ šƷƯvjus table:LOC plate:ACC.PL put:PAST.3 ‘With tight lips, she put the plates onto the table.’ (C134.2) Body part (DoƼuleja 316) PaulƯne sƝdƝja uz skapja sit:PAST.3 on cupboard:GEN Pauline:NOM malas lepni atmest-u galv-u proudly throw:PTCPL-ACC head-ACC edge:GEN ‘Pauline was sitting on the edge of the cupboard with her head proudly tilted backwards.’ Even if no body part proper is mentioned, the bare instrumental may be understood as specifying a certain temporary aspect of the participant’s physical state, cf. (C135). (C135)

Latvian – Ornative/manner (DoƼuleja 312) ErnestƯne pavƝrƗs meitƗ Ernestine:NOM turn:PAST.3.REFL daughter:LOC noguruš-u skatien-u look-ACC tired-ACC ‘Ernestine turned to her daughter with a tired look [on her face].’

In addition to these cases, bare instrumentals are also relatively common with other kinds of ornatives involving inanimate nouns; some of these instances may be interpreted as part-whole relations, cf. (C136). (C136) Latvian – Ornative/part-whole (C136.1) Side-walls of boats (DoƼuleja 336) tur bija nelielas airu there be:PAST.3 small:NOM.PL oar:GEN.PL

kƗrbas boat:NOM.PL

320 Case studies slaidi izliekt-iem sƗn-iem side-DAT.PL lean:ADV bend:PTCPL-DAT.PL Ɲrtas un vieglas light:NOM.PL convenient:NOM.PL and ‘There were small rowing boats with slender concave sides, convenient and light.’ (C136.2) Stable with walls and roof (DoƼuleja 319) toties PƝteris bija uzcƝlis be:PAST.3 build:PTCPL however Peter:NOM knjti betona sien-Ɨm un concrete:GEN wall-DAT.PL and stable:ACC šindeƺu jumt-u roof-ACC shingle:GEN.PL ‘Peter, however, had built a stable with concrete walls and a shingle roof.’ Similarly often, the bare instrumental occurs on NPs which represent some kind of material, cf. (C137). (C137) Latvian – Material (C137.1) Material filling a room (DoƼuleja 314) mƝsl-iem un velƝn-Ɨm apkrautajƗ manure-DAT.PL and turf-DAT.PL load:PTCPL.DEF.LOC bnjdƗ tiem vietas ne-bija NEG-be:PAST.3 hut:LOC this:DAT.PL place:GEN ‘In the hut, filled with manure and turf, there was no space left for them.’ (C137.2) Face covered with sweat (DoƼuleja 338) sƗrtƗ bƝrna sejiƼa bija face:DIM:NOM be:PAST.3 rosy:DEF.NOM child:GEN norasojusi sƯk-Ɨm tiny-DAT.PL cover_by_dew:PTCPL sviedru lƗsƯt-Ɲm drop:DIM-DAT.PL sweat:GEN.PL ‘The rosy face of the child was covered with microscopic beads of sweat like dew.’ (C137.3) Grave covered with plants (DoƼuleja 338) kaps bija apstƗdƯts puƷ-Ɲm grave:NOM be:PAST.3 plant:PTCPL:NOM flower-DAT.PL un mƝtr-Ɨm un vienmƝr tƯri izravƝts and creeper-DAT.PL and always neatly weed:PTCPL:NOM

B-Type: Latvian 321

‘The grave was adorned with flowers and creepers and always neatly weeded.’ To some extent, the bare instrumental still competes with the PPs headed by ar. This is especially the case with expressions of ornatives, part-whole relations and material – less so with body part instruments. However, in all these contexts, the bare instrumental is an optional choice. The examples quoted so far could all be accompanied by ar. The preposition is clearly the primary marker of the relations under scrutiny, as it can substitute for the bare instrumental and has the more general domain. The replacement of the bare instrumental by the PP is blocked for a restricted number of cases. Lötzsch (1978: 669) mentions three functional subdomains which favour the bare instrumental, namely, manner (“Instrumental der Art und Weise”), time (“Instrumental der Zeit”) and measure (“Instrumental der Maßeinheit”). The word forms used in the bare instrumental are all in the plural: DienƗm ‘for days on end’, metriem ‘by the metre’, lƝkšiem ‘galloping’, etc. Besides the fact that ar is not entirely banned from expressions of manner and time, this functional niche of the bare instrumental does not pose any special problems to the classification of Latvian as a B-type language: Those domains where the bare instrumental – whatever its exact status – is still strong are domains which fall outside the areas of Comitativity and Instrumentality as defined above. Wherever the latter two are involved, ar is the dominant relator. Owing to the fact that animacy does not have a say in the organisation of the distributional profile of ar, it is only the distinction of accompaneeorientation vs. companion-orientation which occasions some kind of formal distinction. Table (C138) summarises our description of the Latvian facts. For the sake of comprehensiveness, we also specify where the bare instrumental is a potential competitor and where lƯdzi has to be used to encode companion-orientation. Grey shading indicates the domain of ar. (C138) Distribution of ar, bare instrumental and lƯdzi Instrumental Tool Other instruments Body part Means of transportation Material Part-whole

bare INS bare INS bare INS bare INS

Ornative Confective Cause Manner Time Accompaniment Coordination

bare INS

bare INS bare INS lƯdzi

Interestingly, there are no spatial categories involved in the distributional profile of the relators. What the Latvian evidence teaches us is that there

322 Case studies may be enough variation hiding behind an apparently very homogeneous B-type system to call for caution in the interpretation of the facts. We will reconsider the Latvian situation in Part D, when we discuss the historical changes that make a language move from one type to the other. 12.3. C-Type: Icelandic 12.3.1. Generalities In contradistinction to the other members of the Germanic phylum, where Type-B reigns, Icelandic is a Type-C language. On superficial inspection, Icelandic seems to employ the same patterns as all the other Germanic languages. The preposition með ‘with’295 is a cognate of German mit and is etymologically identical with similar prepositional relators in the vast majority of the Germanic language – except in English and Faroese, where with and við are used, which are nevertheless prepositions, too. Just like these relators elsewhere in the Germania, Icelandic með is used for the purpose of encoding both the comitative and the instrumental. However, Icelandic is a case language, and með belongs to the large group of Icelandic prepositions which are characterised by differential case government (Kress 1982: 194–209). Með takes its complement either in the dative or in the accusative.296 The choice of the appropriate case on the noun or pronoun depends largely on the semantics of the entire construction. Both prototypical comitatives and prototypical instrumentals require the dative, whereas confectives, ornative, and possession require the dependent NP of the PP headed by með to be in the accusative.297 However, the domains of the constructions overlap to some extent – and this overlap is semantically motivated (cf. below).298 Thus, the Icelandic relator is itself complex because it does not only consist of a preposition, i.e., a free morpheme, but also regularly involves a certain semantically determined morphological case on the complement i.e. a bound morpheme, cf. (C139). (C139)

Icelandic – complex relator (Meira 22) ƒ A

B

ég fer oft með I go often with

han-a she-ACC

út out

X ‘I often take her out [to play in the playground].’

C-Type: Icelandic 323

Henceforth, the combination of með + NP[dative] will be glossed X/Y (on the preposition) and X (on the case-marked complement), whereas the combination með + NP[accusative] will be glossed X/Y (again on the preposition) and Y (on the noun or pronoun). On definite nouns, case inflection regularly occurs twice, with few exceptions, so that the special glosses will also appear more than once in these cases. These glossing convention also apply to those words which morphologically agree with the noun representing the participant of the situation. The borderline separating the domain of the two constructions, however, is subject to some variation. We address this issue in due course below. In the subsequent sections, we also inspect more closely the possibility of replacing með + DATIVE or með + ACCUSATIVE with other only partially synonymous constructions. 12.3.2. The functional domain of með + DATIVE When með combines with a dependent NP in the dative, the PP can be used to link an instrumental adjunct to a lexical verb. Barðdal (2001: 72) observes that the instrumental counts among the three semantic roles which are consistently marked by only one grammatical case (= dative for the instrumental). However, she also states that among the grammatical cases, the dative has the widest range of semantic roles. Indeed, the functional domain of the dative PP covers the entire range of instrumentality. We find examples galore for each and every kind of instrumental relation being encoded by með + DATIVE, cf. (C140). (C140) Icelandic – Instrumental relations (C140.1) Primary instrument/tool (Morð 121) þegar mat-ur-inn kom fékk when meal-NOM.SG-DEF come.PRET.3SG get.PRET.3SG Daria sér bit-a af glóðarsteikt-u roast:PTCPL-DAT Daria REFL:DAT bite-ACC of eggald-i-n-i með gaffl-i-n-um X/Y fork-X-DEF-X eggplant-DAT-DEF-DAT ‘When the meal came, Daria got herself a piece of the roasted eggplant with the fork.’ (C140.2) Secondary instrument (weapon) (Morð 242) þad kæmi jafnvel til greina að ég to branch that I it come:IRR:3SG even

324 Case studies draepi han-a með bys-u-nn-i þin-ni your-X kill:IRR:1SG she-ACC X/Y gun-X-DEF-X ‘Maybe I should even kill her with your gun.’ (C140.3) Ununusal instrument (Náttvíg 152) hann hrin-ti mér með byssuskef-in-u he push-PRET.3SG I:DAT X/Y rifle_butt-DEF-X ‘He pushed me with the rifle-butt.’ (C140.4) Body part instrument (Háska 31) hann klóra-ð-i sér í nef-in-u nose-DEF-DAT he claw-PRET-3SG REFL.DAT in með óhrein-ni hend-i-nn-i svo að smurolía that oil X/Y dirty-X hand-X-DEF-X so dreif-ð-i-st yfir andlit-ið over face-ACC:DEF run-PRET-3SG-MED ‘He scratched himself in the nose with the dirty hand so that the oil ran over his face.’ In (C140.1), the instrument is the fork with which the protagonist is eating the eggplant. In (C140.2), a potential killer of two people is weighing the pros and cons of using the gun of one of her victims to kill the other. As in a variety of similar examples from other languages (cf. above and below), the gun is only a secondary instrument because the deadly wound is actually caused by the bullet entering the victim’s body. In (C140.3), a criminal in a hold-up forces his victim to move by pushing him latter with the rifle butt, which is not a typical tool, but a somewhat unusual means to achieve a certain goal. (C140.4) describes an everyday situation where a body part instrument is involved: The oil-smeared hand touches the nose of the agent causing the oily liquid to run down his face. However, for body part instruments (but not exclusively for these), there is an alternative construction, considered archaic by Kress (1982: 206), although it is quite frequent in the written register. It is a remnant of the old dative-instrumental. This alternative construction lacks the preposition með and consists only of a dative-marked NP, cf. (C141). (C141) Icelandic – Body part instrument Prepositionless dative (Háska 7) Strák-ar-n-ir tveir horf-ð-u two:NOM.M look-PRET-3PL boy-NOM.PL-DEF-NOM.PL á hljómsveit-in-a opn-um mun-ni open-X mouth-X on band-DEF-ACC ‘The two boys looked open-mouthed at the band.’

C-Type: Icelandic 325

It is important to note that the “pure” dative can only fulfil the task of an instrumental299 and never those of a comitative! We will come back to this issue in the final part of the present section. The use of the full dative PP extends to encoding means of transportation, which in Icelandic may alternatively also be treated as containers, and thus that spatial prepositions come to be used as relators (cf. [C150.4]). However, in (C142), the tram is encoded by the full dative PP, just like the other kinds of instruments discussed above. (C142) Icelandic (C142.1) Means of transportation (Náttvíg 12) hann var að fara með to go:INF X/Y he be.PRET.3SG sporvögn-u-n-um upp í bæ up in town tram-X-DEF-X ‘He was going up town by tram.’ (C142.2) Means of transportation (Krummi 61–62) þaðan getiði tekið hvaða vagn sem take:PTCPL what bus REL thence can:2PL er á Hlemm og síðan með Hlemm and then X/Y be.3SG on Seltjarnesvagn-i-n-um Seltjarnes_bus-X-DEF-X ‘From there you can take any bus that is there on Hlemm and then (you continue on) the Seltjarnes bus.’ In both cases, the participant whose actant serves as the subject of the clause(s) is a passenger travelling by a means of public transportation. The person is not a proper agent because the movement is done by the tram or bus, and the passengers are not really active. In addition, the choice of the appropriate case on the noun depends on further criteria. The dative in the above examples depicts the situation as a clear case of “instrumentality”, i.e., the passengers of the tram and bus make use of these means of transportation with a view at being moved from one location to the other. Nevertheless, we also find examples where the dependent noun is in the accusative. In such cases, the relation between the participants (agent and means of transportation) is a different one: Now, it is the means of transportation that is being moved from one place to the other, cf. (C143), where the women themselves move the car to a different place. The issue of the semantics involved in the differential case government is discussed in some detail in Section 12.3.3 below; cf. also (C150).

326 Case studies (C143)

Icelandic Accusative (Háska 46) morgun-inn eftir fór-u go.PRET-3PL morning-DEF after kon-ur-n-ar með bíl-inn í viðgerð in repair woman-NOM.PL-DEF-NOM.PL X/Y car-Y ‘The next morning, the women drove the car to the mechanic.’

The instrument used to achieve a certain goal does not necessarily have to be a concrete object. Abstract instruments are encoded by exactly the same construction as bona fide concrete instruments, cf. (C144). (C144) Icelandic (C144.1) Abstract instrument (Bestu 51) hann lag-ð-i-st á hurð-in-a með he lay-PRET-3SG-MED on door-DEF-ACC X/Y öll-um sín-um þung-a all-X his-X weight-X ‘He pushed the door with all his weight.’ (C144.2) Imaginary instrument (Krummi 11) það var sem náttúra-n sjálf hef-ð-i it be.PRET.3SG REL nature-DEF self have-PRET-3SG lostið þá með töfrasprot-a sín-um her-X hit:PTCPL them X/Y wand-X ‘It was like nature herself had touched them with her wand.’ In (C144.1), the bodily weight of the person pushing in the door is of course not a primary instrument. Rather, it is the body that is used for the purpose of opening the door or moving away a concrete obstacle. The actual weight is just indicative of the strength of the agent. Likewise, imaginary instruments as the metaphorical wand of nature in (C144.2) are far from being concrete objects. In this case, the charming atmosphere of a sunny summer afternoon is reified and turned into an object that could be used as an instrument or tool. Nevertheless, imaginary instruments of this kind are still related to prototypical instruments in so far as properties of the latter are metaphorically transferred to the former. Með + DATIVE is also the usual construction for the material that is employed to produce something or change its appearance. In (C145.1), an old towel is used as stuffing in a wooden box that is meant to serve as a bedstead for a pet animal. The towel not only provides the material for the stuffing but is identical to the stuffing itself. In (C145.2), a woman paints

C-Type: Icelandic 327

her fingernails and the red varnish changes the appearance of the nail. In (C145.3), the name of a caterer is written on his delivery van for advertisement. The written version of the name is made up of the large-sized letters – the letters being the material of which the name label is composed and the same time forming the component parts of the name itself. (C145) Icelandic – Material (C145.1) Stuffing (Krummi 37) uppi á háaloft-i fund-u þeir grunn-an shallow-ACC up on loft-DAT find:PRET-3PL they trékass-a sem þeir fóðru-ð-u REL they stuff-PRET-3PL wood_chest-ACC að innan með göml-u slitn-u handklaeði to inside X/Y old-X torn-X towel ‘Up on the loft, they found a shallow wooden chest which they stuffed inside with an old torn towel.’ (C145.2) Colour (Bestu 26) [hún] tók pensil-inn upp úr out [she] take.PRET.3SG brush-DEF.ACC up glas-in-u og strauk með brush.PRET.3SG X/Y glass-DEF-DAT and fagur-rauð-um lit yfir ein-a nögl-in-a nail-DEF-ACC nice-red-X colour over one-ACC ‘She took the brush out of the glass and stroke over one of her nails with beautiful red colour.’ (C145.3) Letters (Bestu 79) nafn-ið Hallvarður Björnsson var be.PRET.3SG name-DEF.NTR Hallvarður Björnsson skrifað á hlið-ar-n-ar með X/Y write:PTCPL on side-ACC.PL-DEF-ACC.PL stór-um hvít-um stöf-um. big-X white-X letter-X ‘The name “Hallvarður Björnsson” was written on the sides in big, white letters.’ As can be expected from the discussion of other case studies, með + 300 DATIVE is a construction which has a lot of modal functions as well. The randomly picked examples under (C146) only give a very vague idea of how widespread and common the use of the dative PP is in modal contexts. All the examples listed below involve such a dative PP which modifies the semantics of the verb by way of specifying that the action was not easily

328 Case studies carried out (C146.1), that the speech act was done in a way that indicated the speaker’s attitude (C146.2 and C146.3), and that the agent looked at another person in a peculiar way (C146.4). (C146) Icelandic – Modal (C146.1) (Lalli 30) með erfiðismun-um tók-st hon-um he-DAT X/Y difficulty-X take.PRET-MED að setj-a-st upp up to sit-INF-MED ‘With difficulty he managed to sit up straight.’ (C146.2) (Meira 67) nei sag-ð-i Jón Oddur með sannfæring-u Jón Oddur X/Y conviction-X no say-PRET-3SG ‘No, said Jón Oddur with conviction.’ (C146.3) (Krummi 36) þeir kvöddu dýralækn-inn með virkt-um they say_goodbye:PRET.3PL vet-ACC.DEF X/Y grace-X ‘They parted from the vet with a heartfelt goodbye.’ (C146.4) (Háska 76) Björk horf-ð-i hins vegar á Þorstein on Þorstein Björk look-PRET-3SG however með mikl-u trúnaðartraust-i X/Y much-X trust-X ‘Björk, however, looked very trustfully at Þorsteinn.’ In addition to the modal readings of the construction, með + DATIVE also serves the function of encoding the prolative, i.e., a spatial concept of parallel position or, preferably, movement of a trajector along a prototypically static ground; e.g. the American coastline along which the ships moved on which the protagonist used to work as a sailor in (C147.1) and the side of the hill over which the shadow of some entity is moving in (C147.2). (C147) Icelandic – prolative (C147.1) Along the shores (Náttvíg 231) hann haf-ð-i ver-ið í sigling-um og he have-PRET-3SG be-PTCPL in sailing-DAT and var skiptstjóri stundum með Ameríkuströnd-um at.times X/Y American_shore:PL-X be:PRET.3SG captain ‘He had been to sea and at times served as captain along the American shores.’

C-Type: Icelandic 329

(C147.2) Parallel to a limit at the side (Krummi 29) sjá-ðu dökk-u skugg-a-n-a see-you dark-ACC shadow-ACC.PL-DEF-ACC.PL með hlið-i-nn-i X/Y side-X-DEF-X ‘Look at the dark shadows on the side!’ Kress (1982: 205) mentions the prolative as the first in the list of usages of með + DATIVE in Icelandic. His choice can hardly be motivated by frequency, because in terms of token frequency the prolative cannot compete with the other and much more frequent functions of the dative PP. Therefore, it is legitimate to assume that the idea behind Kress’s hierarchy of functions is based on qualities. Probably, the spatial category was more basic for him – perhaps even in the diachronic perspective. In contradistinction to this preference for the spatial function of med, Böðvarsson (1985: 631–632) lists confective and comitative first (in this order), whereas the prolative function is mentioned as function number 8 – the last one before the residual, i.e., non-productive usages. We now approach an area where the distribution characteristics of the construction under scrutiny become more complicated, because its competitor, með + ACCUSATIVE, comes into play as well.301 For a start, consider some uncontroversial cases of prototypical comitatives, cf. (C148). (C148) Icelandic (C148.1) Static (Náttvíg 231) hann sat með kon-u-nn-i X/Y wife-X-DEF-X he sit.PRET.3SG sin-ni yfir glas-i his-X over glass-DAT ‘He had a drink with his wife.’ (C148.2) Reciprocal action (Lalli 54) eftir miklar fortölur tók-st succeed.PRET.3SG-MED after many:PL persuasion Mon-a og Davíð loksins að fá Lall-a finally to get Lalli-ACC Moni-ACC and David til þess að mæta á ein-a to meet:INF on one-ACC to that:GEN körfuboltaæfing-u með liði-n-u X/Y team-DEF-X basketball_training-ACC ‘After a lot of persuasion, Mona and David succeeded eventually to get Lalli to meet with the team during basketball training.’

330 Case studies (C148.3) Partnership (Bestu 124) [ég] hitt-i aðr-a döm-u og [I] meet.PRET-1SG other-ACC lady-ACC and slysa-ð-i-st til að eig-a own-INF accidentally_do-PRET-1SG-MED to to barn með hen-ni líka child X/Y she-X too ‘I met another woman and I accidentally also produced a child with her.’ (C148.4) Joint action (Meira 75) hún fór að klippa jólaskraut she go.PRET.3SG to cut:INF Christmas_adornment með bræðr-u-n-um X/Y brother:PL-X-DEF-X ‘She started to cut out paper adornments for the Christmas tree together with the brothers.’ (C148.5) Membership (Háska 57) ég held að ég sé með be.IRR.1SG X/Y I think.PRET.1SG to I ríku-st-u manneskj-um á Island-i on Iceland-DAT rich-SUPER-X man-X ‘I think that I belong to the richest people on Iceland.’ The situations described in the examples in (C148) each involve two participants who act together and are more or less on equal terms: In (C148.1), a man and his wife are having a drink together. Both actively take part in the event of drinking (or sitting [down to have a drink] for that matter). There is nothing that would prevent a reformulation of (C148.1) in which the coordinating conjunction og ‘and’ is used instead of the subordinating preposition með (of course, the coordinated subject would require plural agreement on the finite verb). In (C148.2), the situation is one of a reciprocal action: If the protagonist Lalli meets the members of the basketball team, the team automatically meets the potential new member Lalli. Again, both participants are active, although one cannot simply substitute og for með in this case. It is clear that the two participants in (C148.3) – the womaniser and his new acquaintance – both had been active in the event of getting the woman pregnant. Shared parenthood is regularly expressed as having a child with someone – just like other cases of joint ownership.302 In (C148.4), the mother tries to calm down her twin sons by participating jointly in the preparations for Christmas. The mother and the two brothers are equally active in the event of producing paper clippings for the Christmas tree. As in (C148.1) before, a switching of með and og would be

C-Type: Icelandic 331

possible, as long as the appropriate morphological conventions for agreement are observed.303 In (C148.5), no such substitution is admissible. The situation described is static. The speaker – a woman from a poor family – allocates herself socially in a certain stratum (although this is only done in a very metaphorical way: She considers herself rich not because she owns a lot of money, but because she has made friends with a group of nice people). Therefore, any aspect of activity is absent from the situation. Note that in (C148.4) the finite verb form fór ‘(she) went’ is not an instance of the motion verb fara ‘to go’ but rather reflects the grammaticalised use of this verb as inchoative auxiliary – meaning ‘to start, to begin’. This observation is important, because the case government of með oscillates between dative and accusative if the main verb is a motion verb (or a positional).304 In situations where motion applies, the case marking on the second actant may follow the pattern of the prototypical comitatives in (C148), cf. (C149). (C149) Icelandic – Verbs of motion (C149.1) Dative (Lalli 93) hann var hættur að fara í bíltúr go:INF in ride he be.PRET.3SG positive to með foreldr-um sín-um X/Y parents-X his-X ‘He was considering going on a drive with his parents.’ (C149.2) Dative (Lalli 39) má ég fara með þér you:X may I go:INF X/Y ‘May I go with you?’ In both examples in (C149), the accompanee is in the dative. As human beings, both participants in the situations described are highest on the animacy hierarchy. Their participation in the events is active. In both sentences, the protagonist – Lalli in (C149.1) and the speaker in (C149.2) – consider joining somebody else on their own account. Thus, their volition is high. However, this is markedly different with the examples in (C150). (C150) Icelandic – Verbs of motion (C150.1) Bring along (Bestu 158) um tíuleytið kom Sossa með about ten:DEF come.PRET.3SG Sossa X/Y dreng-inn sem hét Símon REL be_called:PRET.3SG Símon boy-DEF:Y ‘At 10 o’clock, Sossa came with (= brought along) the boy whose name was Símon.’

332 Case studies (C150.2) Take somewhere (Lalli 36) við verð-um að fara með þig to go:INF X/Y you:Y we will-1PL strax í rannsókn immediately in check ‘We will take you immediately to check (you medically).’ (C150.3) Walk the cat (Lalli 61) við get-um farið í göngutúr með falleg-a walk X/Y nice-Y we can-1PL go:PTCPL in kött-inn þinn your:Y cat-DEF:Y ‘We can go for a walk with your nice cat.’ (C150.4) Transport someone by plane (Bestu 167) Það var maður sem ætla-ð-i REL want-PRET-3SG that be.PRET.3SG man að fljúga með mig í lítil-li I:Y in small-DAT to fly:INF X/Y flugvél en mig langa-ð-i meira að more to plane but I:ACC long-PRET-3SG fara í bíl car go:INF in ‘That was a man who wanted to fly back home with me in a little plane, but I rather wanted to go by car.’ In all of these cases, the second actant is in the accusative, notwithstanding the fact that the participants have the property [+animate]. In (C150.1), the character Símon is an eight month old baby who, in the accompanying context, is said to be asleep for most of the time. In (C150.2), the protagonist, Lalli, shows evident signs of being ill – according to his parents. The parents therefore consider taking their son to the hospital for a medical check. In (C150.3), the children talk about taking the cat for a walk in order to have some privacy for conversation. In (C150.4), a little boy gets offered by a stranger to be taken home by air, although the boy has different plans. The baby, the temporarily invalid Lalli, the cat, and the little boy (though less so) have one thing in common. As to control, they are not on an equal footing with the other participant – the accompanee. The accompanee indeed is the agent who has control over the other participant and thus the latter, the companion, becomes more like a patient than a full blown companion (in the sense of “second” or “co-agent”). The fact that the little boy in (C150.4) refuses to accept the offer is a completely different story: The frustrated pilot

C-Type: Icelandic 333

would nevertheless have been the one who had control over the plane, the flight, and the passengers (only that the airborne travel was never realised). In all of the examples in (C150), the companion is being taken along by the accompanee. That is why an appropriate English translation of koma/fara með + ACCUSATIVE would require a verb of transportation meaning ‘to bring along, to take along’, whereas koma/fara með + DATIVE corresponds neatly to English together with (or the coordinating conjunction and). The dative is also the appropriate case forms in those cases where a reflexive pronoun governed by með is coreferential with the accompanee as in (C151). (C151) Icelandic – Reflexive (C151.1) Take along (Krummi 54) Þeir tók-u hrafn-inn út með sér they take.PRET-3PL raven-DEF.ACC out X/Y REFL:X ‘They took the raven out with them.’ (C151.2) Have around (Meira 39) eftir að snjór-inn kom vor-u come.PRET.3SG be.PRET-3PL after to snow-DEF strák-ar-n-ir latari að hafa lazy:COMPL to have:INF boy-NOM.PL-DEF-NOM.PL litl-u stelp-ur-n-ar með sér girl-ACC.PL-DEF-ACC.PL X/Y REFL:X small-ACC ‘After the snow had come, they boys became more reluctant to have the little girls around.’ Evidently, með + ACCUSATIVE is responsible for encoding companions in confective constructions independent of animacy, cf. (C143) above. This is tantamount to saying that comitatives proper fall within the functional domain of með + DATIVE. We focus again on this functional difference between the two constructions in Chapter 3 below. A competition of the two constructions is also attested when it comes to marking ornatives. Með + DATIVE is very often used to link an attribute to a head noun, cf. (C152). (C152) Icelandic – Attribute dative PP (C152.1) (Lalli 9) hún haf-ð-i búið í litl-a ljót-a she have-PRET-3SG live:PTCPL in small-DAT rotten-DAT hús-i-n-u með ryðgað-a þak-i-n-u house-DAT-DEF-DAT X/Y rust:PTCPL-X roof-X-DEF-X ‘She had lived in the small, derelict house with the rusty roof.’

334 Case studies (C152.2) (Krummi 17) þetta var rauð-ur Range Rover red-NOM.SG.M Range Rover that be.PRET.3SG með R-númer-i X/Y R_number-X ‘That was a red Range Rover with a Reykjavík number plate.’ (C152.3) (Krummi 27) þá settu Árbæjar-og Breiðholtshverfi they see:PRET.3PL Árbæjar-and Breiðholts_quarter sérstæðan svip á höfuðborg-in-a capital-DEF-ACC extraordinary:ACC likeness on með svipmikl-um sambýlishús-um X/Y impressive-X apartment_block-X ‘They saw the Árbæjar and Breiðholts-quarters’ extraordinary similarity to the capital, with its impressive apartment blocks.’ (C152.4) (Háska 10) Í botn-i-n-um á hverri þeirra on each:DAT they:GEN in bottom-DAT-DEF-DAT vor-u hvít blóm með flower X/Y be.PRET-3PL white blóðrauð-um krónublöð-um blood_red-X petal:PL-X ‘On the bottom on each of them, there were white flowers with blood-red petals.’ (C152.5) (Háska 12) hún stakk aum-um fingr-i-n-um sore-DAT finger-DAT-DEF-DAT she put.PRET.3SG ofan í glas með köld-u vatn-i upper_part in glass X/Y cold-X water-X ‘She put her sore finger into the glass with cold water.’ The five sentences in (C152) all follow the same pattern. There is an inanimate head noun to which an attribute PP is added. In some of the cases, the relation between head noun and attribute is of the part-whole type. In (C152.1), the rusty roof forms part of the building that it characterises at the same time. In (C152.3), the high-rise buildings which are typical for the skyline of Reykjavík are also part of the capital city itself. In (C152.4), the red petals are likewise a constitutive part of the flower to which they belong. The part-whole relationship is less obvious with (C152.2): The license plate of course belongs to the vehicle to which it is attached – how-

C-Type: Icelandic 335

ever, it is not a component part of the car’s material structure. In a way, the license plate is an additional extra. Moreover, (C152.5) is not a proper instance of a part-whole relationship because the ice-cold water is not a part of the container in which it happens to be at the time of the event described. The water is easily dissociable from the glass. According to Kress (1982: 205), attribute PPs headed by með normally have their dependent noun in the dative, provided that the head noun refers neither to an animate being nor to a container.305 In Kress’s wording, this distribution is a tendency or preference – meaning that there are also examples deviating from the pattern. The rule passes the test with animate head nouns, cf. (C153). (C153) Icelandic – Accusative (C153.1) Human being (Háska 11) þetta vor-u ekki börn eins og hún this be.PRET-3PL NEG child:PL one:GEN and she heldur strák-ar í matrósföt-um in sailor_dress-NOM.DAT rather boy-NOM.PL og stelpur með hvít-ar blúndusvunt-ur X/Y white-Y lace_apron-Y and girl-NOM.PL ‘These weren’t children like her but boys in sailors’ uniforms and girls with white lace aprons.’ (C153.2) Animal (Krummi 50) í landvar-i-n-u undan klöpp-i-nn-i in country_side-DAT-DEF-DAT before cliff-DAT-DEF-DAT beint neðan við Suðurnesvörð-u-n-a directly below at Suðurnes_observatory-ACC-DEF-ACC syn-t-i æðarkoll-a með X/Y be_visible-PRET-3SG eider_duck-NOM ung-a sín-a í langr-i halaróf-u row-DAT young-Y its-Y in long-DAT ‘In the countryside in front of the cliff directly below the Suðurnes outlook appeared an eider-duck with her young in a long row.’ In both sentences in (C153), the head nouns to which the attribute NPs are linked via með display the property [+animate]. The girls in (C153.1) are of course [+human], and the eider-duck in (C153.2) is [+animate]. The animacy of the attribute NP is of no importance for the choice of the appropriate case, notwithstanding the fact that (C153.1) is an uncontroversial instance of an

336 Case studies ornative – someone is wearing a certain dress – whereas (C153.2) may also be understood as a confective – a duck leading her young somewhere. The use of the accusative PP is fairly consistent in combination with animate head nouns. However, contrary to what Kress depicts as a tendency towards using með + ACCUSATIVE for attributes of a head noun that refers to a container, our sources in the vast majority of cases opt for the dative PP as well; cf. (C152.5) above and (C154), where we encounter a variety of containers – bottles of different size and envelopes. (C154) Icelandic – Containers (C154.1) Bottle (Meira 95) flaska-n með vaxhrúg-u-nn-i var wax_heap-X-DEF-X be.PRET.3SG bottle-DEF X/Y nefnilega ekki til lengur NEG to long namely ‘The bottle with the wax on it was no longer there.’ (C154.2) Envelope (Meira 83) hún fann umslag með pening-um inni envelope X/Y money-X inside she find.PRET.3SG í bók-inn-i in book-DEF-DAT ‘She found the envelope with money in the book.’ (C154.3) Small bottle (Bestu 27) hún loka-ð-i ekki ein-u sinn-i one-DAT time-DAT she close-PRET-3SG NEG glas-in-u með naglalakki-n-u nail_varnish-DEF-X glass-DEF-DAT X/Y ‘She didn’t even close the bottle with nail varnish once.’ More light will be shed on the functional domain of með + ACCUSATIVE in the subsequent Chapter 3. With a view to closing the present section, two more phenomena have to be mentioned for the sake of comprehensiveness. A number of verbs take a complement dative PP headed by með. The use of the dative PP is valency-bound. Note, however, that there are also instances of valency-bound accusative PPs (cf. below). Typical examples of með + DATIVE are given in (C155). (C155) Icelandic – Valency-bound dative PP (C155.1) (Háska 100) Gunnar hörfa-ð-i til Gunnar step_back-PRET-3SG to

baka og back and

C-Type: Icelandic 337

fylg-d-i-st áhyggjufull-ur með X/Y follow-PRET-3SG-MED concerned-NOM stimping-u-n-um fight-X-DEF-X ‘Gunnar stepped back and observed the fight with concern.’ (C155.2) (Morð 174) eins og dýr hneppt í snör-u one:GEN and animal button:PTCPL in trap-DAT hugsa-ð-i ég með mér I X/Y I:X think-PRET-1SG ‘Like an animal caught in a trap, I thought by myself.’ (C155.3) (Háska 48) maður átti að halda með hold:INF X/Y man:NOM.SG have:PRET:3SG to vinkon-u sinn-i girl_friend-X her-X ‘One should be on one’s girlfriend’s side.’ In addition, there is some residual competition of með + DATIVE and við + ACCUSATIVE when it comes to expressing some relations which otherwise can be considered to belong to the realm of the prototypical comitative. Kress (1982: 209) enumerates those cases where við + ACCUSATIVE tends to be used. Among these we find partnership and family relations. In the rest of the cases, við + ACCUSATIVE seems to be valency-bound in combination with certain verbs (including speech act verbs). Outside this subdomain, the preposition við, whose case government likewise oscillates between the dative and the accusative, has a variety of spatial meanings reaching from ‘against’ over ‘next to’ to ‘towards’, cf. (C156). (C156)

Icelandic – við (Morð 133) mig lang-ar til að tala um Pepper við to talk:INF about Pepper to I:ACC long-3SG to þig við tækifæri opportunity you:ACC at ‘I want to talk with/to you about Pepper when the opportunity comes up.’

Með and við are not in free variation. The distribution of the prepositions is subject to restrictions and their domains are for the most quite distinct. Furthermore, the use of við in a comitativelike sense is clearly not the principle function of this relator. Nevertheless, this limited competition of með and

338 Case studies við is interesting, because it only applies to the comitative, whereas the prepositionless dative (cf. above) competes exclusively with the instrumental, although both comitative and instrumental receive the same standard encoding by með + DATIVE. What this means is that, irrespective of syncretism on the expression side, the partial synonymy (and therefore occasional competition) with other distinct relators demonstrates that the semantics of the construction með + DATIVE is indeed complex, allowing for several foci which may lead an at least semi-autonomous life, in a manner of speaking. A multi-level model is called for in order to describe the given facts more adequately. 12.3.3. The functional domain of með + ACCUSATIVE As the discussion in the previous section demonstrates, the domains of the two constructions under scrutiny are not always neatly separated. To some extent at least, this is also true of the expression of predicative possession. Icelandic has a rather intricate system of possession in which alienability/inalienability, temporary/permanent and abstract/concrete are the most important parameters.306 In predicative possession, three constructions are in use: The two transitive verbs hafa ‘have’ (mostly for abstract possession) and eiga ‘own’ (mostly for social relations and permanent possession), and the construction vera með + ACCUSATIVE (mostly for temporary possession, illnesses and body parts). (C157) provides a list of examples for the regular use of vera með + ACCUSATIVE. (C157) Icelandic – Possession (C157.1) Body parts (Háska 84) Þíð er-uð með svo lík-t nef og even-Y nose and you:PL be-2PL X/Y so falleg blá aug-u beautiful blue eye-Y:PL ‘You have such a straight nose and beautiful, blue eyes.’ (C157.2) Temporary mental state (Lalli 25) strák-ur-inn var með grátstaf-inn boy-NOM-DEF be.PRET.3SG X/Y tear_pole-Y:DEF í kverk-u-n-um in throat-DAT-DEF-DAT ‘The boy was almost crying (lit. was with the pole of tears in the throat).’

C-Type: Icelandic 339

(C157.3) Temporary physical state (Náttvíg 197) hann sá að hún var she be.PRET.3SG he see.PRET.3SG to með opin aug-u-n X/Y open eye-Y:PL-DEF ‘He saw that she had her eyes open.’ (C157.4) Alienable actual possession (Morð 70) veit-stu hverjir er-u með lykil who:NOM.PL be-PL X/Y key.Y know-2SG:you ‘Do you know who has a key?’ (C157.5) Abstract possession (Krummi 140) þú var-st með þá hugmynd X/Y that:Y idea you be.PRET-2SG með hverfistein-inn X/Y grindstone-Y:DEF ‘You had the idea with the grindstone.’ (C157.6) Abstract possession (Bestu 150) það stóð í hon-um að svara og to answer:INF and it stand.PRET.3SG in he-DAT þó var hann með all-t þetta úrval X/Y all-Y that selection though be.PRET.3SG he á stór-u spjald-i fyrir framan sig for in_front REFL:ACC on big-DAT menu ‘He didn’t manage to answer although he had all these options on the list right in front of him.’ The bulk of the attestations of predicative possession with vera með have, in fact, accusative government, as in the examples in (C157) above. There are, however, idiomatic expressions where dative government applies instead. Kress (1982: 205) emphasises that these are residues of a formerly more widespread use of the dative in predicative possession. A typical instance of an idiomatic expression with dative PP is (C158). (C158)

Icelandic – Residual dative [Bestu 74] það er eins og enginn nobody it be.3SG one:GEN and rétt-u ráð-i hérna right-X sense-X here ‘It’s like everybody is half crazy here.’

sé með be.IRR.3SG X/Y

340 Case studies Synchronically, the dative is no longer productive in predicative possession. Owing to the fact that many of the idiomatic expressions in which the dative is still obligatory refer to physical or mental capabilities of human beings where the possessor has no control over the possessee, the attestations form a small functional subdomain focusing on one segment of inalienable possession, whereas vera með + ACCUSATIVE covers the larger part of the possessive categories. At this point, it is in order that we come back to the issue of competition between the two PPs. As pointed out above, með oscillates between dative and accusative government in two cases. Recapitulation: First, the change from dative government to accusative marks the transition from a full blown comitative relation to one of a confective with motion verbs. Second, in attribution, the dative PP normally occurs with an inanimate head noun, whereas the accusative PP is reserved for attributes of animate head nouns. The comitative-confective distinction is not really a problem as long as motion verbs are involved. The dependent noun is in the accusative if the companion does not participate in the situation with the same degree of volition, initiative or control as the accompanee. Involuntary or guided participation of animate beings is treated on a par with the participation of inanimates, which per definition lack control. In addition to (C150) above, (C159) is a typical instance of guided participation: The wolfhounds need someone to walk them regularly through the neighbourhood. The woman guides her wolfhounds, i.e., she has control over the action and the companions (no matter how much the wolfhounds are accustomed to, and therefore long for, the regular walk). (C159)

Icelandic – Accusative (Morð 64) Þarna var kona sem alltaf hljóp always run.PRET.3SG there be.PRET.3SG woman REL á sam-a tím-a og ég á time-DAT and I on on same-DAT morgn-a-n-a og önnur sem and other REL morning-ACC.PL-DEF-ACC.PL gekk oft fram hjá hús-i-n-u at house-DAT-DEF-DAT walk.PRET.3SG often ahead mín-u með tvo úlfhund-a X/Y two:Y wolfhound-Y my-DAT ‘There was a woman who always jogged at the same time in the morning that I did and another who often walked by my house in the afternoon with two wolfhounds.’

C-Type: Icelandic 341

On superficial inspection, things seem to become a bit more complicated if the situation is not dynamic. Consider the pair of sentences in (C160). (C160) Icelandic (C160.1) Accusative (Meira 7) hvað ertu búinn að gera af do:INF of what be:2SG:you finish:PTCPL to kon-u-nn-i spur-ð-i ung kona young woman woman-DAT-DEF-DAT ask-PRET-3SG sem var þarna með lít-inn strák REL be.PRET.3SG there X/Y small-Y boy ‘What have you done to your wife?, asked a young woman who was there with a little boy.’ (C160.2) Dative (Meira 7) þess vegna er ég hér með because be.3SG I here X/Y this:GEN mín-um strák-um my-X boy-X ‘That’s why I am here with my boys.’ (C160.1–2) occur in the same paragraph where the first school day of the twin brothers Jón Bjarni and Jón Oddur is described. Parents accompany their children to the school. As a matter of fact, the protagonists’ father is the only man around, all the other children being in the company of their mothers. This is why the woman in (C160.1) inquires about what happened to the twins’ mother. This concerned young woman is said to be there (at the school) with her little son. The NP referring to the son is in the accusative – governed by með. The same preposition is used in (C160.2), again with the static copula vera ‘to be’. But this time the preposition governs the dative. As a matter of fact, the change in the pattern of case-government reflects a change of orientation. Notwithstanding the fact that the parents guide their children to school, only (C160.1) is in accordance with this asymmetry in control, volition and initiative between the participants. The accusative used on strákur ‘boy’ in (C160.1) appropriately reflects the parent-child relation. In (C160.2), however, emphasis and emotion come into play. The woman who wanted to know why the twins’ mother was not there goes on commenting about her absence. (C160.2) is the father’s direct reply to the woman’s reproachful statement (C161).

342 Case studies (C161)

Icelandic (Meira 7) mér finn-st nú maður verði að fara must:3SG to go:INF I:DAT find.3SG-MED now man með litl-u greyj-u-n-um fyrsta dag-inn X/Y small-X poor-X-DEF-X first:ACC day-DEF:ACC ‘Now, I think that one has to accompany (lit. go with) the poor little ones on their first day.’ Owing to the ambiguity of the Icelandic word maður, which may either be used as a full autosemantic noun meaning ‘male person’ (= first option) or ‘human being’ or as an indefinite pronoun equivalent to English one, the twins’ father is able to make a witty reply: If the woman deems it necessary that “a man” accompanies the children, then that is exactly what the father is doing. The dative (plural) on the noun strákur ‘boy’ is motivated on the one hand by the immediate context: In (C161), the woman depicts the situation as one of accompaniment with two egalitarian participants. Notwithstanding the fact that the children need someone to accompany them, they act independently (they could easily go without the parental escort). Thus, the twins’ father adopts this pattern when he explains his presence in (C160.2). On the other hand, foregrounding and backgrounding are important as well. In (C160.1), the boy with whom the woman is there at the school is less important, the less important participant is in the background. In (C160.2) contrastive focus (both on the father and the boys) and foregrounding (of the boys) contribute to the choice of the dative. Thus, we may conclude that með + ACCUSATIVE is appropriate in constellations where, in addition to a lesser degree of volition, initiative and control, the second participant is backgrounded, whereas með + DATIVE is the best choice when the degree of volition, initiative and control is similar to the one of the first participant and/or foregrounding applies. It remains to be seen whether or not backgrounding/foregrounding is of any explanatory value for the competition of the two constructions in the realm of ornatives. The accusative PP is used with animate head nouns, cf. (C162). (C162) Icelandic – Accusative (C162.1) Direct attribute NP (Krummi 81) niðri á mó-u-n-um mátti sjá down on heath-DAT-DEF-DAT may:PRET:3SG see.INF ær með fullvax-ið lamb sitt mother_sheep X/Y grown_up-Y lamb her:Y ‘Down on the heath he might have seen a mother-sheep with its lamb.’

C-Type: Icelandic 343

(C162.2) Indirect attribute NP (Lalli 81) hann lá í rúm-i-n-u með bed-DAT-DEF-DAT X/Y he lay.PRET in kött-inn ofan á sér above on REFL:DAT cat-DEF:Y ‘He lay in bed with the cat on top of him.’ No matter whether the head noun refers to a human being or other animates, the attribute NP takes the accusative independent of its syntactic distance to the head noun: In (C162.1), the attribute is adjacent to the head noun whereas in (C162.2), the accusative PP is almost a secondary predication with a certain degree of independence (iconically marked by its distance from the subject pronoun hann on the surface). In addition to these standard cases, there are also instances where the accusative PP is used in combination with an inanimate head noun, cf. the passage marked by single underlining in (C163). (C163)

Icelandic – Accusative (Krummi 27) niðri á þjóðveg-in-um var bíl down on country_road-DEF-DAT be.PRET.3SG car ekið með ofsahrað-a í austurátt X/Y high_speed-X in eastwards drive:PTCPL og minnti þetta ökutæki með X/Y and remind:PRET:3SG this vehicle lang-an rykmökk aftur úr sér á smoke_tail behind out REFL:DAT on long-Y gríðarstór-an gráleit-an drek-a með greyish-ACC dragon-ACC X/Y huge-ACC gljáandi haus sem veifa-ð-i hal-a-n-um shining head REL wave-PRET-3SG tail-DAT-DEF-DAT ‘Down on the country-road, some one drove this car at high-speed and this vehicle with its long tail of smoke on the back was reminiscent of a huge grey dragon with a shining head wiggling the tail.’

In this longish example the attribute to the noun ökutæki ‘vehicle’ is in the accusative: með langan rykmökk ‘with a long smoke-tail’. Kress (1982: 206) remarks that actual usage of the two cases is subject to variation exactly because of the coexistence of the dative PP for inanimate head nouns and the accusative PP for animate head nouns. Analogy could be responsible for the occasional employment of the accusative replacing the dative (i.e. the accusative PP qualifies as an attractor for all attributive constructions). As far as we

344 Case studies can judge, there is no semantically transparent motivation for cases like (C163) to lean towards the accusative PP.307 Despite the fact that Icelandic often treats means of transportation as containers, the potential conceptualisation of a vehicle as a container is of no significance in the present context because the attribute – the long smoke-tail – is said to be visible on the outside of the car. Thus, the situation is different from the typical constellation where some substance is inside of a container. In addition, pace Kress (1982: 205), we have not been able to find any substantial number of cases where head nouns designating containers require their attribute to be an accusative PP. Quite the opposite: The only attestations of container head nouns with attribute PPs involve the dative PP. This fact may be occasioned by our relatively modest corpus from which we draw our examples. Nevertheless, the accusative PP does not seem to be a very frequent option for attributes of container NPs. Of course, there are also cases of valency-bound or otherwise obligatory með + ACCUSATIVE. For the sake of brevity, we make do with just enumerating some pertinent example in passing, cf. (C164). (C164) Icelandic – Accusative (C164.1) (Bestu 62) heyr-ðu hvernig er eiginlega hear-you how be.3SG properly ‘Listen, what is the matter with you?’ (C164.2) (Háska 48) hvað með þig what X/Y you:Y ‘What about you?’ (C164.3) (Bestu 53) og er hann ánægður content:NOM.M.SG and be.3SG he ‘And is he content with that?’ (C164.4) (Háska 16) Solla hélt áfram með X/Y Solla hold.PRET.3SG ahead ‘Solla continued with the story.’

með X/Y

þig you:Y

með X/Y

það that:Y

sög-u-n-a story-Y-DEF-Y

In (C164.1–2), the idiomatic form of the inquiries about someone’s opinion or actual state most probably is elliptic – the copula is missing. In (C164.3), the predicative adjective ánægður requires the accusative PP (this pattern is relatively common in Icelandic), whereas in (C164.4) it is the complex verb

C-Type: Icelandic 345

halda áfram ‘to go on’ that calls for a complement PP. There is a variety of verbs which combine with the accusative PP headed by með. To conclude this section we mention a case where the competition between the two PPs is again involved to some extent. Discounting some minor alternatives, the English verb to play has two translation equivalents in Icelandic, namely the verb leika and its reflexive variant leika sér. Both allow for a prepositional complement headed by með – however with different case government each, cf. (C165). (C165) Icelandic – To play (C165.1) Dative (Lalli 96) hann lofa-ð-i því að leika they:DAT to play:INF he promise-PRET-3SG með liði-n-u X/Y team-DEF-X ‘He promised to play with the team.’ (C165.2) Dative (Lalli 33) hún vildi helst fara strax she will:PRET:3SG rather:SUPER go:INF immediately út að leika X/Y með hon-um X/Y he-X out to play:INF ‘She wanted to go out immediately to play together with him.’ (C165.3) Accusative (Meira 20) það er alltaf svo spenanndi að leika always so exiting to play:INF it be.3SG sér með leikföng REFL:DAT X/Y toy:PL.Y ‘It is always so exiting to play with toys.’ The first two examples (C165.1–2) describe a situation that is typically associated with a comitative, because the participants are all human beings and take actively part in the event of playing. In (C165.1), the protagonist says to be inclined to play on the basketball team, i.e., to participate with the other members of the team in the scheduled match in which the entire team including the new member is actively involved. In (C165.2), a girl would like to run outside to play with her boyfriend, i.e., both would be actively engaged in children’s joint activities. In both cases, the lexical verb is simply leika and the adjunct PP has the dependent noun in the dative – as could be expected in a comitative-like context. In contradistinction to these two cases, (C165.3) contains a sentence in

346 Case studies which children express their opinion about playing with toys in general. The toys as the patient of the action described are represented by a noun in the accusative – dependent on the preposition með, whereas the lexical verb is reflexive this time: leika sér. The accusative appropriately reflects the fact that the toys have no control of the event of playing and that they are manipulated by the agents. 12.3.4. Summing up the Icelandic situation This differential behaviour of dative and accusative is fully in line with their functions outside the PPs. Icelandic is a language with a particularly rich inventory of non-canonical subjects, i.e., with NPs marked for oblique cases (in stead of the nominative) functioning as subject or subject-like categories (Andrews 2001: 87–93). According to Barðdal (2001: 56), a subject may be marked for accusative or dative – besides nominative as the most frequent and genitive as a further minor option. However, when used to encode fundamental grammatical relations the accusative is normally associated with the less agentive roles – such as affected object, patient, theme – whereas the dative very often takes a position in-between actor and undergoer roles. Barðdal’s statistics (2001: 86–87) support this assumption.308 The close association of the accusative with non-agentivity is also indicative of the absence of control on the part of the participant whose actant is in the accusative. Lack of control over a situation is characteristic of inanimate entities, whereas the reverse is true of animates especially those with the property [+human] which normally are ascribed the capability to control a situation. However, the distribution of cases cannot be reduced to the absence or presence of [+animate]. It is true that animacy is an important parameter – but it is hardly sufficient to explain the employment of the dative and accusative PPs. Table (C166) surveys the correlation between animacy, site of morphological marking, and choice of case. Grey shading marks the participant whose actant bears the morphological marking. A black cell indicates the incompatibility of a case with a given constellation.

C-Type: Icelandic 347

(C166)

Animacy correlation

Participants’ animacy Accompanee Companion + + + +

dative + +

með + accusative

+ +

¬ ¬

+ +

+

¬ ¬ ¬ ¬

+ + ¬ ¬

+ + + +

(+)

+

Comments only dative is admissible dative = autonomous participants accusative = no control/ ornative/ valency-bound only dative is admissible dative = valency-bound accusative = instrumental/ animate ornative/valencybound dative = prolative dative = ornative dative = prolative dative = inanimate ornative (accusative = containers [if at all!])

The dative gets the maximum number of plusses because it is admissible in all of the constellations, whereas the accusative occurs much more restrictedly. Discounting the sometimes rather idiosyncratic instances of valencybound PPs, the following pattern emerges from (C166): The accusative is excluded if the actant representing the accompanee receives the morphological marking. Likewise, if the actant representing the accompanee is the dependent noun of the PP, it automatically is in the dative. This correlation holds irrespective of the animacy of the participants involved. On the other hand, the accusative is employed only if the site of the morphological marking is on the actant representing the companion. However, the correlation between marking-site and morphological case is not that straightforward as with accompanee and dative. If the companion is [+animate] (especially [+human]) then chances are that the dative will be used and not the accusative. The dative is the appropriate case on an animate companion if and only if the companion himself also has some control over the event (or acts on his own account, etc.). If control is not exercised by the companion, the dative is ruled out and the accusative has to be used instead. One problematic case remains to be explained, viz. the dative used for inanimate ornatives. Of course, it is clear why the accusative is used for ornatives of animate nouns: The referents of the attributes are considered entities subordinate to the referent of the head noun. This subordinate rela-

348 Case studies tion is independent of the animacy of the attribute, although most often the attribute lacks the property [+animate]. Attributes of inanimate head nouns may be either [+animate] or [¬animate]. In both cases, the appropriate morphological case is the dative. In the former constellation, i.e., an animate attribute of an inanimate head noun, the use of the dative maybe caused by the higher degree of animacy on the part of the companion though this explanation cannot be extended to the latter constellation. Here, animacy is clearly not decisive. Rather, it is factual or virtual autonomy of the participants that triggers the dative (irrespective of the fact that some of the attribute-noun combinations replicate the part-whole pattern). The dative signals that the two participants are egalitarian as to their control of the situation – two inanimates exercise an equal control of a situation, viz. no control at all. Put differently, control is the major criterion when it comes to deciding which of the two PPs – dative PP or accusative PP is to be employed in a given context. Table (C167) sums up the distribution of the two PPs headed by með over functional subdomains. Grey shading marks those functions which are expressed by með + DATIVE. Unmarked cells belong to the functional domain of með + ACCUSATIVE. (C167) Distribution of labour among dative PP and accusative PP instrumental tool secondary instrument body part means of transportation abstract material modal temporal

spatial prolative comitative animate ornative inanimate ornative animate confective animate confective inanimate actual possession abstract possession body part possession

As (C167) suggests, Icelandic, too is a language where comitativeinstrumental syncretism applies. In this respect, Icelandic behaves exactly like its Germanic next-of-kin. However, this similarity diminishes as soon as we take the intermediate categories into account for which the cognates and translation equivalents of Icelandic með are normally employed elsewhere in the Germanic-speaking world. (Part of the) ornatives and all confectives together with possession do not go along with the prototypical functions. Thus, we find two syncretistic patterns: INS = COM = MOD = LOC = ORNINANIMATE z ORNANIMATE = CONF = POSS.

Across the types 349

12.4. Across the types In the previous sections, we have looked at representatives of the three types A, B and C. None of the three languages chosen for the exemplification can be said to instantiate the ideal of their respective type because Maltese, Latvian and Icelandic also display each a remarkably high dose of individualism when it comes to assigning functions to their relators. Thus, the solutions adopted by Maltese cannot simply be generalised across the board for all members of the A-type, ditto Latvian and Icelandic for the B-type and the C-type respectively. Nevertheless, these peculiarities do not exhaust the functional domains of the markers under scrutiny. Quite to the contrary: there is considerable common ground to base sound generalisation on. The most striking commonality of the three languages is the formal distinction between accompanee-orientation and companion-orientation. Maltese, Latvian and Icelandic employ different expressions for these two kinds of orientation. In Maltese, this is indirectly reflected by the fact that the autonomous companion requires the preposition ma’ whereas a companion with reduced autonomy combines with bi (cf. [C106]). The choice of the appropriate relator is tantamount to a shift of viewpoints: bi depicts the situation from the vantage point of the accompanee who is presented as being in control of the action whereas ma’ concedes equal share of control to both participants. In Latvian (cf. [C138]), ar and lƯdz co-vary in a similar way with the two viewpoints. Simplifying, the use of ar corresponds to the one of ma’, i.e., both participants are granted the same amount of control and autonomy. LƯdz however indicates that the companion is in a clearly subordinate position to the accompanee who again controls the situation. What is achieved by the oppositions of Maltese ma’ z bi and Latvian ar z lƯdz, is a matter of differential case-government in Icelandic: með + DATIVE characterises the companion as autonomous participant as opposed to með + ACCUSATIVE which goes along with reduced autonomy of the companion (cf. [C166]). In Maltese and Icelandic, the expression employed for the subordinate companion is identical to the one used for prototypical instruments. This possibility is blocked of course for Latvian as B-type languages generally have Comitative-Instrumental syncretism. Thus, Latvian makes use of an additional marker which has no other function in the whole system – lƯdz exclusively serves the purpose of marking the subordinate companion. This is noteworthy because it shows that even in a purportedly homogeneous language of the B-type, there may be functional niches where additional formal distinctions thrive unexpectedly.

350 Case studies With a view to determining whether this common preoccupation with the autonomy/dependence of the companion is more than just a funny incidence, we check a selection of our European sample languages for the distinction of accompanee-orientation vs. companion-orientation. In (C168), we present one additional example for each type. (C168) Languages with distinct markers for companion-orientation (C168.1) A-type: Finnish (Karlsson 1978: 220) men-en poik-i-en mukaan boy-PL-GEN along go-1SG ‘I go along with the boys.’ (C168.2) B-type: Estonian (Hasselblatt 1995: 21) hea küll ma tul-en siis kaasa then along alright I come-1SG ‘It’s ok, I come along then.’ (C168.3) C-type: Lithuanian (Dambrinjnas, Klimas, and Schmalstieg 1980: 281) 1263 met-ais Mindaugas buvo 1263 year-Y Mindaugas:NOM be:PRET.3SG nužudytas kartu ir jo du snjnnjs also his two son:NOM.PL murder:PTCPL along ‘In 1263, Mindaugas was murdered and (lit. along also) his two sons too.’ In the A-type language Finnish, the regular marker for the Comitative is either the inflectional comitative -ine- or the postposition kanssa with the genitive. Neither of these two relators can be used when the perspective in which the situation is depicted is the one of the companion. In this case, mukaan (or a variant thereof) is the appropriate relator – again a postposition with genitive government (C168.1). Similarly, the closest kin of Finnish, the B-type language Estonian, does not employ the syncretistic inflectional comitative -ga in contexts of this kind (C168.2) but resorts to the postposition/adverb kaasa which happens to be a cognate of -ga (and Finnish kanssa for that matter) all being derived from a noun meaning ‘company’ cf. below Chapter 13. In (C168.3), Lithuanian represents the C-type. In lieu of the complex expression of the Comitative (= su + N[INS]), the adverb kartu is used which is itself the lexicalised inflectional instrumental of the common noun kartas ‘time, event, occurrence’. English is of course another example of a language which tends to distinguish the two orientations by formal means. However, the English adverb along has a property which it shares with many other similar cases

Across the types 351

throughout Europe: If the accompanee is explicitly mentioned in the sentence with companion-orientation, the adverb cannot function on its own but always combines with the primary relator of the Comitative and/or Instrumental, i.e., for English along with. In actual fact, the occurrence of the marker of companion orientation is often dependent upon the co-occurrence of the primary relator of the central categories. This also holds for Lithuanian, cf. (C169). (C169)

Lithuanian (Balkevièius and Kabelka 1977: 351) einam kartu su man-im X/Y I-Y go:1PL along ‘Come along with me!’

The imperative is a typical context in which companion-orientation can be activated. However, if we look more closely at the example (C169), we realise that this is by no means a regular imperative of the 2nd person but an adhortative of the 1st person plural which is meant as a translation of the Latvian regular imperative phrase nƗc man lƯdz(i) ‘come (along) with me!’. This is discrepancy between the Latvian original and its Lithuanian translation is indicative of another property of markers used for companion-orientation. Lithuanian kartu can have two readings, namely either ‘along’ or ‘together’. This semantic pattern recurs in a variety of other languages as well – among which we find Albanian bashkë, Slovak spolu, Turkish berarber, etc. The relator itself does not distinguish between companion-orientation proper and egalitarian participation: together invites a reading which assigns equal importance to both participants whereas along identifies a hierarchy of strong accompanee and weak companion. Thus, the essence of the two functions combined in these relators is nothing but stating that there is another participant involved in addition to the accompanee. The formal distinction of the orientations is rather strict for the three languages discussed in Sections 12.1–12.3; it is also (almost) compulsory in Finnish and Estonian. German and Dutch display peculiar solutions: In both languages, a mobile verbal prefix is used if companion-orientation is explicitly marked. In German (cf. Section 4.1.1 above), the prefix mit- and the preposition mit are phonologically identical whereas, in Dutch, the verbal prefix comes in the shape of mee as opposed to the preposition met, cf. (C170). (C170)

Dutch (LLP Dutch XXVI.101) en je zult met me mee willen lachen and you will:2SG X/Y me along want:INF laugh:INF ‘And you will want to laugh along with me.’

352 Case studies The mobile prefix mee- belongs to the lexical verb meelachen ‘laugh along with s.b.’. It is also used in situations where the companion is construed as the direct object of a given verb as, e.g., in (C171). (C171)

Dutch (LPP Dutch XXVI.140) ik kan dit lichaam niet NEG I can this body ‘I cannot take this body along.’

mee-nemen along-take:INF

The explicit marking of orientation is not altogether compulsory in every instance. In both German and Dutch, the mobile prefix may be left out and the orientation is thus understood. In other languages, too, the distinction of accompanee-orientation and companion-orientation is less strictly observed. In the majority of our European sample languages, there is no formal distinction at all, meaning: the very same relator is used for both orientations, cf. (C172). (C172) Languages with neutralisation of orientation (C172.1) A-type: Basque (Kühnel 1999: 159) gu-rekin etorri da come AUX.3ABS 1PL-X ‘He has come along with us.’ (C172.2) B-type: Swedish (Karlsson 1978: 220) jag går med pojk-ar-na X/Y boy-PL-DEF:PL I go:PRES ‘I go along with the boys.’ (C172.3) C-type: Welsh (Ifor 14) roedd deng milltir gyda Ifor a Iolo DEC:be:PRET.3SG ten soldier X Ifor X/Y Iolo i fynd i Gaerdydd to go to Cardiff ‘There were ten soldiers along with Ifor and Iolo going to Cardiff.’ In the A-type language Basque (C172.1), the inflectional comitative -(r)ekin is used indiscriminately for both purposes. It is simply a matter of which NP hosts the suffix: if the NP representing the accompanee bears the comitative marker, then this is a case of companion-orientation and vice versa. More or less the same can be said of the PPs headed by med in the B-type language Swedish (C172.2): Depending on the lexical verb and the NP which serves as complement of the preposition, accompanee-orientation and companionorientation may be distinguished from each other – but only in a rather indirect way. In the C-type language Welsh (C172.3), the preposition gyda fulfils

Across the types 353

both functions: It encodes the regular/default Comitative relation (i.e. the one with accompanee-orientation) and is also employed in cases with a clear companion-orientation. In all these languages, the distinctions are neutralised on the level of morphology. The orientation is recoverable from the context, if at all. The following picture emerges: There are two classes of languages, viz. those which make a formal distinction as opposed to those which do not. Both patterns – distinction vs. non-distinction – are attested in the three types A, B and C. Thus, there is no typologically-induced distribution of the classes identifies above. There are however genealogical and areal factors which come into play: Romance languages do not generally have a regular formal device to distinguish the two orientations, nor do Scandinavian languages including Faroese. In the European west, Celtic languages and Basque conform to this pattern of non-distinction. Those languages which keep the two orientations apart by formal means, be it by obligatory rules or with a certain degree of stylistic freedom, are mostly located in the Eastern half of Europe (though not exclusively). The explicit marking of companion-orientation is thus an albeit relatively strong minority solution with a certain regional and genealogical bias. It does not qualify as a Europeme. Its simultaneous attestation in the three languages chosen for the exemplification of the types in the previous sections is attributable to chance. However, this does not imply that we cannot generalise over the structural solutions we encounter in the languages of Europe. Starting from a comparison of (C108), (C138) and (C167) above, we see that the functional domains of the three languages are each unevenly sub-divided. In Maltese, Latvian and Icelandic, there is one relator/expression which dominates either because it covers the greater number of contexts/functions or because the functional domain of the competing relator is a sub-set of the functional domain of the primary marker. This asymmetry seems to hold wherever several expressions are available, no matter whether they are primary markers or other. As far as we can judge from our data-base, the dominant or larger functional domain always involves the relator which encodes the Instrumental either solely or syncretistically with the Comitative. The Instrumental is the category which lends itself to being employed for adverbials of the MANNER, TIME and CAUSE types. Thus, the Instrumentals tend to carry the greater functional load. Interestingly, spatial concepts are not everywhere associated with the Instrumental whereas there are several instances of a closer relationship between Comitativity and spatial con-

354 Case studies cepts.309 Occasionally, the preposition which is involved in the formation of the complex expression for the Comitative has differential casegovernment. When combined with an inflectional case different from the instrumental, the new combination has a spatial meaning, cf. (C173). (C173)

Polish (LPP Polish III.13) spadáeĞ z nieb-a X sky-GEN fall:PAST:2SG ‘Did you fall down from heaven?’

In practical all Slavic languages where case inflection is still operative, the preposition s/z can also combine with an inflectional case other than the instrumental.310 There are two patterns, namely genitive-government and accusative-government both of which occur from Russian via Slovak to Serbian in the bulk of the phylum. In (C173), the Polish preposition z requires the dependent noun to be in the genitive. This combination of z + N[GEN] has a spatial meaning corresponding to the ablative/elative. This spatial function is widespread and is attested in the entire phylum with the exception of the two Balkan Slavic languages, Macedonian and Bulgarian. Superficially, this runs counter to the more general observations made in connection with Section 9.3 above. However, the spatial functions are not expressed by the primary marker alone but by a combination of two different markers which together fulfil a certain task. This is also true of combinations of the inflectional instrumental with other prepositions, a constellation which again recurs throughout the case-inflecting majority of the Slavic phylum, cf. (C174). (C174)

Slovenian (LPP Slovenian IV.27) ali ima njegov oþe kaj father what whether have:3SG something pod palc-em under finger-Y ‘Whether his father has some money (lit. has something under the finger).’

Slovenian allows for several prepositions to combine with the instrumental. These combinations have spatial functions and to a lesser extent also temporal ones. Put the other way round, the Comitative function is based on a very specific combination of elements: s/z needs the inflectional instrumental (and vice versa) when it comes to encoding the Comitative. Every other inflectional case would give rise to a completely different interpretation. Thus, the Comitative is encoded by a complex expression whose compo-

Across the types 355

nents may have deviant meanings outside this combination. Whether or not this is indicative of a closer relation to spatial concepts in general is discussed in Part D. The more extended functional domain of the Instrumental is of course indicative of the advanced stage of grammaticalisation as opposed to the functionally more restricted and a tiny bit less abstract Comitative. Nevertheless, between Comitativity and Instrumentality, there are no absolutely clear-cut boundaries as the various relators may diffuse into the territory of the neighbouring categories. This also applies to the distinction of dominant and sub-dominant functional domains. In Maltese, the borderline cuts across TIME adverbials where the distinction of punctual (marked by bi) and extended (marked by ma’) is crucial. For Icelandic, the split affects the ORNATIVE and is determined by animacy. In Latvian, the boundary is of course only virtual as the bare instrumental is largely a stylistic option. This caveat notwithstanding, we observe that ORNATIVES and CONFECTIVES are treated differently and thus the boundary runs right through the categories which are pre-dominantly NP-internal ones. In general, if a marker is used for the less extended of two functional domains or even restricted to one sole function, this marker tends to be the one which encodes accompaniment of some kind. This includes the animate CONFECTIVE of Icelandic as well. In sum, the European languages behave almost predictably in one respect: Their primary markers are employed in more or less the same way as long as prototypical situations of Instrumentality and Comitativity are involved. Beyond these prototypes, the functional domains of the markers show a number of individual traits. Irrespective of these peculiarities, it is possible to accommodate the various functional domains with the idea of pan-European traits. If we map the functional domains onto one another, a certain pattern emerges: Never are the exponents of Instrumentality and Comitativity confined to the prototypes. There is normally a plethora of other functions some of which are only remotely reminiscent of ACCOMPANIMENT and TOOL in the narrowest sense of the terms. Moreover, if there are two relators (as in A-type and C-type languages), the division of labour between them yields asymmetrically partitioned functional domains. This asymmetry privileges the Instrumental in the sense that it normally covers the larger functional domain. Thus, the Instrumental is also the category which is especially apt to fulfil those functions which are the periphery such as, e.g., MANNER, TIME and CAUSE. The languages differ most in the exact spell-out of the functional subdomains (for instance, the question of whether or not spatial concepts are

356 Case studies included). Diagram (C175) is meant to reflect the general tendency found among the languages of our European sample. (C175) Preferred extensions of functional domains CAUSE

MANNER

INSTRUMENTS

Prototype[TOOL]

N[INANIMATE]

TIME

SPATIAL

NP-internal relations

Prototype[ACCOMPANIMENT]

N[ANIMATE]

The diagram indicates that the prototypical case of Instrumentality may be functionally extended to cover a variety of other functions from less prototypical instances of Instrumentals to adverbials of various kinds. The relator used for accompaniment is less prone to display such a wide range of functions unless it is syncretistic with the Instrumental. Spatial concepts are situated in a kind of no-man’s land where they are accessible for both the Instrumental and the Comitative. Independent of the many individual traits, European languages display a certain degree of similarity. This similarity is not restricted to the prototypical cases but can also be observed on the periphery although they are far from identical even there. One of the issues raised by this similarity in principle which seems to be largely independent of typological factors is a diachronic question: How has this similarity come about? We tackle this problem in Part D where the major lines of the dynamics of Comitatives and Instrumentals are discussed.

Part D Something better change! Origins, life-cycle, contacts: The dynamics of comitatives

13. Grammaticalisation Irrespective of the recently expressed severe criticism of the notion of grammaticalisation (Campbell 2001), we continue to use this handy term throughout this chapter which is devoted to giving a bird’s eye view of the diachronic patterns typical of Comitatives. We start from a laboratory-like constellation pretending that the processes we observe are instances of languageinternal developments. In Chapter 14, this perspective is complemented by a closer look at contact-induced phenomena in the realm of Comitatives. As mentioned above (Chapter 8), Comitatives play an important role in the shaping of ideas within the wider framework of grammaticalisation research and theory starting with Lakoff and Johnson (1980) and culminating in Heine, Claudi, and Hünnemeyer (1991). In the World Lexicon of Grammaticalization Heine and Kuteva (2002: 319 and 329) not only mention the attested etymological sources of the primary markers of Comitatives but also provide a list of targets, i.e., of categories whose expressions derive from erstwhile comitative relators. The latter aspect – the grammaticalisation beyond the stage of the Comitative proper – has been amply documented and discussed in the previous chapters although the concept of grammaticalisation was not always explicitly invoked. It makes sense therefore to focus on the prehistory of Comitatives in Section 13.1. In addition, Section 13.2 contains also a cursory discussion of some points of interest concerning the later stages of grammaticalisation. 13.1. On sources Comitatives do not develop out of the blue. Going by the indications in Heine and Kuteva (2002: 329), there are only three sources from which markers of Comitatives usually originate311, viz. a nominal concept corresponding more or

358 Grammaticalisation less closely to the one instantiated by English comrade alongside two verbal concepts which are at the basis of English to follow and to take (whose semantic range is admittedly rather broad). Two things are remarkable about these etymologies: On the one hand, the mere fact that there are recurrent patterns of grammaticalisation suggest that the Comitative is not an absolutely primitive category in the sense that it always requires a distinct basic and unchangeable expression of its own and, on the other, the number of source concepts is astonishingly small (other categories in the same lexicon display a much wider variety of etymologies). Moreover, one of the two verbal sources looks a bit suspicious as one intuitively associates translation equivalents of English to take with Instrumentals as evidenced in many serialising languages (Heine and Kuteva 2002: 288–289). It is not entirely clear from the examples given in Heine and Kuteva (2002: 287) to which of our three types the languages discussed belong. Chances are that they represent the B-type and thus the use of an equivalent of English to take makes sense in terms of an extension of the domain of the Instrumental to include the one of the Comitative, too. This however is at odds with the idea of unidirectionality of grammaticalisation processes which, in our case, presupposes that a Comitative may be grammaticalised into an Instrumental but never vice versa, cf. below. Tables (D1–2) survey the major source concepts312 for Comitatives and Instrumentals mentioned in the extant literature; we divide these source concepts into verbal and nominal ones (Stolz 1993: 42)313 The target categories may belong to different word classes, most often adpositions or adverbs. (D1)

Verbal sources of Comitative and Instrumental markers

Quoted from Givón (1975: 93; 1979: 261–264)

Sources TAKE/HOLD/USE BE AT/JOIN/MEET

Targets Ins Com BE WITH Com/Ins Paul (1982: 124) USE Ins FOLLOW Com Hagège (1993: 211) USE/TAKE Ins ACCOMPANY/FOLLOW Com HOLD/USE/MEET Ins Heine et al. (1993: 265–272) BE INCLUDED AMONG/FOLLOW/BE Com ONE WITH/ SEIZE TAKE/JOIN Ins Heine and Kuteva (2002: 329, 332) FOLLOW Com TAKE Com/Ins

On sources 359

(D2)

Nominal sources of Comitative and Instrumental markers

Quoted from Heine et al. (1993: passim), Hagège (1993: 214)

Sources COMPANION/COMRADE/COMPANY BODY/SIDE WAY

Targets Com Com/Ins Ins

First of all, it is interesting to see that both verbal and nominal source concepts exist. Even more interesting is the fact that there are source concepts which appear to be ambiguous as to the target categories for which they provide the expressions. We suspect that these multiple target options do not constitute genuine alternatives but are the result of later processes of secondary grammaticalisation i.e. of Comitatives acquiring properties of Instrumentals – meaning: The Instrumental is not directly derived from the source concept but only via the intermediate stage of the Comitative. Thus, these cases of multiple targets are of a different kind than the ones with a single target in the above tables. Furthermore, in one and the same language, several source concepts can give rise to a variety of largely synonymous expressions as in the Dravidian A-type language Kannada, cf. (D3). (D3)

Kannada (Jensen 1969: 40–41)

Comitative o²ane omdige knj²a samgada

Meaning ‘together with’

‘with’

Source o²a omdu knj²a Sanskrit

Meaning ‘union’ ‘one’ ‘company’ ‘allied with’

What the Kannada data also show is that we have to reckon with additional source concepts. BE ONE WITH is mentioned only among the also ran in Heine et al. (1993: 162). In actual fact, the numeral ONE314 recurs all over the place as a source concept of Comitatives: Old Turkic bir-‘one’ forms part of the etymological predecessor of present-day ile ‘with’ in Turkish, birlä (Stolz 1993: 38).315 In Samoan, fa’atasi ma ‘together with’, the reinforced construction of the Comitative, contains the multiplicative numeral fa’atasi ‘once, one time’ derived from cardinal tasi ‘one’ by regular prefixation of multipurpose fa’a-, etc. Notably, the numeral serves especially often as a source concept when it comes to reinforcing the expression used for the comitative (cf. Section 13.2). Estonian ühes ‘together’ is a case in point, cf. (D4).

360 Grammaticalisation (D4)

Estonian (Hetzer 2003: 97) ta lounastab ühes Reinu-ga together Rein-X/Y s/he eat:3SG ‘She is eating (together) with Rein.’

Ühes is the regular inessive singular of the cardinal numeral üks ‘one’ – lexicalised as a preposition governing the inflectional comitative with the meaning of ‘together with’, cf. below. It comes as no surprise to find similar phenomena in the distantly related Hungarian where együtt ‘together’ is based on the numeral egy ‘one’: It reflects a residual inflectional locative -tt (nowadays scarcely used with toponyms) and combines with a noun inflected for comitative-instrumental (Tompa 1972: 45–46). This widespread pattern depicts the relation of the two participants as a close union in which the two entities become one. It is perhaps feasible to subsume such source concepts as ONE, UNION, COMPANY etc. under one higher level concept UNITY (Stolz 1993: 43). Moreover, Haspelmath (1993: 225–226) argues that the two postpositions which are used to express Comitativity in Lesgian are grammaticalized converbal forms of various locative copulas: galaz ‘with’ < gala ‘to be behind something’ and gwaz ‘with’ < gwa ‘to be at’. Consider also the obligatorily possessed relational noun ibb- ‘vicinity’ in Marshallese which functions as the regular Comitative marker in (D5) (Stolz 1993: 46). (D5)

Marshallese (Zewen 1977: 136) Bnjjen ej kwalkoƺ kein mane ibbe-n Rosi tool eat X-3SG.POSS Rosi Bujen 3SG:PROG wash ‘Bujen is doing the dishes with Rosi.’

Stroh (1998) demonstrates that in the course of the history of several Romance languages (French, Catalan, Gascon, Provençal, etc.) relators with a clear spatial function became markers of Comitativity (derived from Latin apud, ab hoc) and only later on integrated Instrumentality in their functional domain. It is clear that spatial concepts also have a say in the genesis of Comitative expressions although this is either not acknowledged in the literature or somehow understood as self-evident. Notwithstanding the rich inventory of source concepts, it is still possible to give a unified account of the origins of Comitatives. (D6) is a necessarily crude attempt to demonstrate that all the various etyma we encounter in the diachrony of Comitatives belong to a tightly-knit network of concepts.316

On sources 361

(D6)

Neighbourhood relations of source concepts IDENTITY

GROUP FORMATION

one accompany

body meet

companion follow

side be near to SPATIAL PROXIMITY

company

friend

be allied

SOCIAL CLOSENESS

Four parameters determine the choice of the concept which serves as a source of the grammaticalised item: Identity (= companion and accompanee are conceptualised as being one), group (= companion and accompanee are conceptualised as forming a higher level unit composed of two independent entities), social closeness (= companion and accompanee are conceptualised as being in a relation of mutual trust and friendship) and, last but not least, spatial proximity (= companion and accompanee are conceptualised as occupying contiguous sub-regions in space). All this boils down to relative closeness of the two participants – ranging from extreme closeness (= identity) to abstract closeness (= be allies). Since the abstract or social kind of proximity is so prominent among the potential source concepts, one cannot simply claim that (D6) lends itself to a straightforward localist interpretation. Space in this sense is not topologically concrete but derived and abstract. However, more tangible forms of spatial relations also count among the inventory of sources of the Comitative and thus the best one can do is postulate a continuum of source concepts oscillating between concrete and abstract spaces, in a manner of speaking. We daresay that Comitatives escape being classified as just another instance of a concrete local case.317 This corresponds nicely with our observation that syncretism of the Comitative with expressions of spatial concepts is mainly restricted to those constellations in which the Instrumental partakes in the syncretistic pattern (cf. Section 9.3 above). If the Comitative was a purely spatial concept, then there is no need for an intermediate bridge-category to serve as a linker between Comitative and other spatial concepts.

362 Grammaticalisation There is of course a general methodological problem with the identification of grammaticalisation paths. Owing to the fact that, for the vast majority of the world’s languages, the time depth of documentation is not sufficient to prove by fact that what linguists assume to be the source is attested earlier than what linguists consider to be the derived category. This is particularly problematic in the case of syncretism: Several categories which otherwise are good candidates for being distinct are expressed by the same morphosyntactic means in a given language. There is no reliable information about the earlier stages of this language – and thus we are forced to interpret the facts in the light of what we know from the history of better documented languages. Palancar (2002: 148) tries to overcome this difficulty by distinguishing between extensional paths and clines: While the former recount the entire diachronic semantics of a given syncretistic expression, the cline constitutes the final step of the process of extension – meaning the relation between a category A and the category B which directly follows A as the next case of semantic extension. This differentiation may prove helpful for instances in those cases in Tables (D1–2) where Comitative and Instrumental seem to share a common source concept: The relation that holds between source concept and Comitative is a cline whereas it is only indirect for source concept and Instrumental, i.e., it is part of the extensional path. The relation between Comitative and Instrumental in turn may be again an instance of a cline. Note however that this distinction of cline and extensional path relies heavily on the hypothesis that synchronically attested syncretism has a diachronic explanation determined by universally valid patterns of semantic development. Moreover, we already know from previous sections that the Instrumental and the Comitative have markedly different preferences when it comes to selecting partners in syncretistic patterns (cf. Chapter 9 above). Space in the narrow sense is much more important for Instrumentals than for Comitatives. The extensional paths and the clines for Instrumentals may thus consist of categories different from the ones reported for the Comitative. Where the two actually meet the evidence seems to speak for a unidirectional derivation of the Instrumental function from a previous Comitative one318, i.e., for a cline Comitative > Instrumental. Heine, Claudi, and Hünnemeyer (1991: 158) claim that wherever there is historical information available, it turns out that the COMITATIVE use preceded the INSTRUMENT use in time, not vice versa (small caps in the original).

Going almost full cycle 363

The diachronic relation between Comitative and Instrumental is unidirectional. If the two categories are diachronically connected to each other in a given language, this connection must be such that the Instrumental function derives from the Comitative one. Various attempts have been made to demonstrate that unidirectionality – one of the essential principles on whose validity the entire framework of grammaticalisation theory rests – does not apply without exceptions (and that the counter-examples call for a thorough revision of the whole approach). As far as we can judge, the counterevidence discussed so far has not shattered the foundations of grammaticalisation theory. This is partly due to the fact that the relevant data are either dubious themselves because they either stem from unattested or reconstructed stages of languages or they are isolated examples which do not join to constitute a full-blown trend in the opposite direction (Lehmann 1995: 16–19).319 All in all, the general lines of argumentation of grammaticalisation theoreticians are corroborated by the observable facts. However, Section 13.2 highlights an especially intriguing case of apparent violation of unidirectionality – a reversal which occurs much too often to be neglected. 13.2. Going almost full cycle In the diachronic development of categories, syncretism, bleaching, phonological reduction etc. may conspire in such a way that an erstwhile thriving grammatical category comes close to vanishing from the system as a distinct entity. This is the predictable course of life of linguistic signs (Lüdtke 1980). However, there are mechanisms which are often activated in order to safeguard the survival or revival of a given category (if such teleological wording is acceptable, at all). Lehmann (1995: 20–24) discusses the concepts of renovation, innovation and reinforcement. The label innovation covers diachronic processes which involve the introduction of a new category which was still unknown on earlier stages of development (this may include the use of inherited means of expression for completely novel functions). Renovation is the term used for the replacement of expression A by expression B both of which fulfil the same function whereas reinforcement presupposes that the original expression continues to be used – but with the addition of further morphosyntactic material. These three process types are also frequently attested for Comitatives and Instrumentals.

364 Grammaticalisation Especially B-type languages display a certain predilection for allowing for a secondary formal differentiation of the various readings their syncretistic relators invite. Consider example (D7) from Estonian. (D7)

Estonian (Hetzer 2003: 97) ta elab selles toas koos kahe õe-ga s/he live:3SG this:INESS room:INESS together two sister-X/Y ‘She lives in this room together with two sisters.’

Koos ‘together with’ like ühes in (D4) above is traditionally understood as a preposition governing the inflectional comitative -ga in Estonian. The word class membership of koos is perhaps less clear as it may also functions as an adverb and thus constitutes a borderline case.320 Be that as it may, what is essential here is the fact that koos and ühes are used in combination with the fully inflected noun. Since the noun itself is already inflected for comitative, the sentences would be grammatically correct even without koos and ühes. Add to these two relators the adverbial kaasa ‘along’ in companion-oriented constructions (cf. Section 12.4), there are three additional morphemes which may be used in the wider context of Comitativity but never in the one of Instrumentality. The inflectional comitative -ga is syncretistic and covers both Comitative and Instrumental – and thus the so-called prepositions serve as disambiguating elements which prevent an Instrumental interpretation (no matter how nonsensical such an interpretation would be). Neither koos nor ühes is obligatory. That is the reason why Estonian still is a good exemplar of a B-type language. However, the fact that it is the Comitative and not the Instrumental which is reinforced by the prepositions is telling. The more so since the pattern is not restricted to Estonian but recurs frequently in Europe and elsewhere. Before we look beyond Estonian, it is in order to scrutinise the history of the inflectional comitative. Habicht (2000: 43–48) summarises the findings of the extant literature on the diachrony of this morphological case. In the earliest written sources dating back to the 16th century, the functional equivalent of modern -ga was the adposition kaas which gradually became a case-suffix via morphologisation during the 17th–18th century. Already early on in the history of Estonian, this adposition displayed a functional domain which corresponds to the one of the modern case-suffix, i.e., kaas fulfilled the full range of tasks of a syncretistic Comitative-Instrumental. Owing to the fact that in the very beginning of Estonian literacy native speakers of German and translations from German played the most important role, it is very likely that the syncretistic behaviour of kaas was facili-

Going almost full cycle 365

tated by the fact that German mit – a typical syncretistic relator – served as a pattern, cf. Section 14.2 below. Leaving the issue of superstrate influence aside for the time being, we agree with Habicht (2000: 44) that on the initial stage of grammaticalisation, the use of kaas was restricted to prototypical functions of the Comitative. This is plausible because kaas, via the expected processes of phonological attrition typical of grammaticalisation (Lehmann 1995), derives from the common noun kansa ‘people, crowd’ whose semantics clearly presuppose a maximum degree of animacy of the participants involved. Note that the Estonian development kansa ‘people, crowd’ > kaas ‘with’ is a parallel of the development of cognate Finnish kansa(ssa) ‘(in the) people/crowd’ > kanssa ‘with’ – except for the important detail that, in Finnish, the denominal postposition is still confined to Comitativity (cf. [A57] above). Since literary Finnish, more often than not, is conservative in contrast to the rather progressive Estonian, it is safe to assume that the contemporary functional range of Finnish kanssa reflects the domain of Estonian kaas in pre-literary times. Having established the origin of the postposition kaas, there remains the problem of determining into which system the new relator intruded. First of all, the introduction of kaas as a grammatical means to express Comitativity and Instrumentality is an instance of renovation because there must have been distinct case-forms for the two functions. No traces of the alleged old inflectional comitative (a look-alike of the Finnish -ine-) remain as vowel apocope and the subsequent weakening of the nasal in final position deprived this case together with the genitive and the instructive of their characteristic morphological forms.321 Kaas entered the scene as a kind of saviour of the Comitative, in a manner of speaking. While there is a functional motivation for the renovation of the old Comitative, the situation is different for the Instrumental because there is no evidence for a similarly precarious state-of-affairs. Occasional residues in the modern language of a formerly more wide-spread use of the inflectional adessive for the purpose of encoding Instrumentality (Tauli 1983: 87–88), suggest three things: – At an early stage, the Estonian adessive had the same functional range as the adessive of modern literary Finnish including the expression of Instrumentality. – In accordance to the unidirectionality hypothesis, the adessive of old literary Estonian could not replace the former inflectional comitative which was far advanced on its way towards disintegration. – Again in accordance with the ideas of grammaticalisation theory, the new adpositional comitative not only underwent morphologisation

366 Grammaticalisation (kansa > kaas > ka > -ga) on the formal level but also extended its functional domain beyond the boundaries of Comitativity to include the entire realm of Instrumentality (to the detriment of the inflectional adessive). Thus, there is good reason to assume that Estonian moved from the A-type to the B-type in the course of the last 500 years (cf. Section 14.2 below). This change of typological class is not terminal. What we observe presently are the incipient stages of a process which, in the end, might cause Estonian to move once again from the B-type to the C-type and perhaps back to the A-type. Presently, the reinforcement strategies optionally serve to disambiguate the Comitative reading from the Instrumental one. There are no equally conventionalised and systematic reinforced constructions reserved for the Instrumental nor is there reinforcement which leaves the syncretistic status unaffected. These are optimal circumstances in which the optional reinforcement might become compulsory and thus give rise to a new obligatory distinct expression of the Comitative. In (D8), the development from pre-literary Estonian via the attested intermediate stages towards the futuristic speculation about a potential state-of-affairs in times to come is outlined. The variable X stands for the inherited Balto-Finnic inflectional comitative, Y represents the inflectional adessive and Z for any form that is identical neither to the former nor to the latter. (D8)

From pre-literary Estonian to science-fiction

Stages Stage I Stage II Stage III Stage IV Stage V Stage VI Stage VII Stage VIII

Comitativity X X ~ Z kaas Z kaas Z kaas

Instrumentality Y Y Y Z kaas ~ Y Z kaas Z-ga koos Z-ga ~ Z-ga Z-ga koos Z-ga Z-ga

Development A-type Optional reinforcement Renovation C-type B-type Morphologisation Optional reinforcement > C-type A-type

There is nothing that prevents the process to continue beyond stage VIII for instance with a remake of what happened on stage IV i.e. the extension of the expression used for the comitative to cover Instrumentality as well. A return to the B-type via the C-type is again a possibility. Despite its speculative undertones, this cyclic model is more than mere fantasy. As a matter of fact, other languages have already gone through more stages than have been attested for Estonian. Consider first a small selection

Going almost full cycle 367

of parallel cases of reinforcement. In a variety of Indo-European languages, patterns of reinforcement similar to the Estonian ones are attested. Generally, reinforcement is most strongly conventionalised for the Comitative (function) whereas the Instrumental (function) either makes do without any reinforcement or only allows for reinforcement in marked registers or styles. In B-type languages such as German, for instance, it is more frequent and stylistically neutral to disambiguate the syncretistic relator mit ‘with’ by means of the adverb zusammen ‘together’ highlighting Comitativity instead of using a strategy which focuses on Instrumentality (most of which have the feel of officialese and written registers). This disambiguation is still largely optional – German is therefore a bona fide case of a Btype language just like Estonian. In Latvian, the southern neighbour of Estonian, the situation is almost the same. There is an optional reinforcement strategy involving the adverb kopƗ ‘together’ which is exclusively used for Comitativity – and it is hard to find anything remotely similar on the side of Instrumentality, cf. (D9). (D9)

Latvian (Doòuleja 224) Alise nožƝloja ka nav PREV:regret:PAST:3 that NEG.be Alise nofotografƝjusies kopƗ ar zƝn-u PREV:take_photo:PTCPL.F:REFL together X/Y boy-ACC ‘Alise regretted that she had not taken a picture of herself together with the boy.’

KopƗ is the regular locative singular of the noun kopa ‘entirety, collective’. In 12.2, we have summarily described the synchrony of the B-type language Latvian. In this sketch, we also mentioned that there is evidence for a different constellation in the pre-literary period because there still are some mildly productive instances of the so-called bare instrumental.322 The observable facts suggest that Latvian has experienced the same development as Estonian, i.e., it moved from the A-type of stage I via reinforcement and class-membership in the C-type to the B-type stage of today. Just in Estonian, there is nowadays the possibility of optionally disambiguating the two competing readings of the PP ar + N[ACCUSATIVE] by reinforcing the Comitative one with the adverb kopƗ. With a view to getting tangible proof for this hypothesis, we have a look at what Lithuanian – the closest relative of Latvian – has on offer. Lithuanian is a C-type language. It has an inflectional instrumental whose functional domain covers all shades of Instrumentality. In addition, there is the

368 Grammaticalisation preposition su which governs the instrumental (su itself or a form related to it, can be reconstructed for Proto-Indo-European). If it comes to expressing Comitativity, su + N[INSTRUMENTAL] is the prescribed solution. The bare instrumental may not be used to express Comitativity323, whereas the PP headed by su is syncretistic, cf. (D10). (D10)

Lithuanian (Ulvydas 1971: 641) su rykšt-e vaiko ne-užmui NEG-PREV:hit:FUT:2SG X/Y stick-Y child:GEN ‘You won’t kill (lit. hit off) the child with a stick.’

Granted that there was a time when the inflectional instrumental was without competitors and Lithuanian or one of its predecessors counted among the B-type languages, we may assume that the preposition su was introduced first for the purpose to disambiguate optionally Comitative and Instrumental readings. At that time, su was associated with Comitativity alone (= stage II). The disambiguation became mandatory and Lithuanian entered stage III as a fresh A-type language. The process has not stopped here: Instrumentality proved to be a strong attractor for the extension of the domain of su + N[INSTRUMENTAL] and thus Lithuanian moved to stage IV to become (once again) a C-type language. In analogy to the relation between conservative Finnish and progressive Estonian, modern Lithuanian often lags behind as to innovative changes when compared to Latvian. Thus, it is legitimate to assume that Lithuanian displays a state-of-affairs which Latvian has long since left behind. This also means that Latvian ar + N[ACCUSATIVE] at an earlier period reflected the distributional properties of su + N[INSTRUMENTAL] as they are reported for modern Lithuanian. In many Slavic languages such as Polish, for instance, there are formally identical patterns: A bare inflectional instrumental coexisting with a PP whose head governs the inflectional instrumental. However, the two constructions do not overlap as to their functions: While the bare instrumental in Polish is confined to the expression of Instrumentality324, the PP z + N[INSTRUMENTAL] is reserved for the expression of Comitativity (Laskowski 1972: 74–75). Polish reflects stage III and is thus an example of the A-type. In the western part of Europe, Welsh is a representative of the C-type. The constellation (Stolz 1998a) attested in Welsh is of the following kind (cf. also Sections 11.2.2 –11.2.3 above): There is a syncretistic relator â/ag and a specialised Comitative relator gyda(g). Historically, â/ag represents the oldest layer as the relator is cognate to the Irish conjunction ac ‘and’ – a fact that points to a common origin in Proto-Celtic times. Gyda(g) however

Going almost full cycle 369

is a relative latecomer to the Welsh system. The formation itself is the product of univerbation of a reinforced construction which consisted of the noun cyd ‘union, junction’ and the preposition â/ag. This reinforcement developed exactly in the same contexts as the ones discussed above, i.e., it served to disambiguate the conflicting readings of the syncretistic preposition by highlighting Comitativity. Currently, gyda(g) seems to be gaining ground to the detriment of â/ag because there are still relatively marginal examples of gyda(g) being used to express Instrumentality. These examples being atypical, it is safe to classify modern Welsh as a C-type language (which might develop into a B-type language if the process of extension of the functional domain of gyda(g) continues). When gyda(g) came up in the 16th century (Evans 1977: 124–125), it began to occupy the position of the inherited preposition gan (a cognate of modern Breton gant ‘with’). In modern Welsh, gan its still around with a high token frequency – but it has lost its erstwhile share of functions in the realm of Comitativity (and Instrumentality). Nowadays, gan is used predominantly in predicative possession and as the marker of the passive agent in the newly introduced personal passive. In spoken Welsh, gyda(g) has also made inroads into the territory occupied by gan in predicative possession although the replacement of gan by gyda(g) to mark the possessor is still an exception in the written register. Gyda(g) renovates the Comitative which formerly belonged to the functional domain of gan. Reinforcement and disambiguation start with Comitativity. This pattern of disambiguation and reinforcement recurs so often325 that we are impelled to interpret it as evidence of the special status of the Comitative: It is marked because it is so often subject to change but it is also constantly renovated. If the Comitative looses its distinct primary marker, secondary means of distinction are almost immediately available to make sure that the meanings of the remaining relator can be disambiguated if necessary. Why is this fact interesting for the proponents of unidirectionality? The widely accepted hypothesis states that a development Comitative > Instrumental is in accordance with the general ideas of grammaticalisation, whereas the reversal of the diachronic schema, i.e., Instrumental > Comitative is ruled out by the same principles. These schemes however simplify the facts because they seem to suggest that the Instrumental does not retain functions of the former Comitative. In actual fact, what usually happens when Comitatives develop into Instrumentals is that the functional domain of the Comitative is extended to also cover Instrumentality. Extending the domain normally means that the original domain is included in the new more extended version, i.e., the new Instrumental is more of a Comitative-

370 Contact Instrumental combining both areas of Comitativity and Instrumentality. Admittedly, the older functions can be lost subsequently but normally there is a phase in which (at least some chunks of) Comitativity and Instrumentality coexist in the functional domain of the relator. Reinforcement and disambiguation of the kind discussed above operate on the basis of a Comitative-Instrumental (and not on the one of a neatly distinct exclusive Instrumental as the grammaticalisation formula pretends). From this basis, a new formally distinct Comitative may arise. Thus, these processes do not violate the principle of unidirectionality, at least not in the way they seemed to do. In lieu of a combination of the processes Comitative > Instrumental > Comitative, we are facing something markedly different, viz. Comitative > Comitative-Instrumental > Comitative. Wherever Comitatives derive their expression historically via reinforcement from those which also express Instrumentality, these expressions are syncretistic per se. To the best of our knowledge, there are no convincing examples of a process leading from a distinct Instrumental to a distinct Comitative by semantic change alone. The attested dynamics of Comitatives prove the hypotheses right which grammaticalisation theoreticians have been defending over the last decade. The unidirectionality principle can be upheld although the recurrence of renovation of Comitatives via disambiguation/reinforcement calls for a revision of the handy schemes used to pinpoint the typical grammaticalisation paths. Becoming an Instrumental is something else than adding functions of Instrumentality to a functional domain. In Chapter 14, we look at a selection of cases in which exclusively language-internal explanation fails to account for the changes that affect Comitatives and Instrumentals. What we observe in language contact largely corroborates the findings from language-internal diachrony. 14. Contact We distinguish two kinds of contacts, viz. overt borrowing and covert convergence. Since the former is a relatively straightforward phenomenon which is almost self-explanatory, we start our discussion of the role of Comitatives and Instrumentals in language contact with a brief treatment of relator-borrowing (cf. Section 14.1). As for Comitatives and Instrumentals, relators are only seldom subject to overt borrowing. What seems to happen more often though is the re-adjustment of functional domains of the inherited relators of language A according to the model of language B – both of

Overt borrowing 371

which have been long term partners in language contact. Several examples of this process of convergence are dealt with in Section 14.2. 14.1. Overt borrowing Spanish discourse particles and other function-words such as, e.g., conjunctions and prepositions have been borrowed by an astonishingly large number of indigenous languages of the former Spanish colonial empire which comprised vast territories in Latin America and Austronesia.326 What strikes the eye most is the fact that the inventories of borrowed items resemble each other no matter where the indigenous languages belong genealogically, areally and typologically. Among the frequently borrowed Spanish words, we encounter the prepositions con ‘with’ and sin ‘without’. In contradistinction to many other cases of function-word borrowing, however, con seems to be largely confined to Mesoamerican recipient languages whereas it is only occasionally found outside this area. In Spanish, con is of course the syncretistic marker of Comitative and Instrumental, cf. (D11). (D11) Spanish (D11.1) Accompaniment (LPP Spanish XXI.9) ven a jugar con-migo327 come.IMP to play:INF X/Y-me ‘Come and play with me!’ (D11.2) Tool (LPP Spanish I.7) logré trazar con un lápiz de color X/Y a pencil of colour manage:PAST.1SG draw:INF mi primer dibujo my first drawing ‘I was able to draw my first drawing with a coloured pencil.’ As a matter of fact, con covers the whole range of Comitativity and Instrumentality with the exception of container-like Instruments where the spatial preposition en ‘in’ is preferred over con. Otherwise, Spanish is a good representative of the B-type as one expects of a Romance language spoken in the European South-West. The Spanish preposition is attested in at least five languages in Mesoamerica, namely the three Otomanguean languages Chochotec, Otomí, and Zapotec alongside Uto-Aztecan Nahuatl and the Totonacan-Tepehuan language Totonac, all of which are spoken in Mexico. In addition, there is one

372 Contact controversial case in South America which we skip for simplicity’s sake: Colloquial Guaraní – the major representative of the Tupi-Guaranian phylum. If we disregard the Spanish-derived con for a moment, the following observation holds: With the exception of Zapotec which displays evident signs of a B-type language (with a leaning towards the C-type), all of these Amerindian languages belong to the A-type.328 This typological classmembership has not been altered through language contact for the original A-type languages Nahuatl, Otomí, Totonac and Guaraní. Zapotec utilises Spanish con in a way which does not interfere with the typological borderline status of Zapotec. Thus, we are entitled to say that the borrowing of con from Spanish has not led to typological convergence with the donor language. Only in the case of Chochotec are the extant descriptions suggestive of a wholesale adoption of the functional domain of the Spanish preposition which turns Chochotec into a B-type language. A small number of examples should suffice to give us an idea of the services to which the borrowed preposition is put in the recipient languages. The use of con in Totonac seems to be rather rare. (D12) is one of the few examples of this preposition in everyday discourse.329 For convenience, we transmorphemise the borrowed preposition as LOAN in the examples from the recipient languages. The primary relators for Comitative and Instrumental used in Totonac are exemplified in (A65.1–2). (D12)

Totonac (Jones 1979: 132) a’xni’ca’ cha’-lh con ix-patron LOAN POSS.3SG-boss when arrive:3SG-PAST ‘And when he arrived at his boss’s place…’

As is immediately evident, Totonac employs the preposition for a function which is not prominent within the functional range of con in the donor language: In (D12), con has a clear spatial function. Despite the looks, the sentence does not mean that the boss is the companion of the subject participant! There is of course the marginal possibility in Spanish to use con with a spatial meaning, viz. in the construction estar con N[+human] ‘to be with N[+human] = to be at N’s place’. In Spanish, this use of con is restricted to static situations. In Totonac however, this restriction does not hold as borrowed con may freely combine with motion verbs – in these combinations, con marks the GOAL of the movement (provided the GOAL is N[+human]). It is clear that this kind of borrowing from a minor segment of the preposition’s functional domain and its subsequent re-analysis and functional extension simply cannot affect the status of Totonac as an A-

Overt borrowing 373

type language. At the same time, the motivation for the borrowing of con lies outside prototypical Comitativity and Instrumentality. Zapotec has an inherited preposition len ‘with’ which is a multipurpose function word covering not only Comitativity and Instrumentality but also beneficiary etc. The examples in (D13) show how len is used to introduce either the COMPANION in an ACCOMPANIMENT situation or the TOOL. This distribution characterises Zapotec as a B-type language. (D13) Zapotec (D13.1) Accompaniment (Butler 1988: 247) yogu6’ ža cheje’ scuel len bi’i che’en6’ school X/Y child POSS.3SG every day go:3SG.HON ‘Every day she goes to the school with her little child.’ (D13.2) Tool (Butler 1988: 247) yagu6 n’ len ya guagu6 n’ gwchogue’ COMPL:cut:PAST:3SG tree:DET X/Y axe:DET ‘He cut the tree with the axe.’ However, as in many other languages of the Otomanguean phylum, Instrumentals may also remain formally unmarked, i.e., they may come in the shape of a direct object of the verb, cf. (D14). (D14)

Zapotec (Butler 1988: 185) na’ ca do chop ši’iljan’ gwdina’ ya guagu6n’ COMPL:hit:1SG axe:DET and as in two time ‘and when I had hit one or two times perhaps with the axe…’

Owing to the fact that len is optional (though highly frequent) with instruments but compulsory in situations of Accompaniment330, one may interpret this as evidence of Zapotec being close to the C-type. The borrowed preposition con has a peculiar distribution, cf. (D15). (D15) Zapotec (Butler 1988: 247) (D15.1) Body part instrument/means of transportation gwyaje’ le’ej con ni’en6’ COMPL:go:3SG Natividad LOAN foot:PL:DET ‘He went to Natividad on foot.’ (D15.2) Reciprocal/adversity ben6’ byo con no’ol6 na’ ch6 s 6’6dile’ and 3PL:CONT:fight person male LOAN female ‘and the men were having a fight with the women’

374 Contact In both cases, Spanish would use different prepositions, viz. a píe ‘on foot’ and contra ‘against’. Incidentally, Butler’s (1988: 247–248) indications suggest that the meaning range of con in situations with two human participants is restricted to (reciprocal) adversity and thus con covers the same contexts as contr ‘against’ – another loan from Spanish (< en contra de). Totonac and Zapotec are comparable: In both cases, con is used in the recipient language not with one of its prototypical functions but rather as a highly specialised marker focusing on less prototypical areas of Comitativity and Instrumentality (although the Spanish equivalents of the verb in [D15.2], pelear/luchar ‘fight’ can also collocate with con). The principle properties of Otomí are already known from our discussion of the facts in connection with (B108–109): The relationship of Companion and Accompanee is predominantly expressed via number and person inflection while the prototypical Instrumental counts among the functions of the so-called circumstantial mood. It is however possible at least for native speakers of Otomí with a certain competence in Spanish to utilise con as in (D16). (D16) Otomí (Voigtlander and Echegoyen 1985: 304) (D16.1) With con cadena bi dąt’i con yÀ IMPERF.3SG fasten:OBJ LOAN DET.PL chain ‘He fastened it with chains.’ (D16.2) Without con cadena bi dąt’ yÀ IMPERF.3SG fasten DET.PL chain ‘He fastened it with chains.’ In any case, the use of con is optional. If it used at all than always in the realm of Instrumentality and never for the purpose of expressing Comitativity.331 In (D16.1), the verb carries the marker for a direct object (= the things being fastened by chains) whereas in (D16.2), in the absence of con, the verb appears in the unmodified base form. As far as we can tell, the introduction of con does not alter the existing typological properties of Otomí as it remains faithful to the A-type.332 What one might suspect as a motivation for the borrowing of con is perhaps explicitness because the circumstantial mood covers a wide range of meanings (Voigtlander and Echegoyen 1985: 224–229) and con characterises one of these functions (= Instrumental) as a distinct category. However, the validity of this assumption cannot be verified in this study.

Overt borrowing 375

In Chochotec, the preposition kù (< Spanish con) has taken over the functions of the expressions used in pre-contact times of which there are no traces left, i.e., their system would have to be reconstructed via comparison with the sister-languages, cf. (D17). (D17) Chochotec (Veerman-Leichsenring 2000: 80) (D17.1) Accompaniment suá rí cuá rí kù CORR come CORR LOAN he ‘He is coming (along) with me.’ (D17.2) Tool rí kù ngnj bËárxà CORR LOAN a hit.PAST ‘He hit with a stick.’

má I

rí CORR

ndà stick

The borrowed preposition has been fully integrated into the morphosyntax of Chochotec that it is also univerbated with a small number of verbs of whose stems it now forms the final component and thus carries the inflection (Veerman-Leichsenring 2000: 51), cf. (D18). (D18)

Univerbated verb-forms

Translation ‘to carry’ ‘to fetch’ ‘to play with’

3rd person dì-kù þí-kù citá-kù

2nd person dì-ku-à þí-ku-à citá-ku-à

Subject 1st person dƯ-ki-á šƯ-ki-á cƯtá-ki-á

collective dƯ-ku-í šƯ-ku-í cƯtá-ku-í

These forms are reminiscent of similar structures in Zapotec where the inherited preposition len may also be incorporated in the verb to form a new stem: chej-len ‘to accompany (lit. go-with)’, cho’e-len ‘to have a conversation (lit. talk-with)’ and ch6 s 6’6 dil6-len ‘to have a fight (lit. strugglewith)’ (Butler 1988: 253). This incorporation of the preposition seems to be possible only for Comitative functions. In Chochotec, chances are that bound kù is repeating the history of an inherited marker which has been ousted by the Spanish loan element. In Nahuatl, the A-type system of the classical language described in (A8) above is still in vigour. Modern varieties usually distinguish Comitative wan ‘with, and’ from Instrumental ica ‘with, by’. Occasionally, Spanish-derived con appears on the scene as well. However, it is attested mainly outside the prototypical contexts of Accompaniment and Tool, cf. (D19).

376 Contact (D19)

Nahuatl (Lastra 1980: 104) o-k-ilpi kon ANT-OBJ.3SG.DEF-fasten LOAN ‘He fastened it with difficulties.’

trabahos difficulties

In (D19), con is used to introduce a Manner adverbial. As the complement of the preposition is itself a noun borrowed from Spanish (< trabajos ‘work’), it is likely that the entire collocation was borrowed en bloc and not the component parts separately.333 Note however that in Spanish, the collocation would normally be con dificultad ‘with difficulties’. One cannot expect occasional borrowings of such quality to trigger major changes of the system of Comitativity/Instrumentality. Of all the Mesomerican languages surveyed in this section, only Chochotec has undergone a serious Hispanicisation in the sense that it not only employs the Spanish-derived relator but also has become a B-type language like Spanish. It must be borne in mind though that, irrespective of the many homologies that exist in principle, the extent to which function words are borrowed from Spanish and integrated into indigenous discourse may differ considerably from one local variety to the other, from one individual to the other, and from one communicative act to the other. The above scenarios invite the interpretation that even though morphological material is borrowed from a prestigious language, the systems of Comitativity/Instrumentality appear to be largely immune to contact-induced change. Discounting Chochotec for once, what changes there are constituent marginal phenomena only. However, this idea of an apparent immunity to external pressure has to be revised if we look at the covert changes triggered by language contact. 14.2. Contact-induced typological change As our discussion of the differences between closely related languages in Europe (cf. Section 13.2) suggests systems of Comitativity/Instrumental like all other linguistic phenomena are subject to change in the course of time. If they were not, there should be no structural dissimilarities between languages of the same genealogical stock. Generally speaking, changes may have two driving forces: On the one hand, they may be determined by language-internal factors (which in turn are functionally explicable as governed by the communicative purposes and needs of the language users), on the other hand, in multilingual settings, the patterns of one of the languages

Contact-induced typological change 377

which are involved in a contact constellation may prove to be an attractor for the other languages and thus the systems of the latter are re-shaped according to the foreign model. All this can happen without any overt borrowing of form, meaning: Those systems which experience change continue operating with inherited morphosyntactic expressions. As to grammaticalisation processes, Heine and Kuteva (2005) convincingly argue that external and internal forces often conspire when it comes to contactinduced changes. Under external pressure, languages may introduce certain innovations which, on closer inspection, only reflect an advanced stage of a logically possible grammaticalisation process of the very same structures which changed according to the patterns of the model language. Among their case studies, Heine and Kuteva (2005) repeatedly discuss examples of Comitatives and Instrumentals. In the remainder of this section, we take up the issue of covert changes of systems of Comitativity/Instrumentality in the languages of Europe because this is the continent whose linguistic history is the most accessible world-wide. If we go by the oldest records of languages spoken on European soil, it is reasonable to assume that in ancient times the distribution of the three language types was markedly different from the one we observe today. There must have been a period when the A-type or the C-type dominated although we do not maintain that this putative dominance of the A-type or C-type is the starting-point of everything. A/C-type dominance is a possible constellation which has a pre-history of its own and is not stable over time. For the oldest stages of the Indo-European languages, the neogrammarians assumed a common syncretic Comitative-Instrumental which came in the shape of the so-called inflectional instrumental.334 Early on the history of the various branches of the macro-phylum, many of the individual languages developed strategies which disambiguated the Comitative reading from the Instrumental one and eventually led to the grammaticalisation of a distinct expression for the Comitative. As long as the bare inflectional instrumental retained parts of its former functions in the realm of the Comitative, the languages remained in the C-type – and later, after the inflectional instrumental lost its ties with Comitativity completely, became members of the A-type. This process is relatively easy to follow on the basis of the documented history of Indo-European languages: wherever disambiguation and reinforcement took place, they affected the Comitative and, in the vast majority of cases, the new grammaticalised expression of the Comitative involved a preposition governing the instrumental (or one of the inflectional cases which succeeded the erstwhile instrumental) on its (pro-)nominal com-

378 Contact plement (Brugmann and Delbrück 1911: 518–547). For the two classical languages of Europe, Latin and Greek, the evidence speaks in favour of their C-type membership in the pre-Christian era, cf. (D20) and, for Latin, Chapter 2 above. (D20) Greek Latin

Classic cases of C-type membership Comitativity (syn +) N[dative]/metà + N[genitive] cum + N[ablative]

Instrumentality N[dative] (cum +) N[ablative]

In both languages, one of the two reported constructions may be used for both Comitativity and Instrumentality whereas the other one is confined to only one of the two functional subdomains. Interestingly, the distribution is not identical: In Greek, the PP headed by syn is not absolutely compulsory and thus the bare inflectional dative may also be used for Comitativity whereas, in Latin, the PP has already been making inroads into the territory of the bare inflectional ablative (Bornemann and Risch 1978: 193–206). The earliest records of other Indo-European languages of Europe attest to similar patterns: Old Irish and Gothic advanced on the same path as Latin where the use of the bare instrumental is already severely restricted whereas the PP dominates almost everywhere (Lewis and Pedersen 1961: 162–164; Braune and Ebbinghaus 1973: 125). In Old Church Slavonic, the grammaticalisation of the PP had not yet led to ousting of the bare inflectional instrumental rather, the distinction between Instrumental without and Comitative with preposition is robust. Thus, at the beginning of the Middle Ages, many IndoEuropean languages seem to have been on their way from an earlier A-type membership to the B-type via a transitional stage during which they formed part of the group of C-type languages. Some languages completed this change earlier than others. Still others have not progressed too far yet whereas a small number of languages – among them also non-Indo-European ones – are probably preparing for a second round on the course from A to B via C and back. The observations made by Braune and Mitzka (1959: 180) about the instrumental in Old High German are symptomatic for the situation in a variety of not necessarily closely related languages335 (the developments reported for Old Norse/Old Icelandic, Old Saxon, Gothic etc. follow the general lines of Old High German scenario; differences mainly occur on the lower levels): – The inflectional instrumental is only attested for masculine and neuter nouns; it is absent from the paradigm of feminine nouns.

Contact-induced typological change 379

– In the 8th/9th century, the inflectional still has a relatively high text frequency. – Only in very beginning of written documentation does the bare inflectional instrumental occur; already early on, so-called prepositional support by mit ‘with’ is more frequent – by the 10th century, mit + N[dative] replaces mit + N[instrumental]. – Nominal word-forms inflected for the instrumental are occasionally attested into the 11th century. – Later these forms become lexicalised adverbials. – The inflectional instrumental survives longest (well into the Middle High German period) in the paradigm of a small set of pronouns where it takes over the functions of the erstwhile dative and is synchronically the dative of modern High German. At the turn of the 9th and 10th century, Old High German had completed the change from B-type via C-type and A-type back to the B-type. Romance languages must have been several centuries ahead of German as it is reasonable to assume that most features of the A-type and C-type disappeared from the emerging languages when the erstwhile case systems of Latin disintegrated and the inflectional ablative could no longer fulfil the task of encoding the Instrumental. In Greek, the old dative was lost from the case paradigm and thus the basis of the prepositionless expression of the Instrumental disappeared in the early Byzantine period, i.e., between the 6th and the 11th century (Adrados 2001: 230) leaving metà + N[genitive] (later > me + N[genitive]) as the sole (and syncretistic) survivor of the classic period. Brythonic Celtic gave up case inflection already before the earliest sources of Old Welsh were produced and thus the prerequisites for a system which operated on the basis of reinforcement were no longer fulfilled (Morris-Jones 1955: 6) while Goidelic has preserved an admittedly reduced inventory of case distinctions which are still functional even nowadays. With the notable exception of Old French where various renovations of the Comitative occurred alongside relatively stable distinct expressions of the Instrumental (Stroh 1998), the bulk of the Romance languages has been in the B-type for their entire documented history. Viewed from this angle, the spread of the B-type in early Medieval Europe could be understood as a wave of innovation triggered by RomanceGreek-Germanic-Celtic language contacts. Tempting as it may seem this hypothesis must remain speculative because we lack substantial proof of covert borrowing or calquing of the necessary kind. Moreover, there is nothing that precludes internally-motivated parallel developments especially because the development Comitative > Instrumental (cf. Section 13.2 above) is

380 Contact taken for granted not only by grammaticalisation theory. If this change is part of the evolutionary program of the categories involved, in a manner of speaking, there is of course no reason for surprise when we see what we expect to see. On the other hand, it is indeed remarkable that so many languages undergo the same kind of change at almost the same time. The emphasis is on the attenuating adverb “almost”. Owing to the unique nature of the earliest documentation phase of the languages under scrutiny, we are not in a position to take the chronology of their first texts simply at face value and interpret it in terms of diachronically successive steps of development. Nevertheless, the possibility cannot be excluded that the relatively late attestation of a phenomenon reflects a relatively late introduction into a given system, although this cannot be made into a rule. Be that as it may, we cannot conclusively solve the chronological problems for the first millennium of the Christian era. There are however chances to come to grips with the spread of B-type features in Europe when we look at subsequent developments which started in the early modern times (from the late Renaissance onwards). We being with the assumption that languages with a common genealogical background should ideally share the same pre-conditions for the dynamics of their systems of Comitativity/Instrumentality, i.e., their common ancestry presupposes a common original state of affairs from where the subsequent developments of the individual languages started. What this means is that if two relatively closely related languages display similar traits in the area of Comitativity/Instrumentality, this might be evidence for common heritage, whereas dissimilarities suggest that at least one of the languages compared has developed in such a way that it deviates from the pattern all of the languages shared at an earlier stage. This is of course a simplification of the facts as we will see soon. However, this schematic approach saves us the trouble of finding the often unattainable philological proof of type-membership of a given language at a given point in time. In Old Church Slavonic – attested from the mid-9th century onwards – the noun declension comprised an inflectional instrumental with a wide array of functions among which the expression of Tool is prominent. However, for the expression of Comitativity (especially Accompaniment) it was already common at that time to use the inflectional instrumental together with the preposition su (Trunte 1990: 130–131). In the earliest phase of documentation, Old Church Slavonic was still a C-type language as the bare inflectional instrumental was occasionally used to express Comitativity. On the later stages, the functional domains of the PP and the

Contact-induced typological change 381

bare instrumental were neatly separated with the latter being restricted to Instrumentality and the former to Comitativity. With minor modifications, this pattern is preserved in the majority of the modern Slavic languages (West Slavic, East Slavic and several South Slavic languages). Only a minority of the Slavic languages differ more substantially from the Old Church Slavonic regulations. Bulgarian and Macedonian have lost their case inflection and thus the inflectional instrumental disappeared for good in the course of their history. What remained of the Old Church Slavonic system are the successors of the preposition su which came to take over the functions of both Comitativity and Instrumentality. These South Slavic languages are presently B-type languages. One could understand this development as nothing particularly remarkable because the loss of case distinctions is one of the best known phenomena of diachrony. This could happen anywhere at any time as similar processes in Brythonic Celtic, Vulgar Latin, mainland North Germanic, West Germanic (without German) etc. amply demonstrate. However, the changes which have affected the southernmost Slavic languages also contribute to convergence with their contact-partners in the Balkan Sprachbund: Albanian, Greek, Romanian and Turkish too belong to the B-type. Middle Greek and (Proto-)Romanian (as developed from Balkanic Vulgar Latin) had completed the change from A/C-type to the B-type several centuries before the two South Slavic languages started to follow their lead. Thus, there is the possibility that what happened to Macedonian and Bulgarian was at least partly reinforced by external influence. External influence is an even more convincing argument in the two other cases from the Slavic phylum which diverge from the old pattern. In Sorbian and Slovenian, the inflectional instrumental is still alive but always occurs in combination with a preposition. For Sorbian, Lötzsch (1996: 56) observes that it has experienced very strong Germanic influence which he holds responsible for the change of both varieties of Sorbian from the original A-type to the B-type, cf. (D21) where the ubiquitous preposition z cooccurs with the inflectional instrumental. The parallel facts from Slovenian have been discussed already in Section 11.2.2 above.336 (D21) Upper Sorbian (Lötzsch 1996: 56) (D21.1) Body part instrument ja dĨČáam z ruk-u X/Y hand-X/Y I work:1SG ‘I work by hand.’

382 Contact (D21.2) Reciprocal ja rČþu z X/Y I speak:1SG ‘I talk with/to a friend.’

prČüel-om friend-X/Y

The thorough Germanicisation of the Slavic territory between the rivers Elbe and Oder began in the late 10th century and became especially intensive from the 12th century onwards. Extensive bilingualism with German became the rule for many native speakers of Slavic varieties, i.e., the probability of contact-induced change was very high (Lötzsch 1996: 51). At the same time, it is also clear that the B-type features were first acquired by Old High German (cf. above) and then passed on to Sorbian via Middle High German. A largely similar scenario can be postulated for Slovenian: The territory inhabited by this speech-community came under the influence of Germanic and Romance speaking groups at almost the same period as Sorbian. The importance of the Germanic impact for the shaping of Slovenian is evident already in the earliest texts from 10th/11th century which were written in a peculiar way reflecting the conventions of Old High German orthography (Svane 1958: 16). For about a millennium, Slovenia remained under direct control of Austria and German had ample opportunity to leave its mark in Slovenian. One of the results of these contacts is the adoption of the crucial B-type patterns in Slovenian. Thus, the two Slavic languages which are spoken on the Germanic-Slavic border and have been subject to most intensive influence on the part of German (and some Italian in the case of Slovenian) are also the ones which changed from the A-type to the B-type whereas the larger Slavophone communities, especially those less exposed to German influence because of numbers and geographical distance, have remained faithful to the A-type. Bulgarian and Macedonian, on the one hand, and (Upper and Lower) Sorbian and Slovenian, on the other, have become B-type languages at least partly because the languages of their prestigious and powerful neighbours (adstrates and superstrates) happened to be B-type languages. The driving force is areal convergence, the models after which the Slavic systems were re-shaped are not identical: Greek, Romanian, Albanian and Turkish influenced the Balkan Slavic languages whereas German provided the pattern for Slovenian and both varieties of Sorbian. In any case, co-territoriality of languages and multilingualism helped the diffusion of the B-type beyond the original boundaries. The diffusion is even more tangible in the case of the linguistically diverse Circum-Baltic region (Stolz 2001c).337 Given that the members of the Ger-

Contact-induced typological change 383

manic phylum were firmly established in the B-type by the turn of the 11th/12th century, the following scenario gains in plausibility: – In a period immediately pre-dating the advent of Germanic speaking conquerors, the Balto-Finnic languages338 and probably Proto-Saami too belonged to the A-type.339 – The languages of the Baltic branch of Indo-European, spoken in territories directly adjacent to the ones occupied by Finnic-speaking peoples, most probably also represented the A-type. – From the 12th century onwards, the presence of speakers of Germanic languages (Old Swedish, Old Danish, Middle Low German) on the eastern shores of the Baltic began to have an effect on the languages of the per-Germanic autochthonous population. – When the first written documents in the Baltic and Balto-Finnic languages were produced in the 15th/16th century, the situation was as follows: – Finnish proper conformed to the A-type;340 – Estonian was already a B-type language (Habicht 2000), cf. Section 13.2 above; – Latvian displayed remnants of a former C-type stage (GƗters 1993: 161–183);341 – Lithuanian was still an A-type language (Schmalstieg 1988: 243–257 and 294).342 – During the incipient stages of literacy, writing in the autochthonous languages of the Baltic region was predominantly exercised by clerics with a German/Swedish cultural and linguistic background and thus the possibility cannot be counted out that these churchmen inadvertently introduced Germanicisms into the local languages exactly during the important formative years of an early form of written Estonian, Latvian etc. – Latvian became an AND-language via the borrowing of the coordinating conjunction un from Low German un(d). The modern preference for the use of ar in contexts where high animacy of the participants represented by the conjuncts is involved is evidence of a former WITH-language status. – The fact that Latvian and Lithuanian make use of two etymologically different prepositions ar and su points to the possibility that the respective constructions developed independently of each other, whereas Finnish, Estonian and semantically also Lule Saami draw on one and the same grammaticalisation channel to develop new markers of the Comitative.

384 Contact – Subsequently, what remnants there were in Latvian of a former Ctype stage underwent further marginalisation and Latvian became a full-blown member of the B-type. – During the same period, Lithuanian acquired C-type features by way of allowing certain facets of Instrumentality to be encoded optionally by su + N[instrumental]. This relative chronology shows how the B-type features introduced via language contact with Germanic slowly intruded in the autochthonous languages of the Baltic region. It is noteworthy that the earliest records we have of these languages – with the exception of Finnish – display a similar situation as the one given for Old High German above: The best one can get are relics of a different structure of the system of Comitativity/Instrumentality at an earlier stage. This is also indirectly indicative of a relatively late change because the same state of affairs was reported half a millennium earlier for Old High Germanic. Finnish has escaped the typological change induced by contact with Germanic languages. However, Nau (1995: 134–135) discusses a variety of examples which attest to some recent processes in modern Finnish (and Swedish-influenced old written Finnish too) which suggest Swedish influence because they have not become integrated part of common colloquial Finnish but are strongest in those areas where the Swedish minority in Finland has its strongholds.343 Another case of diffusion of B-type features is described by Haase (1992: 67–71) for colloquial varieties of Basque which have been under Romance influence – especially by French and Gascon. Standard unified Basque is a typical representative of the A-type (cf. [B83] above). Diatopically, Basque is fragmented into a number of regional varieties which form the basis of spoken colloquial Basque. From the second century B.C. onwards, Basque must have been exposed to heavy influence from Latin as the many Basque words with a Latin etymology suggest. When the individual Romance languages developed from Vulgar Latin, Basque was surrounded exclusively by B-type languages – a situation which has lasted for centuries and is still the same today. The B-type features of the Romance languages are transferred to colloquial Basque via interference of multilingual speakers. What is especially interesting is the fact that the spread of Btype features involves the expansion of the functional domain of the Basque inflectional comitative to the detriment of the instrumental and not vice versa. Haase (1992: 70) explains this preference for the comitative in terms of the higher analysability of its forms which still resemble postpositions to some extent. If the comitative can be “understood” as a kind of postposi-

Contact-induced typological change 385

tion, says Haase (1992: 70), then it is also possible to equate it with Romance preposition in analogy to other Basque postpositions which are treated as functional equivalents of Romance prepositions. This is perhaps only a side-aspect of the process pace Nau (1995: 138). What seems to be more important is the fact – also mentioned by Haase (1992: 70) that the functional domain of Basque -ekin i.e. the comitative marker is fully included in the functional domains of the Romance counterparts such as, e.g., Gascon dab or French avec, whereas the Basque instrumental -z has a much wider domain in which we encounter numerous functions which are absolutely alien to the Romance prepositions. Moreover, Comitativity is said to constitute the prototypical segment of the functional domains of the Romance prepositions. In other words, where the comitative of colloquial Basque diffuses into the territory of the instrumental, these new uses of the comitative correspond with those areas of Instrumentality which are also covered by the prepositions of Gascon and French (and Spanish, for that matter). We summarise the various cases of areal diffusion in map (D22). The broken lines indicate two different things: either it is unclear whether there was contact-induced transference of the features or the impact has not been strong enough to trigger a wholesale typological change to the B-type (Finnish, Basque allow B-type features in their non-standard varieties only and Lithuanian, Welsh have become C-type languages). The pattern of diffusion corresponds nicely with the general ideas of Haspelmath (2001) about the genesis of the European linguistic area in which Germanic and Romance played the most important roles. However, this is not the end of the story: The map clearly shows that it is only one type that is actually expanding its territory, namely the B-type. In the catalogue of case studies discussed in this section, there is not a single instance of the reverse process triggered by language contact. Admittedly, the diffusion of B-type features might yield an increase of C-type languages. Nevertheless, the first impression one could get from looking superficially at the map is that there is something about the B-type itself which facilitates its transferability in language contact situations. This idea is not altogether wrong, but a caveat is in order: It is a historical coincidence that the languages spoken by those people who gained cultural, military and/or economic supremacy over others in the Middle Ages and early modern times were representatives of the B-type. Chances are that if they had been A-type languages their features had spread nevertheless. Since changes from one type to the other and back

386 Contact can be ascertained for many individual languages, there is nothing that impedes the contact-induced genesis of a new A-type language. However, we do not deny that the pattern of the B-type lends itself more easily to borrowing because the syncretism transferred reflects the general semantic potential of the relators. Given that grammaticalisation theory is right in postulating the unidirectional development Comitative > Instrumental, syncretism is the only logically possible next stage for a relator in a language which does not belong to the B-type yet. Moreover, this observation does not apply to any marker of Comitativity or Instrumentality. Rather, it is only the Comitative which has this potential because, in A-type languages, the Instrumental should ideally be confined to Instrumentality – from where, as we know, there is no purely semantic way back to expand the domain to include Comitativity.344 Thus, it is unsurprising to find that wherever we can be sure of contact-induced change towards the B-type it is the marker of Comitativity which is generalised. (D22)

Spread of B-type features in Europe Saami Finnish Estonian Latvian

Welsh

Lithuanian Sorbian Germanic

Basque

Slovenian Macedonian/Bulgarian Romance

Greek

Turkish

This is the end 387

15. This is the end… In the previous chapters, we have highlighted as many aspects of the Comitative as possible. This attempted comprehensiveness notwithstanding, there surely remain areas which we hardly touched upon or even overlooked completely – either because of ignorance or because time and space limitations did not allow us to introduce and elaborate on more topics. We regret for instance that, for various reasons, we had to skip practical everything that is connected to Comitativity in the wider realm of verb grammar including converbs, subordination and clause-combining. Thus, what we have said so far might turn out not to be the final word on our subject matter. Future investigations by us or others will reveal how much more the Comitative has on offer for linguistics. Nevertheless, our study is the first detailed typological treatment of this issue which goes beyond the usual size of a journal article. We have attempted to cover the most important controversial issues of Comitative research. In this final chapter, we start with a necessarily short review of our major arguments and findings. For obvious reasons, we have to leave aside numerous side issues and details for which the reader is referred back to main body of this monograph. For the same reasons, we also abstain from quoting and citing our sources and references upon which we have drawn extensively in Chapter 1 through 14. In Part A, it was shown that methodology is crucial. Many of the points where students of Comitatives disagree can be accounted for in terms of different methodological frameworks and science-philosophical convictions. It actually makes a difference whether one looks at the system of language from the point of view of the linguistic sign as a unity or from one of its components. If one goes by semantic criteria, the outcome of our analyses must of necessity be different from the one which is produced by those who adopt a form-first approach to the problem. The most challenging task consists in not to take apart the linguistic sign and achieve valuable results nevertheless. We argued that the slogans according to which taking form seriously is a must require the strict application of the form-first criterion and not an alternation between form and function in turns dictated by convenience or preconceived ideas. Methodological strictness however leads us to the insight that extant studies of the Comitative or other categories have shunned the ultimate consequences: – If form is the starting point, then the smallest meaning-bearing unit used to encode the Comitative must be taken into consideration chiefly.

388 This is the end... – If it is taken into consideration, the full range of its distributional properties has to be accounted for, too. The usual picture however is different mainly because of two reasons: On the one hand, syntactic form more often than not is taken as hierarchically superior to morphological form and thus the latter is believed to provide secondary evidence at best, and, on the other, markers display distribution profiles which go beyond or do not exhaust the area of grammar about which the linguist is eager to say something relevant. Therefore, research into the nature of the Comitative is often cut short when the danger is imminent that form-function mapping is not as neat as expected. Two paradoxes arise: First, if one kind of form overrules the other, why bother at all with the inferior one at all; second, if the result of form-function mapping is clear beforehand, why pretend to follow the lead of form in the first place? The intra-participant relations we postulate are a case in point. One encounters numerous instances of Comitative/Instrumental relators being used NPinternally – and thus falling outside the scope of most of the studies devoted to semantic case. However, if form guides the observer then intraparticipant relations have to be accounted for as well when it comes to determining the Gesamtbedeutung of the Comitative/Instrumental. Note also that both the presence and the absence of the relevant relators in the realm of adnominal relations may be typologically meaningful. We admit that it is much easier to criticise the usual practice than to go to the trouble of showing the ideally-conceived methodology is feasible. As a matter of fact, we are in a fix. We insist on the assumption that only a full distributional analysis can reveal the semantics of the Comitative to us, but, at the same time, absolute comprehensiveness of corpus-independent distributional analyses is illusory for an individual language and technically impossible when it comes to comparing languages (no matter how small or large the sample). All occurrences of the same morphological (or morphosyntactic) expression are taken to be indicative of a content- or functionbased connection of the various contexts or constructions in which it is attested. The sum of these contexts and their common functional denominator define the “meaning” of the Comitative in a given language. For typology and universals research, the sum of the sum of contexts and the common functional denominator of the common functional denominators of all sample languages provide us with the necessary clue which helps us postulating universals. Regrettably, humans are not perfect and so are linguists. We must make do with what can be done within a lifetime. Meaning: There are technical problems which impel us to operate with generalisations, in-

This is the end 389

formed guesses, hypotheses etc. on the basis of more or less extended fragments of distribution profiles. As long as we acknowledge that our postulates thus remain hypotheses themselves, this lamentable state of affairs is not as devastating as is sometimes insinuated. However, we must bear in mind that if, under these circumstances, one comes forward with the semantic map of the Comitative or the like, what he or she presents merely zooms in on a segment of a potentially much larger map. Comitatives of this kind represent selected aspects of the Comitative. Since the Sisyphean workload of mapping literally hundreds of distribution profiles of Comitatives and Instrumentals for our sample languages unto one another would have overstrained us immensely, viability induced us to fall back upon the usual crosslinguistic methodology in Part B. Given that the exact semantic extension of Comitatives and Instrumentals of individual languages is beyond our reach to determine, we still are in a position to use the prototypes and primary markers to carry out large-scale comparison of languages. The guiding question was whether or not Comitatives and Instrumentals are two of a pair. With a view to finding an answer to this question, we made the most of a critical review of the supposed universal Companion-metaphor according to which expressions of prototypical Comitatives and Instrumentals must be identically. This assumed universal syncretism of Comitative and Instrumental was disproved by a thorough statistical evaluation of the data drawn from 323 languages world-wide. Our check revealed various important facts: – In global perspective, Comitative-Instrumental syncretism is a minority solution whereas languages clearly prefer to keep the two functions distinct by formal means. – The only area where Comitative-Instrumental is statistically overrepresented is Europe and thus the supposed universal turns out to be an areal preference, a Europeme/Euroversal. – In addition, there is a certain genetic bias which shows that Comitative-Instrumental syncretism clusters in Indo-European languages especially in the western half of the European continent and thus, this syncretistic pattern is also typically attested in the SAE-languages. We have identifies three types A, B and C which differ as to the syncretistic behaviour of their primary markers. Whether the C-type (i.e. the one which could also be labelled mixed) could be dispensed with is a matter of perspective: If we talk about languages, the C-type must obviously be kept as it is based on a particular constellation of structural facts. However, if we

390 This is the end... focus on relators, then the C-type becomes dysfunctional because relators have only two options – either being syncretistic or asyncretic, two choices which do not leave anything in between. Despite the fact that the universal as such is a failure, it nevertheless reflects a relatively frequently recurring syncretistic pattern. However, Comitatives and Instrumentals display an intricate network of largely disjunct domains of syncretism. Instrumentals go together well with categories bearing the feature [+agentivity] or [+locative] whereas the Comitative is more prone to syncretise with the AND-conjunction or the marker of possession. Wherever there are syncretistic patterns which seem to disobey these rules, Comitative and Instrumental are also syncretistic among themselves and thus function as bridge categories which allow the association with otherwise dispreferred partners in syncretism. These facts add to the two-facedness of Comitatives and Instrumentals: On the one hand, they appear to be categories which are largely autonomous of each other while, on the other, they may join if syncretism exceeds certain limits. It is difficult to capture this vacillatory behaviour in terms of conceptual affinity or averseness. Moreover, it can be shown that Comitatives and Instrumentals tend to interact in a systematic way such that treating them as absolutely dissociated entities would not be appropriate. Both categories are often neutralised under negation. Their formal expressions sometimes bear similarities which suggest at least historical ties. It is even possible to apply the criteria of markedness theory to this pair of categories which identify the Comitative as the marked member of the opposition and the Instrumental as the unmarked one – although many problems remain to be solved. Thus, Comitatives and Instrumentals are neither insolubly tied to each other nor are they complete strangers to each other. As an absolute universal the Companion-metaphor is not tenable, however, it is a very good candidate for the status of a potential universal in the sense that it is one of the obvious options speakers of languages have at their disposal. The competitors are mainly those which conceptually underlie the other syncretistic patterns in which Comitative or Instrumental partake. All this calls for closer scrutiny of that region of the world where languages have a significantly higher leaning towards Comitative-Instrumental syncretism than elsewhere. In Part C, we have looked more closely at the European situation with the idea in the back of our minds to determine how similar languages are beyond the mere fact that they employ syncretistic or asyncretic markers for Comitative and Instrumental. Of course, overall (dis)similarity of languages is a concept which requires a detailed compara-

This is the end 391

tive analysis of distribution profiles. The only way to keep such an undertaking within reasonable limits is a corpus-based approach. We have compared the distribution profiles of the primary markers of Comitative and Instrumental for 64 European languages and regional varieties using the translations of Le petit prince as reference frame. This was done both quantitatively and qualitatively. We discussed the factors which determine the degree of similarity of pairs of languages and in doing so we made the following observations: – Similarity across the board is highest in the case of prototypical functions, i.e., languages tend to employ their Comitatives/Instrumentals in a parallel fashion when the core meanings are involved in contexts where Accompaniment, Tool or Ornative play a role. – Similarity becomes more genetically-induced if non-prototypical functions are expressed – ditto for areality. – Languages may behave unpredictably as individuals again in the realm of non-prototypical functions and the particulars of their employment according to grammatical or stylistic rules. Members of all three types behave similarly in so far as they indiscriminately tend to employ one of their markers of Comitativity/Instrumentality for the prototypical functions. Dissimilarities arise when the functional domains of markers of A- and C-type languages are compared. In general, genetic affiliation proved to be a relatively reliable prognostic criterion. Furthermore, our scrutiny of the European state of affairs revealed a geolinguistic cline which is based on the gradual increase/decrease of properties from the South to the North and from the West to the East of the continent. This asymmetric distribution of properties largely corroborates other studies of the European linguistic landscape which emphasised that there is an areally-based disparity on the West-East axis of Europe. With a view to getting a better idea of the extent similarity and dissimilarity may have among the languages, we reviewed one representative of each of the three types in separate case studies independent of the textual basis used for the corpus analysis. It must be borne in mind that the three languages chosen for these case studies are individuals, meaning: There are representative of their type only to the extent that generalities and higherlevel phenomena are involved. Besides the expected dissimilarities, the case studies also identified a number of remarkable similarities across the typological classes. One of the especially interesting phenomena of this kind is the wide-spread distinction of Companion-orientation vs. Accompanee-

392 This is the end... orientation which prior to our investigation seems to have largely escaped notice. This distinction resembles the better known opposition of Agentorientation vs. Patient-orientation – a fact which renders Comitatives even more interesting for a thorough-going linguistic analysis. One of the followup studies to this one could thus focus on the differential orientation in the realm of Comitativity which is a typologically promising subject as the formal distinction of orientations is not attested in each and every language of our sample. Part D looks into the dynamics of Comitatives and Instrumentals. We devoted one main chapter to grammaticalisation proper where we tried to identify the lexico-semantic sources from which Comitatives derive their expressions. Superficially, the etymological sources appeared to be heterogeneous. However, on closer inspection, it turned out that the various source categories all belong to a kind of lexical field or frame model, for that matter, which is based on metaphorical space – though not exclusively. The major differences between Comitatives and Instrumentals as to their grammaticalisation channels are on the one hand the higher degree of tolerance of the latter for spatial concepts and, on the other, the fact that the Comitative seems to draw more often on more strictly lexical sources whereas the Instrumental also makes use of already grammaticalised expressions. This is in line with the general idea that the Comitative is less grammaticalised as opposed to the Instrumental. More particularly, this is also compatible with the often expressed idea that the Comitative may develop into an Instrumental or, the other way round, the Instrumental may develop out of a Comitative. Our findings suggest however that this has to be taken cum grano salis because of two reasons: – For the vast majority of languages of the world, the diachronic pattern Comitative > Instrumental is based on conjecture as there are no philologically reliable documents of earlier stages of the language. – The pattern itself is a simplification because it suggests a full-blown functional change including the loss of all of the erstwhile functions the relator fulfilled in the realm of Comitativity – a suggestion which contradicts the attested facts. We have shown that what actually happens when a Comitative develops into an Instrumental is that at least temporarily the relator is syncretistic and thus functions as a marker of both the Comitative and the Instrumental. Moreover, it is exactly this constellation which more often than not gives rise to the development of a new distinctive Comitative. Very frequently,

This is the end 393

this new Comitative comes into being via reinforcement/disambiguation of the Comitative reading of the syncretistic relator. As a consequence, a grammaticalisation paradox arises: The principle of unidirectionality precludes the development of Comitatives from Instrumentals but reinforcement/disambiguation strategies prepare the ground for the “extraction” of the Comitative from an erstwhile syncretistic Comitative-Instrumental (which in turn is the “Instrumental” in the pattern Comitative > Instrumental used in grammaticalisation research). We demonstrated that this happens time and again in the history of languages. Accordingly, a cyclic development of the types A > C > B > C > A etc. can be postulated. This cycle may be operative with or without external stimuli. In language contact, Comitatives and Instrumentals are of course subject to borrowing and calquing just as the majority of other linguistic phenomena. Where overt borrowing is attested, one should not expect to find the relators to be used with their prototypical functions in the recipient language. Often, they are made use of in the periphery of the functional domain of Comitativity/Instrumentality. Covert contact-induced change however can trigger a wholesale typological change of the recipient languages. We have gathered evidence for a spread of B-type features in the course of the linguistic history since the Roman Empire came down. These features have begun to establish themselves in the languages of the eastern half of Europe only relatively recently (15th/16th century). Since there are evident signs for languages to have gone through the cycle several times either with or without influence by another language, it is safe to say that, all things being equal, there will never be a complete standstill in the system of Comitativity. What this also seems to imply is that none of the three types is more natural than the others. World-wide, the A-type presently dominates – however this contemporary situation is in principle not immune against being turned upside down. Going by the present distribution of features, the A-type is the unmarked option – and it is tempting to try to explain this typological preference of languages in terms of conceptual systems etc. However, this explanation would be panchronic or achronic – meaning universally valid independent of time. What is needed to prove this explanation right or wrong is a compatible crosslinguistic study of a previous or subsequent period of time – wherever possible complemented by data from child language acquisition, creolisation and sundry fields of interest. We are confident that one cannot simply reverse the present distribution and numerical strength of the types for times gone by or the distant future – and this not only because languages do not change at the same speed and within

394 This is the end... the same intervals. There is something about the A-type which qualifies it as an especially good option. However, there must also be something about the B-type that allows B-type features to win out against A-type features every once in a while. Trivially, the range of typological options (= types A, B and C) is so small that the following holds: under the premise that the system of Comitativity/Instrumentality of a given language is subject to change at all and this change is not simply the renovation of the already existing systems, then the change must be either A > B or B > A (each of the two going via a C-type stage).345 In the absence of crosslinguistic studies of two or more different diachronic stages, the best we can do is assume that both A-type and B-type have advantages and disadvantages of a functional nature which determine the choice of the speakers of languages. The A-type focuses on the differences of two categories potentially associated to each other whereas the Btype focuses on the commonality that links the one to the other. The A-type is more explicit, the B-type more implicit when it comes to marking categorical distinctions. In a way, the A-type respects the needs of the hearer as the asyncretic markers do not burden the hearer with two much decoding tasks. On the other hand, the B-type conforms more to the speaker’s interests according to which it is easier in speech production to operate with a small number of rules and categories. Admittedly, the interests of the participants in a communicative act cannot be reduced to the preference for redundancy on the part of the hearer and economy on the part of the speaker. However, the apparently constant change to and fro the A-type/Btype nicely corroborates Gabelentz (1891: 182–183) who in an early functionalist model identified two major forces which determine to considerable extent what can happen in language change: The necessity of making oneself understood (“der Zweck der Verständlichkeit”) and the tendency to reduce the effort in speech-production (“die Neigung zur Kraftersparniss”). A final word about the Comitative itself: This study has started from the assumption that there are several possibilities of approaching the topic at hand. Most of the time, researchers – often tacitly – adopt a restricted definition of what is and what is not a Comitative. Thus, the extant literature overwhelmingly deals with selected aspects of the Comitative leaving other aspects aside. What this amounts to is that under the same label relatively different phenomena are highlighted which only in combination could yield a “realistic” picture of the Comitative. In Part A, we have emphasised this problem by way of demonstrating that the usual practice results in a methodological paradox. However, the prototype model allows us to consider

This is the end 395

some aspects of the Comitative more central than others – and so far, it has been the periphery of the Comitative which has escaped notice by linguists. In the peripheral areas, languages sometimes start to behave idiosyncratically and thus do not lend themselves to being compared with a view of finding common traits among languages. The core area however is characterised by a much higher degree of crosslinguistic comparability as the individualism of languages is drastically reduced leaving only a handful of options which in turn constitute the basis for the identification of types. At the heart of the Comitative there is the task of encoding the relation between Accompanee and Companion in an Accompaniment situation. All other uses to which the expression of the Comitative is put are directly or indirectly connected to the central one by metonymy and metaphor. The further away from the centre, the lesser the degree of predictability for the qualities which are also expressed by the same relator. In the end, this insight is almost a repetition of our initial ideas. This is not trivial because we have reached this conclusion only after a thorough and extensive investigation of (almost) all possible facets of the subject at hand. The Comitative is a permanent borderline case to which the vague characterisation of “more or less grammaticalised” can be applied without doing it to much injustice.

Appendix I Sample languages

Our previous samples The following table gives the glossonyms of the 371 languages which formed part of at least one of our four earlier cross-linguistic samples (marked by x). The language names appear in alphabetical order, alternative glossonyms and/or further specifications of varieties are given in brackets. Sometimes different varieties are distinguished by Roman numbers because our sources were not always explicit as to the proper designation of the variety they described. For the distribution over types, areas and phyla we refer the reader to the original publications (ditto for the bibliographic details of the sources from which we have drawn the data).

Abkhaz Accadian Achagua Acoma Acooli Afrikaans Ainu Alamblak Albanian Aleut Alyawarr Amele Andoke Anguthimri

2

3

4

5

6

7

8

9

10

11

12

13

14

Language

1

No.

Mpakwithi

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

Stolz 1996

X

X

X

X

X

Stolz 2001b

Sample languages 397

Language Apalai Arabic (Egyptian) (Gulf) Arabic Aramaic Aranda Armenian (Eastern) Avar Awtuw Aymara Badiotto Bagirmi Bahuana Baka (in Sudan) Balese Bambara

No.

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

Stolz 1996

X

X

X

X

X

X

X

X

X

Stolz 2001b

398 Appendix I

Language Barasano Bari Basque Belarusan Bikol Binumarien Bisa Blackfoot Brahui Breton Bukiyip Bulgarian Burushaski Cakchiquel Cambodian

No.

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

Stolz 1996

X

X

X

X

X

Stolz 2001b

Sample languages 399

Language Carib Catalan Cemuhî Chemehuevi Chinantec (Lealao) Chinese (Mandarin) Chiriguano Chochotec Chol Chukchi Cocopa Comanche Comorian Comox Coos (Hanis)

No.

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

Stolz 1996

X

X

X

Stolz 2001b

400 Appendix I

Language Coptic Cora Czech Dagur Dahalo Dani (Lower Grand Valley) Danish Desano Dimili Drehu Dullay Dumi Dutch Efik Enga

No.

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

X X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

Kurdish (Zaza)

Degano

X

X

X

Daur

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

Stolz 1996

X

X

X

X

X

X

X

Stolz 2001b

Sample languages 401

Language English Estonian Even Evenki Ewe Faroese Feroge Fijian (Boumaa) Finnish Fon French Frisian Fulani Fur Futuna-Aniwa

No.

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

X X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

Stolz 1996

X

X

X

X

X

X

X

X

Stolz 2001b

402 Appendix I

Language Gadsup Galla Garífuna Gbeya Bossangoa Ge’ez Georgian German Goajiro Gola Gondi Gooniyandi Greek (Modern) Greenlandic (West) Guaraní Guarijío

No.

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X X X X

X X X X

X

X

X

Stolz 2001b

X

X

X

Stolz 1996

Sample languages 403

Language Gumbaynggir Gunin Guugu Yimidhirr Haida Hausa Hawaiian Hebrew (Modern) Hindi Hixkaryana Hmong Ho Hua Huave (San Mateo del Mar) Huichol Hungarian

No.

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

Stolz 1996

X

X

X

X

Stolz 2001b

404 Appendix I

Language Hunzib Hupa Icelandic Ila Imonda Indonesian Inga Iraqw Irish Italian Itelmen Jakaltek Japanese Kaingáng Kamass

No.

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

Stolz 1996

X

X

X

X

Stolz 2001b

Sample languages 405

Language Kambera Kanjobal (Western) Kannada Kanuri Karok Kâte Kayardild Kera Ket Khalaj Khalkha Khanty Kharia Khasi Khoekhoe

No.

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

Nama

X

X

X

Mongolian

X

X

X

X

X

X

X

X

Acatec

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

Stolz 1996

Nama

X

X

X

X

Stolz 2001b

406 Appendix I

Language Kihunde Kilivila Kinyarwanda Kisi Koasati Kobon Kogui Koiarie Komi-Zyrian Korean Koryak Kot Koya Koyra Chiini Krongo

No.

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

X

Songhai

X

X

X

X

X

X

Cágaba

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

Stolz 1996

X

X

X

Stolz 2001b

Sample languages 407

Language Kunama Kurukh Kwamera Kwami Kxoe Ladakhi Ladin Lahu Lak Lakhota Lango Languedocien Latin Latvian Laz

No.

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

X X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

Stolz 1996

Occitan

X

X

X

X

Stolz 2001b

408 Appendix I

Language Lezgi Limbu Lithuanian Luangiua Lugbara Luiseño Luo Luvale Maidu (Northeast) Malay Malayo Maltese Mamvu Manchu Mangarayi

No.

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

Wiwa

X

X

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

Malaiisch

X

X

Stolz 1996

X

X

X

X

X

Stolz 2001b

Sample languages 409

Language Maninka (Western) Manjaku Mansi Maori Mapuche Marathi Margi Mari (Hill) Marind Marshallese Martuthunira Masalit Mataco Mazatec Mbalanhu

No.

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Malinke

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

Stolz 1996

X

X

X

X

X

Stolz 2001b

410 Appendix I

Language Meithei Mixe Mixtec (Atatlahuca) Mixtec II Mogol’ Mokilese Mongor Mono-Alu Mundari Mupun Muruwari Nahali Nahuatl (Mecayapan) Nandi Navajo

No.

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

X

X

X

X

X

X

X

X

X

X

X

Mixtec I

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

Stolz 1996

X

X

Stolz 2001b

Sample languages 411

X X X X X X X X X X

Newari (Dolakha) Ngalakan Ngiti Ngiyambaa Nivkh Nobiin Norwegian Nukuoru Nyawaygi Nyulnyul Oromo (Harar) Oromo (Mechaa) Otomí (Mezquital) Ouargla

226

227

228

229

230

231

232

233

234

235

236

237

238

239

Wargla

X

X

X

X

Nenets

225

Stolz, Stroh, and Urdze 2005

Language

No.

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

Stolz 1996

X

X

X

X

Stolz 2001b

412 Appendix I

Language Ovambo Paez Paiwan Palauan Pali Pashto Pengo Pero Persian Piaroa Pipil Pirahã Pitta-Pitta Polish

No.

240

241

242

243

244

245

246

247

248

249

250

251

252

253

X

X

X

X

WoɃtiheh

Farsi

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

Stolz 1996

X

X

X

X

X

Stolz 2001b

Sample languages 413

Language Pomo Popoloca (San Juan Atzingo) Popoloca II Portuguese Qawasqar Quechua (Ayacuchano) Quechua (Imbambura) Quiche Rapanui Romani (Bugurdži) Romanian Romansch Russian Saami (North) Saliba (in Papua New Guinea)

No.

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

X X X?

X X X

X X X X X X X X

X X X X X X X X

X

X

Popoloca I

X

Stolz 1997a

X

X

Stolz, Stroh, and Urdze 2005

X X

X

X

X

X

Stolz 2001b

X

X

X

X?

X

Stolz 1996

414 Appendix I

Language Salt-Yui Samoan Sango Sanskrit Saramaccan Sardinian Scots-Gaelic Selepet Serbo-Croatian Shinassha Shona Shushwap Siuslawan Slave Slovak

No.

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

X X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

Stolz 1996

X

X

X

X

X

X

Stolz 2001b

Sample languages 415

Language Slovenian So Somali Sora Sorbian (Lower) Spanish Sre Sumerian Supyire Susu Swahili Swedish Tabassaran Tagalog Tahitian

No.

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

X X X X X X X

X X X X X X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

X

X

X

Stolz 2001b

X

X

X

X

X

Stolz 1996

416 Appendix I

Language Tairora Takelma Tamazight (Ayt Ndhir) Tamil Tarascan Tashlhiyt Tauya Telugu Tera Thai Tigak Tikar Timucua Tinrin Tlapanec

No.

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

X X

X

X

X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

X

X

Shilha

X

X

X

X

X

Stolz; Stroh, and Urdze 2005

X

X

Stolz 1996

X

X

Stolz 2001b

Sample languages 417

Language Tlingit Tokelauan Tokharian I Tokharian II Tongan Toqabaqita Totonac Triqui (Copala) Trumai Tshimshian Tuareg Tubu Tupi Turkish Ubykh

No.

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

X X X X X X X X X X X

X X X X X X X X X X

X

X

X

X

Stolz 1997a

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

Stolz 1996

X

X

X

Stolz 2001b

418 Appendix I

Udi Udmurt Ugaritic Ukrainian Umbundu Una Uradhi Urak Lawoi’ Urdu Urubu-Kaapor Usarufa Usarufa Uzbek Vai Venda

330

331

332

333

334

335

336

337

338

339

340

341

342

343

Language

329

No.

X X X X X X

X X X X Auyana X

X X

X X

X

X

X

X

X

Stolz 1997a

X

X

X

Stolz, Stroh, and Urdze 2005

X

X

X

X

X

Stolz 1996

X

X

X

Stolz 2001b

Sample languages 419

X X

Votic Wangkumara Warao Wargamay Warrwa Waskia Wayampi Welsh Wintu /Xam Xhosa !Xun (Ekoka) Yakut Yaqui

345

346

347

348

349

350

351

352

353

354

355

356

357

358

X X X

X X

X

X

X

X

X

X

X

X

X

X

Stolz 1997a

X

X

X

X

X

X

X

X

X

X

Vietnamese

344

Stolz, Stroh, and Urdze 2005

Language

No.

X

X

Stolz 1996

X

X

X

X

X

X

X

X

Stolz 2001b

420 Appendix I

X X

X X

Yiddish Yir Yoront Yokuts Yoruba Yucatec Yukaghir (Kolyma) Yukulta YupɃik (Central) Zapotec (Yatzachi) Zapotec I Zoque (Chimalapa)

361

362

363

364

365

366

367

368

369

370

371

X

Zapotec II

X

X

X

X

X

Mayo

X

X

X

X

X

X

X

X

X

Yessan-Mayo

360

X

Yelî Dnye

Stolz 1997a

359

Stolz, Stroh, and Urdze 2005

Language

No.

X

X

X

Stolz 1996

X

X

X

X

Stolz 2001b

Sample languages 421

Language family Arawakan, Maipuran, Northern Maipuran, Inland Indo-European, Germanic, West, Low Saxon-Low Franconian, Low Franconian Language Isolate Mayan, Kanjobalan-Chujean, Kanjobalan, Kanjobal-Jacaltec Sepik-Ramu, Sepik, Sepik Hill, Alamblak Indo-European, Albanian, Gheg Indo-European, Albanian, Tosk Eskimo-Aleut, Aleut Indo-European, Germanic, West, High German, German, Upper German, Alemannic Australian, Pama-Nyungan, Arandic Indo-European, Italic, Romance, Italo-Western, Western, PyreneanMozarabic, Pyrenean Indo-European, Armenian

Language

Achagua

Afrikaans

Ainu (Japan)

Akateko

Alamblak

Albanian (Gheg)

Albanian (Tosk)

Aleut

Alsatian

Alyawarr

Aragonese

Armenian

Appendix II Genetic affiliation

Language family Australian, Pama-Nyungan, Arandic Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, West Iberian, Asturo-Leonese Trans-New Guinea, Main Section, Central and Western, East New Guinea Highlands, Eastern, Gadsup-Auyana-Awa Altaic, Turkic, Southern, Azerbaijani Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, GalloRomance, Gallo-Rhaetian, Rhaetian Niger-Congo, Atlantic-Congo, Volta-Congo, North, Adamawa-Ubangi, Ubangi, SereNgbaka-Mba, Ngbaka-Mba, Ngbaka, Western, Baka-Gundi Niger-Congo, Mande, Western, Central-Southwestern, Central, Manding-Jogo, Manding-Vai, Manding-Mokole, Manding, Manding-East, Northeastern Manding, Baman Basque Indo-European, Slavic, East North Caucasian, East Caucasian, Tsezic, East Tsezic Trans-New Guinea, Main Section, Central and Western, East New Guinea Highlands, Eastern, Tairora

Language

Arrarnta (Western)

Asturian

Awiyaana

Azerbaijani (North)

Badiotto

Baka

Bamanankan

Basque

Belarusan

Bezhta

Binumarien

Genetic affiliation 423

Language family Algic, Algonquian, Plains Afro-Asiatic, Omotic, North, Gonga-Gimojan, Gonga, North Indo-European, Celtic, Insular, Brythonic Indo-European, Slavic, South, Eastern Language Isolate Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, East Iberian Uto-Aztecan, Northern Uto-Aztecan, Numic, Southern Sino-Tibetan, Chinese Oto-Manguean, Popolocan, Chocho-Popolocan, Chocho Chukotko-Kamchatkan, Northern, Chukot Uto-Aztecan, Northern Uto-Aztecan, Numic, Central Indo-European, Italic, Romance, Southern, Corsican

Language

Blackfoot

Boro

Breton

Bulgarian

Burushaski

Catalan

Chemehuevi

Chinese (Mandarin)

Chochotec

Chukot

Comanche

Corsican

424 Appendix II

Language family Indo-European, Slavic, South, Western Indo-European, Slavic, West, Czech-Slovak Trans-New Guinea, Main Section, Central and Western, Dani-Kwerba, Southern, Dani Indo-European, Germanic, North, East Scandinavian, Danish-Swedish, DanishRiksmål, Danish Austronesian, Malayo-Polynesian, Central-Eastern, Eastern Malayo-Polynesian, Oceanic, Central-Eastern Oceanic, Remote Oceanic, Loyalty Islands Australian, Djamindjungan Altaic, Turkic, Northern Sino-Tibetan, Tibeto-Burman, Himalayish, Mahakiranti, Kiranti, Western Indo-European, Germanic, West, Low Saxon-Low Franconian, Low Franconian Indo-European, Germanic, West, English Uralic, Finnic Altaic, Tungus, Northern, Even

Language

Croatian

Czech

Dani (Lower Grand Valley)

Danish

Dehu

Djamindjung

Dolgan

Dumi

Dutch

English

Estonian

Even

Genetic affiliation 425

Language family Indo-European, Germanic, North, West Scandinavian Indo-European, Indo-Iranian, Iranian, Western, Southwestern, Persian Uralic, Finnic Niger-Congo, Atlantic-Congo, Volta-Congo, Kwa, Left Bank, Gbe, Fon Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, GalloRomance, Gallo-Rhaetian, Oïl, French Indo-European, Germanic, West, Frisian Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, GalloRomance, Gallo-Rhaetian, Rhaetian Indo-European, Celtic, Insular, Goidelic Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, West Iberian, Portuguese-Galician Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, Oc Kartvelian, Georgian

Language

Faroese

Farsi

Finnish

Fon

French

Frisian (Western)

Friulian

Gaelic (Scottish)

Galician

Gascon

Georgian

426 Appendix II

Language family Indo-European, Germanic, West, High German, German, Middle German, East Middle German Language Isolate Niger-Congo, Atlantic-Congo, Atlantic, Southern, Mel, Gola Indo-European, Greek, Attic Indo-European, Greek, Attic Eskimo-Aleut, Eskimo, Inuit Tupi, Tupi-Guarani, Subgroup I Afro-Asiatic, Chadic, West, A, A.1 Indo-European, Indo-Iranian, Indo-Aryan, Central zone, Western Hindi, Hindustani

Carib, Southern, Southern Guiana Sino-Tibetan, Chinese Uralic, Finno-Ugric, Ugric, Hungarian

Language

German (Standard)

Gilyak

Gola

Greek

Greek (Ancient)

Greenlandic

Guaraní (Paraguayan)

Hausa

Hindi

Hixkaryána

Hokkien

Hungarian

Genetic affiliation 427

Language family North Caucasian, East Caucasian, Tsezic, East Tsezic Indo-European, Germanic, North, West Scandinavian Austronesian, Malayo-Polynesian, Malayic, Malayan, Local Malay Indo-European, Celtic, Insular, Goidelic Indo-European, Italic, Romance, Italo-Western, Italo-Dalmatian Chukotko-Kamchatkan, Southern Japanese Altaic, Mongolian, Eastern, Oirat-Khalkha, Oirat-Kalmyk-Darkhat Dravidian, Southern, Tamil-Kannada, Kannada Nilo-Saharan, Saharan, Western, Kanuri

Language

Hunzib

Icelandic

Indonesian

Irish

Italian

Itelmen

Japanese

Kalmyk

Kannada

Kanuri (Central)

428 Appendix II

Language family Altaic, Turkic, Western, Ponto-Caspian Uto-Aztecan, Northern Uto-Aztecan, Numic, Southern Altaic, Turkic, Western, Aralo-Caspian Yeniseian (Isolate) Muskogean, Eastern Trans-New Guinea, Main Section, Central and Western, East New Guinea Highlands, Kalam, Kalam-Kobon Trans-New Guinea, Main Section, Eastern, Central and Southeastern, Koiarian, Koiaric Language Isolate Creole, Portuguese based Nilo-Saharan, Unclassified Indo-European, Indo-Iranian, Iranian, Western, Northwestern, Kurdish

Language

Karaim

Kawaiisu

Kazakh

Ket

Koasati

Kobon

Koiarie

Korean

Korlai Creole Portuguese

Krongo

Kurdish (Northern)

Genetic affiliation 429

Language family Niger-Congo, Atlantic-Congo, Volta-Congo, Benue-Congo, Bantoid, Southern, Narrow Bantu, Central, D, Lega-Kalanga (D.20) Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, GalloRomance, Gallo-Rhaetian, Rhaetian Siouan, Siouan Proper, Central, Mississippi Valley, Dakota Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, Oc Indo-European, Italic, Latino-Faliscan Indo-European, Baltic, Eastern North Caucasian, East Caucasian, Lezgic, Nuclear Lezgic, East Lezgic Indo-European, Germanic, West, High German, German, Middle German, West Middle German, Rhenisch Franconian Indo-European, Baltic, Eastern

Language

Kwami

Ladin

Lakota

Languedocien

Latin

Latvian

Lezgi

Limburgisch

Lithuanian

430 Appendix II

Language family Uto-Aztecan, Northern Uto-Aztecan, Takic, Cupan, Luiseno Indo-European, Germanic, West, High German, German, Middle German, West Middle German, Moselle Franconian Indo-European, Slavic, South, Eastern Creole, Portuguese based Austronesian, Malayo-Polynesian, Barito, East, Malagasy Austronesian, Malayo-Polynesian, Malayic, Malayan, Local Malay Chibchan, Aruak Afro-Asiatic, Semitic, Central, South, Arabic Altaic, Tungus, Southern, Southwest Niger-Congo, Mande, Western, Central-Southwestern, Central, Manding-Jogo, Manding-Vai, Manding-Mokole, Manding, Manding-West

Language

Luiseño

Luxembourgeois

Macedonian

Malaccan Creole Portuguese

Malagasy (Plateau)

Malay

Malayo

Maltese

Manchu

Maninkakan

Genetic affiliation 431

Language family Austronesian, Malayo-Polynesian, Central-Eastern, Eastern Malayo-Polynesian, Oceanic, Central-Eastern Oceanic, Remote Oceanic, Central Pacific, East FijianPolynesian, Polynesian, Nuclear, East, Central, Tahitic Araucanian Sepik-Ramu, Sepik, Sepik Hill, Bahinemo Austronesian, Malayo-Polynesian, Central-Eastern, Eastern Malayo-Polynesian, Oceanic, Central-Eastern Oceanic, Remote Oceanic, Micronesian, Micronesian Proper, Marshallese Mixe-Zoque, Mixe, Western Mixe Indo-European, Italic, Romance, Eastern Altaic, Mongolian, Eastern, Oirat-Khalkha, Khalkha-Buriat, Mongolian Proper Uto-Aztecan, Southern Uto-Aztecan, Aztecan, General Aztec, Aztec

Language

Maori

Mapudungun

Mari (Eastern)

Marshallese

Mixe (Totontepec)

Moldavian

Mongolian

Nahuatl (Classical)

432 Appendix II

Language family Khoisan, Southern Africa, Central, Nama Pidgin Australian, Gunwingguan, Rembargic Australian, Djamindjungan Indo-European, Germanic, North, East Scandinavian, Danish-Swedish, DanishBokmål Indo-European, Indo-Iranian, Indo-Aryan, Eastern zone, Oriya Afro-Asiatic, Cushitic, East, Oromo Indo-European, Indo-Iranian, Iranian, Eastern, Northeastern Oto-Manguean, Otopamean, Otomian, Otomi Oto-Manguean, Otopamean, Otomian, Otomi

Language

Nama

Ndyuka-Trio Pidgin

Ngalakan

Ngaliwuru

Norwegian (Bokmål)

Oriya

Oromo

Osetin

Otomí (Mezquital)

Otomí (Querétaro)

Genetic affiliation 433

Language family Austronesian, Malayo-Polynesian, Palauan Indo-European, Indo-Iranian, Iranian, Eastern, Southeastern, Pashto Indo-European, Slavic, West, Lechitic Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, West Iberian, Portuguese-Galician Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, Oc Quechuan, Quechua II, C Quechuan, Quechua II, B Quechuan, Quechua II, B Indo-European, Indo-Iranian, Indo-Aryan, Central zone, Romani Indo-European, Italic, Romance, Eastern

Language

Palauan

Pashto

Polish

Portuguese

Provençal

Quechua (Ayacuchano)

Quechua (Ecuadoriano)

Quechua (Columbiano)

Romani

Romanian

434 Appendix II

Language family Indo-European, Slavic, East Uralic, Sami, Eastern Uralic, Sami, Western, Northern Austronesian, Malayo-Polynesian, Central-Eastern, Eastern Malayo-Polynesian, Oceanic, Western Oceanic, Papuan Tip, Nuclear, Suauic Austronesian, Malayo-Polynesian, Central-Eastern, Eastern Malayo-Polynesian, Oceanic, Central-Eastern Oceanic, Remote Oceanic, Central Pacific, East FijianPolynesian, Polynesian, Nuclear, Samoic-Outlier, Samoan Indo-European, Indo-Iranian, Indo-Aryan Creole, English based Indo-European, Italic, Romance, Southern, Sardinian Uralic, Samoyed Niger-Congo, Atlantic-Congo, Volta-Congo, North, Gur, Senufo, Suppire-Mamara

Language

Russian

Saami (Akkala)

Saami (North)

Saliba

Samoan

Sanskrit

Saramaccan

Sardinian (Logudorese)

Selkup

Senoufo (Supyire)

Genetic affiliation 435

Language family Indo-European, Slavic, South, Western Hokan, Salinan-Seri Indo-European, Slavic, West, Czech-Slovak Indo-European, Slavic, South, Western Niger-Congo, Atlantic-Congo, Volta-Congo, Benue-Congo, Bantoid, Southern, Narrow Bantu, Northwest, C, Kele (C.60) Afro-Asiatic, Cushitic, East, Somali Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, IberoRomance, West Iberian, Castilian Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, GalloRomance, Gallo-Rhaetian, Rhaetian Niger-Congo, Atlantic-Congo, Volta-Congo, Benue-Congo, Bantoid, Southern, Narrow Bantu, Central, G, Swahili (G.40) Indo-European, Germanic, North, East Scandinavian, Danish-Swedish, Swedish

Language

Serbian

Seri

Slovak

Slovenian

So (Congo)

Somali

Spanish

Sursilvan

Swahili

Swedish

436 Appendix II

Language family Austronesian, Malayo-Polynesian, Meso Philippine, Central Philippine, Tagalog Trans-New Guinea, Main Section, Central and Western, East New Guinea Highlands, Eastern, Tairora Penutian, Oregon Penutian, Takelma Altaic, Turkic, Western, Uralian Austronesian, Malayo-Polynesian, Central-Eastern, Central Malayo-Polynesian, Timor, Nuclear Timor, East Unclassified Austronesian, Malayo-Polynesian, Central-Eastern, Eastern Malayo-Polynesian, Oceanic, Central-Eastern Oceanic, Remote Oceanic, New Caledonian, Southern, South, Zire-Tiri Creole, English based, Pacific Totonacan, Totonac Oto-Manguean, Mixtecan, Trique

Language

Tagalog

Tairora (North)

Takelma

Tatar

Tetun

Timucua

Tiri

Tok Pisin

Totonac (Papantla)

Triqui (Copala)

Genetic affiliation 437

Language family Language Isolate Austronesian, Malayo-Polynesian, Sulawesi, Muna-Buton, Tukangbesi-Bonerate Altaic, Turkic, Southern, Turkish Austronesian, Malayo-Polynesian, Central-Eastern, Eastern Malayo-Polynesian, Oceanic, Central-Eastern Oceanic, Remote Oceanic, Central Pacific, East FijianPolynesian, Polynesian, Nuclear, Samoic-Outlier, Ellicean Mayan, Cholan-Tzeltalan, Tzeltalan North Caucasian, East Caucasian, Lezgic, Udi Uralic, Permian Indo-European, Slavic, East Indo-European, Italic, Romance, Italo-Western, Western, Gallo-Iberian, GalloRomance, Gallo-Rhaetian, Rhaetian Uralic, Finnic

Language

Trumaí

Tukang Besi

Turkish

Tuvaluan

Tzotzil (Chamula)

Udi

Udmurt

Ukrainian

Vallader

Vod

438 Appendix II

Language family Language Isolate Tupi, Tupi-Guarani, Subgroup VIII Indo-European, Celtic, Insular, Brythonic Eskimo-Aleut, Eskimo, Inuit Altaic, Turkic, Northern Indo-European, Germanic, West, High German, Yiddish Australian, Pama-Nyungan, Yidinic Eskimo-Aleut, Eskimo, Yupik, Alaskan Oto-Manguean, Zapotecan, Zapotec Niger-Congo, Atlantic-Congo, Volta-Congo, Benue-Congo, Bantoid, Southern, Narrow Bantu, Central, S, Nguni

Language

Warao

Wayampi

Welsh

West Greenlandic

Yakut

Yiddish (Eastern)

Yidiny

Yupik (Central)

Zapotec (Yatzachi)

Zulu

Genetic affiliation 439

Appendix III Questionnaire

This appendix contains the full text of the English version of our questionnaire which we distributed among specialists of individual languages in 1997 in order to build up a data-base independent of the extant literature. The questionnaire came in six translations (English, French, German, Russian, Spanish, and Turkish) and was sent out to about 120 linguists who had reacted positively to our request on the Linguist List. We received 67 completed questionnaires within a period of 30 months. Unfortunately, among these 67 answers, there were many which dealt with the same language. Originally, it was intended to have several follow-up questionnaires. This idea however had to be abandoned not only because of time but also because several of the contacted specialists complained about the European (cultural) bias of the situations to be described linguistically (bicycles for instance were said to be non-existent in some parts of the world and thus [10a] would not make much sense to the native speakers). Since our project was not primarily about the perfect questionnaire design, we deemed it more promising to go on with our data hunt on the basis of extant written material. *** Encyclopaedia and Thesaurus of WITH- and WITHOUT-Relations: Comitative and Related Subjects from a Typological Point of View A project of the University of Bremen, FB 10, supported by the DFG Priv.-Doz. Dr. Thomas Stolz (head), Dr. Suphi Abdülhayo÷lu, Dr. Cornelia Stroh, Traude Gugeler M.A., Oxana Jarovaia, Tobias Mahlow, Susanne Müller, and Anna Sabater Fuentes QUESTIONNAIRE No. 1 January 1997 Dear colleagues, Since January 1996 we have been conducting at the department 10 of the University of Bremen a linguistic research project supported by the Deutsche Forschungsgemeinschaft. The subjects of our research work are case relations, which in the linguistic literature are usually described as Comitative, Instrumental, Sociative, and Abessive. For reasons of conciseness, we have introduced the expression WITH- and WITHOUT-relations as a generic term for the large number of categories which are interesting for our investigation.

Questionnaire 441 It is one of the essential goals of the project to reconstruct to what extent withand without-relations are subject to systematic behaviour peculiar to the relations themselves and functionally distinguishing them from other relations. We intend to track this system empirically mainly by examining the typological spectrum of WITH- and WITHOUT-relational structures with regard to constants and variations. For this purpose we set ourselves the task of gathering the relevant data from as large a number of languages as possible. Using this comprehensive data base as a starting point we intend to inventory the phenomena relevant to our research, as well as to develop a theory of WITH- and WITHOUT-relations. Since the beginning of the project, we have developed a number of hypotheses which now require testing. Up till now, during the course of our empirical work we’ve had to rely on our own competence, grammars, dictionaries and special investigations. Of course these procedures cannot answer all our questions. Hence, we’ve decided to standardize our comparative basis and to obtain certain specific information by consulting the experts directly in standardized form. Thus, we have developed the enclosed questionnaire. Therefore, we would like to ask you for your cooperation. Your contribution would consist of filling out the questionnaire and returning it to us as soon as possible. By means of this questionnaire we hope not only to collect general information about the language for which you have sufficient expertise. The questionnaire also contains a few standardized elicitation sentences which form the most important part of the questionnaire. Above all, we are interested in receiving your translation of the elicitation sentences into the language you are working on. If you believe the elicitation sentences to be too unnatural, feel free to present more natural sentences, as far as they reflect the topic of investigation in an appropriate manner. Please bear in mind that we require in any case a grammatical morphematic analysis of the examples you’ve supplied. Please indicate the source of your examples (either complete bibliographical references or statement: self-construed example, information from native speakers etc.). Provided you haven’t any objections, we intend to publish in the acknowledgements of the planned final publication the names of all those who have provided us with in-formation. It goes without saying that we are happy to receive any additional information you are willing to provide. Advices concerning the peculiarities of WITH- and WITHOUT-relations in your object language (dialects, sociolects, children’s language etc.), relevant publications, or researchers who are concerned with our topic are more than welcome. In addition, further questionnaires in intervals of several months are planned. Thus, we would like to ask you, whether you wish to receive questionnaire No. 2 as well. Your name will be included in our address list only if you mark yes in the corresponding box on the questionnaire No. 1. You can send us your answers in the following ways: (a) e-mail: [email protected] (b) fax: Komitativ-Projekt, Thomas Stolz international ++49-421-218 4283; national 0421-218-4283

442 Appendix III (c) letters:

Komitativ Projekt Priv.-Doz. Dr. Thomas Stolz Universität Bremen FB 10: Linguistik PF 330 440 D-28 334 Bremen

With kindest regards and many thanks for your efforts Thomas Stolz (in the name of the project team) Explanations and advice For your own sake, we would ask you to only reply to those questions to which you can give a certain answer without a great deal of effort. For example, if you don’t want to make a statement with respect to the typological classification of your language, it’s not necessary to fill out the corresponding fields. Should the need arise, please insert an additional sheet wherever you need more room for your answer. Where we have given alternatives for the translation of the elicitation sentences (sentences 10–11), it is necessary to translate these alternatives only if there are actually differences in the use of the relators. Please use a Latin-based scientific transcription for languages that don’t have a Latin-based orthography. We also willingly accept questionnaires for evaluation that are filled out only partially, in case time doesn’t permit the translation of all the elicitation sentences. We just ask you to return the questionnaire as soon as possible. It would be very helpful for our further plans if you could return it by July 7th, 1997 at the latest. In questionnaire No. 2 we will ask questions about the etymology of the relators. Syncretisms which are characteristic for the WITH- and WITHOUT-relators of your language will be the subject of questionnaire No. 3. Further questionnaires are under development. Should you be interested in further general information about our project, feel free to contact us under the above addresses and telephone/fax numbers. The comitative project team QUESTIONNAIRE No. 1 Filled out by: Last Name ..................................... First Name ........................................................ Address ..................................................................................................................... Fax ................................................. E-mail ................................................................ Source language: Language name.......................................................................................................... Language family ........................................................................................................

Questionnaire 443 Continent ...................................... Country ............................................................ Status writing system yes ˆ no ˆ official yes ˆ no ˆ endangered yes ˆ no ˆ extinct yes ˆ no ˆ Pidgin yes ˆ no ˆ Typology predominant fundamental relations Accusative pure ˆ mixed ˆ Ergative pure ˆ mixed ˆ Active pure ˆ mixed ˆ predominant constituent order SOV ˆ VOS ˆ free ˆ SVO ˆ OVS ˆ VSO ˆ OVS ˆ predominant marking type head-marking pure ˆ mixed ˆ dependent-marking pure ˆ mixed ˆ double-marking pure ˆ mixed ˆ predominant morphological technique agglutinating ˆ inflecting ˆ isolating ˆ polysynthetic ˆ incorporating ˆ Relators Which are the relators used in your language as translation equivalents for the following English relators? Please indicate also the morphemic class /lexical category of the relator. In case there are various possibilities of translation, please list them and specify the context in which the relators are used; if necessary, use a separate sheet. WITH

(1) ............................................................................................................. (2) ............................................................................................................. (3) ............................................................................................................. (4) ............................................................................................................. WITHOUT (1)................................................................................................................ (2)................................................................................................................ (3)................................................................................................................ (4)................................................................................................................

444 Appendix III Elicitation sentences Please translate the following sentences (a) into your object language (b) and add a morpheme analysis. For all languages not written in a Latin alphabet please use a Latin-based transcription system. Please use a separate sheet for further comments. (1a) The teacher goes to school with the students (1b) .................................................................................................................. (1c) .................................................................................................................. (2a) The teacher works with the students at school (2b) .................................................................................................................. (2c) .................................................................................................................. (3a) The teacher speaks with the students at school (3b) .................................................................................................................. (3c) .................................................................................................................. (4a) The teacher and the students are going to school (4b) .................................................................................................................. (4c) .................................................................................................................. (5a) The teacher talks to the students in a loud voice (5b) .................................................................................................................. (5c) .................................................................................................................. (6a) The teacher is writing on the board with chalk (6b) .................................................................................................................. (6c) .................................................................................................................. (7a) The teacher is writing words on the board with his right hand (7b) .................................................................................................................. (7c) .................................................................................................................. (8a) The teacher points with his finger to a letter (8b) .................................................................................................................. (8c) .................................................................................................................. (9a) The teacher leaves school with his bag (9b) .................................................................................................................. (9c) .................................................................................................................. (10a) The teacher goes to school on foot/by bus/by bike* (10b) .................................................................................................................. (10c) .................................................................................................................. (11a) The teacher comes to school on foot/on his horse* (11b) .................................................................................................................. (11c) .................................................................................................................. (For the sentences 10–11 marked with * please give various translations only if your language differentiates with regard to the use of relators depending on the means of transportation.) (12a) The teacher comes to school with a dog (12b) .................................................................................................................. (12c) ..................................................................................................................

Questionnaire 445 (13a) (13b) (13c) (14a) (14b) (14c) (15a) (15b) (15c) (16a) (16b) (16c) (17a) (17b) (17c) (18a) (18b) (18c) (19a) (19b) (19c) (20a) (20b) (20c) (21a) (21b) (21c) (22a) (22b) (22c) (23a) (23b) (23c)

The teacher comes to school with a friend .................................................................................................................. .................................................................................................................. The teacher comes to school with a broken arm .................................................................................................................. .................................................................................................................. The teacher comes to school with a new hat .................................................................................................................. .................................................................................................................. The teacher goes to school without the students .................................................................................................................. .................................................................................................................. The teacher works at school without students .................................................................................................................. .................................................................................................................. The teacher draws a straight line on the board without a ruler .................................................................................................................. .................................................................................................................. The teacher leaves school without his bag .................................................................................................................. .................................................................................................................. The teacher comes to school without his dog .................................................................................................................. .................................................................................................................. The teacher comes to school without his friend .................................................................................................................. .................................................................................................................. The teacher comes to school without his bike .................................................................................................................. .................................................................................................................. The teacher comes to school without a hat, but with glasses. .................................................................................................................. .................................................................................................................. .................................................................................................................. References The examples in the object language are in/from ................................................................................................................................... ................................................................................................................................... ................................................................................................................................... ...................................................................................................................................

446 Appendix III Further comments (please add additional sheets if necessary) ................................................................................................................................... ................................................................................................................................... ................................................................................................................................... I would like to participate in further questionnaire campaigns of the comitative project: Please send me questionnaire No. 2 yes ˆ no ˆ

Notes

1.

2.

3.

4.

The two terms are not synonymous: To our mind, crosslinguistics is a less systematic way of tracing the distribution of certain phenomena among the world’s languages, whereas typology proper aims at correlating various phenomena in such a way that their interrelation (or the lack thereof) can be used to classify languages in a consistent way. Throughout this study, we employ three different homophonous but heteronymous terms which are distinguished from one another by means of graphic representation (outside this footnote, the relevant terms will appear without the inverted commas used below): – The term ‘comitative’ in lowercase and in plain type refers to the (surface) case category of an individual language (often accompanied by an attribute identifying the marking strategy used to express this category [e.g. ‘morphological/inflectional/adpositional comitative’ etc.]) and is largely independent of semantic parameters. – The term ‘Comitative’ capitalized and in plain type refers to the universal function of expressing accompaniment by grammatical means. – The term ‘COMITATIVE’ in small capitals refers to the languageindependent thematic role of NPs in relation to other NPs and/or the predicate of the same clause/sentence. – The term ‘comitative’ in boldface refers to the category networks which result from an analysis of the distributional profile/functional domain of comitative markers in individual languages. These conventions are also applied to other categories as well, and thus the reader will find, for instance, ‘instrumental’, ‘Instrumental’ and ‘INSTRUMENTAL’ alongside each other. In order to avoid repeating data and arguments, we refrain from discussing at length the languages already covered by Lehmann and Shin (2005). Thus, the reader will sometimes search in vain for detailed information about Hmong, Japanese, Kambera, Kayardild, Khmer, Korean, Lezgian, Mandarin Chinese, Thai, Turkish, Vietnamese, Yidiny, Yucatec Maya and Yukaghir – all of which are painstakingly analysed by Lehmann and Shin (2005). We consider this decision of ours a practical contribution to the necessary division of labour among linguists. Many aspects of the grammar of German mit have been addressed already in Seiler (1974a, 1974c), Eroms (1976), Kunze (1992) and the comprehensive study by Koch (1978). Notwithstanding the fact that we disagree with these four pioneers on a number of issues, we use German only occasionally as an appetizer for the things to come in the remainder of the present study. Where

448 Notes

5.

6.

7.

8.

9.

German examples are not accompanied by bibliographical references, we have made up examples based on our competence as native speakers of German. The same holds for Latvian examples for which one of us, Aina Urdze, can claim native-speaker competence. In earlier publications, we often used the handy term WITH-relation (German MIT-Relation) as a convenient cover for Comitatives and Instrumentals. Owing to the inherent danger of invoking associations with English (or German) and the morphosyntax and functional domain of the preposition, we avoid our erstwhile favourite term in this publication. Throughout this study we use the term participant indiscriminately for both central and peripheral participants, i.e., we do not use the otherwise very useful Tesniere’ian label circumstant(s) suggested by Lazard (1994: 15) and Lehmann and Shin (2005). It goes without saying that we acknowledge that actants and participants belong to two different levels, namely the latter to the semantic level and the former to the morphosyntactic one. The è-criterion is a pertinent example of this assumed ‘biuniqueness’ (Fanselow and Felix 1993: 83). As a matter of fact, the validity of this principle has been contested by various authors of different theoretical convictions. Schlesinger (1995: 70) for instance claims that “there is no motivation for the principle that every noun phrase has to be assigned case”. He introduces the concept of caseless NPs. The two divergent views depend crucially on what exactly is meant by the term case. Fortunately, we do not have to solve this problem because our Comitatives clearly are not cases in the usual sense. However, we insist that all NPs of a clause/sentence partake in relations not only between them and the predicate nucleus but also among themselves. For our purpose, it is irrelevant whether or not one chooses to call these relations or a section of them cases, role, relation or the like. Henceforth, boldface marks those morphemes and their equivalents in the transmorphemizations and English glosses which encode the Comitative or any category that is closely associated with the Comitative or in the focus of the accompanying discussion. X, Y and X/Y are conventionally used in the transmorphemizations as placeholders of the labels Comitative, Instrumental and Comitative-Instrumental, respectively. They are employed according to the etymological principle, i.e., wherever the morph occurs that is also used as the principle encoding for the Comitative, it is rendered by X in the transmorphemization no matter exactly what semantic relation applies to a structure in a specific context. Wherever we deem it necessary for expository purposes, we employ X, Y and X/Y also as transmorphemizations of grammemes which otherwise are the primary markers of categories other than the ones in focus here. Further conventions used in later examples and diagrams are explained in due course ‘on the spot’. Croft (1990: 9–10) presents the polysemy of English with as a telling case for certain “causal relations between participants and properties of an event” with-

Notes 449

10.

11.

12.

13.

out going into much detail. He also adduces evidence from Hausa and Classical Mongolian – two languages which display similar patterns of polysemy of their translation equivalents. This similarity, according to Croft (1990: 10), cannot be explained by mere chance. This observation is fully acceptable. However, one should not forget that there are numerous languages which do not follow the same pattern, as we are going to demonstrate in this book. Superficially, it may seem odd to postpone the discussion of methodological issues. However, we have good reasons to reserve the ultimate clarifications as to methodology for a separate section. As the question of how one can make the best of the ubiquitous form-function dilemma is also crucial to the feasibility of our own approach, we do not go the easy way of simply stipulating methodological axioms. With a view to better understand why the semasiological approach has certain advantages and disadvantages at the same time, we deem it necessary to present as many data as possible first before drawing conclusions as to the best method of interpreting them. This is important especially because we will see that the phenomenon to be scrutinised decides how far the form-orientation of our identification procedure may go. Owing to the fact that quite a few science-theoretical aspects are at issue, methodology itself deserves a treatment of its own after the reader has become familiar with the most characteristic phenomena. The discussion of methodology automatically involves a review of other theoretical issues, and thus it makes sense to address the history of research in the realm of Comitatives too in a later section. For the time being, we rely on traditional terms or those which are easily translatable into the current linguistic models. In Chapter 4, we introduce our own terminology and the necessary explanations. We quote from the paperback edition published in 2003 (the original of A Share in Death dates back to 1993): (Share 68) Balancing the squirming toddler on one hip and her shopping on the other, she kicked the Escort’s door shut with unnecessary spite. The German translator preferred mit-PPs over gerunds (= present active participles) because the latter reek of bureaucratic officialese in German. The unusual accumulation of mit-tokens in one and the same sentence, however, is much more acceptable to the native speaker. For matters of glossonyms of individual languages and their genetic affiliation we largely follow the suggestions of the Ethnologue – with the proviso that wherever a better established traditional term appears to be simpler than the one in the Ethnologue we adhere to the more economic alternative solution. Especially in Chapter 8 (Part B), however, we adopt Ruhlen’s (1987) classification into macrophyla as it allows us to make large-scale statistical observations about language groups. Again, we prefer the term Indo-European over the much debated Indo-Hittite, which for this reason does not show up in our

450 Notes

14.

15.

16.

17.

18.

19.

20. 21.

22. 23.

24.

tables. In the Appendix II, we provide a concordance to relate the two systems of classification to each other. The situation is slightly more complicated because there also is the associative -stul/-stül which is exclusively used in contexts for which the label accompaniment is appropriate. It therefore is a partial synonym of -val/-vel. Cf. (B5). Note that -huƗn is not a full-blown postposition of the kind represented by ic! Rather, -huƗn behaves like a coordinating conjunction – an issue which we will come back to in Part B, Section 9.2. In this chapter, we make do with enumerating our agenda without referring to the pertinent literature in order to save space. The discussion of the extant ideas, models and theories follows in due course in the subsequent sections wherever appropriate. Admittedly, zero-marking, parasitic marking (= a category is never encoded separately) and the use of verb-based relators in non-serialising languages pose a number of problems, cf. Part B, Chapter 10. Seiler (1988: 54) argues that “the participants are manipulated upon the basis of their relative importance in the said relation”. The term “manipulation” can also be understood in the sense that speakers employ participants in such a way that the description of a state of affairs yields a picture which conforms to the communicative intentions of the speaker. There is absolutely no need for an objectivist correspondence between what happened in the real world and how we choose to describe what happened. As a matter of fact, there is even no need for anything to happen in order to make speakers talk about events in the first place. Human minds are creative, inventive and deliberately deceiving (Keller 1991). Whether or not it makes sense to distinguish a third type of relations, namely intra-participant ones, is an issue which we will only allude to in passing when discussing adnominal phrases in the subsequent chapters. We owe this term to the creative mind of Christel Stolz who came forward with this terminological suggestion as soon as 1992. In logical terms, we are dealing with a two-place predicate – the relator having predicative function itself. However, this two-place predicate may be embedded in a more extended predication where the various (higher and lower) predicators fuse. The subsequent discussion is meant to highlight some of the major difficulties with the “valence” implied in our formula. It is an instance of a satellite in Talmy’s terminology (2000: 223). The two instances of mit(-), however, do not form a pair of the original-pluscopy type. As the subsequent discussion demonstrates, they represent two different distribution classes and are even functionally distinct, as one is used in COMPANION-oriented clauses and the other in ACCOMPANEE-oriented ones. Mobile prefixes are also typical of German’s next of kin, Low German and Dutch. In Dutch, however, the preposition met ‘with’ and the mobile prefix mee- are segmentally distinct: (LPP Dutch XXVI, 101) en je zult met me mee

Notes 451

25.

26.

27.

28.

29.

30.

31.

32. 33.

34.

willen lachen ‘and you would like to laugh together with me’, where the infinitive of the verb is meelachen ‘to laugh together with somebody’. The NPs representing ACCOMPANEE and COMPANION can both be fronted for the purpose of topicalization or focusing: [mit Robert] topic/focus kommt Peter (mit) ‘it is Robert (a) whom Peter takes along/ (b) whom Peter is accompanying/as for Robert, it is Peter who (a) takes him along/ (b) accompanies him’. In the (a)-version, there is no slot for the sentence-final mit, whereas the mobile prefix has to be there in the (b)-version. Zero-marking, material incorporation, and direct argument status are somewhat difficult to accommodate with this radical solution. They will each be discussed in turn in the subsequent chapters. Coseriu (1970) has put forward a very similar idea which is also recognisable in Seiler’s principle of CONCOMITANCE (1974b: 218). Jolly (1993) introduces the term SAME-SPACE for this phenomenon. Iconicity is also indicative, as the NPs representing ACCOMPANEE and COMPANION are never adjacent in German main clauses, provided that the COMPANION is not an NP-internal attribute (cf. [A32]). The distance between the two NPs could then be interpreted as an iconic mapping of their conceptual distance (Haiman 1985, Behagel 1923–1932). However, for a language like German where discontinuous constituency is not rare, this topological argument may easily be stretched too much. Nevertheless, such nominalizations and other complex formations as, e.g., phrasal compounds, are frequent, especially in spontaneous spoken German and also in contemporary journalistic prose (Meibauer 2003). It is tempting to extend this argument to the micro-roles the NPs represent by claiming that they must be close neighbours on the scale provided by Lehmann, Shin, and Verhoeven (2000: 19). On this scale, the instrument is farther away from the AGENT than the COMITATIVE. This fits our German example. However, this nice compatibility of micro-role topology and participant relations in the sentence under scrutiny is the product of a simplification of the facts. Section 4.2 demonstrates that the neighbourhood-relation of AGENT and COMITATIVE is more complicated. This and other uses of ar clearly show that the distribution and functional domain of the Latvian preposition is not equivalent to German mit, pace Eckert, Bukeviþinjtơ, and Hinze (1994: 370). Note that there is a common form for the 3rd person, irrespective of number (cf. Chapter 9). As in the case of the formula used for the ACCOMPANIMENT situation, we could translate (A17) into a three-place logical predication (with the same problems of relator-[dif]fusion). The famous animacy hierarchy is an example of this anthropocentricity. Linguistic units associated with human referents are often grammatically privileged in the sense that certain categorial distinctions are only possible for

452 Notes

35.

36.

37.

38.

them (e.g., very often, only nouns with a human referent are sensitive to number distinctions, etc.). Grammaticalization theory has a whole thesaurus of pertinent examples: The expressions for certain concepts are grammaticalised from lexical sources that are closely associated with human beings (e.g., the development of adpositions out of body part nouns), cf. Stolz (1991, 1992a). In terms of orientation theory, the central participant in a given situation is privileged to the detriment of the less central or peripheral co-participants (Serzisko 1991: 274, Stolz 1997b: 525). The latter term has been in use in Indo-European studies, Classical philology and other branches of our disciplines which have been under the influence of traditional (i.e. Neogrammarian-inspired) comparative grammar, cf. Brugmann and Delbrück (1911: 520–523). As Croft (1991) does not provide a definition nor a circumscription of what is meant by Associative, we dare to put forward a definition that is patterned according to Croft’s model (note that this proposal is not meant as a substitute for the term COMITATIVE, which we favour despite the terminological confusion it tends to create): Associative: An entity that participates in a causal chain at the same point and in the same role as the NP to which it is related. It is likely that the associative role does not require that the pertinent NP be the initiator of an act of volitional causation […]. This definition comes close to the one proposed by Wilkins’ (1989: 211) for the so-called ‘simple’ comitative in Arrarnta: No subject-orientation but rather a same-role relation. It remains to be seen what implications sharing one and the same role (on which level?) has for our notion of roles in general. Incidentally, Wilkins (1989: 202) uses the label associative for a different case in Arrarnta, namely the one which encodes association of a person or entity with place of origin, groups of kin, etc. This usage come close to what is commonly called associative plural (cf. Daniel and Moravcsik 2005). The discussion of the notion of Associative requires a chapter of its own. With a view to keeping our presentation within reasonable limits, however, we relegate the description of the Associative to a future following study. From the descriptions given in our sources, it is not altogether clear whether the NPs marked for Comitative can be considered full-blown (COMITATIVE) objects. Therefore, we prefer to classify them as not absolutely obligatory but nevertheless preferred complements. Native speakers invariably considered the Turkish version with Comitative NP more acceptable, although the intransitive use of the nuclear predicate was not totally ruled out either. According to the valence dictionary of Turkish verbs (Abdülhayo÷lu 1990: 300), the Comitative complement of evlenmek ‘to marry’ is next to compulsory. For the Saami examples, Nickel (1990: 499) prefers a classification as proper objects. The Indonesian case is still different, as the use of the preposition is not obligatory. The whole construction can be replaced by ia mengawini seorang

Notes 453 janda ‘he married a widow’ (where mengawini < meN- + kawini), i.e., the verb hosts a prefix that makes the preposition superfluous and promotes the postverbal NP to the status of direct object (Kwee 1981: 122–123). This may be indicative of a reduced degree of transitivity of the construction involving the preposition dengan. 39. For convenience, we add pertinent examples of these other uses of the relators in the three languages (the different vowel quality in the Turkish suffix -le/-la is determined by the rules of vowel harmony): (i) Indonesian (Sneddon 1996: 225) (a) INSTRUMENTAL dia me-motong kayu dengan parang machete he TRANS-chop wood X/Y ‘He chopped the wood with a machete.’ (b) ACCOMPANIMENT dia pergi ke Medan dengan istrinya He go to Medan X/Y wife ‘He went to Medan with his wife.’ (ii) Saami (a) INSTRUMENTAL (Nickel 1990: 499) Máhtte vujii biila-in Mattis drive:PAST.3SG car-X/Y ‘Mattis went by car.’ (b) ACCOMPANIMENT (Bartens 1989: 94) mun vuolggán dál Biera-in guolástit I set_out:1SG now Peter-X/Y fish:INF ‘I am going to fish with Peter now.’ (iii) Turkish (a) INSTRUMENTAL (Kornfilt 1997: 227) kapı-yı bir çekiç-le kır-dı-m hammer-X/Y break-PAST-1SG door-ACC a ‘I broke the door with a hammer.’ (b) ACCOMPANIMENT (Kornfilt 1997: 228) konser-e Hasan-la git-ti-m concert-DAT Hasan-X/Y go-PAST-1SG ‘I went to the concert with Hasan.’ 40. Again, the same relator -ga is used for the purpose of marking the companion and the tool in prototypical constellations: (i) Estonian (Hasselblatt 1992: 95) (a) INSTRUMENTAL kirjuta-n pliiatsi-ga write-1SG pencil-X/Y ‘I am writing with a pencil.’

454 Notes (b) ACCOMPANIMENT ma lähe-n sinu-ga I go-1SG you.SG-X/Y ‘I go with you.’ 41. From a Portuguese perspective, this use of the preposition goes far beyond those cases where com introduces a complement in European Portuguese. The use of cognates of Portuguese com as a marker of human object NPs is widespread in Portuguese-based Creoles of the Indian Ocean (for instance, in Korlai Creole Portuguese [spoken in between Goa and Bombay], Clements 1996: 160). However, Baxter (1985: 167–168) convincingly argues that the local adstrate/substrate languages are the likely source for the pattern of the functional domain of Malaccan Creole Portuguese ku. The Austronesian-based lingua franca, Bazaar Malay, employs a marker sama, which has almost the same range of functions as ku – with the notable exception of the INSTRUMENTAL (which is expressed by dengan, just as in standard Malay and Indonesian), cf. (i). Hokkien (a southern variety of Chinese widely spoken in South-East Asia) is likewise a possible source. In this language, the relator kap is used indiscriminately for COMITATIVE, INSTRUMENTAL and (human) object marking, cf. (ii). The examples are from Baxter (1985: 167–168). (i) Bazaar Malay (a) Human object gua tengok sama lu I see X you ‘I see you.’ (b) ACCOMPANIMENT gua pergi utang sama dia I go jungle X he ‘I went to the jungle with him.’ (ii) Hokkien (a) Human object guà kăp î khuă I X/Y he look ‘I saw him.’ (b) INSTRUMENTAL guà yǀng chhá kăp î phhă I use stick X/Y he hit ‘I hit him with a stick.’ (c) ACCOMPANIMENT guà kăp î khhi khiƗ I X/Y he go walk ‘I went for a walk with him.’ Notwithstanding the potential substrate origin of the object-marking function of ku in Malaccan Creole Portuguese, it is evident that we are not dealing with a

Notes 455

42. 43.

44. 45.

46.

47.

linguistic singularity in the languages of the world. Note, however, that in their survey of COMITATIVES in Creoles, Michaelis and Rosalie (2000) do not mention a single instance of PATIENT/THEME-marking by the Comitative relator. Baxter (1985: 157) points out that the object in (A28) could also be zeromarked – a version the native speaker informants considered “slightly odd”. At this point, we neglect most of the prerequisites of the prototype because animacy is only marginally involved in the Romanian examples (though elephants and boas are of course animate). For the sake of comprehensiveness, we also include examples of the use of the preposition cu in the prototypical situations (Iordan and Robu 1978: 522): (i) Romanian: ACCOMPANIMENT Ion vine cu Gheorge John comes X/Y George ‘John is coming with George.’ (ii) Romanian: INSTRUMENTAL taie lemne cu toporul cuts wood:PL X/Y axe:DEF ‘He cuts the wood with the axe.’ As adnominal attributes may themselves be modified recurrently, the number of levels is potentially unlimited. However, there is no reason to assume any relations holding between the adnominal attributes of different participants. As a rule, only participants of the same level which are directly related to the same higher-order unit can be in an inter-participant relationship. Objects marked for Comitative can also be encountered in German, for instance. Especially reflexive verbs bearing the (erstwhile applicative) prefix be- take a prepositional object introduced by mit as, e.g., sich befassen mit ‘to occupy oneself with something’, sich belasten mit ‘to burden oneself with something’, sich bescheiden mit ‘to content oneself with’, etc. In none of these cases is the mit-phrase optional, i.e., mit introduces an argument of the verb – an argument that certainly has the macro-role UNDERGOER. It is in no way detrimental to our argument – pace Pinkster (1988: 96–97), that the use of the morphological ablative alone (i.e. without additional preposition) to encode SOURCE is largely restricted to toponyms (names of cities and small islands). There is a lexico-semantically well-defined group of expressions for which the ablative (syncretistic with the dative) is compulsory. The ablative has an additional function with exactly the same group of words, namely to encode LOCATION: The ablative plural AthƝnis could occur in combinations with static verbs (as, e.g., in AthƝnis sum ‘I am in Athens’) or dynamic verbs (as, e.g., in AthƝnis veniǀ ‘I come from Athens’). In other cases, the residual locative is used to encode location (as, e.g., in rurƯ sum ‘I am in the countryside’).

456 Notes 48. The Irish preposition le ‘with’ is also used as the primary marker of COMITATIVE and INSTRUMENTAL: (i) Irish: ACCOMPANIMENT (Ó Siadhail 1985: 104) tá Cáit anseo le Máirtín be Cáit here X/Y Máirtín ‘Cáit is here with Máirtín.’ (ii) Irish: Tool (Ó Siadhail 1985: 104) tá mé ag scríobh le peann mór be I at writing X/Y pen big ‘I am writing with a big pen.’ 49. In Malayo, there is only syncretism between INSTRUMENT and GOAL, whereas the COMITATIVE has a distinct expression of its own (Amaya 1989: 58–61): (i) Malayo: TOOL duka gua-ga kauchu-mba dove kill-3 cauchera-Y ‘He killed the dove with a cauchera.’ (ii) Malayo: ACCOMPANIMENT ra kugu-bi nak-un I man-X come-PROG ‘I am coming with the man.’ 50. In Arrarnta, Instrumental and Comitative are encoded by distinct morphological means. The instrumental (formally identical with the locative [cf. example (A37)] and the ergative, cf. first case-marked NP in [i] below) is used in prototypical situations of instrumentality: (i) Arrarnta: TOOL (Wilkins 1989: 173) Ampe urreye-le thipe we-rlepe-ke pwerte-le child boy-Y bird pelt-do_along-PAST.COMPL rock-Y shanghai-le sling_shot-Y ‘The boy went along pelting birds with stones, with a sling shot.’ Note, however, that the morphological ablative is used to mark mechanical MEANS OF TRANSPORTATION (a context from which the instrumental seems to be banned): (ii) Arrarnta: Means of transportation (Wilkins 1989: 186) ayenge ingwenthe irrepelane-nge lhe-tyenhe ngkwenge-werne go-NONPAST.COMPL you:DAT-ALL I tomorrow plane-ABL ‘Tomorrow, I’ll be travelling by plane to where you are.’ 51. The sentence should not be misunderstood as meaning ‘the man is just coming with his six sisters’! 52. Some uses of the relator –for ABSTRACT INSTRUMENTS etc. for instance – will only be touched upon in passing. Likewise, we do not elaborate on valencybound Comitatives and Instrumentals in this chapter. This topic is postponed until the case studies in Part C. For the discussion of Comitatives and Instru-

Notes 457

53.

54. 55. 56.

mentals employed for MANNER, LOCATIVE and other categories, we refer the reader to Part B where we take up the issue of syncretism. As a matter of fact, many languages worldwide employ their Instrumental or Comitative categories to regularly mark the CAUSEE in causative constructions. Hungarian is a case in point (Palmer 1994: 220–221): (i) Hungarian level-et ira-tt-am a fiú-val letter-ACC write-PAST:CAUS-1SG DET boy-X/Y ‘I made the boy write the letter.’ We will come back to this issue in Part B when we discuss the peculiar relationship that holds between agentivity and Comitatives/Instrumentals. For the notion of body part instrument, cf. Biboviü (1976). In lieu of the prepositional phrase introduced by von, mit Hilfe may also require a dependent NP to be marked for genitive. Sentences such as, e.g., Agnes samt/nebst Werner trinkt/trinken Kaffee would best translate into English as ‘Agnes is drinking coffee and Werner too’, meaning that the two persons may act independently of each other in completely different locations. Note also that native speakers disagree as to whether samt/nebst require plural agreement on the finite verb or not (the latter being the solution prescribed by normative grammar), although both prepositions govern the dative, and thus plural agreement would be ruled out as it depends on a subject marked for nominative. However, samt and nebst semantically behave like the coordinating conjunction und – a similarity that leads some native speakers astray, especially because both prepositions are overwhelmingly (though not exclusively) used NP-internally, which is also reminiscent of und. Karl Philipp Moritz (1782) observed that mit and nebst have different meanings: “Nebst drückt nur so viel aus, daß ich mit jemandem einerlei Handlung verrichte, mit aber bezeichnet so viel, daß ich eine einzige Handlung mit jemandem unternehme (original italics).” [‘What nebst expresses is that I and somebody else carry out actions of the same kind, whereas mit indicates that I and somebody else carry out one and the same action.’]

57. Sentence (A40.2) can be reformulated in two versions: Either subject-oriented Agnes samt/nebst Werner unterhalten sich or object-oriented Agnes unterhält sich samt/nebst Werner. In the former case, the same observation made with respect to (A40.1) in the preceding footnote holds. Both participants are having conversations with anonymous third parties – but not with one another: Agnes is having a conversation and Werner too. In the latter case, however, the meaning of sich unterhalten changes from ‘to have a conversation’ to ‘to entertain oneself’. Agnes is the sole AGENT acting for the benefit of herself and Werner: Agnes is entertaining herself and Werner.

458 Notes 58. Thus, sentence (A40.1) for instance may be changed to Agnes trinkt Kaffee mit Hilfe von Werner ‘Agnes is drinking coffee aided by Werner’, where Werner is helping Agnes, who is likely physically disabeld to reach her cup of coffee and drink from it. 59. (A40.5) turns into something like Agnes geht mit Hilfe eines Regenschirms nach draußen ‘Agnes is going outside using an umbrella (to walk)’. The umbrella is no longer an object carried along by Agnes but the substitute for a walking stick she needs to perform the act of leaving the house. 60. Cf. for instance, Die Bundesregierung will die Arbeitslosigkeit mittels/durch besondere Maßnahmen reduzieren ‘The federal government plans to reduce unemployment by means of special measures’, where the special measures to be taken are more abstract rather than concrete tools etc. Mit Hilfe von is also acceptable in this context, whereas, unsurprisingly, samt, nebst and zusammen mit are absolutely inadmissible. 61. The dissociation of the two participants is probably strongest in the case of (A40.7): Agnes trinkt immer Kaffee und Milch ‘Agnes always drinks coffee and milk’, which is more suggestive of an enumeration of independent substances than of a mixture of liquids. 62. Disambiguation of the two readings under coordination can be achieved by adding the adverbial modifier zusammen either to the VP (preferred option: right-adjacent to the finite verb) or to the NP (under focus): (i) German Agnes und Werner (zusammen)trinken (zusammen)Kaffee (zusammen) Agnes and Werner (together) drink:PL (together) coffee (together) ‘Agnes and Werner are drinking coffee together.’ 63. It is tempting to interpret the syntactic variability as determined by iconicity. The closer the NP marked for comitative is positioned to the periphery of the clause, the more autonomy the participant represented by this NP is assigned. Admittedly, this hypothesis still needs to be corroborated empirically. 64. Unfortunately, we have not been able to get a clear picture of the NP-internal structures of this variety of Guaraní. Chances are that both ndive and reheve can be used in both adverbial and adnominal functions, although indisputable evidence is wanting. 65. The vowels alternate according to the rules of Mongolian vowel harmony. 66. The examples provided by our source do not include instances of TOOL but from the further explanations given, we deduce that TOOL (and MATERIAL) count among the functions fulfilled by the instrumental marker. 67. The inflectional comitative comes in two varieties, conventionally distinguished as comitative I and II (the latter often being composed of the markers of comitative I and instrumental). However, this distinction has no bearing on the semantic functions fulfilled by the two morphemes. As far as we understand the extant descriptions, it is more a matter of word class membership of the noun/pronoun hosting the comitative marker.

Notes 459 68. This statement is not entirely correct for the instrumental, as there are adjectives/participles governing the instrumental. Thus, a complex NP such as (i) Kalmyk (Benzing 1985: 63) antropolog-in kerg-är mergn kün knowledgeable man anthropology-GEN science-Y ‘anthropologist (lit. a man familiar with/versed in the science of anthropology)’ contains an instrumental as part of the attribute of the head noun. However, the instrumental itself is not the relator which establishes the attributive relation, as -är depends on mergn and not on kün. 69. Of course, Even also employs a regular instrumental case which can also be used for animate and human instrumentals: tar-ic b6j-ic {this}-{Y}# {man}-{Y}# ‘through this man’, nen-ac {dog}-{Y}# ‘with the dog’ (Benzing 1955: 63) (i) Even: Instrumental (Benzing 1955: 63) Iwan halka-c b6jcin Iwan hammer-Y work ‘Iwan is working with a hammer.’ Non-sentential examples for the various comitatives taken from Benzing (1955: 64–65) are: – Comitative I: nen-Ƽun {dog}-{X1}# ‘with the dog’, herka-Ƽu-mu {knife}-{X1}-{my}# ‘with my knife’ (= confective) – Comitative II: aman-ci-l {father}-{X2}-{PL}# ‘with his father’ – Comitative III: Petr-gali {Peter}-{X3}# ‘together with Peter’, oror-gali-w {tame_reindeer}-{X3}-{ACC}# ‘(wild reindeers) together with tame reindeers’ – Comitative IV: Pale-ja {Pelageja}-{X4}# ‘with Pelageja’ The comitatives may also be employed in adnominal attribution to cover ORNATIVE and CONFECTIVE, and possibly even possessive functions. 70. Owing to the exemplification practice of our source, the animacy criterion can only be tested for the companion, as the data only involve human and other animate ACCOMPANEES. 71. There is also a distinct instrumental -nohai which – like the bulk of the case markers – is not sensitive to number. (i) Tairora (Vincent 1973: 534) aiqu-nohai-ro ari-ro hit-3SG leg-Y-3SG ‘He hit it with his foot.’ 72. Some Papuan languages have similar number-dependent pairs of comitative morphemes, cf. genetically related Binumarien where -ndiri is used in the singular and -imbaqa in the plural. The markers in the two languages are obviously cognates. The same phenomenon is attested in the genetically and areally more remote Koiari, in which the two allomorphs of the comitative are specialised for number: The postposition vore is confined to ACCOMPANEES in the singular, whereas ruhuta can only be combined with plural nouns/pro-

460 Notes

73.

74.

75. 76.

77.

78.

nouns (Dutton 1996: 52). Note that in none of the cases is there any evidence for an etymological link between singular and plural comitatives! This lack of historical ties suggests that individual participants and groups of participants are conceptually differentiated. Number sensitivity of this kind is probably an areal trait of Papuan languages. There is a multi-purpose postposition which, inter alia, also covers instrumentality. This postposition has the two allomorphs ‘unai and ‘udiedi – the former is used with singular NPs, the latter with plural NPs (Mosel 1994: 15–16). (i) Saliba: singular instrument weku habulu-na ‘unai se kaiheya Y.SG 3PL play stone small-3SG ‘They played with a small stone.’ (ii) Saliba: plural instrument weku habulu-di haiyona ‘udiedi se kaiheya three Y.PL 3PL play stone small-3PL ‘They played with three samll stones.’ The two comitative markers are unique within the adpositional system of Saliba, as they are prepositions, whereas the majority of adpositions are postpositions. In addition, postpositions never co-encode alienability distinctions. Mosel (1994: 16) remarks that the plural subject pronoun se is motivated by the content of the proposition and not on formal grammatical criteria. In contradistinction to the comitatives, there is only one instrumental morpheme, namely, -tal (with a small set of phonologically conditioned allomorphs). Our source is definite about the fact that -tal is used for TOOLS and BODY PART instruments alike (Kroeber and Grace 1960: 91). An example for the former is: (i) Luiseño (Kroeber and Grace 1960: 181) poË-a.max mi • xuk kut • upi-tal pohu • -tal pó • -kun bow-Y arrow-Y 3SG-QUOT 3SG-hunting was po-še’ax mi • xuk 3SG-shooting was ‘It is said that he hunted with a bow that he shot with arrows.’ The Auyana instrumental morpheme is the invariable -nei: (i) Auyana (McKaughan and Marks 1973: 186) nointantaaq-nei tubuqmai what_thing-Y hit ‘What thing did he use to hit (with)?’ We adopt the term “focus” as it is used in the Philippinianist literature, although it does not correspond to the uses the term “focus” is put to elsewhere in general linguistics. The instrumental has a distinct encoding of its own. In (i), the instrumental focus morphology is exemplified (Drossard 1984: 41): (i) Tagalog i-p-in-utol niyá ang gulok DET knife Y-cut-PAT-cut he

Notes 461 ‘He cut with the knife.’ 79. Drossard (1991) makes a terminological distinction: the more UNDERGOERlike category is called “Komitativ”, whereas the ACTOR-oriented one bears the label “Soziativ”. 80. For our purpose, it is irrelevant to settle the question of whether these comitatives can be considered proper case-forms of the nominal paradigm (= traditional analysis) or special predicative forms, as suggested by Kämpfe and Volodin (1995: 54). 81. There is a distinct inflectional instrumental in -(t)e whose status as nominal case is not disputed: (i) Chukchi (Kämpfe and Volodin 1995: 102) kejn6-n tur-milger-e t6 m-nen new-rifle-Y kill-OBJ.3SG bear-ABS.SG ‘He killed a bear with the new rifle.’ 82. The comitatives are also used for ACCOMPANIMENT by human/animate companions as in (i) Chukchi (Kämpfe and Volodin 1995: 89) ge-ninqej-ine-mirg-e X1-boy-POSS-grandfather-X1 ‘together with the boy’s grandfather’ 83. Besides the frequently employed kanssa, there is an array of other postpositions, all of which fulfil some kind of Comitative-like functions, as, e.g., mukanaan etc. (Stoebke 1968, Stolz 1993: 60). However, we consider these to be secondary markers at best. The same applies to the extraordinary use of the preposition kera ‘with’: (i) Finnish (Sulkala and Kirjalainen 1992: 225) hän tuli kera koira-nsa dog-POSS.3 he come:IMPERF X3 ‘He came with his dog.’ This rich inventory of partially synonymous postpositions notwithstanding, kanssa has all characteristics of the unmarked case. According to Toivainen (1980: 145), it is acquired and mastered early on in child-language development and surpasses all other means of case-/role-marking as to frequency in Finnish child language. 84. Sulkala and Kirjalainen (1992: 226) give an example for the instructive but emphasise that it is a separate adverbial (= not a NP-internal attribute): (i) Finnish sie-llä oli mies avopä-in bare_head-Y2 there-Y1 be:PAST man ‘There was a bare-headed man there.’ If it is indeed an adverbial, then the literal reading would be something like: ‘There was a man there – bare-headed’.

462 Notes 85. In a rather polemic short note, Hakulinen (1949: 173) mentions comitative attributes only as examples of bad style in newspaper adds: (i) Finnish huone alkove-ine-en vuokrattavana room alcove-X1-POSS.3 search ‘searching for a room (provided) with alcoves’ Interestingly, Nau (1995: 134–135) also discusses ‘bad newspaper style’ in her treatment of the inroads kanssa has been making into the former territory of the inflectional comitative: Where Hakulinen refuted to accept the attributive comitative case, it is now the postposition which, form the point of view of normative grammar, is out of place – but is nevertheless frequently employed, especially in regions where the Swedish adstrate has been strong. 86. Besides adjectival attribution and nominal compounding, the elative and the partitive case are appropriate for the encoding of MATERIAL: (i) Finnish (Sulkala and Karjalainen 1992: 230) talo on tiil-tä/ tiile-stä house is brick-PART/ brick-ELA ‘The house is made of brick.’ Moreover, the case of the HUMAN/ANIMATE INSTRUMENT is a bit tricky, as a complex postpositional structure is commonly used instead of the simple adessive. Normally, one would choose N[+animate]-GEN + avulla – a construction which contains the noun apu ‘help’ in the appropriate adessive as a denominal postposition in statu nascendi. 87. The use of the inflectional adessive is restricted to animate possessors. Inanimate possessors bear the inessive suffix instead: (i) Finnish (Sulkala and Karjalainen 1992: 227) puu-ssa on oks-i-a branch-PL-PART tree-INESS is 88. Kross (1983) surveys the remnants of the instructive in the smaller Baltic Finnic languages. She claims that the instructive has been artificially revived and its functional domain extended in standard Finnish. Nevertheless, Kross (1983: 244–245) too considers the marking of BODY PART instruments the most prominent function of the instructive. 89. Dani employs the relators eppetak – if both participants are human – and inom – irrespective of animacy, for ACCOMPANIMENT and various Comitative-like functions (Bromley 1981: 90–91). (i) Dani: Human participants n-eppetak hakot lvok-a INCL-X1 later go:ADHORT-INTERR ‘Shall we two go together later?’ 90. Animacy also determines the choice of avec or à although perhaps only on a moderate scale:

Notes 463

91.

92.

93.

94.

(i) French: Animate (Klein and Kleineidam 1983: 175) j-ai traduit ce passage avec l-aide de mon frère I-have.1SG translated this passage X/Y DET-help Y5 my brother ‘I have translated this passage with the help of my brother.’ (ii) French: Inanimate j-ai traduit ce passage à l-aide d-un dictionnaire I-have.1SG translated this passage Y4 DET-help Y5-a dictionary ‘I have translated this passage with the help of a dictionary.’ In this respect, en competes with de and à (Klein and Kleineidam 1983: 174–175): (i) French: MATERIAL I je me suis acheté un pull de laine sweater Y5 wool I REFL.1SG be:1SG bought a ‘I have bought myself a woollen sweater.’ (ii) French: MATERIAL II j-ai trouvé une montre en or dans le parc watch Y3 gold in DET park I-have:1SG found a ‘I have found a golden watch in the park.’ (iii) French: MATERIAL III Eve fait des dessins à l-encre Eve makes Y5:PL pictures Y4 DET-ink ‘Eve draws pictures with ink.’ Par has even a much wider functional domain, but in many cases par is only a secondary marker. This is different for ABSTRACT INSTRUMENTS (Klein and Kleineidam 1983: 176): (i) French: ABSTRACT INSTRUMENT il veut arriver à ses fins par tous les moyens DET:PL means he wants arrive:INF Y4 his:PL goal Y2 all ‘He intents to get to his goals by all means.’ In many other languages we reviewed, spatial expressions are of course widely used with MEANS OF TRANSPORTATION. German, for instance, allows for similar choices as French with in ‘in’ being used for container-like vehicles and auf ‘on’ for those which provide an open surface. Note, however, that riding on a bicycle can also be expressed by (alongside à) en ‘in’ in French (and auf in German). In the absence of any reliable frequency counts, we are not in a position to decide whether the use of spatial prepositions in French is more normal/unmarked in French as opposed to German, where – according to our native-speaker intuition – the spatial expressions are secondary markers only. For reasons of space, this issue cannot be settled here. Follow-up studies may take up this problem in the years to come. For the sake of readability, the table simplifies the facts a bit, as not all details are given. We have also refrained from overinterpreting our Dani source and have thus decided to leave three cells empty (or better: put a question mark there). The numerical indices on the Finnish case exponent -in are meant to

464 Notes distinguish the genitive (= -in1) from the instructive (= -in2), as they are formally identical for many Finnish nouns in the singular. 95. On superficial inspection, this observation seems also to hold for French en. This is, however, a false impression occasioned by our selective list of contexts: en is used adnominally too as example (ii) in footnote 91 clearly demonstrates. It could either be subsumed under part-whole relationships, or an additional INGREDIENT relation would have to be created. 96. This simple method of course has various shortcomings. First of all, native speakers may not be available. Secondly, languages may have several markers specialised for different categories of comitativity or instrumentality. Thus, the test procedure has to be repeated with every possible context. 97. The lexicographers may not be native speakers of one (or more) of the languages involved in their project. Thus, the internal organisation they impose on a lexicon entry may inadvertently be determined by the patterns they are familiar with from their native language. Moreover, there are traditions in lexicography with a long history or succession of dictionary projects which either draw extensively on the preceding ones or make an effort to describe the facts in a distinct way. 98. We were surprised to see how many descriptive grammars either make do with just a remark in passing or even keep silent about Comitatives (and offer even less about Instrumentals). Often, it turned out to be necessary to gather the data directly from original texts. 99. This usually applies to overt coding strategies – among which we include the processual ones of the incorporation kind, cf. (A71). There are of course other strategies corresponding to what Whorf (1988: 88) termed covert categories, namely distributional properties such as combination restrictions, etc. As Martínez (2001: 337) rightly remarks, their strict application would ultimately lead to denying the existence of syncretism. To our mind, this problem disappears as soon as the primary marker concept is employed: Distributional facts are secondary properties of primary linguistic signs. The latter thus take precedence in our analysis. 100. In Mari, the pertinent relators for Comitative and Instrumental are postpositions (retrieved from the glossary in Alhoniemi [1993: 195–248], namely ðene, pelen, p}6rl’a, saB a ) . However, their distribution over instrumentality and comitativity is still largely unclear to us (at least ðene seems to be a good candidate for Comitative, as it is also used as the coordinating conjunction ‘and’). 101. We do not discuss the morphological and morphosyntactic differences of Comitatives and Instrumentals whose relators may belong to different distribution classes and whose hosts may also belong to different word classes. These issues, along with zero-marking, will be dealt with in Part B, where we turn our attention to markedness relations.

Notes 465 102. Guessing from the few lines that Blake (1994) devotes to comitatives, as opposed to the much more frequent references to instrumentals, Comitatives do not often come in the shape of bound case-morphemes, whereas Instrumentals are less of a rarity in inflecional case-paradigms. This striking difference is of interest for the determination of markedness relations, cf. Part C. 103. This is trait is common not only to numerous North American Indian languages of different genetic stock but reaches also far beyond the boundaries of the North American area into Mesoamerica (Palancar 1999: 160). We find evidence for similar phenomena for instance in Totonac (Levy 1992) and Popoloca (Veerman-Leichsenring 1992). 104. For a discussion of related phenomena in Classical Nahuatl and modern Yucatec Maya, cf. (Stolz 1994: 68–75; Lehmann and Shin 2005). 105. The fact that there are often (= not only in Takelma) serious difficulties when it comes to distinguishing spatial and Instrumental/Comitative categories is a further topic for our discussion of syncretisms in Part B. 106. Wa is also used as a free postposition encoding instrumental relations such as, e.g., wayƗ’a wa ‘with his knife’ (Sapir 1922: 68). 107. Double instrumentals are not uncommon in Takelma. In these cases, the body part prefix precedes the general instrumental prefix (Sapir 1922: 69). (i) Takelma dan k!ama p!ai-İƯ-wa-sgƗ’ak’sgig-i’İn rock tongs down-hand-Y-pick_up-1SG ‘I pick up the rocks with the tongs and put them down.’ In this and similar cases, the body part prefix indicates what is sometimes called the ‘primary instrument’: If you use tongs for tools, you automatically also use your hand(s) to hold the tongs in the appropriate way. The use of the tongs presupposes the use of the hands. 108. There are of course more issues of a morphological nature that could be discussed at length here. As this study is not primarily devoted to morphology, however, we refrain from going into the details of too many side issues. Suffice it to mention that the word class membership of the morphological host of the bound relators is also interesting. In our presentation of the facts, nouns or NPs are by far the most common host for markers of Comitativity/Instrumentality. However, this does not exhaust the possibilities. Comitative and Instrumental are often encoded by bound morphemes on the finite verb, especially in headmarking languages. It is not important in this respect whether the language distinguishes the two categories formally or not. In Aleut, there are two possibilities of encoding the Comitative, namely either by the (de)verbal relator agiita-, the particle as(ix) or the verbal derivation in -Vsa-. Only in the latter case is the Comitative identical to the Instrumental (Bergsland 1994: 488). 109. For an extensive review, cf. Stolz (1997d). 110. The fact that Palmer (1994) hardly ever mentions COMITATIVES in his book on grammatical relations is misleading because there is in fact an enormous

466 Notes amount of linguistic publications devoted to comitatives, Comitatives, COMITATIVES and CONCOMITANCE, covering a wide range of topics from historical morphology to language philosophy. 111. One of the earliest attestations of the terms comitative and instrumental as distinct categories is Schleicher (1861: 454). Winfried Boeder (p.c.) brought to our attention a long passage from the writings of Basilius of Caesarea, *AD 330, (Blum 1967: 89–93), in which the semantics of Comitative-Instrumental relators, spatial relators and the coordinating conjunction as used in the Bible are insightfully discussed at length. 112. It goes without saying that Hjelmslev’s (1972) classic treatment of linguistic case published in the 1930s also covers instrumentals and comitatives. 113. Many of these early contributions were published in the philosophy-oflanguage oriented journal Foundations of Language, whose very title underlines the fact that a clarification of the problems posed by COMITATIVES and INSTRUMENTALS was considered of much relevance to linguistics in general. In the same period, and partly also in the same journal, a number of studies were published which focussed on the affinities between Comitatives and coordination, cf. Shou-Hsin (1970) and Hetzron (1973). 114. Reviewed in detail by Stolz (1997e). 115. Haspelmath (2000) discusses the most common misunderstandings which make communication between functionalists and formalists so difficult. On the one hand, we support Haspelmath’s attempt at clarifying a number of problems. However, his notion of functions is still too narrowly grammaroriented to do justice to the full range of characteristics of human language. 116. Leiss’s radical ideas have been put into practice by Bittner (1995: 123–124). 117. We have coined this term expressly for the present purposes. It is meant to account for the fact that linguistic signs interact with each other according to principles of form and content resemblance. A Zeichenverband captures these relations and turns the associated signs into a kind of higher-order unit. 118. Haspelmath (2003: 239) even mentions psycholinguistic experiments which shatter the foundations of the widely held belief that temporal and spatial uses of local prepositions such as English in (February/Leipzig) share their representations. 119. In this quote, Nilsen alludes to an unpublished dissertation by Shroyer (1969), which we have not consulted ourselves, as we deemed it unnecessary for the present purpose. 120. This is even less so with suggestions based on assumptions of economy to impose an upper limit to the number of case distinctions in order to keep the model operative, cf. the discussion of the various proposals in Rauh (1988). 121. For the term co-subject, cf. Hansen (1971). 122. In Finnish philology, infinitives are labelled numerically: The 1st infinitive serves as the lexical entry of verbs, the 2nd infinitive infinitive functions as a

Notes 467 gerund, the 3rd infinitive is used for various purposes (e.g. purposive) and the 4th infinitive is a verb-noun (Karlsson 1978). 123. Much to our surprise, Seiler’s (1974a–c) publications must have escaped Schlesinger’s attention, for nowhere in his study does he acknowledge the many similarities of his own approach and reasoning with the Principle of CONCOMITANCE. 124. CAUSE: to shiver with cold, LOCATION: to stay with somebody, TIME: to mellow with age, so-called SPECIFICATIONS OF THE ACTION: to go on with one’s work, obligatory (= valency-bound) with-phrases: to deal with something, and miscellaneous: The storm broke with loud peals of thunder. We argue that it makes sense to search for a common denominator of these cases as well, instead of discarding them as belonging somewhere else. 125. In various places, Schlesinger (1995: 26) talks of “syntactic rules” instead of the unspecific “linguistic regularity”. 126. Consider also Werner (1991), who demonstrates that a microscopic analysis of language structures inevitably reduces to a minimum the possibilities of forming larger classes of phenomena defined by similarity. The problem of incomparability on the microscopic level can only be solved by restricting the comparison to the mesolevel (the macrolevel being less suited as it is the level where similarities tend to dominate). 127. Croft’s idealistic claims cannot be realised because language involves potentially unlimited creativity on the part of the language users. If we do not opt for a corpus-based approach (which is certainly not the best solution for what Croft has in mind), we will have to find a way of determining that, at a certain point, the full list of possible contexts is complete. However, the very nature of spoken language renders this decision an artificial imposition of the descriptive linguist – because the next distributionally relevant fact may always be waiting for us around the next corner. 128. There is another very important idea that Schlesinger and Croft (perhaps unknowingly) share: Schlesinger (1995: 70 [cf. quote above]) downplays the occurrence of with outside the domain of his C-case as evidence for historical processes, meaning diachronic semantic extensions which at present may no longer be felt (by the native speaker) to belong together. In various places, Croft (2001) emphasises that the co-presence of evidence showing various stages of of a given element’s grammaticalisation should not be confounded with elements which are synchronically associated with each other by the speakers. To his mind, earlier stages of grammaticalisation are part of language history. Their use is conventionalised, and the semantic motivation is no longer retrievable for the speakers. Both are surely right, as long as the grammaticalised and source elements differ on the level of phonological form. However, if grammaticalisation has not (yet) affected the phonological shape of the element, speakers may still associate the various variants with one an-

468 Notes other. This potential association suffices to reject the idea that we are dealing with distinct but homophonous entities. 129. One could follow Croft’s own practice and use indexed labels as, e.g., ComitativeKorean, ComitativeEnglish, ComitativeTurkish and so on. The corresponding domains of Comitatives in individual languages could then be compared by way of identifying their zones of overlap. Ultimately, a language-independent general Comitativehuman language emerges. This is of course one of the most common strategies in typologically-inspired research to come to empirically well-founded generalisations. In the case of Comitatives, the situation suffers from two shortcomings. Firstly, (semantic/conceptual) networks as we understand them have not been the object of many crosslinguistic studies. Secondly, the extant candidates for the status of descriptions of language-specific COMITATIVES/INSTRUMENTALS are often fragmentary and equally often only partly compatible because their proponents adhere to different theoretical frameworks. Thus, we cannot simply add up the various results to gain new and valid insights. On the one hand, there has always been a certain emphasis on English as the object language in studies on Comitatives. However, especially in the later years, other languages have come into focus as well. We mention a few of the publications devoted to individual languages (or minimal samples) that we have published since the mid-1970s. Other pertinent titles are mentioned in due course in the remainder of this book. It is worth noting that about 50% of all titles on our subject matter in Schwarz, Stroh, and Urdze (2001) are devoted to the instrumental in Slavic languages. For our purposes we have drawn, amongst others, on Faarlund (1974) on Norwegian, Lötzsch (1978) on Latvian, Wierzbicka (1980) on Russian, Craúoveanu (1980) on Romanian, Ariste (1983) on Vod, Szabó (1984) on Kola Saami, Dyla (1988) on Polish, Aissen (1989) on Tzotzil, Müller (1990) on Old Irish, Menges (1991) on Turkic and Altaic languages, Janda (1993) on Russian and Czech, Stachowski (1994) and (1995) on Yakut and Dolgan, Girke (1995) on Russian, Endruschat (1997) on Afro-Portuguese, Obst (2002) again on Russian, Luraghi (2001b) on Indo-European, Martínez (2001) on Greek, Fonseca (1994) on Portuguese, etc. The emphasis on Malagasy in Kempf (1974) is also notable. Before the 1990s, comparative attempts are rare. Nilsen (1973) and Schlesinger (1979) also mention languages other than English without basing their research and results on a decent analysis of these additional languages. Their samples are small-sized ones – and Nau (1995: 188) criticises their interpretation of the non-English data as faulty. In addition to the above mentioned publications and many others, there are also long lists of studies on Comitatives and Instrumentals with a strong orientation towards the needs of foreign language teaching (the contrastive grammar approach), such as the numerous contributions to the problem of German mit and its equivalents in Polish, cf. Schröder (1989). Psycholinguistic contributions to the problem at hand are not as nu-

Notes 469 merous. Clark and Carpenter (1989), for instance, give an account of the use the prepositions by and with are put to in English child language. 130. One might ask how it is possible to accept the results of typological language comparison if the data input stems from descriptions which all suffer from the disadvantage of methodological opportunism: If, for this reason, they are automatically only fragmentary and unreliable, the simple adding up of their results and the identification of a common core is also dubious at the very least. Otherwise, one would have to accept that a lot of methodological opportunism is less than a bit of it. 131. Croft, Shyldkrot, and Kemmer (1988), Kemmer (1993) and Michaelis and Rosalie (2000) are exemplary studies in which semantic mapping is applied for crosslinguistic research. On the theory of semantic maps, cf. van der Auwera and Plungian (1998) and Haspelmath (2003). We come back to the issue of semantic maps in Part B. 132. Owing to the fact that Croft’s suggestions have triggered a lively discussion about syncretism, its cognitive foundations and its importance for linguistic theory in general, we will re-examine in Part B the two classes of roles postulated by Croft, where we also consider the contributions by Croft’s discussants. 133. We are not sure whether Croft (2001) still subscribes to this opinion, given the fact that he now advocates a radicalised version of construction grammar, according to which there is no place for so-called atomic elements, in other words, elements below the construction level have no independent existence. 134. An example of this variety is of course the early version of construction grammar as described in Fillmore, Kay, and O’Connor (1988). 135. Generally speaking, invoking the context as the decisive factor is tantamount to demoting the component parts of constructions to the status of distinctive features or the like. At the same time, context in itself is not self-explanatory. If one does not axiomatically require the upper limit of a construction to coincide with, say, the sentence, we are entitled to consider higher-level units on the discourse and text levels to be constructions (including entire texts – and one could even try to extend this boundary in such a way that it comprises genres and perhaps the entire [and panchronic!] universe of communicative acts realised in a given speech-community). It is probably hard to find naturally defined limits for the notions of construction and context. Therefore, limits will have to be imposed on theoretical grounds or by simple stipulation. 136. This is particularly clear if context/construction goes beyond the sentence level. A whole text may be seen as a linguistic sign having unique form and function (one may also think of indirect speech acts and similar phenomena). At the same time, this text is made up of several utterances each with an internal structure, etc. which can also lay claim to the status of linguistic signs. Ergo: One sign contains a wide variety of other signs. As long as there is no convincing suggestion for a more restrictive definition of linguistic sign (which identifies the intermediate level of constructions as the locus of sign

470 Notes formation), there is no way of denying sign status to grammemes embedded in larger units, because we also assign sign status to constructions which are part of larger units, no matter whether the meaning of the largest unit can actually be interpreted in terms of compositionality. 137. Times have changed, however: Kilian-Hatz and Stolz (1993) still take issue with the idea of the form-first principle, which is now at the bottom of our approach. 138. For a critical evaluation of her thoughts, cf. the contributions by her discussants (Dryer 2000; Perkins 2000; Schweiger 2000). 139. There are various publications which deal with Comitatives/Instrumental from a dialectological perspective, cf. Obenland (1967). As far as we can see, these studies do not call for any revisions of non-dialectal based hypotheses about the categories under scrutiny. 140. In the end, instead of the usual hundreds of SVO languages, we would get thousands of SVO varieties or even several hundred millions of SVO idiolects. This does not seem to be a desirable result. Perhaps, dialectal and idiolectal evidence gains relevance only if there are indeed varieties which do not conform to the trend typical of the standard language and/or the majority of its non-standard varieties. 141. To our mind, it is almost a contradiction in terms to use terms such as, e.g., fine-grained typology for the detailed description of an individual language. Typology is not co-extensive with language description and cannot be reduced to treatments of single languages, as it by definition requires comparison – meaning comparison of several languages on equal terms. Thus, it is not sufficient to check a given language for the realisation of typologically-based parameters. This approach to language description can be called typologicallyinspired description instead of plain and simple typology. 142. For detailed information about the text and the translations, cf. Part C. 143. To mention only two examples for the wide acceptance of the universal: Heine, Claudi, and Hünnemeyer (1991: 103–105) put the Lakoff and Johnson universal at the heart of their own conceptual framework of grammaticalisation. Even Schlesinger (1995: 69–70, footnotes 6–7) claims that the universal speaks in favour of his C-case model. 144. Heine, Claudi, and Hünnemeyer (1991: 36) operate with so-called basic propositions which bear some likeness to the propositionally organised metaphors. These propositions are conceived of as complex cognitive structures “that appear to be basic to human experience”. Among them we find the Comitative Proposition. 145. The languages forming part of this sample are enumerated in the Appendix I. 146. The main difference between the two papers lies in the fact that the earlier one allowed for extinct languages such as, e.g., Latin, Old Aramaic, Classical Nahuatl, etc. to be part of the sample, whereas the more recent article accounts exclusively for languages which are still actively spoken by native speakers.

Notes 471 147. For simplicity’s sake, we have looked only at large areas of continental expansion. Oceania comprises the Pacific islands, Australia and the large island of Newguinea, whereas the Philippines and the Indonesian archipelago are part of Asia. The dividing line between Europe and Asia runs along the heights of the Ural down to the Caspian Sea and along the boundaries between the Southcaucasian republics and Iran. The entire Turkish territory belongs to Europe. Malta and Greenland are also ‘incorporated’ into our concept of Europe. Armenian, Georgian, Azerbaijani, Turkish, Kurdish, Maltese and Greenlandic are meant to make the linguistic map of Europe more interesting. 148. Notwithstanding our many reservations against the lumping approach in genetic linguistics, we have adopted Ruhlen’s (1987) list of macrophyla etc. because it is a relatively handy inventory. We are of course aware of the fact that several of his genetic super-families have not found approval by the specialists. As the statistics to be provided below only allow statements on the macro-level, there is no need for a more sophisticated genealogy. 149. In previous publications, we have used the labels COHERENT, INCOHERENT and MIXED extensively for the language types we identify. These labels were meant to be suggestive of either conformity (= COHERENT) or nonconformity (= INCOHERENT) of a given language with the universal based on the Companion Metaphor (Stolz 1997a: 125–126). Owing to the fact that these terms can easily be misunderstood, we have replaced them by less imaginative labels. 150. For the sake of brevity, we skip those cases where three markers coexist, of which two are specialised to either Comitative or Instrumental and one is syncretistic. Where such constellations are reported, further investigations are called for to clarify whether all of the competing markers deserve the designation primary marker. 151. Its token frequency is relatively low. For some native speakers, -stul/-stül are stylistically marked and classified as typically rural. It remains to be investigated whether it would be more appropriate to consider the sociative to be a secondary marker. 152. Again, the question arises as to which status can be ascribed to téì and ku. The latter is surely a primary marker in the realm of Comitativity. This is less evident with téì in contexts where Instrumentality applies. 153. Veenstra (2002: 337) classifies one case of téì as an example of the Comitative. (i) Saramaccan mi téì hen gó a dí sikóutu I Y him go LOC DET police ‘I took him to the police.’ As gó is also used as a heavily grammaticalised non-initial member of serialverb constructions corresponding to a directional marker (Veenstra 2002: 328–330), we consider this example a direct translation of the English to take sb. to LOC and thus do not classify it as evidence for a Comitative use of téì.

472 Notes 154. This picture is also corroborated by those languages which are additionally represented in our questionnaire. Of these eight languages, four (Karaim, Kazakh, Tatar and Ndyuka-Trio Pidgin) represent type B – the three first mentioned being members of the Altaic macrophylum and spoken in Europe or on its borders. Ndyuka-Trio Pidgin is an English-based Creole. All non-A languages (Selkup, Tetun, Tukang Besi and Zulu) are spoken outside Europe. A much broader range of macrophyla (Uralo-Yukaghir, Austronesian, NigerKordofanian) is represented. Tukang Besi is the only represenative of type C. The other three languages belong to type A. 155. According to the semantic maps provided by Michaelis and Rosalie (2000: 92–97), many Creoles are B-type languages with a few representatives of the C-type. Owing to the fact that Michaelis and Rosalie (2000) survey only a small number of languages, this generalisation of ours is tentative. 156. In connection with verba dicendi, the complex relator -kƝ sƗth can be replaced by the otherwise Instrumental postposition se. This is, however, the only context in which the two expressions are in what is presumably free variation (Stolz and Stroh 2001: 399). (i) Hindi (Bender 1967: 82) laDŽkƗ aurtõ kƝ-sƗth bǀltƗ hai talk be.3SG boy woman:OBL.PL X ‘The boy is talking with the women.’ (ii) Hindi (McGregor 1977: 21) maim us-se hameĞƗ hindƯ boltƗ hnjm I he-Y always Hindi speak be:1SG ‘I always talk with/to him in Hindi.’ 157. As to an explanation of why combinations of Instrumental and Causative are missing from the table, cf. below. 158. As to an explanation of why combinations of Instrumental, Causative and additional categories other than the Comitative are missing from the table, cf. below. 159. We have to be cautious though: For practical reasons, Stolz (1997a) only accounts for the first seven categories that he finds to combine in syncretistic patterns with Comitatives and Instrumentals. Other categories were systematically ignored and thus it cannot be ruled out that the picture will change if other categories are taken account of too. 160. The categories are listed in top-down order according to the decreasing percentage of their areal foci – marked grey in the table. 161. Palancar (2002) provides the most comprehensive study of the origins of AGENT markers. INSTRUMENTALS are admittedly a major source of AGENT markers (Palancar 2002: 88–114) but are by no means the only one, as will be shown below. However, Palancar’s meticulous study amply corroborates that the distinct COMITATIVE itself is not a viable direct source for the development of AGENT markers.

Notes 473 162. Originally, the list of unattested combinations also included InstrumentalCausative. This pair of categories has been cancelled from (B29), as we now have sufficient evidence disproving this special facet of the former claim of Stolz (1997a), whose problems were caused by the difficulties which arise when having to tell cases of human instrument (A40.10) and cases of causatives proper apart. The absence of ternary and more extended syncretistic patterns involving Instrumental and Causative in (B23) is largely due to the same problems and must therefore be revised. According to Comrie (1985: 339), causees in Finnish causative constructions based on transitive verbs are usually marked by the adessive, i.e., by the same inflectional case which is also employed to encode Instrumentality, cf. (A57) above. (i) Finnish minä rakennut-i-n talo-n muurare-i-lla I build:CAUS-PAST-1SG house-GEN bricklayer-PL-Y ‘I had the bricklayers build the house.’ As Finnish has accusative alignment and only knows an impersonal passive, the causative is the only category where [+agentivity] comes into play for syncretism. 163. The member languages of this sample are identified in the Appendix I. 164. It is interesting to see that of the 50 European languages, the non-IndoEuropean ones opt overwhelmingly against pattern II (eight out of twelve languages). Of the 38 Indo-European languages spoken in Europe, those belonging to the Romance and Germanic phyla in their entirety choose pattern II (nine and ten languages, respectively). Only Slavic and various IndoEuropean languages do not show a predilection for the same pattern (two thirds of the Slavic languages prefer pattern III, whereas the minor IndoEuropean languages are evenly divided between patterns I, II and IV), cf. (Stolz 2001b: 162). The statistical weight of pattern II in the sample therefore rests to considerable extent (19 out of 40 languages = 47.5%) on typical SAE languages of the Romance and Germanic phyla. 165. Ergative-Instrumental syncretism is particularly well attested in Australia (cf. [B66]). However, Ergatives and Instrumentals often syncretise in ergative languages of other continents: In Caucasian Udi, for instance, Ergative and Instrumental are both encoded by -en, whereas the Comitative receives a distinct marker -Vxol (Harris 2002: 24–25); and in the Kiranti language Dumi (Nepal), Ergative and Instrumental share the expression -/a, whereas the comitative suffix is -k´y (Van Driem 1993: 62–75). On Dumi, cf. also below. 166. The Cushitic language Oromo is also characterised by syncretism of the Instrumental and the passive Agent, as both are encoded by -dhaan ~ -tiin ~ -n (Griefenow-Mewis and Bitima 1994: 163–164). However, the information as to Comitativity is too scarce (actually nonexistent) in our source to allow for any definite statements. We therefore can only speculate that Oromo represents the A-type and perhaps belongs to the class of AND-languages. The co-

474 Notes ordinator is -f(aa), which is different from a number of adverbials like watiin meaning ‘together’. 167. As in many languages worldwide (Siewierska 1984), the passive is normally agentless in Farsi, though avoiding the agent is not compulsory (Mahootian 1997: 143). There are also two distinct causatives, in neither of which the marking of the causee involves ba (Mahootian 1997: 225–226). 168. In other varieties of Quechua, -wan may also mark the passive agent. This is the case in Ecuadorian Quechua, where -wan is rendered -huan orthographically (Catta Quelen 1987: 238): (i) Quechua (Ecuadoriano) llama-ca atug-huan micu-shca eat-PASS lama-tOP lobo-X/Y ‘The lama was eaten by the lobo.’ 169. In §§186–187, Loogman (1965: 124–125) states that “[t]he means or instrument is normally indicated by kwa, and always so when the personal agent is mentioned in the same sentence. If however the agent is not mentioned, then na may be used to indicate the means or instrument. […] Both the personal agent and the instrument may be indicated by the particle na; but when na is used to introduce the agent, it should not be used a second time in the same sentence to indicate the instrument.”

On the basis of this rule, the following correct sentences can be produced: (i) Swahili a-li-pig-wa na fimbo a-li-pig-wa kwa fimbo a-li-pig-wa 0 fimbo a-li-pig-wa 0 fimbo na baba stick X father 3SG-PERF-beat-PASS (Y) a-li-pig-wa na baba a-li-pig-wa na baba kwa fimbo father Y stick 3SG-PERF-beat-PASS X ‘He was beaten (by his father) with a stick.’ Whereas the logically possible combination *alipigwa kwa baba na fimbo is ruled out, since the passive Agent is marked by the wrong relator (Loogman 1965: 125). At present, we have no evidence for coordinated passive Agents or for coordinated Instruments, and thus we cannot ascertain whether the restrictions upon the multiple use of na also extends to these cases. Irrespective of this empirical gap, we are in a position to classify Swahili as a language which tends to avoid homophony. While this is explicable from a functional perspective, other languages suggest that homophony avoidance is not compulsory. In Dehu, an Austronesian A-type WITH-language of the Loyality Islands, Ergative and Instrumental are syncretistic, as they are both encoded by the preposition hnen (Moyse-Faurie 1983: 151):

Notes 475 (ii) Dehu hnen la joxu hna humuth la kuli hnen la sine-sinöe DET chief PAST kill DET dog Y DET piece-wood Y ‘The chief killed the dog with a stick.’ In the above example, hnen occurs twice, fulfilling two distinct functions. The co-occurrence of several instances of hnen in one and the same sentence is rather common in Dehu. The only rule to be observed is the non-adjacency of the NPs introduced by hnen: “L’agent est déjà marqué par le fonctionnel hnen; il vient en tête d’énoncé, et l’instrumental arrive après le prédicat (et l’objet). La présence d’un auteur virtuel (instrument) implique la thématisation de l’auteur réel (agent)” (MoyseFaurie 1983: 151). [The agent is already marked by the function word hnen, it is in utterance-initial position and the instrumental is placed after the predicate (and the object). The presence of a virtual actor (instrument) requires the topicalisation of the actual actor (agent)].

For a similarly ‘tolerant’ attitude towards multiple employment of polysemous markers in Aymara, cf. (B56). 170. Haspelmath (2003: 228) introduces this term explicitly for the “comitativelike participant that takes active part in the action”. The co-agent is a virtual category to be located between the roles recipient and comitative. It remains to be seen what is left of the comitative if the co-agent is singled out as a distinct category. 171. On the basis of a very large sample (= about 800 languages), Hagège (1986: 35) estimates that languages in which Comitative(-Instrumental) and AND are syncretistic cover about 19% of the world’s languages. With 97 relators out of 645, Stolz’s (1997a) attests to a percentage of almost 15% of all relators. 172. Wälchli’s (2003: 4) distinction of natural vs. accidental coordination may also be an important parameter, although we have to postpone a detailed study of these aspects for a later time. 173. In Stolz (1997a), one language (Bamanankan) was reported to display Instrumental-AND syncretism without parallel cases involving the Comitative. This classification was based on fragmentary data. Our new data support a reclassification of this exceptional case among COM-INS + AND. 174. Leger (1994: 147) explicitly describes the functional domain of káa as predominantly Instrumental-like, although he also mentions the following example: (i) Kwami yì wòttu káa yìË ì he come Y him ‘He came with him.’ The exact status of the two participants is unclear. The reading could perhaps be ‘he brought him along’ or ‘he went along with him’.

476 Notes 175. As far as we can tell, there is no convincing evidence for functionally specialised AND conjunctions besides the distinction of NP-conjunction vs. clause-/sentenceconjunction. We have not encountered cases of coordinating conjunctions which are reserved for semantically defined classes of NPs, such as more instrument-like ones vs. more comitative-like ones. But cf. Comrie (1982), who discusses casemarking on conjunctions in Huichol. 176. Just to show how far group formation under coordination can go grammatically speaking, we add examples from Afrikaans (other languages display similar phenomena). Afrikaans is a typical AND-language. In coordinated NPs, if the first constituent is a personal pronoun and the second a proper noun, the usual morphosyntactic rules determining the use of the oblique case form of the pronoun become optional. The coordinator is en ‘and’. In (a), the usual possessive pronoun of the first person singular should be my instead of *ek se, the oblique forms of ek in (b)–(c) would equally be my. (i) Afrikaans (Ponelis 1978: 616) (a) Possessor ek en Jan se metode werk die beste I and Jan POSS method work DET best ‘The method of Jan and me works out the best.’ (b) Direct Object hulle het ek en Koos vermy they have I and Koosavoid ‘They have avoided me and Koos.’ (c) Complement of a preposition hulle spot met ek en Koos. they jest X/Y I and Koos ‘They make fun of me and Koos.’ Ponelis (1978: 616) explains this behaviour as follows: The coordinated NPs are interpreted as forming a unit and not a combination of two distinct units, i.e., syntax cedes to semantics. 177. In standard Russian, singular vs. plural agreement with NP s NP (NP with NP) constructions is regulated according to the criterion of status equality: If the NPs are equal in status, the plural is used, otherwise the singular is used (Girke 1995). 178. That is the reason why -ntin-/-ndin- is also considered as a kind of ‘secondary’ number marker with a collective reading (Catta Quelen 1987: 222–223). 179. Both nouns belong to class 6. However, as they are personified in the story, they are treated like nouns of class 1. Thus the use of wa- instead of ya-/zi- is a matter of style. For the intricate rules of class-agreement on verbs for coordinated NPs with different class membership of their constituents, cf. Loogman (1965: 327–328). 180. For the Comitative relator and its functional domain in Welsh, cf. below Section 9.4.

Notes 477 181. The two possible “correct” versions of the sentence are: (i) Maltese: WITH fl-Italja kien marr l-Kumissjarju Axisa Axisa in:DET-Italy be.PERF go.PERF DET-commissioner li kien Gƫawdxi mas-Supretendent Calleja REL be.PERF Gozitan X:DET-superintendent Calleja ‘The Commissioner Axisa, who was a Gozitan, had gone to Italy with Superintendent Calleja.’ (ii) Maltese: AND fl-Italja kien-u marr-u l-Kumissjarju Axisa in:DET-Italy be.PERF-3PL go.PERF-3PL DET-commissioner Axisa li kien Gƫawdxi u s-Supretendent Calleja REL be.PERF Gozitan and DET-superintendent Calleja ‘The Commissioner Axisa, who was a Gozitan and Superintendent Calleja had gone to Italy.’ 182. Native speakers preferred singular agreement (bijusi) to plural agreement (bijuši) in sentences like (i) Latvian grƗmata ar zƯmul-i bij-usi zem galda book:NOM.F X/Y pencil-ACC.M be-PTCPL.F.SG under table:GEN ‘The book and the pencil were under the table.’ 183. The fact that the relators of languages like Quechua and Swahili indeed have full blown coordinating functions results from the employment of the relators outside the context of NP coordination. In Swahili, for instance, na is frequently used as a sentence connector: (i) Swahili (Möhlig and Heine 1999: 97) bila ya kuku hu-pat-i mayai na bila mayai without hen 2SG.NEG-get-NEG PL:egg X without PL:egg kuku wa-ta-kwisha hen 3PL-FUT-finish ‘Without hens you don’t get eggs and without eggs the hens will die out.’ 184. The final vowel of the suffix is often dropped in the syntactic chain. Note also that Hardman, Vásquez, and Yapita (1988: 210) distinguish two homophonous relators -mpi1 (for Comitative and coordination) and -mpi2 (for Instrumental and Causee) in order to capture the functional range of the morpheme under scrutiny. 185. In his survey of systems of instrumentality, Nilsen (1973: 74) mistook the Aymaran causee marker -mpi for a general agentive marker (corresponding more or less to an Ergative) (Stolz 2001a: 157 note 5). 186. Within the Localist framework, Anderson (1977: 122–124) pondered the idea of subsuming the Comitative under the broader category of local cases/relations.

478 Notes 187. For Perlative and Prolative as typical syncretistic partners and grammaticalisation sources of Instrumentals and Agent markers in Indo-European and European languages, cf. Luraghi (2001a: 38 note 7) and Palancar (2002: 206). 188. As far as we can tell from the extant sources, Wayampi is probably a representative of the WITH-languages. Example (B58.1) displays dual agreement in the subject pronominal slot of the finite verb, and thus the Comitative relator may also be seen as a kind of coordinator of NPs. 189. In distantly related Osetin, the inflectional ablative is used to encode Source and Instrument: arv-æj zæxmæ ‘from heaven to earth’ and færæt-æj amajyn ‘to trim with an axe’, whereas the Comitative receives a distinct encoding by the suffix -imæ (Abaev 1964: 19). 190. The association of Superessive and Instrumental recurs in many languages of different genetic and areal background. Accordingly, this pattern is reported for the isolated A-type AND-language Burushaski (Berger 1974: 22), the Caucasian A-type AND-language Lezgi (Haspelmath 1993: 327) and the Iranian B-type AND-language Pashto (Lorenz 1982: 55). 191. Exactly the same pattern as in Bezhta is reported for Hunzib (van den Berg 1995: 43–48). 192. Kibrik and Testelets (2004: 236) ascribe to the Comitative-Locative the meaning: “next to but in nonspecific position relative to inanimate X, together with animate X”. 193. It is worth noting that neither Blake (1977: 44–50) nor the pertinent contributions to Dixon (1976: 313–420) all of which investigate the meaning range of Instrumentals and their relation to Locatives and Ergatives in Australian languages mention Comitative-like functions of the Instrumental, cf. below Section 10.6. 194. For an evaluation of syncretism of spatial cases in Australian languages, cf. Stolz (1992b: 77–84). 195. In a very small number of accusative languages, the Instrumental may be used as the marker of the passive Agent (Blake 1977: 62). 196. From a crosslinguistic perspective, this is remarkable as there are many cases of [+locative]-Agent syncretism which do not involve the Instrumental, nor the Comitative (Palancar 2002: 205–208). Thus, we are perhaps facing an areal feature of Australian languages. 197. When discussing the possibility of object-NP deletion, Vaquero (1965: 103) mentions in passing a sentence synonymous to (B68.1) which contains the postposition aisiko glossed con ‘with’. However no further explanations are given as to the functional domain of this relator (chances are that it is specialised for Comitativity – although it is derived formally from aisía, the Instrumental relator): (i) Warao aisiko naruae X go ‘[He] went away with [his mother].’

Notes 479 198. Unfortunately, we were unable to determine the functions of a fourth postposition glossed Spanish con, detrás ‘with, behind’, viz. eyamo (Vaquero 1965: 101). 199. For obvious reasons, this is not the right place to enter into a discussion of whether or not the label possession is an adequate terminological cover for all the functions which are subsumed under it in the bulk of the literature. 200. For a more detailed study of the relationship between Comitatives/Instrumentals and Possession, we refer the reader to the forthcoming monograph of our working-group Stolz, Kettler, and Urdze (forthcoming). 201. In Dixon (1976: 203–311), over 100 pages are devoted to the discussion of the so-called having-affix (= proprietive) in Australian languages. Sutton (1976: 300–304) provides examples from 50 Australian languages and points out some etymological ties that may exist between Comitatives, Instrumentals and Proprietives in some of these languages. 202. Again, caution is necessary when it comes to interpreting the information given in descriptive grammars. One can be mislead easily by the terminology employed there. For Gulf Arabic for instance, Holes (1990: 95) translates the prepositional marker of alienable possession Ë ind as ‘with’. However, Ë ind is never used for Comitativity nor for Instrumentality which require the relators wiyya and bi, respectively (Holes 1990: 127). As a matter of fact, Ë ind translates only the spatial reading of English with where it corresponds to at! 203. For NP coordination, the correlative -a…-a is used on all coordinated NPs. In contradistinction to the associative marker -a (cf. below), the coordinator does not trigger gemination of a stem-final consonant, i.e., it has different morphonological effects (Cyffer 1991: 70). Notwithstanding these differences, we are convinced that coordinator and associative marker are etymological cognates. Note however that Cyffer (1991: 97) often gives an alternative translation of the associative: alongside ‘with’ we encounter ‘and’. 204. Admittedly, Heine (1997) cannot be made responsible for potential shortcomings of this hypothesis as he cautiously cites Kilian-Hatz and Stolz (1993) and Stolz (1993) as proponents of the conceptual correlation between Comitatives and inalienability. 205. Arithmetic operations are perhaps not the best examples of coordination. Nevertheless, we have chosen this example because coordination of normal NPs is not discussed sufficiently in our sources. The reason for this negligence is obvious: Hausa is a WITH-language – thus there seems to be no need to give separate examples of Accompaniment and coordination. 206. In Michaelis and Rosalie (2000: 96), Bamanankan is depicted as an A-type language with a slightly different distribution of the markers over categories. As it is not clear to us, on which variety of Bamanankan their semantic map is based, we do not include this proposal in our discussion of the data. 207. The initial component of this bipartite structure is identical with the short form of the Comitative relator (a)ni. Ye is perhaps identical with the ho-

480 Notes mophonous postposition ye ‘for, to’. The discontinuous ni…ye is also used in comparative constructions where it marks the standard of comparison (Brauner 1974: 110). 208. For practical reasons, we have slightly modified Kastenholz’ graphical rendering of the Bamanankan examples according to the principles employed by Brauner. Both authors take the standardised variety of Mali as their point of reference. Nevertheless, differences in their analyses may be due either to language reforms which affected the standard in the 15 years between the publication of Brauner (1974) and Kastenholz (1989) or to the fact that their data are inadvertently biased towards different dialectal bases. 209. This is of course tantamount to reclassifying Bamanankan from the C-type to the A-type. Whether this is appropriate remains to be seen. To complicate things a bit further, we give an example from Brauner (1974: 73), where the postposition fè unexpectedly combines with ani in a discontinuous formation which is reminiscent of ni…ye: (i) Bamanankan ekipu fla dònne balòn tan kènè kònò kick place inside team two go:PTCPL ball ani arbitri Monsieur Sany Mady Diallo fè X referee Mr. Sany Mady Diallo Y Joliba galeya a ka kanta Joliba advantage he POSS side ‘After the two teams together with the referee, Mr. Sany Mady Diallo, had entered the football ground, the team from Joliba won the toss.’ 210. In Maninkakan, a close relative of Bamanankan, spoken in Guinea, the following system is attested: The Instrumental postposition lá is also employed to encode inalienable possession (especially of the part-whole type with the exception of body parts which are marked as possessed by the general spatial postposition má) whereas alienability is signalled by the relational noun bólo ‘hand’ (Friedländer 1992: 66–67). Both lá and bólo are used to encode a [+human] passive Agents (non-human Agents being marked by fè) (Friedländer 1992: 122). The domains of lá and ní – the relator for Comitativity and coordination – do not intersect. This evidence from a member of the same phylum runs counter to the idea that there is a conceptual link between Instrumentality and Alienability! 211. The evidence for syncretistic patterns involving the BENEFACTIVE in Stolz (1997a) is too scarce to lend itself to a full-blown evaluation. However, we will nevertheless assign the BENEFACTIVE a region on our semantic maps because the few examples identified in Stolz’s study – replicated here in (B21– 24) – suggest that COMITATIVE-BENEFACTIVE syncretism is exceptional. INSTRUMENTALS outnumber COMITATIVES by 5-to-1 when it comes to syncretising with BENEFACTIVES. This is additional evidence of the greater propensity of INSTRUMENTALS to associate with categories which are bona fide represen-

Notes 481 tatives of so-called semantic cases (cf. below). Therefore, it can hardly surprise to find syncretistic patterns involving the BENEFACTIVE typically in Atype languages (cf. [B25]). The cases discussed in Michaelis and Rosalie (2000) corroborate this trend as only one among 20 languages of their sample (the English-based stable pidgin Tok Pisin) displays identical encoding for COMITATIVE-INSTRUMENTAL and BENEFACTIVE. However, besides the multifunctional, even all-purpose preposition long, COMITATIVE-INSTRUMENTAL may also be encoded by wantaim whereas BENEFACTIVE does not allow for secondary markers. In addition, the semantic maps provided by Michaelis and Rosalie (2000) mostly attest to the syncretistic patterns discussed in the previous chapters. This priority is of course not to be understood as exclusiveness, cf. syncretism with PATIENT and THEME as evidenced by Malaccan Creole Portuguese and other languages, cf. (A26–28). For the sake of comprehensiveness may it suffice to mention the Amerindian A-type AND-language Malayo where Instrumental-Benefactive syncretism is attested. Malayo encodes the Comitative by a suffix -bi whereas Instrumental, Locative and Benefactive are all expressed by -mba (Amaya 1989: 58–61), cf. also Part A, footnote 49: (i) Malayo (a) Locative (cf. also [A35.2]) urraga-mba house-Y ‘in the house’ (b) Benefactive ra-ga suzu ra-n-zhe ade-mba gek-uga backpack 1SG-LIG-GEN father-Y hand_over-1SG I-AG ‘I give the backpack to my father.’ Other examples of the Benefactive syncretising with the Instrumental (and the Comitative) stem from Maori where the secondary (!) marker of the Instrumental, the preposition maa, is also the primary marker of the Benefactive (Bauer 1993: 282–285). On superficial inspection, West Greenlandic seems to be an example of Comitative-Benefactive syncretism. However, the verbal affix -utiis in fact a multi-purpose morpheme and “whether this morpheme has an instrumental, a benefactive, or some other related sense [= comitative] depends on the stem concerned” (Fortescue 1984: 214). Thus, this is rather a case of Comitative-Instrumental-Benefactive syncretism. 212. To some extent, the combination of Instrumental with Possession goes along with a distinction of alienable vs. inalienable or temporary vs. permanent possession in our sample languages. As the previous chapter shows, however, the correlation is at best a tendency (Stolz 2001a: 340). Therefore, we refrain from distinguishing several kinds of possession on our maps. Future investigations have to reveal whether or not the tendency can be confirmed by additional language data.

482 Notes 213. Only within the domain reserved for the Instrumental do we find some recurrent patterns. For instance, Instrumental-[+agentive] syncretism often involves [+locative] as syncretistic partner as well, cf. Section 9.3 above. However, this pattern accounts for just a fifth of all cases of Instrumental-[+agentive] and a tenth of those of the type Instrumental-[+locative]. One may interpret this as evidence of a bridge-function of the Instrumental which makes it possible for [+agentive] and [+locative] to syncretise. Similarly, the Comitative bridges the gap between Possession and coordinator, in a way. It accounts for the same percentage of cases of AND-Possession syncretism as the ComitativeInstrumental. By the way, all instances of AND-Possession imply Comitative or Comitative-Instrumental, never the distinct Instrumental alone. 214. Among other things, Croft (1991: 191–192) argues against the traditional notion of PATIENT. 215. For an attempt to extend the model of causal chains to adnominal contexts, cf. Croft (1991: 228–231). 216. More often than not, animacy functions as a negative selector: There is for instance a relatively frequent pattern according to which languages distinguish two types of Passive Agents, namely animate/human vs. inanimate ones. In both varieties of Armenian and Itelmen – no name just a few examples – the Instrumental syncretises only with the inanimate Passive Agent whereas the animate equivalent operates on a different basis (Kozintseva 1995: 19–20; Georg and Volodin 1999: 82). For East Armenian, the facts are more complicated as the instrumental case is used with natural phenomena as Passive Agent and optionally with “collective human agents” (Kozintseva 1995: 19). 217. Except for the grammars which follow the questionnaire of the Lingua Descriptive Series (Comrie and Smith 1977: 30), many descriptions of individual languages do not provide information about how ‘absence of X’ is expressed. To the best of our knowledge, there is as yet no crosslinguistic study of categories such as, e.g., Abessive, Caritative and Privative (but cf. Ivanov 1994). In the pertinent literature on case relations and especially on Comitatives/Instrumentals, one comes across the occasional sideways remark (Noreen 1923: 339–345; Lehiste 1969: 335; Kunze 1992: 113; Fonseca 1994: 51–52; Blake 1994: 196). Seiler (1974c: 220) notes that there is a property, hardly ever noticed in the grammatical literature, that distinguishes mit from all other prepositions: mit can be negated and shows a proper negative form: ohne ‘without’. For Seiler (1993: 167), the opposition of WITH vs. WITHOUT counts among the basic pairs of categories of his “Dimension der Opposita”. 218. There is evidence that neutralisation can also affect further categorical distinctions that might exist for Comitativity and/or Instrumentality. In Itelmen for instance, the two inflectional comitatives have a common negative form, the inflectional abessive (Georg and Volodin 1999: 87). Unfortunately, our source of

Notes 483 information does not provide examples of negative Instrumentals and thus we cannot be sure whether the neutralising effect also includes the Instrumental. 219. This observation has to be understood cum grano salis: Owing to the fact that pertinent information is often hard to come by in the descriptive grammars and dictionaries, many strategies of encoding going beyond the level simple morphemes may have escaped our notice. With a view to pinpoint the difficulties that arise when one is searching for information on negative Comitatives and Instrumentals, we mention Maori because it uses a strategy that is widely diffused not only among Austronesian languages. This strategy consists of a double predication, one of which states which action is carried out and the other states the absence (= non-existence) of a certain instrument. Note that there is no grammaticalised expression for the negative Comitative; speakers are free to choose from a large inventory of complex constructions (Bauer 1993: 285–286). (i) Maori I patu-a e ia kaahore he maaripi a knife T/A kill-PASS by he NEG ‘He killed it without a knife (lit. he killed it and there wasn’t a knife).’ In West Greenlandic, both Comitative and Instrumental are negated by negative forms of verbs which indicate the prototypical actions associated with Comitativity and Instrumentality, viz. ilagi- ‘to accompany’ and atur- ‘to use’ (Fortescue 1984: 214–216). These verbs are not necessarily the primary means of expression in positive contexts! (ii) West Greenlandic (a) Instrumental: positive nanuq savim-mi-nik kapi-vaa polar_bear knife-his-Y stab-INDIC:3>3 ‘He stabbed the polar-bear with his knife.’ (b) Instrumental: negative savi-ni atur-nagu nanuq tuqup-paa knife-his use-4>3.NEG.CONT polar_bear kill-INDIC:3>3 ‘He killed the polar-bear without (= not using) his knife.’ (c) Comitative: positive Hansi-lu aqagu aalla-ssa-agut Hans-X tomorrow leave-FUT-INDIC:1PL ‘Hans and I will leave tomorrow.’ (d) Comitative: negative Hansi ilagi-nagu aallar-puq Hans accompany-4>3.NEG.CONT leave-INDIC:3 ‘He set off without Hans.’ As a matter of fact, West Greenlandic has a wide variety of more or less grammaticalised expressions for the categories under scrutiny – no matter

484 Notes what polarity applies. This variability adds to the difficulties of coming to grips with the translation equivalents of English without. 220. Admittedly, the Hixkaryána example (B82.4) does not state the absence of a certain tool but rather contrasts one potential utensil with another. Unfortunately, we have not been able to find uncontroversial examples of an Abessive the extant descriptive material of this language. 221. The instrumental marker -(e)z can also co-occur with the negator, if contrastive focus applies. Sentence (B83.3) contains an afterthought ez laban-ez {NEG} {knife}-{Y} ‘(and/but) not with a knife’ (Saltarelli 1989: 158). 222. According to many sources, the use of -stul/-stül is more restricted also in terms of token frequency. Some native speakers associate this suffix with anachronistic or rural style. Thus, -val/-vel – at least for some speakers – may always replace -stul/-stül but not vice versa. A typical example of the associative case is drawn from Bánhidi, Jókay, and Szabó (1975: 345): (i) Associative: Accompaniment csónak-ház-ak sport-és játszóter-ek vár-ják család-o-stul boat-house-PL sport-ADJ playground-PL wait-3PL family-LIG-X gyerek-e-stül az ember-ek-et DET man-PL-ACC child-LIG-X ‘Boathouses and sports grounds are waiting for the people with family and children.’ Accordingly, the Tool-NP and Companion-NP in our examples under (B84) are to be rendered as kalapác-csal ‘with hammer’ and either fegyver-rel (= Comitative-Instrumental) or fegyver-e-stül (= associative) ‘with gun’ if positive polarity applies. 223. First of all, it must be stated that Comitatives/Instrumentals and their negative counterparts do not show a one-to-one mapping of distribution profiles over the polarities, i.e., not every negative Comitative/Instrumental has a positive Comitative/Instrumental in the same context with different polarity and vice versa. As a matter of fact, the negative forms are very used in combinations with conjunctions for the purpose of clause-combining while a positive variant is non-existent. A typical case is German ohne daß/zu ‘without (V-ing)’ which lacks the positive equivalent *mit daß/zu ‘with/while (V-ing)’: (i) German (a) negative er geht vorbei ohne uns zu grüßen he goes past X/Y.NEG us to greet:INF ‘He passes by without caring to greet us.’ (b) positive *er geht vorbei mit uns zu grüßen he goes past X/Y.NEG us to greet:INF ‘He passes by and greets us.’

Notes 485 An acceptable positive rendering would require a full-blown PP with a complement noun (mit einem Gruß an/für uns) or a coordination of two full predications (und grüßt uns) or the somewhat stilted participle converb construction (uns grüßend). Nevertheless, the relationship between positive and negative categories is relatively systematic. In anticipation of what we are going to say about the interaction of Comitatives and Instrumentals in Sections 10.2–3, we point out that, in the overwhelming majority of cases, the expressions of negative Comitatives/Instrumentals are more complex than the ones used for the positive versions (English without vs. with is a typical example: two syllables vs. one, five segments vs. three); in A-type or C-type languages negative forms tend to be more complex than at least the segmentally simplest positive marker (in Kot, negative -fun has more segments than Comitative -oš and Instrumental -ô; in Dullay, negative male is co-extensive with Comitative ‘ale but exceeds Instrumental -kká by one syllable/two segments [Stolz 1996: 33–34]). This is of course hardly surprising as the negative forms have to incorporate semantically the feature [negative] (which makes them semantically more complex – and this complexity is iconically mapped unto the expression if this happens to be more complex than the positive counterpart; note however that this negative component does not necessarily correspond to an identifiable and distinct morph!). The negative forms are also more often derivationally dependent on the Comitative or Instrumental (cf. again the English example where without is formally based on with). Under negation, there is often room only for a reduced inventory of distinctions, negative forms are more affected by language change and borrowing (cf. the numerous examples of Spanish sin ‘without’ borrowed into many indigenous languages of the Americas [Stolz 1996: 44–45]). Likewise, they often display a lesser degree of grammaticalisation than positive Comitatives and Instrumentals. The negative forms thus behave very much like the marked case as opposed to the positive ones – a behaviour that is predictable from the point of view of markedness theory (Stolz 1996: 46–48). Interestingly, translation equivalents of English without behave towards their positive counterparts in much the same way as Comitatives behave towards Instrumentals in A-type and C-type languages. In adnominal constructions, special derivations with a meaning component [absence] are quite common among the languages of the world. Privatives/ Caritives marked by the suffixes -los, -frei or the prefix un- (often added to an ‘applicative’ wordform with an initial be- morpheme) are frequently used in contemporary German for instance: eine kinderlose Frau ‘a childless woman’, atomwaffenfreie Zone ‘nuke-free zone’, ein unbehaartes Bein ‘a hairless leg’, etc. are frequent in everyday conversation. These formations cover the functions of Confective and Ornative – and emphasise permanent properties, meaning: In almost every case, these formations could be re-phrased using a PP headed by ohne ‘without’ but not every instance of ohne can be covered by

486 Notes -los/-frei/un-. These derivations are much more lexicalised and thus often show a tendency to move away from semantic compositionality whereas ohne-PPs in the majority of cases can be understood compositionally. Nevertheless, it remains to be seen how exactly these and other derivations are related to Comitativity and Instrumentality. 224. One could claim that (B86) contains examples which attest to a variant of the syncretistic pattern found in B-type languages: In the latter, Comitative and Instrumental are indistinguishable formally and in the former the similarity on the expression side is such that it comes close to formal identity. Irrespective of the fact that we do not subscribe to this view, adding the languages of (B86) to the B-type would not significantly alter the crosslinguistic predominance of the Atype which still accounts for much more than 50% of the cases. 225. It is likely that the mora would have been a more reliable basis for the calculation of the degree of complexity because it allows calculating vowel length etc. However, it was not possible to adequately interpret many of our data as to their moraic structure (as indication of length distinction and other relevant information is not always available). Taking stock of syllables however has proved to be a rather straightforward procedure. 226. The first four patterns already exhaust the logically possible combinations because those patterns which involve a minus are only the inverse representations of the ones which involve plusses. 227. This observation is less trivial than it appears: A long chain of segments does not automatically imply an increase in the number of syllables whereas one vocalic segment alone makes a difference as to the number of syllables of the relator. 228. In these statistics, we take account only of the most grammaticalised primary markers of the categories involved. If there is allomorphy affecting the complexity of the relator, we generally opt for the shorter allomorph. Thus, the number of Comitatives and Instrumentals is reduced to one relator for each of the two categories per language. 229. Where the Instrumental marker is not syllabic or zero, the ratio would be Xto-0 with X being a variable ranging from one to seven. Only for practical reasons do we treat these cases as a group of its own and count them separately. We are aware of the fact that this is a rather heterogeneous agglomeration of different degrees of complexity on the part of the Comitative – and minimal complexity of the Instrumental. 230. For the complete list of data, cf. Stolz 1996, 1997a, 2001b, 2003a and Stolz, Stroh, and Urdze 2005. 231. Pertinent examples of these additional functions which meet our expectations perfectly are (i) Possession (Carlin 1993: 69) ka ir n8 k 8-sa X house be-1SG ‘I have a house.’

Notes 487 (ii) Coordination (Carlin 1993: 169) ga o-ac mam-bi ka kug-bi ka go call-VEN uncle-POSS.2SG X grandfather-POSS.2SG X dad-bi grandmother-POSS.2SG ‘Go and summon your uncle, your grandfather and your grandmother.’ 232. Carlin (1993: 129) glosses the second occurrence of -Ák as goal-case marker. Goal case and the so-called circumstantial case have phonologically similar markers which differ only in their susceptibility to vowel-harmony (Carlin 1993: 130). As far as we can judge, Carlin’s differential glossing of the two instances of -Ák in the example are erroneous – both are cricumstantial-case markers. 233. Hixkaryána is one of two doubtful cases. In example (B82), we have seen that the Comitative marker is always marked for possessor whereas the Instrumental relator never bears any additional morphological elements. We did not want to waste too much time and space discussing the issue of compulsory possessor marking and boundedness of the possessor-marked adposition. In all other cases, we have adopted a relatively straightforward and traditional criterion: The host of a bound morpheme has to be a lexical morpheme. For other borderline cases, cf. Otomí below. 234. It is also interesting to check which of the two morphemic statuses prevails among B-type languages, too, although this does not directly tell us anything about the possible asymmetry which characterises the relationship of Comitative and Instrumental. Of the 79 B-type languages in Stolz (1997a), 24 (= 30.37% of the B-type) have bound morphemes for their syncretistic Comitative-Instrumental whereas 55 (= 69.63%) employ free morphemes. This contrasts clearly with the preference for bound morphemes attested for A-type and C-type languages. If we add up the sub-totals, we get 172 bound relators and 162 free relators, i.e., an almost even score of 51.5% bound vs. 48.5% free morphemes. It remains to be seen whether the following generalisation can stand future tests: If a language keeps Comitative and Instrumental formally distinct then with more than simple chance frequency at least one of the relators is a bound morpheme. If in a language Comitative and Instrumental syncretise then there is a two thirds probability that their relator is a free morpheme. 235. Our criteria for deciding whether a given relator counts as bound or free have the character of a rule of thumb. Clitics are subsumed under bound forms as long as they do not govern a certain case to whose exponent they attach, i.e., the class of bound morphemes is not restricted to the classic cases of inflectional morphology. 236. Their classification as A-type languages is based on the fact that the phonologically realised Comitative relator interacts with an Instrumental which happens to lack an overt realisation.

488 Notes 237. In this example, the finite verb may also have dual (in other cases plural) agreement. 238. As a matter of fact, this pronominal strategy deserves to be studied in a much wider perspective. Our Latvian example (A15) point to a close connection between Comitatives and number. Bergsland (1947) has shown that constructions of the type PRONOUN + (X/Y) + PROPER NOUN/PRONOUN are areally prominent in North and East Europe (and beyond). Bhat (2003: 104) mentions several examples from elsewhere on the globe and alludes to a forthcoming publication on coordination by Haspelmath (forthcoming) in which the geolinguistic distribution of the phenomenon is addressed in Chapter 6 under the headings “Comitative conjunction” and “Inclusory conjunction”. Equally interesting is Gehling’s (2004: 203–306) concept of con-sociative (“Konsoziativ”). 239. There are numerous loans from Spanish among which we find almost the entire set of prepositions of the donor language. The expression of the Instrumental is also affected by language contact. We discuss this issue separately in Part D when we look more closely at the dynamics of Comitatives and Instrumentals. 240. The genetic isolate Gilyak is an A-type WITH-language which shows that the interconnection of number and Comitative need not be mediated by pronouns. According to Mattissen (2003: 8–9), Gilyak has an associative number which seems to cover the functions of coordination and Comitativity whereas there is a regular inflectional instrumental. The associative number is different from singular, dual and plural. 241. Senoufo is a C-type AND-language which makes use of a discontinuous expression for the syncretistic Comitative-Instrumental, viz. na…i (na is also the regular NP coordinator), alongside a serial verb construction involving the serial verb taha. According to Carlson (1994: 179–180 and 296), the circumposition is the default option for Accompaniment whereas it is less often used for Instrumentality; the serial verb seems to be exclusively employed for the Instrumental – for which it also is the preferred expression. 242. We have not yet found out about the division of labour of the two cases. That is the reason why we give examples for both. 243. The heuristics of these approaches comprise among other things evidence drawn from first-language acquisition, language loss, pidginisation, creolisation, language change, speech errors, etc. There is thus a relatively solid psycholinguistic component (Dressler and Mayerthaler 1987: 12–15). 244. Our notion of Europe is a modified version of the one adopted by Mayerthaler, Winkler, and Fliedl (1993). All evidence stemming from languages other than the ones used for the corpus-based analysis is located within theses artificially imposed boundaries. 245. In actual fact, not all translations were easy to find and lay our hands on. The full number of 64 translations was reached only after three years of intensive search of libraries, book-shops, collector’s shelves worldwide. However, these

Notes 489 logistic problems were created only by a handful of rare items (Tosk Albanian, Armenian, Azerbaijani, Kurdish, and Welsh). The Irish translation unfortunately made it to our office only after the statistics had been completed. Lamentable as it is, many other translations were published too late to be included in our investigation (among these latecomers we find a large number of translations into regional varieties of Italian and German, the first Tatar version, an Udmurt translation and several parallel translations into Saami languages). 246. We use the terms standard, non-standard and regional variety rather liberally. As there is no official standard norm for the varieties of Romani for instance, Lovari (spoken in Hungary) appears in the column reserved for non-standard and regional varieties. Some of our non-standard and regional varieties have a tradition of writing. Thus the distinction made in Table (C1) follows a rule of thumb and thus should not be taken too seriously. 247. Owing to the fact that we take a written text as our frame of reference, we have adopted a very simplistic but nevertheless absolutely legitimate definition of the sentence: Everything that occurs in between two punctuation marks of the type full-stop, question mark, exclamation mark or, in the absence of a punctuation mark on the left, between one of these and the left boundary of a paragraph. With this procedure, we have been able to determine the number of sentences for each of the texts. The French original is made up of 1,652 sentences – a number which is matched exactly only by six other languages (Friulian, Languedocien, Romanian, Serbian and Hungarian). The remaining 57 languages oscillate between the minimum of 1,582 sentences attested for Azerbaijani and the maximum of 1,663 sentences in the Greek version. Except for Greek, the vast majority of the languages reaches a number of sentences that amounts to slightly less (= n ” 20) than the number of sentences of the original. 248. In addition, the negative form again behaves like the marked partner in an opposition. If there are differences in complexity on the expression side, it is usually the equivalents of English without that receives the more complex expression. Of 57 negative relators, 30 contain at least one syllable more and 44 have at least one segment more than the simplest of their positive counterparts. The reverse is true of only three cases! In all other instances, positive and negative forms are equally complex. Note that in Azerbaijani and Turkish, the negative form is generally a bound morpheme whereas the positive relator oscillates between bound and free status. 249. In Table (C4), we have classified syncretistic Comitatives-Instrumentals of Btype languages encoded by an adposition which governs a certain inflectional case on the noun as instances of marking by adposition alone because the case-form of the noun has no distinctive function. 250. For determining the frequency values of relators, we have adopted the following conventions: NP-internal agreement morphology is not counted separately

490 Notes (meaning: all agreeing elements are considered to be one instance only). The same applies to coordinated NPs: if relators are repeated with every conjunct, they are counted as one. 251. We have not tried to compare different translations into the same language for two reasons: (a) only a small number of languages have more than one translated version and (b) time restrictions prevented us from looking into the stylistic potential of the languages. 252. Throughout Section 11.1.2, except otherwise stated, the term typology refers to the classification in one of the three types identifies for Comitatives and Instrumentals (A-type, B-type, C-type). 253. The division of the European area into quadrants is of course arbitrary (and perhaps reflects a subconscious Germany-centred internalised map). One could discuss various problematic cases especially of languages which are situated on the very boundaries. Accordingly, (C7) depicts Europe in a slightly different way as our older map published in Stolz, Stroh, and Urdze (2003: 83). 254. What we can do at the moment is to point to the fact that the frequency of the Welsh relator too is rather high for a language of the North West quadrant, cf. Breton below. In addition, Irish, which for practical reasons had to be excluded from our statistics, makes use of its Comitative-Instrumental relator le 506 times in The Little Prince and thus would rightly claim the first rank on our frequency list (Stolz, Stroh, and Urdze 2003: 74–75). Unfortunately, we do not have any reliable frequency count for Gaelic handy (there is as yet no translation of our sample text into Gaelic). We therefore let the hypothesis stand as is and postpone a close-up check to a later occasion. 255. Two sentences are said to be the same in two or more translations, if and only if they translate the same part of the corresponding sentence of the original. The sentences of the original are consecutively numbered and their translations receive the same number as the one of the original independent of the position they happen to have within the translation itself. In case the sentence boundaries of original and translation do not map, those parts of a sentence in a translation which correspond to a sentence of the original are either isolated from the context in the translation or re-united over sentence-boundaries of the translation. 256. The context is considered to be identical if and only if the relator, morphologically or syntactically, has scope over elements which, in terms of their semantics, correspond largely to the elements in the other language(s) and/or depends upon a syntactically higher element which, in terms of their semantics, correspond largely to the elements in the other language(s). 257. It would be interesting to see whether the methods of dialectometry (Goebl 2001) are of help in our special case. Since a cartographic representation of the individual coefficient values is not among our major goals, we leave this issue to future studies.

Notes 491 258. This and the next four tables give the coefficient for the five highest ranks (1 through 5 in the top row) and the five lowest ranks (59 through 63 in the top row). In the leftmost column, the language with which the top-ranking and low-ranking ones of each row form pairs. For this language, the average coefficient for all of its 63 pairings is indicated. This average will become important below. A full account of the statistics will be published separately as a quantitative study of its own (Stolz, Stroh, and Urdze, in preparation). 259. In the late middle ages, Malta was under Aragonese and Spanish rule for some time (Brincat 2000). However, it cannot be proved that the potential contacts in the past are responsible for the present similarities. To our mind, the coefficient values simply mirror the general degree of Romanicisation of Maltese. 260. It should not go unmentioned that Yiddish participates uninterruptedly in the 42 largest isoglosses from 64 languages down to 34. This fact makes Yiddish unique among our sample languages as every other of them fails to participate in several of these 42 isoglosses. These 42 cases account for a third of all attestations of the Yiddish relator (total: 124 tokens). Accordingly, Yiddish has a rather small number of isolates (just six cases = 4.9% of all attestations). In a way, Yiddish is thus one of the most association-friendly languages of our sample. 261. Even though it may seem idiosyncratic, we use the term isogloss here as a label for cases of identical usage of Comitatives/Instrumentals in the same context in the same sentence of more than one translation. 262. A typical example for these unsystematic gaps is the absence of Italian from the isogloss for sentence I.7a to which 57 languages belong. The crucial context is of the TOOL-kind: The narrator reports that he drew his first picture with a coloured pencil. While the vast majority of the Romance languages use their prepositional relator for Comitative-Instrumental in this context to introduce the TOOL, Italian does not – simply because the translator did not deem it important to mention the TOOL in the first place, i.e., the expected PP introduced by con was left out. However, this single instance does not imply that Italian systematically diverges from its sister languages. In other sentences with a TOOL-context, Italian forms of course part of the isogloss, cf. below. 263. There are two ways for a language to be missing from an isogloss: (a) the sentence or the particular part of a sentence has not been translated by the translator or (b) the construction does not involve a Comitative, Instrumental or combination of the two. We do not distinguish these two types terminologically because in the vast majority of cases, it is (b) that applies. 264. The absence of Basque is incidental. In sentence XXI.144, the same BODY PART INSTRUMENT turns up – and this time Basque participates in the isogloss without any difficulty. The relator is the suffix -z, i.e., the one which is also employed with TOOLS: Bas XXI.144 [Bihotzaz] BODY PART baizik ez dira gauzak garbi ikusten. ‘It is only with the heart that one can see rightly.’

492 Notes 265. Again, these gaps are not motivated by general structural differences. All three absentees also employ their relators in contexts of Comitativity: Geo IV.50 uk’ve ekvsi c’eli ikneba, rac chemma megobarma [tavisi k’ravit] COMPANION-urt mimatova ‘It is already six years since my friend has gone away with his lamb.’ Ice I.28 Ég hefi mikið verið með [fullorðnu fólki] COMPANION ‘I have spent a lot of time with grown-up people.’ Lan II.73 Aital faguèri coneissença amb [lo princilhon] COMPANION ‘In this way I got to know (= made acquaintance with) the Little Prince.’ Only Georgian makes use of a construction different from the on used in the Instrumental contexts: The postposition urt governing the instrumental case on the noun distinguishes this instance of Comitativity from the ones of Instrumentality, cf. (C50). 266. Sentence IV.30 suggests that the absence of Ladin from the isogloss of the ORNATIVES is not of a systematic nature, cf. the second PP in the following example: Gher IV.30 É udú bëndebò na bela cësa cun [ziedli cueceni] MATERIAL, cun [geranies dan i vieresc y tupes sun tët] ORNATIVE ‘I saw a beautiful house made of rosy brick, with geraniums in the windows and doves on the roof.’ 267. We have already mentioned the so-called inflection sociative of Hungarian which is absolutely marginal in our sample text: -stul/-stül is a candidate for those cells which are marked by the asterisk – but this candidature is more of a potential than a preferred replacement of -val/-vel. The single occurrence of the sociative is in sentence X.12: Hun X.12 A kis herceg körülnézett, hová ülhetne le, de a bolygót [mindenestül] COMBINATION beborította a pompás hermelin palást ‘The Little Prince looked for a place where he could sit down, but the magnificent ermine coat covered the entire planet.’ Mindenestül ‘with everything’ however is a lexicalised word-form of minden ‘all’ with the meaning of ‘entire(ly), complete(ly)’. Thus, the sociative is less productive than its competitor, the comitative-instrumental. 268. This is the case with sentence VIII.8 which describes the vanity of the flower which the Little Prince grew on his planet: Gal VIII.8 Elixía con coidado as súas cores Eng VIII.8 She chose her colours with the greatest care In principle, the English translator could also have used an adverbial like very carefully. Tentatively, we may hypothesise that English is more ready to employ the PP-strategy if the MANNER-indicating expression is itself modified. 269. The Moldavian equivalent of XIII.5 is thus: Mol XIII.5 trei plus doi fac cinci ‘Three plus two make five.’

Notes 493 270. Accordingly, the Tosk version of (C75) makes do with only one token of me – which happens to encode an abstract body part instrument (sight, look): Alb-T I.25 mësova të dalloja, me një të parë, Kinën nga Arizona ‘I learned to distinguish China from Arizona at the first look.’ 271. In a variety of other languages of our sample, the Comitatives and/or Instrumentals may have functions which remotely resemble the ones discussed for Gheg Albanian. In Swedish, for instance, some phasal verbs such as, e.g., börja ‘to start’, hålla på ‘to keep/go on’ and slutta ‘to terminate’ combine often with med which introduces their complements: Swe VIII.15 hon höll på med sina hemlighetsfulla toalettbestyr ‘she continued with her mysterious adornment’ 272. Cognates of instrumental bi are to be found all over the place in the Arabicspeaking world, cf. Moroccan Arabic bi (Maas 1999: 66) and Gulf Arabic bi (Holes 1990: 197). Ma’ seems to have a different fate because it is less often preserved in modern regional varieties, although Modern Standard Arabic still employs ma’ as a comitative preposition. Notwithstanding the occasional disappearance of ma’, most of the regional varieties of Arabic still keep comitative and instrumental formally distinct with other prepositions having taken over the former functional domain of ma’. 273. Several factors support the view that the negative prepositions are more marked than their positive counterparts. First, bla and mingƫajr are secondary developments from former syntagms whereas bi and ma’ are Semitic prepositions of long standing. Second, bla and mingƫajr both are segmentally more complex than bi and ma’. Third, bla and mingƫajr lump exactly those functions together which are crucial for the distinction of bi and ma’, i.e., bla and mingƫajr negate not only comitative but also instrumental. Fourth, neither bla nor mingƫajr cliticise to a following NP. Fifth, in contradistinction to bi, ma’, and mingƫajr, bla does not have special bound pronominal forms. In addition, bla and mingƫajr display different distributional properties. Semantically, they are synonymous but bla is excluded from contexts with pronominal complements. Bla and mingƫajr often function as part of a bipartite conjunction with ma as second component, e.g., in (Qrempuƛu 41) U Qrempuƛu, bla ma tilef sekonda oƫra fit-tpaƛpiƛ, deffes geddumu ƥol-ƫamrija u beda jƫaffer. ‘And without losing another second by talking, Qrempuƛu buried his snot in the ground and began to dig.’ and (L’Imsella 48) David gƫamel kopja talpoeĪija mingƫajr ma wrieh b’xejn ‘David made a copy of the poem without letting anybody notice.’ Elsewhere, the domain of mingƫajr is that of a typical preposition: (Kappa 44) Fl-elezzjoni, mingƫajr l-iƛken dubju u b’maƥƥoranza mhux negliƥibbli, rebaƫ il-Partit Demokristjan. ‘In the election, the DemoChristian Party won without the least doubt and with a solid majority.’ An example of bla as a preposition is [L’Imsella 6] Ƥera l’aƫƫar biƛƛa triq u wasal ir-razzett bla nifs. ‘He ran for the last bit of the street and arrived breathless at the farm-house.’

494 Notes 274. The cliticisation applies only to monosyllabic prepositions whose coda is either simple or empty [= (C)V(C(:))]. One important exception is bla ‘without’ – a preposition that never cliticises although it has the appropriate phonological structure for the coda, cf. footnote 2. 275. Cf. (Avventura 11) Sakemm sebaƫ Anna u bintha m’gƫalqux gƫajn m’gƫajn. ‘Anna and her daughter did not close their eyes until dawn.’ Orthographic in all positions except word-final and when preceding does not correspond to any phonological segment, but may indicate the lengthening or diphthongisation of neighbouring vowels. 276. Cf. (Ulied 35) Xi ƫadd kien qed jipprova jiƥbidlu l-attenzjoni billi jwaddab ilƥebel mat-tieqa. ‘Somebody was trying to attract his attention by throwing the stone at the window.’ 277. Cf. (Ƥrajjiet 76) malli ƫarist f’wiƛƛha, intbaƫt li kienet fl-inkwiet. ‘As soon as I looked into her face, I realized that she was in trouble.’ 278. Cf. (L’Imsella 104) ma kienx hemm dawl biĪĪejjed ‘There was not enough light.’ 279. As to orthography, some authors do not follow the widespread practice to write univerbated syntagms involving either bi or ma’ in one word. This may be understood as indicating that the univerbations have not become fully fixed yet. For combinations of ma’ and the noun ƥenb ‘side’, for instance, there are several solutions. The presence or absence of the definite article on ƥenb reflects the distinction of noun-noun syntagms into status constructus (= no article) and marked noun-noun combinations (= article) (Aquilina 1987: 386): (Kappa 37) Bilqiegƫda fuq bank mal-ƥenb tas-sala kien hemm is-segretarja tas-Sur Abela. ‘Sitting at the side of the hall, there was Mr. Abela’s secretary.’ vs. (Kappa 42) Maƥenbu kellu lill-Viƛi Kap, u fuq il-ƥenb l-ieƫor lis-segretarju Ƥenerali, qishom Īewƥ akkolti maƥenb il-Papa. ‘On his one side he had the Vice Chief and on the other the Secretary General like two acolytes accompanying the Pope.’ vs. (L’Imsella 41) Marƛell kien gƫarfu u waqqaf il-mutur ma’ ƥenbhom. ‘Marcello had recognised him and stopped the bike at their side.’ The status constructus combines nouns which are considered to be in a more intimate or inalienable relationship (Fabri 1996: 240). 280. A pertinent example of one of these alternatives is: (Kappa 80) trasferixxa rrabja kollha tiegƫu minn fuq l-applikazzjonijiet formali interni gƫall-paƥni rqaq tal-gazetta, fejn kien ixandar il-kummiedji u t-traƥedji kollha li kien jseƫƫu ƥewwa l-ƫabs, permezz tar-rapporti tiegƫi ‘He transferred his entire rage from the formal applications to the thin pages of the journal where he spread all the comedies and tragedies that took place within the prison through my reports.’ 281. Cf. also (L’Imsella 72) Il-firma fuq l-ittra kienet taƫĪiĪa ma tingƫarafx billinka blu ‘The signature on the letter had been scribbled indecipherably with blue ink’, where the material (= ink) used is only a tertiary instrument – with the hand and the pen being the primary and secondary one, respectively.

Notes 495 282. Note that native speakers of Maltese normally do not volunteer pertinent examples. During fieldwork sessions based on a questionnaire, many Maltese informants rejected such constructions right away as ungrammatical or at least not idiomatical. Others however accepted the constructions (though sometimes only reluctantly). The phenomenon has not been tackled yet in the extant descriptions of the language. Notwithstanding the reaction of the native speakers constructions of this kind can be found in sufficient numbers not only in the written register. 283. There are several candidates for the status of copula verb (Borg 1988: 269– 73), among them qagƫad ‘sit’ which may also be used with a predicate PP headed by bi as, e.g. in (Serafin 38) Mur u oqgƫod bil-gƫaqal ‘Go and be prudent!’, for the attributive equivalent, cf. sentence (C83.1). 284. Note however that the functional domain of bi does not include the marking of the passive agent! This is the task of minn ‘from’ as, e.g. in (Kappa 32) U b’hekk ir-raba’ kaĪ baqa’ qatt ma nstema’ mill-ƥurija popolari ‘And thus, the fourth case was never heard by the lay jury.’ 285. A similar case can be made for examples such as, e.g. (Avventura 64) Sƫab Bastjan resqu lejn it-tfal, qabduhom minn idhom u bdew mixjin bihom lejn ilvann ‘Sebastian’s companions approached the children, grabbed them by the hands and started to walk with them to the van.’ The situation described belongs to a kidnapping story. A gang of three delinquents has kidnapped two children and is about to abduct them in a van. The children are not accompanying their kidnappers voluntarily, so that the criminals use force on the victims to make them walk up to the van. The lack of volition on the part of the children is aptly reflected by the choice of the preposition bi (in bihom ‘with them’) in lieu of ma’ which otherwise would be the preferred option to link the two participants of a motion verb (mixjin ‘be walking’). 286. According to Vella (1994: 11), ma’ and bi differ in the following way: ‘Ma’ is used with persons, whereas bi, as a rule, is used with animals and objects.’ The same rule is given in Moser (1999: 230) in his German-Maltese/MalteseGerman dictionary: ma’ is said to be restricted to combinations with Personen [persons] and bi is reserved for combinations with Tiere [animals] and Objekte [objects]. These rules of thumb oversimplify the complex distribution patterns of the two prepositions. The insistence of the authors on NPs referring to animals to require bi instead of ma’ is an untenable generalisation. 287. Yielding the following sentences whose grammaticality has been checked with native speakers (we are especially grateful to Ray Fabri for his comments on this issue): (C92.1’) U jien dejjem mort ma’/bi wieƫed minnhom hemm ‘And I always went there with one of them (ma’ = the boy wanted to come along/bi = irrespective of the boy’s own intentions)’, (C92.2’) Marru ma’/bi Shark ukoll ‘They went with Shark too (ma’ = the dog was looking forward for his regular walk/bi = the dog had to be taken somewhere no matter whether or not the dog liked to go there)’, (C92.3’) Dakinhar marru bil-ktieb

496 Notes moƫbi fil-basket ‘On that day they went with the book hidden in the basket.’: In the first two examples, both ma’ and bi are legitimate options. If ma’ is used both participants are considered equally active in the event; if bi is used the meaning changes from an egalitarian reading to a hierarchical asymmetry in which the companion is made to participate whereas the control over the action and the companion rests with the accompanee. As long as the companion participates in the action at least partly on his own initiative, bi is not appropriate if the noun it governs is animate, ma’ on the other hand does never combine with inanimate nouns. 288. (C93.3) is in fact a rather peculiar sentence. Ray Fabri (p.c.) explains the choice of ma’ instead of bi as follows: If bi was used, this would imply that the wiping one’s mouth is a (normal) function of the sleeve of one’s shirt. However, this is not the case. Therefore, the occasional instrumental function of the sleeve is turned into a spatial function (the sleeve is the longish surface along which the face is “wiped”). Note that bi would have been absolutely fine with a noun like maktur ‘handkerchief’ because wiping is an expected function of a handkerchief. Of course, with maktur the use of ma’ would be blocked because there is no need to “spatialise” the handkerchief when it is used on one of its canonical functions as an instrument, cf. (Avventura 6) b’maktur ƛkejken mesƫet id-dmugƫ li kien inƥema’ fuq xfar gƫajnejha ‘With a small handkerchief, she wiped away the tears that had gathered on her eyelashes.’ 289. More precisely, it is not only a matter of choice as to the preposition. The verbs themselves are also part of the story. As for ƫallat ‘to mix’ both ma’ and bi are used. According to Ray Fabri (p.c.) the use of ma’ in combination with presupposes the existence of two (or more) independent entities prior to the act of mixing which become one only as a result of the process of mixing. When bi is used there is an asymmetry of the entities involved in the process: In stead of mixing proper it seems to be more appropriate to speak of adding one ingredient to some basic entity. However, the bi used with Īewwaq ‘to mottle’ is an obligatory part of the case-frame of the verb whose semantics presuppose that one of the entities to be combined is less important than the other. 290. Ray Fabri (p.c.) points out that in example (C104.3) the use of bi in stead of ma’ would be grammatically acceptable. However, the reading would then be one of cause – meaning that the character Pipirella could not cope because of all the questions. 291. The preposition bez ‘without’ is the negative equivalent of ar. Bez governs the genitive in the singular and, as Latvian prepositions generally do, the dative in the plural. 292. For a comprehensive synchronic and diachronic study of the prepositional system of Latvian, cf. NƯtiƼa (1978).

Notes 497 293. Grammatical descriptions of Latvian which have been published outside the Baltic countries do not normally follow the Latvian philological tradition. In Nau (1998) there is no mention at all of a distinct instrumental case. Eckert, Bukeviþinjtơ, and Hinze (1994: 292–293) point out that there is no compelling reason for keeping the instrumental in the inflectional paradigms of modern Latvian. In the same vein, Holst (2001: 106) bans the instrumental from the paradigms – but he claims that there are dialects which still have a distinct instrumental. This instrumental is distinct from the accusative and dative because the suffix bears a different tone. Since tone is no longer distinctive in non-initial syllables, this suprasegmental strategy cannot be employed in standard Latvian. 294. Lötzsch (1978: 670 note 9) points to the fact that some of these bare instrumental forms of nouns have become separate entries in dictionaries and may therefore be considered lexemes of their own as, e.g., vietƗm ‘occasionally’ which is also the regular dative plural of vieta ‘place’. 295. The negative counterpart of með is likewise a preposition, viz. án ‘without’ governing the genitive (Kress 1982: 194). Segmentally, án is less complex than með with two segments as opposed to three. Owing to the fact that án always takes the genitive, those distinctions which depend on the differential case-government of með (cf. below) are neutralised under negation. A typical example of án used as preposition is (Bestu 158) Hún slakaði honum til pabbans og lét sig hverfa án athugasemda enda var þetta fastur liður í tilverunni hjá henni. ‘She handed him over to daddy and she kept bragging without end about this being a fixed part of her future.’ 296. Owing to the fact that accusative and dative are often morphologically not distinguished on indefinite nouns in the singular (as e.g. penna dative/accusative/genitive singular = accusative/genitive plural of penni ‘feather’) and on the pronouns for 1st and 2nd person plural (okkur ‘us [dative = accusative]’ and ykkar ‘you [dative = accusative]’), the examples given in the main text only involve uncontroversial cases i.e. mainly definite nouns of both numbers or indefinite nouns in the plural (as e.g. pennanum z penann ‘the feather [dative z accusativ] and pennum z penna ‘feathers [dative z accusative]) as well as pronouns other than the 1st and 2nd person plural. Sometimes dative and accusative can be kept apart even if morphological syncretism applies: Attributes (adjectives, etc.) agreeing with a morphologically underdifferentiating noun normally display distinct case-forms for the two categories under scrutiny. Examples which contain morphologically problematic cases are sometimes added and discussed in the accompanying footnotes. 297. Böðvarsson (1985: 631–632) provides a list of usages of með which largely corresponds to the one in Kress (1982) without being completely identical with the latter. 298. Með is also used as component part of complex conjunctions such as, e.g., með því að ‘because’ in (Meira 15) Sumir voru verstu skúrka og stríddu Jóni

498 Notes Bjarna með því að hann vaeri rangeygður. ‘Some were real bullies and teased Jón Bjarni because he was cross-eyed.’ 299. The use of this construction without preposition also includes some borderline cases which allow some kind of modal reading as, e.g., (Bestu 22–23) Nei, hún verður aldeilis ekki ein, stelpan, sagði Fróðbý hárri röddu ‘No, she absolutely won’t be alone, the girl, said Fróðbý loudly (lit. with a loud voice).’ 300. In addition, með is also employed to introduce a temporal adverbial (with a dependent noun in the dative). This usage is reported in Böðvarsson (1985: 632) although we have not come across pertinent examples in our small corpus (nor are they mentioned in Kress [1982]). According to the dictionary, með kvöldinu ‘in the/by evening’ and með degi ‘daily’ are typical instances of temporal með. 301. Barðdal (2001: 70) claims that the morphological genitive may also be used to mark a comitative relation. However, she does not provide any examples to prove her point – nor have we been able to ascertain her assumption on the basis of our corpus analysis. 302. In Icelandic, kinship terms are regularly combined with the verb eiga ‘to own’ in constructions of predicative possession. 303. It should not go unmentioned that reinforced constructions for comitative or instrumental með are practically unknown in Icelandic. In our corpus, the adverb saman ‘together’ occurs quite frequently – but never as a modifier of með! A typical instance of saman is (Lalli 24) Lalli var búinn að fá algjört ofnæmi fyrir þessum kennara og hefði helst viljað láta hann og Göldru búa saman til æviloka. ‘Lalli had developed an allergy against this teacher and he had rather have him and Galdra live together till the end of time.’ A similar situation can also be described by using með, but then saman seems to be excluded as in (Morð 79) Ég sá aldrei neinn annan sem virtist búa með honum. ‘I never saw anybody else who would have live with him.’ 304. In Barðdal (2001: 62), the term comitative is used in a rather general sense, i.e., it covers all cases of co-presence of two participants no matter whether or not they have equal control over the event. As our subsequent presentation of the facts is meant to demonstrate, this lumping approach adopted by Barðdal does not adequately account for the behaviour of the various constructions and their component parts. 305. Unfortunately, Barðdal (2001: 51–54) excludes PPs headed by með from her discussion of the functional domains of the accusative and the dative. We can only speculate that she deemed prepositions with differential case-government to be an inadequate starting-point for a semantically based analysis. One consequence of Barðdal’s decision to skip this class of prepositions is that in her overview the accusative is not assigned the function of an attribute (whereas the dative is considered to have this function). Of course, there is also a good formal reason for her to come to this conclusion: In her approach, attribute has a rather narrow reading meaning direct noun-noun attributes without a prepo-

Notes 499 sition linking the two NPs (cf. Barðdal 2001: 54). However, when looking at the Icelandic data from the vantage-point of semantic rôles, PPs headed by með are admissible as long as they introduce an instrument or a comitative (Barðdal 2001: 68–69). This inconsistency in treating the morphological cases and their governing preposition is at best unfortunate – methodologically speaking. 306. For a more detailed treatment of possession in Icelandic, cf. Stolz (2003b). 307. There are other instances of “unexpected” accusatives such as, e.g., (Krummi 92) Og hann varð að ríghalda sér og toga sig upp með handriðið, svo kröftugur var vindurinn. ‘And he had to hold on to and pull himself up on the railings so strong was the wind.’ The railings serve as some kind of instrument – which would call for the dative. There is no apparent reason for choosing the accusative in the present example except the possibility that the accusative is somehow part of the case-frame of the verb toga sig upp ‘to pull oneself upp’. This less likely possibility however could not be ascertained. 308. According to the frequency count in Barðdal (2001), the rate of subject-like datives outnumbers the attestations of subject-like accusatives five times whereas the accusative occurs three times more often as the morphological case of an object-NP than the dative. The direct object is most often marked by the accusative. 309. In Latvian, lƯdz(i), the marker of companion-orientation, is etymologically related to the nowadays somewhat archaic lƯdzƗs ‘alongside’ which has spatial meaning of the prolative kind. LƯdz(i) itself however cannot be used to express spatial concepts. 310. We do not look into the diachrony of the preposition which may involve the conflation of several erstwhile distinct prepositions. 311. In the predecessor of Heine and Kuteva (2002), Heine et al. (1993: 267) there were still eleven etymological sources for the Comitative marker. The striking difference in numbers is easy to explain: Heine and Kuteva (2002) achieved a reduction of potential source concepts by way of subsuming the semantically related ones under one more general concept. 312. For a full account of grammaticalisation in the realm of Comitativity/Instrumentality, it would be promising to look more closely at the many-to-manyrelationship between sources and targets: Has a given target concept the monopoly over a given source concept and vice versa? If a source allows for several targets or if a target may draw on various sources, are the competing categories in any conceivable way related to each other? As the subsequent discussion suggests, there is indeed strong evidence for the idea that there are certain restrictions and conceptual bridges which link, on the one hand, sources and targets and, on the other, the various competing sources or the various competing targets among each other. For practical reasons, we cannot go into the details of this otherwise essential area in this publication.

500 Notes 313. Examples of these patterns can be found in Finnish/Welsh IN COMPANY OF > Com, Yucatec Maya/Saami COMRADE > Com, Baka/Nama BODY > Com/Ins, To’aba’ita/Klamath/Vietnamese/Hmong ACCOMPANY > Com, Ainu/Mandarin Chinese FOLLOW > Com, Ewe MEET (REACH) > Com [according to Heine et al. (1993: 148) Ins in lieu of Com] (Stolz 1993: 44 note 96). Suffice it to mention the following examples: – Finnish: The Comitative postposition kanssa is a syncopated form of the regular inessive singular kansassa ‘in the people’ of the common noun kansa ‘people’ with PEOPLE embracing a variety of readings among which we encounter ‘associate(s), comrade(s), partner(s), community’ (Stolz 1994: 66–67), cf. also below; – Yucatec: The syncretistic preposition yetel ‘with, and’ derives from the relational noun et-el ‘companion(ship)’ which bears the possessor prefix – in this case the one of the 3rd person: y-et-el ‘his/her companion(ship)’ (Lehmann 1989b: 12). 314. For other potential sources of the Comitative which are associated with quantities and grammatical number, cf. Section 10.5. 315. For a slightly different interpretation of the Turkish case, cf. Nau (1995: 132). 316. The diagram (D6) does not fully exhaust the catalogue of etymological sources of markers used in the realm of Comitativity. Besides different sources for Confective and Ornative especially in serialising languages (which draw on verbal concepts such as, e.g., LOAD or BE PROVIDED WITH [Stolz 1993: 42]), borrowing may bring about interesting constellations. In Classical Armenian, for instance, the reinforced construction of the Comitative involves the free morpheme handerj ‘together’ which is derived via conversion from the noun handerj ‘clothes; dress’ – possibly a loanword from an Iranian language (Schmitt 1981: 173). Somehow, this case still fits in with the idea of closeness/proximity as the common denominator of the source concepts of Comitatives because the clothes one wears can be understood as an extension of the body or an additional layer which is contiguous with the body. 317. In this respect, we are reluctant to accept Palancar’s (2002: 149) localist stance without reservations. 318. This historical relation does not necessarily imply that the relators are synchronically syncretistic! Instrumentals may derive their grammatical expressions from an erstwhile Comitative with the latter function being renovated by grammaticalisation of a separate distinct Comitative marker. Thus, the pattern Comitative > Instrumental is not identical with a given language belonging to the B-type, cf. Section 13.2. 319. Thus, some of the evidence for a development Instrumental > Comitative adduced by Nau (1995: 128) are of no avail as they partly rely on conjecture. 320. Both koos and ühes may also be used as postpositions and, on top of that, are simultaneously registered as adverbs in Tauli’s (1983: 117 and 214) grammar of standard Estonian. The problem of how to describe the multifarious mor-

Notes 501 phosyntactic properties of koos has given rise to a controversy among normative grammarians (Valge 1988). 321. Nowadays, the morphological genitive is still fully alive although it is sometimes indistinguishable from the nominative in the singular. Elsewhere it displays a characteristic form of consonantal gradation. In the plural, -de/-te are used as genitive exponents. The former -n has disappeared completely. The instructive has long since been lost as a regular member of the case paradigm. What relics remain of the instructive are lexicalised adverbs all of which lack the final -n (Tauli 1983: 126). 322. As explained in 12.2, the bare instrumental cannot fulfil the functions of a prototypical Comitative, i.e., it is excluded from contexts with participants marked [+human]. Interestingly, the distinct inflectional instrumental has survived longest in the paradigm of the personal pronouns of the 1st and 2nd person singular, i.e., exactly there where animacy is highest. According to GƗters (1993: 165), the bare inflectional instrumental of these pronouns could still be used for Accompaniment in the oldest texts though only rather infrequently. In more recent periods, the use of the preposition ar became compulsory also with personal pronouns. 323. According to the Academy Grammar (Ulvydas 1965: 205), the use of the bare instrumental for Comitativity is typical of the Žemaitian dialects. 324. As a matter of fact, Polish also allows for na + N[ACCUSATIVE] and przez + N[ACCUSATIVE] as expressions of prototypical Instrumentality (widzieü na wáasne oczy ‘to see with one’s own eyes’, oglądaü przez lupĊ ‘to observe something through a looking glass’) although the bare inflectional instrumental still seems to be the unmarked case for this function (Laskowski 1972: 68–70). 325. In older Indo-European languages such as, e.g., Sanskrit, Latin and Old Greek the bare inflectional instrumental or a successor of the erstwhile instrumental used to encode Instrumentality whereas Comitativity was the task of a PP with a prepositional head governing the instrumental (or successor case) on the dependent noun. This is exactly the situation we find nowadays in the more conservative Slavic languages and mutatis mutandis in Lithuanian. 326. For more details on this topic, we refer the readers to Stolz and Stolz (1996a–b, 1997) and various other publications of the ongoing project on hispanicization carried out by members of our research crew at the University of Bremen. 327. Spanish has special pronominal forms of the preposition con for the 1st and 2nd persons singular and the reflexive 3rd persons of both numbers: conmigo, contigo, consigo. The final -go of these forms is a reflex of the Latin suffixed adposition cum as in mecum, tecum, secum, i.e., Comitativity is marked twice on the word-forms of modern Spanish, etymologically speaking. Similar forms are also attested in other Romance languages – for instance, in Sardinian. 328. Comitatives and Instrumentals of Amerindian languages in general and Mesoamerican languages in particular have not received the attention they deserve.

502 Notes Lehmann and Shin (2005) on Yucatec and Cuturi and Gnerre (forthcoming) on Huave are recent exceptions. 329. The most serious problem with the Hispanicization of the indigenous languages of the Americas (but less so of Austronesia) is created by the fact that the majority of speakers is bi- or multilingual with Spanish and thus occasional individual code-switching often cannot be told apart properly from borrowing (Voigtlander and Echegoyen 1985: 306). For obvious reasons, we cannot go into the intricacies of the network of factors which determine the use or disuse of borrowed items in the indigenous speech-communities. Owing to the at times considerable dialectal fragmentation of the various indigenous speech-communities, our observations cannot simply be understood as being valid for all the varieties which go by the name of Otomí etc. In the present section, the statements made refer exclusively to the local varieties described in our sources. 330. In these contexts, len may also function as a coordinating conjunction (Butler 1988: 247) – a fact which qualifies Zapotec as a WITH-language. 331. Hekking (1995: 150–161) studies the Spanish-derived prepositions in a variety of Otomí different from the one described by Voigtlander and Echegoyen (1985). In Santiago Mexquititlán, Hekking identifies ko(n) < Spanish con as the most frequent of all borrowed prepositions known to all of his informants – and observes that it is used both for Comitativity and Instrumentality. There is thus reason to assume that this variety of Otomí is moving towards the B-type. 332. In Otomí, con may also be used in contexts where it is not normally employed in the donor language, for instance, when the Instrument is a container, say a kettle or the like – in such cases, Spanish prefers spatial en ‘in’ (Voigtlander and Echegoyen 1985: 304). 333. This is in line with the observation made by Hill and Hill (1986: 188) in their monograph on the Hispanicization of Central Mexican Nahuatl: According to them, the Spanish-derived “prepositions a, con, en, and sin seem to appear mainly in fixed expressions”. 334. Tokharian (still unknown to the neogrammarians) is different in so far as both varieties have developed secondary case-exponents from erstwhile postpositions. Among these new cases, we encounter the inflectional comitative (= Tokharian A -aĞĞäl; Tokharian B -mpa) which is distinct from the inflectional instrumental (= Tokharian A = -yo; Tokharian B -sa [perlative]). The functional domains of these two new cases correspond largely with Comitativity and Instrumentality, respectively (Krause and Thomas 1960: 83–87). 335. It is tempting to assume that Hungarian was an A-type language prior to written documentation as this would be in line with the typological characteristics of other Ugric languages. However, in the famous funeral speech – the first piece of Hungarian in written form which stems from the late 12th century – the inflectional comitative-instrumental has already the functional range of a syncretistic marker (Mátai 2002).

Notes 503 336. Lötzsch (1996: 56) also mentions in passing that similar phenomena can be observed in non-standard varieties of other Slavic languages which have been exposed to particularly strong foreign influence. Heine and Kuteva (2005) also deal with diaspora varieties of Slavic languages such as, e.g., Croatian spoken in the Italian region of Molise. 337. We do not have to prove that language contacts have always been very strong and intricate in the Circum-Baltic region. The various contributions to the two volumes of Dahl and Koptjevskaja-Tamm (2001) give testimony of contactinduced phenomena on every level in dozens of languages. 338. In the light of the comparative evidence from other members of the larger Uralic phylum, this assumption receives additional credibility. In the Permian branch, Zyryan and Permyak distinguish by inflectional means an instrumental -ën from a comitative -kët/-këd. The same holds for the Obugrian branch where Ostyak and Vogul contrast instrumental -t6 ( l) with comitative -näät/naat. In Khanty, the opposition is instrumental -aat vs. comitative -naat, cf. the various contributions in Abondolo (1998). Chances are that this formal distinction can be projected back in time to a pre-historic period. It should not go unmentioned though that there are also other Uralic languages which lack a formal distinction based on inflection. All in all, the history of the Uralic Comitative is rather changeful; however, all Balto-Finnic languages including Saami have developed new expressions of Comitatives by grammaticalisation of (not always cognate) relational nouns (often with subsequent morphologisation) (Laanest 1982: 172–173). 339. In the case of Saami, the evidence is less conclusive as none of the modern varieties seems to make a formal distinction between Instrumental and Comitative although the inflectional comitative has peculiar traits: in the singular, it continues the erstwhile essive while, in the plural of Lule Saami, the expression used reflects the univerbation of a postpositional phrase involving the genitive plural and a relational noun guoibmi ‘companion’. According to Sammallahti (1998: 70), the genitive plural has Comitative functions in all varieties of Saami. We interpret this number-sensitive split as evidence of a former functional distinction of Instrumental vs. Comitative: The genitive reflects a historical genitive-instructive, i.e., a case which also had Instrumental functions. The postpositional phrase from which the modern comitative plural developed invites an interpretation as an erstwhile reinforcement of the Comitative according to the schema discussed above. Going by the short remarks in Laanest (1982: 172), the Finnish inflectional comitative -ine- is historically an instructive obligatorily marked for possessor, the possibility of a diachronic relation of the instructive and genitive in a variety of Uralic languages is mentioned by the same author (Laanest 1982: 167). Owing to the fact that it is again the Comitative which receives the more complex morphological expression (because of the obligatory possessor morpheme), we hypothesise that this is another instance of an erstwhile reinforcement of the Comitative reading of a formerly syncretistic relator.

504 Notes 340. The earliest Finnish texts are of course not free of at times heavy Swedish influence. This is especially case for those texts which were directly translated from a Swedish original. Collinder (1968: 28) mentions the first regal letter composed in Finnish which dates back to November 21, 1555. This letter starts with the formulaic opening Mee Gustaff Jwmalan Armon Kansa ‘we, Gustav, of god’s grace (king of Sweden)’ where the postposition kans(s)a is used in a way which does not conform to its present use as a Comitative marker, cf. below. This use of kanssa is said to be a typical trait of West Finnish dialects in contrast to the more conservative East Finnish ones where the inflectional comitative -ine- is more common (Collinder 1968: 65). Needless to say that the dialects in the western parts of Finland were exposed to Swedish influence for a longer time and because of Swedish settlements in situ also more intensively than the ones spoken in the East. Swedish expansion in Finland started in the mid-12th century. Swedish rule lasted until 1809 but Swedish continued to be the official language of the then Russian GrandDuchy. This is time enough for convergence phenomena to develop. 341. In the Dainas, the old traditional literature, the preposition-less instrumental is still possible as a marker of the Comitative although this is already an infrequent solution (GƗters 1993: 165–166). In the majority of cases, the Comitative goes along with the use of the preposition ar whereas prototypical instances of Instrumentality predominantly require the bare inflectional instrumental (GƗters 1983: 162). 342. In the oldest documents of written Lithuanian, the bare inflectional instrumental was obligatory when it came to expressing prototypical Instrumentality whereas Comitativity triggered the use of the preposition su governing the inflectional instrumental on the dependent noun (Schmalstieg 1988: 243 and 294). Note that Schmalstieg (1988: 243) claims that the “instrumental of means develops from the sociative instrumental” although no evidence of this assumption is provided. What can be stated nevertheless is the fact that in the early days of written Lithuanian, su + N[instrumental] could not be used for TOOL or the like nor could the bare inflectional instrumental be used for ACCOMPANIMENT etc. 343. In Nau’s (1995: 134) list of examples, the Finnish postposition kanssa which is reserved for Comitativity in standard Finnish may serve as a translation equivalent of Swedish med in any context and function: leikkiä auto-j-en kanssa {play:INF} {car}-{PL}-{GEN} {X} ‘to play with cars’ in lieu of standard Finnish auto-i-lla {car}-{PL}-{Y}. In addition, the inflectional adessive -lla/-llä which usually encodes Instrumentality but never Comitativity according to Finnish normative grammar, is relatively widespread in non-standard and journalistic Finnish as a marker of what Nau (1995: 135) calls the qualifying adessive which is equivalent to the Ornative: talo peltikato-lla {house} {tin-roof}-{Y} ‘house with a tin-roof’. However it is never used to express Accompaniment. 344. Nau (1995: 138) points out that her examples of the so-called qualitative adessive demonstrate that the reverse process cannot be excluded categorically.

Notes 505 345. The inventory of types is even smaller than the one Vennemann (1974) used for his well-known hypotheses about the patterns of word order change. This reduction of the number of potential options has the advantage that predictability is higher (though still not 100%).

Sources

Avventura

Casha, Charles (1999): Avventura fl-Gƫar il-Kbir. Blata lBajda: Merlin. Balthasar Kross, Jaan (1987): Kolme katku vahel. Balthasar Russowi Romaan I Köide. Tallinn: Eesti Raamat. Bestu Indriðason, Andrés (1991): Bestu vinir. Reykjavík: Iðunn. Delitti Attard, Edward (2002): Delitti f’Malta. 200 sena ta’ omiƛidji 1800-2000. Malta: BDL. DoƼuleja Gulbis, Harijs (1983): DoƼuleja. RƯga: Liesma. Ƥrajjiet Chetcuti, Ƥuzè (1999): Ƥrajjiet mill-ƫajja ta’ Kappillan. Rabat: Edizzjoni Studia. Háska Baldursdóttir, Heiður (1992): Háska leikur. Reykjavík: VakaHelgafell. Ifor Owen, Ivor (1973): Ifor Bach. Dinbych: Gwag L’Imsella Cauchi, Carmel G. (1994): Il-Misteru ta’ l’imsella msadda. Blata l-Bajda: Merlin. Kappa Schembri, Karl (2002): Taƫt il-kappa tax-xemx. Malta: Minima. Krummi Jónasar, Sigurgeir (1989): Krummi Kleinusnúður. Reykjavík: Fjölva. Lalli Þráinsson, Þorgrímur (1992): Lalli ljósastaur. Reykjavík: Fróði. Les noces barbares Queffélec, Yann (1985): Les noces barbares. Paris: Gallimard. LPP Albanian-Gheg Saint-Exupéry, Antoine (1965): Princi i vogël. Prishtinë: Rilindja. LPP Belarusan Saint-Exupéry, Antoine (1996): Ìàëåíüê³ Ïðûíö. ̳íñê: “Þíàöòâà”. LPP Breton Saint-Exupéry, Antoine (1974): Ar Priñs Bihan. GenvelEbrel: Preder. LPP Dutch Saint-Exupéry, Antoine (1989): De Kleine Prins. Rotterdam: Ad Donker. LPP Faroese Saint-Exupéry, Antoine (1980): Tann lítli prinsurin. Fuglafjørður: Egið Forlag. LPP Finnish Saint-Exupéry, Antoine (1992): Pikku Prinssi. Helsinki: Werner Söderström. LPP Galician Saint-Exupéry, Antoine (81994): O Principiño. Vigo: Editorial Galaxia.

Sources 507 LPP Lithuanian LPP Polish LPP Romanian LPP Russian LPP Slovenian LPP Spanish LPP Swedish LPP Welsh Meira Micmac

Moor

Morð

Náttvíg Paramiþà Qrempuƛu Rajt III Serafin Share Ulied Xindi

Saint-Exupéry, Antoine (1995): Mažasis Princas. Vilnius: Džiugas. Saint-Exupéry, Antoine (1996): Maáy KsiąĪĊ. Warszawa: Instytut Wydawniczy Pax. Saint-Exupéry, Antoine (s.a.): Micul PrinĠ. Bucureúti: Editura Ion Creangă. Saint-Exupéry, Antoine (1992): Maëeíüêèé ïðèíö. Mocêâà: ìåæäóíapoäíüýe îòíîøåíèÿ. Saint-Exupéry, Antoine (1991): Mali Princ. Ljubljana: Založba Mladinska knjiga. Saint-Exupéry, Antoine (1995): El Principito. Barcelona: Sant Vicenç dels Horts. Saint-Exupéry, Antoine (1995): Lille Prinsen. Stockholm: Rabén & Sjögren. Saint-Exupéry, Antoine (1975): Y Tywysog Bach. Liverpool: Cyhoeddiadau Modern Cymreig Cyf. A Gwasg Y Brython. Helgadóttir, Guðrún (1975): Meira af Jóni Oddi og Jóni Bjarna. Reykjavík: Iðunn. Malet, Léo (1987): Micmac moche au boul’mich. Les nouveaux mystères de Paris (Ve arrondissement). Paris: Union générale d’éditions. Crombie, Deborah (1994): Das Hotel im Moor. Aus dem Englischen von Mechthild Sandberg Cileti. München: Goldmann. Jacobs, Jonnie (1995): Morð í hverfinu. Reykjavík: Úrvals bækur. [translated from the English original: Murder among neighbours 1994] Vilhjálmsson, Thor (1989): Náttvíg. Reykjavík: Mál og menning. Andre Barthelemy (s.a.): Paramiþà la balvalàke. S. l. Zahra,Trevor (1991): Qrempuƛu f’Belt il-Ƥobon. Blata lBajda: Merlin. Ganado, Herbert (1977): Rajt Malta tinbidel. It-tielet ktieb 1942–1955. Malta: Interprint. Mifsud, Pawlu (1998): Serafin is-saƫƫar. Malta: Klabb Kotba Maltin. Crombie, Deborah (2003): A share in death. New York: Avon. Agius, Ben (2002): Ulied il-kwiekeb. Malta: Klabb Kotba Maltin. Cutajar, Tony C. (2002): Il-misteru tar-razzett ta’ Xindi. Malta: Klabb Kotba Maltin.

References

Abaev, Vasilij Ivanoviþ 1964 A grammatical sketch of Ossetic. The Hague: Mouton. Abdülhayo÷lu, Suphi 1990 Türkisch-deutsches Valenzlexikon. Hohengehren: Schneider. Abondolo, Daniel (ed.) 1998 The Uralic languages. London: Routledge. Adrados, Francisco R. 2001 Geschichte der griechischen Sprache. Von den Anfängen bis heute. Tübingen: Francke. Aissen, Judith 1989 Agreement controllers and Tzotzil comitatives. Language 65.4: 518–536. Alhoniemi, Alho 1993 Grammatik des Tscheremissischen (Mari). Hamburg: Buske. Altmann, Gabriel, and Werner Lehfeldt 1973 Allgemeine Sprachtypologie: Prinzipien und Meßverfahren. München: Fink. Amaya, María Trillos 1989 Aspecto, modo y tiempo in Damana. In Lenguas aborigines de Colombia. Descripciones 3: Sierra Nevada de Santa Marta, N.N. (eds.), 3–142. Bogotá: Centro colombiano de Estudios en Lenguas Aborigines. Anderson, John M. 1971 The grammar of case. Towards a localistic theory. Cambridge: Cambridge University Press. 1977 On case grammar. Prolegomena to a theory of grammatical relations. London: Croom Helm. Andrews, Avery D. 2001 Non-canonical A/S-marking in Icelandic. In Non-canonical marking of subjects and objects, Alexandra Y. Aikhenvald, Robert Malcolm Ward Dixon, and Masayuki Onishi (eds.), 85–112. Amsterdam: Benjamins. Anttila, Raimo 1989 Historical and comparative linguistics. Amsterdam: John Benjamins. Aquilina, Joseph 1987 Maltese-English dictionary. Vol. I: A–L. Malta: Midsea Books. 1990 Maltese-English dictionary. Vol. II: M–Z and addenda. Malta: Midsea Books.

References 509 Ariste, Paul 1983 Vadja komitatiiv. Fenno-Ugristica 10: 16–25. Balkeviþius, Jonas, and Jonas Kabelka 1977 Latviǐ-Lietuviǐ kalbǐ žodynas. Vilnius: Mokslas. Bánhidi, Zoltán, Zoltán Jókay, and Dénes Szabó 1975 Lehrbuch der ungarischen Sprache. München: Hueber. Barðdal, Jóhanna 2001 Case in Icelandic. A synchronic, diachronic and comparative approach. Lund: Department of Scandinavian Languages. Barnes, Michael 1994 Nyggjar hugsanir um fyrisetingina vid í føroyskum. Málting 10 (1): 23–36. Bartens, Hans-Hermann 1989 Lehrbuch der saamischen (lappischen) Sprache. Hamburg: Buske. Bauer, Winfried 1993 Maori. London: Routledge. Baxter, Alan N. 1985 A grammar of Kristang (Malacca Creole Portuguese). Canberra: The Australian National University. Bechert, Johannes 1988 Konvergenz und Individualität von Sprachen. In Studien zum Sprachkontakt, Karl Heinz Wagner and Wolfgang Wildgen (eds.), 25–41. (BLIcK 1, Schriftenreihe des Instituts für Allgemeine und Angewandte Sprachwissenschaft.) Bremen: Universitätsdruckerei. 1991 The problem of semantic incompatibility. In Semantic universals and universal semantics, Dietmar Zaefferer (ed.), 60–71. Berlin, New York: Foris. Behagel, Otto 1923–32 Deutsche Syntax. Vol. 1: Nomen, Pronomen; vol. 2: Adverbium, Verbum; vol. 3: Die Satzgebilde; vol. 4: Wortstellung, Periodenbau. Heidelberg: Winter. Bender, Ernst 1967 Hindi grammar and reader. Philadelphia: University of Pennsylvania Press. Benzig, Johannes 1955 Lamutische Grammatik. Wiesbaden: Franz Steiner. 1985 Kalmückische Grammatik zum Nachschlagen. Wiesbaden: Harrassowitz. Berg, Thomas 2001 Linguistic structure and change. An explanation from language processing. Oxford: Oxford University Press. Berger, Hermann 1974 Das Yasin-Burushaski (Werchikwar). Wiesbaden: Harrassowitz.

510 References Bergsland, Knut 1947 Omkring my s toboj ‘vi med dig’ (du og jeg) osv. In Festskrif til Olaf Brochs 80-årsdag,1–12. Oslo: Norske-Videnskaps Akademi. 1994 Aleut dictionary. Unangam tunudgusii. Fairbanks: Alaska Native Language Center. Besnier, Niko 2000 Tuvaluan. A Polynesian language of the Central Pacific. London: Routledge. Bhat, Darbhe Narayuana Shanhara 2003 Pronouns. Oxford: Oxford University Press. Biboviü, Ljiljana 1976 On the notion of body-part instrument. Folia Linguistica 9: 311–324. Bisang, Walter 1992 Das Verb im Chinesischen, Hmong, Vietnamesischen, Thai und Khmer. Vergleichende Grammatik im Rahmen der Verbserialisierung, der Grammatikalisierung und der Attraktorpositionen. Tübingen: Narr. 2004 Dialectology and typology – an integrative perspective. In Dialectology meets typology. Dialect grammar from a cross-linguistic perspective, Bernd K ortmann (ed.), 11–46. Berlin: Mouton de Gruyter. Bittner, Dagmar 1995 Affixhomonymie in der Natürlichkeitstheorie – Betrachtung der Form-Inhalt-Relationen bei den -er-Bildungen im Deutschen. In Natürlichkeitstheorie und Sprachwandel, Norbert Boretzky, Wolfgang U. Dressler, and Andreas Bittner (eds.), 123–144. Bochum: Brockmeyer. Bläsing, Uwe 2003 Kalmuck. In The Mongolic languages, Juha Janhunen (ed.), 229– 247. London: Routledge. Blake, Barry J. 1977 Case marking in Australian languages. Canberra: Australian Institute of Aboriginal Studies. 1994 Case. Cambridge: Cambridge University Press. Blum, Martin (ed.) 1967 Basilius von Cäsarea. Über den Heiligen Geist. Freiburg i.Br.: Lambertus. Böðvarsson, Árni 1985 Íslensk orðabók handa skólum og almenningi. Reykjavík: Bókaútgáfa Menningarsjóðs. Boretzky, Norbert 1994 Romani. Grammatik des Kalderaš-Dialektes mit Texten und Glossar. Wiesbaden: Harrassowitz.

References 511 Borg, Albert 1988 Ilsienna. Studju grammatikali. ƪas Sajjied: s.l. Borg, Albert, and Marie Azzopardi-Alexander 1997 Maltese. London: Routledge. Bornemann, Eduard, and Ernst Risch 1978 Griechische Grammatik. Frankfurt a.M.: Diesterweg. Bossong, Georg 1985 Empirische Universalienforschung. Differentielle Objektmarkierung in den neuiranischen Sprachen. Tübingen: Narr. Braune, Wilhelm, and Ernst A. Ebbinghaus 1973 Gotische Grammatik. Tübingen: Niemeyer. Braune, Wilhelm, and Walther Mitzka 1959 Althochdeutsche Grammatik. Tübingen: Niemeyer. Brauner, Siegmund 1974 Lehrbuch des Bambara. Leipzig: Enzyklopädie. Brincat, Joseph 2000 Malti – Elf sena ta’ stoja. Malta: PIN. Bromley, H. Myron 1981 A grammar of Lower Grand Valley Dani. Canberra: Australian National University. Brugmann, Karl, and Berthold Delbrück 1911 Grundriss der vergleichenden Grammatik der indogermanischen Sprachen. 2. Band, 2. Teil. Strassburg: Karl J. Trübner. Buckingham, Hugh W. 1973 The comitative and case grammar. Foundations of Language 10: 1–39. Butler H., Inez M. 1988 Gramática zapoteca. Zapoteco de Yatzachi el Bajo. México: Instituto Lingüístico de Verano. Bybee, Joan, William Pagliuca, and Revere D. Perkins 1991 Back to the future. In Approaches to grammaticalization. Vol. 2: Focus on types of grammatical markers, Elizabeth Traugott, and Bernd Heine (eds.), 17–58. Amsterdam/Philadelphia: John Benjamins. Bybee, Joan, and Östen Dahl 1989 The creation of tense and aspect in the languages of the world. Studies in Language 13 (1): 51–103. Byrne, Francis 1984 Instrumental in Saramaccan. York Papers in Linguistics 11: 39–50. 1987 Grammatical relations in a radical Creole: Verb complementation in Saramaccan. Amsterdam: Benjamins. Cadiot, Pierre 1990 La préposition avec: grammaire et représentation. Le Français Moderne 58: 152–173.

512 References 1991

A la hache ou avec la hache? Représentation mentale, expérience située et donation du référent. Langue Française 91: 7–23. 1993 De et deux des ses concurrents: Avec et À. Langages 110: 68–106. Campbell, Lyle (ed.) 2001 Grammaticalization: A critical assessment. Oxford: Pergamon. Campe, Petra 1994 Case, semantic roles and grammatical relations. A comprehensive bibliography. Amsterdam: Benjamins. Carlin, Eithne 1993 The So language. Köln: Institut für Afrikanistik. Carlson, Robert 1994 A grammar of Supyire. Berlin: Mouton de Gruyter. Catta Quelen, Javier 1987 Gramatica del quichua ecuatoriano. Quito: Abya Yala. Clark, Eve, and Kathie L. Carpenter 1989 On children’s use of from, by and with in oblique noun phrases. Journal of Child Language 16 (2): 349–364. Clements, J. Clancy 1996 The genesis of a language. The formation and development of Korlai Portuguese. Amsterdam: John Benjamins. Collinder, Björn 1968 Finnisch als Kultursprache. Hamburg-Volksdorf: Christoph von der Ropp. Comrie, Bernard 1982 Grammatical relations in Huichol. In Studies in Transitivity, Paul J. Hopper, and Sandra A. Thompson (eds.), 95–115. (Syntax and Semantics 15.) New York: Academic Press. 1985 Causative verb formation and other verb-deriving morphology. In Language typology and syntactic description. Vol. 3, Timothy Shopen (ed.), 309–348. Cambridge: Cambridge University Press. 1993 Language universals and linguistic typology: Data-bases and explanations. Sprachtypologie und Universalienforschung 46 (1): 3–14. 2001 The Slavonic languages. London: Routledge. Comrie, Bernard, and Norval Smith 1977 Lingua descriptive studies: questionnaire. Lingua 42: 1–72. Coseriu, Eugenio 1970 Sprache. Strukturen und Funktionen. XII Aufsätze zur Allgemeinen und Romanischen Sprachwissenschaft. Tübingen: Narr. Craúoveanu, Dumitru 1980 ConsideraĠii privind circumstanĠialul instrumental. Limba Română 29 (6): 591–596. Croft, William 1990 Typology and universals. Cambridge: Cambridge University Press.

References 513 1991

Syntactic categories and grammatical relations: The cognitive organization of information. Chicago: The University of Chicago Press. 1995 Autonomy and functionalist linguistics. Language 71: 490–532. 1998 Linguistic evidence and mental representations. Cognitive Linguistics 9 (2): 151–173. 2001 Radical construction grammar. Syntactic theory in typological perspective. Oxford: Oxford University Press. Croft, William, Hava Bat-Zeev Shyldkrot, and Suzanne Kemmer 1988 Diachronic semantic processes in the middle voice. In Papers from the 7th International Conference on Historical Linguistics, Anna Giacalone Ramat, Onofrio Carruba, and Giuliano Bernini (eds.), 179– 192. Amsterdam: Benjamins. Cuturi, Flavia, and Maurizio Gnerre forthcoming Concomitance in Huave. Unpublished manuscript. Cyffer, Norbert 1991 We learn Kanuri. Köln: Köppe. Cysouw, Michael 2003 Against implicational universals. Linguistic Typology 71: 89–101. Dahl, Östen, and Maria Koptjevskaja-Tamm (eds.) 2001 The Circum-Baltic languages: Typology and contact. Amsterdam: Benjamins. Dambrinjnas, Leonardas, Antanas Klimas, and William R. Schmalstieg 1980 Introduction to modern Lithuanian. Brooklyn/NY: Franciscan Fathers. Daniel, Michael, and Edith Moravcsik 2005 The Associative Plural. In The World Atlas of Language Structures, Martin Haspelmath, Matthew S. Dryer, David Gil, and Bernard Comrie (eds.), 150–153. Oxford: Oxford University Press. Davies, John 1989 Kobon. London: Croom Helm. Derbyshire, Desmond C. 1979 Hixkaryana. Amsterdam: North-Holland Publishing Company. Dixon, Robert Malcolm Ward (ed.) 1976 Grammatical categories in Australian languages. New Jersey: Humanities Press. 1980 The languages of Australia. Cambridge: Cambridge University Press. Dressler, Wolfgang U. 1985 Morphonology: The dynamics of derivation. Ann Arbor: Karoma. Dressler, Wolfgang U. (ed.) 1987 Leitmotifs in Natural Morphology. Amsterdam: John Benjamins. Dressler, Wolfgang U., and Willi Mayerthaler 1987 Introduction. In Leitmotifs in Natural Morphology, Wolfgang U. Dressler (ed.), 3–24.

514 References Drossard, Werner 1984 Das Tagalog als Repräsentant des aktivischen Sprachbaus. Tübingen: Narr. 1991 Kasusmarkierung: zur Zentralität und Peripherizität von Partizipanten. In Partizipation. Das sprachliche Erfassen von Sachverhalten, Hansjakob Seiler, and Waldfried Premper (eds.), 446–481. Tübingen: Narr. Dryer, Matthew S. 1989 Large linguistic areas and language sampling. Studies in Language 13: 257–292. 2000 Counting genera vs. counting languages. Linguistic Typology 4 (3): 334–349. 2003 Significant and non-significant implicational universals. Linguistic Typololgy 7 (1): 108–127. Dutton, Tom E. 1996 Koiari. München: LINCOM Europa. Dyla, Stefan 1988 Quasi-comitative coordination in Polish. Linguistics 26: 383–414. Eckert, Rainer, Elvira Bukeviþinjtơ, and Friedhelm Hinze 1994 Die baltischen Sprachen. Eine Einführung. München: Langenscheidt. Endruschat, Annette 1997 Die Präposition com im Afroportugiesischen: Versuch einer neuen Interpretation ihrer spezifischen Verwendung. In Afrolusitanistik – eine vergessene Dispziplin in Deutschland?, Ruth Degenhardt, Thomas Stolz, and Hella Ulferts (eds.), 243–278. Bremen: Universität Bremen. forthcoming Komitativität und Verben - durch ‚mit’ eingeleitete präpositionale Objekte in den romanischen Sprachen? (Diversitas Linguarum 13.) Bochum: Brockmeyer. EndzelƯns, JƗnis 1951 Latviešu valodas gramatika. RƯga: Latvijas Valsts IzdevniecƯba. Eroms, Hans-Werner 1976 Zu deutschen Präpositionalphrasen mit ‘mit’. Sprachwissenschaft 1: 223–240. Ersen-Rasch, Margarete I. 1980 Türkisch für Sie. München: Hueber. Evans, D. Simon 1977 Gramadey Cymraeg Canol. Caerdydd: Gwasg Prifysgol Cymru. Faarlund, J. Terje 1974 Preposisjonsuttrykk innledet av med in moderne norsk. Norsk Tidskrift for Sprogvidenskap 28: 45–62. Fabri, Ray 1996 The construct state and the pseudo-construct state in Maltese. Rivista di Linguistica 8 (1): 229–244.

References 515 Fanselow, Gisbert, and Sascha Felix 1993 Sprachtheorie 2: Die Rektions- und Bindungstheorie. Tübingen: Francke. Fillmore, Charles J. 1968 The case for case.In Universals in linguistic theory, Emmon Bach, and Robert T. Harms (eds.), 1–88. New York: Holt, Rinehart and Winston. Fillmore, Charles, Paul Kay, and Catherine O’Connor 1988 Regularity and irregularity in grammatical constructions. The case of let alone. Language 64: 501–538. Fonseca, Maria do Ceu 1994 Contribution à l’étude de la preposition en portugais: Le monème fonctionnel com ‘avec’. La Linguistique 30 (1): 45–57. Fortescue, Michael 1984 West Greenlandic. London: Croom Helm. Friedländer, Marianne 1992 Lehrbuch des Malinke. Leipzig: Langenscheidt/Enzyklopädie. Gabelentz, Georg von der 1891 Die Sprachwissenschaft. Ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: Tauchnitz. GƗters, AlfrƝds 1993 Lettische Syntax. Die Dainas. Frankfurt a.M.: Peter Lang. Gauto Bejarano, Miguel Ángel 1990 Introducción a la gramática elemental del guaraní con explicaciones en español y alemán. Asunción: El Foro. Gehling, Thomas 2004 ‘Ich’, ‘du’ und andere. Eine sprachtypologische Studie zu den grammatischen Kategorien “Person” und “Numerus”. Münster: LIT. Georg, Stefan, and Alexander P. Volodin 1999 Die itelmenische Sprache. Grammatik und Texte. Wiesbaden: Harrassowitz. Girke, Wolfgang 1995 Komitative Konstruktionen im Russischen. In Slavistische Linguistik 1994. Referate des 20. Konstanzer Slavistischen Arbeitstreffens, Daniel Weiss (ed.), 97–124. München: Sagner. Givón, Talmy 1975 On the change from SOV to SVO. In Word order and word order change, Charles Li (ed.), 47–112. Austin: The University of Texas Press. 1979 On understanding grammar. New York: Academic Press. 1984 Syntax. A functional-typological introduction. Vol. I. Amsterdam: John Benjamins. 1995 Functionalism and grammar. Amsterdam: Benjamins.

516 References Goebl, Hans 2001 Arealtypologie und Dialektologie. In Language typology and language universals/Sprachtypologie und sprachliche Universalien/La typologie des langues et les universaux linguistiques. Volume 2, Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds.), 1471–1491. Berlin/New York: de Gruyter. Greenberg, Joseph J. 1966 Language universals with special reference to feature hierarchies. The Hague: Mouton. Grenand, Françoise 1980 La langue Wayãpi (Guyane française). Phonologie et grammaire. Paris: SELAF. Griefenow-Mewis, Catherine, and Tamene Bitima 1994 Lehrbuch des Oromo. Köln: Köppe. Haase, Martin 1992 Sprachkontakt und Sprachwandel im Baskenland. Die Einflüsse des Gaskognischen und Französischen auf das Baskische. Hamburg: Buske. Habicht, Külli 2000 Grammaticalization of adpositions in Old Literary Estonian. In Estonian: Typological studies IV, Mati Erelt (ed.), 19–58. Tartu: University of Tartu. Haenisch, Erich 1986 Mandschu-Grammatik. Leipzig: Enzyklopädie. Hagège, Claude 1986 La structure des langues. Paris: PUF. 1993 The language builder. An essay on the human structure in linguistic morphogenesis. Amsterdam: John Benjamins. Haiman, John 1985 Natural syntax. Cambridge: Cambridge University Press. Hakulinen, Lauri 1949 Komitatiiviattribuutin olemattomuudesta. Virittäjä 53: 172–173. Hansen, Erik 1971 Jensen er nede i postkassen med et brev. Konstruktioner med consubjectum i moderne dansk. Danske Studier 66: 5–36. Hardman, Martha James, Juana Vásquez, and Juan de Dios Yapita 1988 Aymara. Compendio de estructura fonologica y gramatical. La Paz: Gramma. Harris, Alice C. 2002 Endoclitics and the origin of Udi morphosyntax. Oxford: Oxford University Press. Hartmann, Roswitha 1985 Rimaykullayki. Unterrichtsmaterialien zum Quechua Ayacuchano. Berlin: Reimer.

References 517 Haspelmath, Martin 1993 Lezgian. Berlin: Mouton de Gruyter. 2000 Why can’t we talk to each other? Lingua 110: 235–255. 2001 The European linguistic area: Standard Average Euopean. In Language typology and language universals/Sprachtypologie und sprachliche Universalien/La typologie des langues et les universaux linguistiques. Volume 2, Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds.), 1492–1510. Berlin/New York: de Gruyter. 2003 The geometry of grammatical meaning: semantic maps and crosslinguistic comparison. In The new psychology of language. Vol. 2, Michael Tomasello (ed.), 211–242. Mahwah/NJ: Erlbaum. forthcoming Coordination. In Language typology and linguistic description, Timothy Shopen (ed.), Cambridge: Cambridge University Press. [quoted from the preliminary version on the net http://email.eva. mpg.de/~haspelmt/coord.pdf] Haspelmath, Martin, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds.) Language typology and language universals/Sprachtypologie und sprachliche Universalien/La typologie des langues et les universaux linguistiques. Volume 2. Berlin/New York: de Gruyter. Hasselblatt, Cornelius 1992 Grammatisches Wörterbuch des Estnischen. Wiesbaden: Harrassowitz. 1995 Lehrbuch des Estnischen. Wiesbaden: Harrassowitz. Heine, Bernd 1997 Possession. Cognitive sources, forces, and grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer 1991 Grammaticalization. A conceptual framework. Chicago: The University of Chicago Press. Heine, Bernd, Tom Güldemann, Christa Kilian-Hatz, Donald A. Lessau, Heinz Rohberg, Mathias Schladt, and Thomas Stolz 1993 Conceptual shift. A lexicon of grammaticalization processes in African languages. Köln: Institut für Afrikanistik. Heine, Bernd, and Tania Kuteva 2002 World lexicon of grammaticalization. Cambridge: Cambridge University Press. 2005 Language contact and grammatical change. Cambridge: Cambridge University Press. Hekking, Ewald 1995 El Otomí de Santiago Mexquititlán: desplazamiento lingüístico, préstamos y cambios gramaticales. Amsterdam: IFOTT. Hetzer, Armin 2003 Estnisch. Eine Einführung. Wiesbaden: Harrassowitz.

518 References Hetzron, Robert 1973 Conjoining and comitativization in Hungarian – a study in rule ordering. Foundations of Language 10: 493–507. Hill, Jane H., and Kenneth C. Hill 1986 Speaking Mexicano. Dynamics of syncretic language in Central Mexico. Tuscon: The University of Arizona Press. Himmelmann, Nikolaus P. 2000 Towards a typology of typologies. Sprachtypologie und Universalienforschung 53 (1): 5–12. Hjelmslev, Louis 1972 La catégorie des cas. Étude de grammaire générale. München: Fink. Hoddinott, William G., and F. M. Kofod 1976 Djamindjungan. In Grammatical categories in Australian languages, Robert Malcolm Ward Dixon (ed.), 397–401. New Jersey: Humanities Press. Höftmann, Hildegard 1993 Grammatik des Fon. Leipzig: Langenscheidt/Enzyklopädie. Holes, Clive 1990 Gulf Arabic. London: Routledge. Hollenbach, Fernando, and Elena Hollenbach 1975 Trique de San Juan Copala. México: El Colegio de México. Holst, Jan Henrik 2001 Lettische Grammatik. Hamburg: Buske. Iordan, Iorgu, and Vladimir Robu 1978 Limba română contemporană. Bucureúti: Editura didactică úi pedagogică. Ivanov, Vyacheslav 1994 Typology of non-possession (Caritivity). I: Indo-European, II: Paleo-Siberian, III: Amerindian. Series of lectures at the Institute of Linguistics/University of Cologne, May 30 – June 6, 1994. Jacobson, Steven A. 1984 Yup’ik Eskimo dictionary. Fairbanks: Alaska Native Language Center. Janda, Laura A. 1993 A geography of case semantics. The Czech dative and the Russian instrumental. Berlin/New York: Mouton de Gruyter. Jensen, Hans 1969 Grammatik der kanaresischen Schriftsprache. Leipzig: Enzyklopädie. Jolly, Julia A. 1993 Preposition assignment in English. In Advances in role and reference grammar, Robert D. Van Valin (ed.), 275–310. Amsterdam: John Benjamins. Jones, Linda K. 1979 Discourse studies in Mesoamerican languages. 2 vols. Dallas: University of Texas at Arlington.

References 519 Jones, Morris, and Alan R. Thomas 1977 The Welsh language: studies in its syntax and semantics. Cardiff: University of Wales Press. Kämpfe, Hans-Rainer, and Alexander P. Volodin 1995 Abriß der tschuktschischen Grammatik auf der Basis der Schriftsprache. Wiesbaden: Harrassowitz. Karlsson, Fred 1978 Finsk grammatik. Helsinki: Suomalaisen Kirjallisuuden Seura. Kastenholz, Raimund 1989 Grundkurs Bambara (Manding). Köln: Köppe. Keenan, Edward L. 1985 Passives in the world’s languages. In Language typology and syntactic description. 3 Vols, Timothy Shopen (ed.), 243–280. Cambridge: Cambridge University Press. Keller, Rudi 1991 Sprachwandel. Von der unsichtbaren Hand in der Sprache. Tübingen: Francke. Kemmer, Suzanne 1993 The Middle voice. Amsterdam: Benjamins. Kempf, Zdislaw 1974 Sur les cas contextuels. Studi italiani di linguistica teorica ed applicata 2 (3): 345–360. Kenesei, István, Robert M. Vago, and Anna Fenyesi 1998 Hungarian. London: Routledge. Kibrik, Aleksandr E., and Ja. G. Testelets 2004 Bezhta. In The indigenous languages of the Caucasus. Vol. 3: The North East Caucasian languages. Part I, Dieter Michael Job (ed.), 217–297. Ann Arbor: Caravan Books. Kilian-Hatz, Christa 1995 Das Baka. Grundzüge einer Grammatik aus der Grammatikalisierungsperspektive. Köln: Institut für Afrikanistik. Kilian-Hatz, Christa, and Thomas Stolz 1993 Grammatikalisierung und grammatische Kategorien. Ein Bericht aus der Pathologie. Essen: Universität GH Essen, FB Sprach- und Literaturwissenschaften. Kircher, Chantal 1985 Emplois et valeurs de la désinence d’ablatif en latin. In Syntaxe et latin. Actes du IIme Congrès International de Linguistique Latine, Aix-en-Provence, 28–31 Mars 1983, Christian Touratier (ed.), 139– 156. Marseille: Jeanne Laffitte. Klein, Hans-Wilhelm, and Hartmut Kleineidam 1983 Grammatik des heutigen Französisch. Stuttgart: Klett.

520 References Koch, Wolfgang 1978 Kasus – Kognition – Kausalität: Zur semantischen Analyse der instrumentalen mit-Phrasen. Lund: Gleerup. Kornfilt, Jaklin 1997 Turkish. London: Routledge. Kortmann, Bernd (ed.) 2004 Dialectology meets typology. Dialect grammar from a crosslinguistic perspective. Berlin: Mouton de Gruyter. Kozintseva, Natalia 1995 Modern Eastern Armenian. München: LINCOM Europa. Krause, Wolfgang, and Werner Thomas 1960 Tocharisches Elementarbuch. Band 1: Grammatik. Heidelberg: Carl Winter. Kress, Bruno 1982 Isländische Grammatik. Leipzig: Enzyklopädie. Kroeber, Alfred L., and George W. Grace 1960 The Sparkman grammar of Luiseño. Berkeley: University of California Press. Kross, Kristiina 1983 Die Grenze zwischen dem Kasus und dem Adverb im Instruktiv der ostseefinnischen Sprachen. Sovetskoe Finno-Ugrovedenie 4: 241–253. Kühnel, Helmut 1999 Wörterbuch des Baskischen. Wiesbaden: Reichert. Kühner, Raphael, and Carl Stegmann 1955 Ausführliche Grammatik der lateinischen Sprache. Satzlehre. 1. Teil. Hannover: Hahnsche Buchhandlung. Kunze, Jürgen 1992 Einige Betrachtungen zum Komitativ und zu verwandten Konstruktionen. In Fügungspotenzen. Zum 60. Geburtstag von Manfred Bierwisch, Ilse Zimmermann, and Anatol Strigin (eds.), 111–131. Berlin: Akademie Verlag. Kwee, John B. 1981 Indonesian. A complete working course. London: Hodder & Stoughton. Laanest, Arvo 1982 Einführung in die ostseefinnischen Sprachen. (Autorisierte Übertragungs aus dem Estnischen von Hans-Hermann Bartens.) Hamburg: Buske. Lakoff, George 1968 Instrumental adverbs and the concept of deep structure. Foundations of Language 4: 4–29. 1987 Women, fire and dangerous things. What categories reveal about the mind. Chicago: The University of Chicago Press.

References 521 Lakoff, George, and Mark Johnson 1980 Metaphors we live by. Chicago, London: The University of Chicago Press. Landmann, Fred 1989 Group I & II. Linguistics and Philosophy 12: 559–605 and 723–744. Lasersohn, Peter 1990 Group action and spatio-temporal proximity. Linguistics and Philosophy 13: 179–206. Laskowski, Roman 1972 Polnische Grammatik. Leipzig: Enzyklopädie. Lastra, Yolanda 1980 Nahuatl de Acaxochitlán. México: El Colegio de México. Launey, Michel 1981 Introduction à la langue et à la littérature aztèques. Tome 1: Grammaire. Paris: L’Harmattan. Lazard, Gilbert 1994 L’actance. Paris: Presses universitaires de France. 1999 Pour une terminologie rigoureuse: Quelques principes et propositions. Mémoires de la Société de Linguistique de Paris N.S. 6: 111–133. Lee, Sang-Hyup 1990 Konfrontative Analyse zwischen dem ungarischen und koreanischen Kasussystem. Wiesbaden: Harrassowitz. Lefevbre, Claire 1989 Instrumental take-serial constructions in Haitian and in Fon. Canadian Journal of Linguistics 34 (3): 319–337. Leger, Rudolf 1994 Eine Grammatik der Kwami-Sprache (Nordostnigeria). Köln: Köppe. Lehiste, Ilse 1969 ‘Being’ and ‘having’ in Estonian. Foundations of Language 5: 324–341. Lehmann, Christian 1983 Rektion und syntaktische Relationen. Folia Linguistica 17: 339–378. 1984 Progress in general comparative linguistics. Studies in Language 8: 259–286. 1989a Language description in General Comparative Grammar. In Reference grammar and modern linguistic theory, Gottfried Graustein, and Gerhard Leitner (eds.), 133–162. Tübingen: Max Niemeyer. 1989b Grammatikalisierung und Lexikalisierung. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42 (1): 11–19. 1992 Yukatekische lokale Relatoren in typologischer Perspektive. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 45 (6): 626–641. 1995 Thoughts on grammaticalization. München: LINCOM.

522 References Lehmann, Christiann and Yong-Min Shin 2005 The functional domain of concomitance. A typological study of instrumental and comitative relations. In Typological studies in participation, Christian Lehmann (ed.), 9–104. Berlin: Akademie Verlag. Lehmann, Christian, Yong-Min Shin, and Elisabeth Verhoeven 2000 Direkte und indirekte Partizipation. Zur Typologie der sprachlichen Repräsentation konzeptueller Relationen. München: LINCOM. Leiss, Elisabeth 1997 Synkretismus und Natürlichkeit. Folia Linguistica 31 (2): 133–160. Levinson, Steven C. 1991 Relativity in spatial conception and description. Nijmegen: MaxPlanck-Institute for Psycholinguistics. Levy, Paulette 1990 Totonaco de Papantla, Veracruz. Mexico City: Colegio de México. 1992 Body-part prefixes in Papantla Totonac. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 45 (6): 530–542. Lewis, Henry, and Holger Pedersen 1961 A concise comparative Celtic grammar. Göttingen: Vandehoeck & Rupprecht. Lötzsch, Ronald 1978 Zur Frage des sog. Instrumentals im Lettischen. Zeitschrift für Slavistik 23: 667–671. 1996 Interferenzbedingte grammatische Konvergenzen und Divergenzen zwischen Sorbisch und Jiddisch. Sprachtypologie und Universalienforschung 49 (1): 50–59. Loogman, Alfons 1965 Swahili grammar and syntax. Louvain: Nauwelaerts. Lorenz, Manfred 1982 Lehrbuch des Pashto (Afghanisch). Leipzig: Enzyklopädie. Lüdtke, Helmut 1980 Auf dem Weg zu einer Theorie des Sprachwandels. In Kommunikationstheoretische Grundlagen des Sprachwandels, Helmut Lüdtke (ed.), 182–252. Berlin: de Gruyter. Luna Alonso, Ana 2000 Contrastes estilísticos: le petit Prince en lingua galega. In La lingüística francesa en España. Camino del siglo XXI, II. Jesús Lago Garabatos in memoriam, Maria Luz Casal Silva et al. (eds.), 647– 653. Arrecife. Luraghi, Silvia 2001a Syncretism and the classification of semantic roles. Sprachtypologie und Universalienforschung 54 (1): 35–51. 2001b Some remarks on Instrument, Comitative, and Agent in Indo-European. Sprachtypologie und Universalienforschung 54 (4): 385–401.

References 523 Maas, Utz 1999

Einführung in das marokkanische Arabische. Zur Struktur einer Sprache im Werden (“Marokkanisch”). Osnabrück: University of Osnabrück. Mahootian, Shahrzad 1997 Persian. London: Routledge. Martínez, Rafael 2001 The cognitive base of case syncretism: conceptual blending of instrumental, comitative, and agent in ancient Greek. Sprachtypologie und Universalienforschung 54 (4): 329–345. Maslova, Elena 2000a A dynamic approach to the verification of distributional universals. Linguistic Typology 4 (3): 307–333. 2000b Stochastic models in typology: obstacle or prerequisite? Linguistic Typology 4 (3): 357–364. 2003 A case for implicational universals. Linguistic Typology 7 (1): 101–107. Mátai, Mária D. 2002 Kleine ungarische Sprachgeschichte. Hamburg: Buske. Mattissen, Johanna 2003 Dependent-head synthesis in Nivkh. A contribution to a typology of polysynthesis. Amsterdam: John Benjamins. Mayerthaler, Willi 1987 System-independent morphological naturalness. In Leitmotifs in Natural Morphology, Wolfgang U. Dressler (ed.), 25–58. Amsterdam: John Benjamins. Mayerthaler, Willi, Günther Fliedl, and Christian Winkler 1993 Infinitivprominenz in europäischen Sprachen. Teil 1: Die Romania (samt Baskisch). (Tübinger Beiträge zur Linguistik 390.) Tübingen: Narr. MacGregor, Ronald Stuart 1977 Outline of Hindi grammar. Dehli: Oxford University Press. McKaughan, Howard (ed.) 1973 The languages of the Eastern family of the East New Guinea Highland Stock. Seattle: University of Washington Press. McKaughan, Howard, and Doreen Marks 1973 Notes on Auyana phonology and morphology. In The languages of the Eastern family of the East New Guinea Highland Stock, Howard McKaughan (ed.), 181–189. Seattle: University of Washington Press. McNally, Louise 1993 Comitative coordination: a case study in group formation. Natural Language and Linguistic Theory 11: 347–379. Meibauer, Jörg 2003 Phrasenkomposita zwischen Wortsyntax und Lexikon. Zeitschrift für Sprachwissenschaft 22 (2): 153–188.

524 References Mel’þuk, Igor 2001 Towards a formal concept ‘zero linguistic sign’: applications in typology. In Morphology 2000. Selected papers from the 9th Morphology Meeting, Vienna 24-28 February 2000, Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer, and Marija D. Voeikova (eds.), 241–258. Amsterdam: John Benjamins. Meléndez Lozano, Miguel Ángel 1989 El nominal en Achagua. In Lenguas aborigines de Colombia. Descripciones 4: Orinoquia, Jon Landaburru (ed.), 5–68. Bogotá: CCELA. Menges, Karl H. 1991 Der Instrumentalis und einige seiner Funktionen (im Türkischen, anderen altaischen und verwandten Sprachgebieten). Turcica 21/23: 321–342. Michaelis, Susanne, and Marcel Rosalie 2000 Polysémie et cartes sémantiques: le relateur (av)ek en créole seychellois. Etudes Créoles 23 (2): 79–100. Miller, J. 1985 Semantics and syntax. Cambridge: Cambridge University Press. Mithun, Marianne 1988 System-defining structural properties in polysynthetic languages. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41 (4): 442–452. Möhlig, Wilhelm J. G., and Bernd Heine 1999 Swahili Grundkurs. Köln: Köppe. Monod-Becquelin, Aurore 1975 La pratique linguistique des indiens trumai. Paris: SELAF. Montan, Per, and Håkan Rosenqvist 1982 Prepositionsboken. Malmö: Skriptor. Moritz, Karl Philipp 1782 Deutsche Sprachlehre für die Damen. Berlin: Wever. Morris-Jones, John 1955 A Welsh grammar. Historical and comparative. Oxford: Clarendon. Mosel, Ulrike 1994 Saliba. München: LINCOM Europa. Mosel, Ulrike, and Even Hovdhaugen 1992 Samoan reference grammar. Oslo: Scandinavian University Press. Moser, Manfred (megƫjun minn Christopher Meilak) (1999) Dizzjunarju Malti-ƤermaniĪ/ƤermaniĪ-Malti. San Ƥwann: PEG. Moser, Mary Beck 1996 Seri de Sonora. México: El Colegio de México. Moyse-Faurie, Claire 1983 Le drehu. Langue de Lifou (Iles Loyauté). Paris: SELAF.

References 525 Müller, Nicole 1990 Zur altirischen Präposition la. In Deutsche, Kelten und Iren. 150 Jahre deutsche Keltologie. Gearóid Mac Eoin zum 60. Geburtstag gewidmet, Hildegard L. C. Tristram (ed.), 115–123. Hamburg: Buske. Nau, Nicole 1995 Möglichkeiten und Mechanismen kontaktbewegten Sprachwandels unter besonderer Berücksichtigung des Finnischen. München: LINCOM Europa. 1998 Latvian. München: LINCOM Europa. Neukom, Lukas, and Manideepa Patnaik 2003 A grammar of Oriya. Zürich: ASAS. Nickel, Klaus Peter 1990 Samisk grammatikk. Oslo: Universitetsforlaget. Nilsen, Don L. F. 1973 The instrumental case in English. Syntactic and semantic considerations. The Hague: Mouton. Nishina, Yoko 2001 Comitative and instrumental in Japanese. Sprachtypologie und Universalienforschung 54 (4): 346–364. NƯtiƼa, Daina 1978 PrievƗrdu sistƝma latviešu rakstu valodƗ. RƯga: Liesma. Noreen, Alfred 1923 Einführung in die wissenschaftliche Betrachtung der Sprache. Beiträge zur Methode und Terminologie der Grammatik. Halle a.d.S.: Niemeyer. Obenland, Roland 1967 Die Präposition MIT in den Mundarten Württembergs und angrenzender Gebiete. Untersucht an den Tonbandaufnahmen der Tübinger Außenstelle des Deutschen Spracharchivs; samt einem Abriß aller in diesem Material vorkommenden Präpositionen. Freiburg im Breisgau. Obst, Ulrich 2002 Kritisches zu den Begriffen ‘komitativ’/‘Komitativität’ und ‘ɫɨɜɦɟɫɬɧɵɣ’/‘ɫɨɜɦɟɫɬɧɨɫɬɶ’ in der Russistik. Die Welt der Slaven 47 (2): 85–106. Olpp, Johann 1963 Einführung in die Namasprache. Wupperthal: s.l. (3rd edition of the original published in 1909). Ó Siadhail, Míchéal 1985 Lehrbuch der irischen Sprache. Hamburg: Buske. Osumi, Midori 1995 Tinrin grammar. Honolulu: University of Hawaii Press.

526 References Palancar, Enrique L. 1999 Instrumental prefixes in Amerindian languages: an overview to their meanings, origin and functions. Sprachtypologie und Universalienforschung 52 (2): 151–166. 2001 Emergent markers: the relation of instruments, causes, and agents in grammatical constructions. Sprachtypologie und Universalienforschung 54 (4): 365–384. 2002 The origin of agent markers. Berlin: Akademie Verlag. Palmer, Frank Robert 1994 Grammatical roles and relations. Cambridge: Cambridge University Press. Paul, Waltraud 1982 Die Koverben im Chinesischen. Köln: Institut für Sprachwissenschaft. PauliƼš, Osvalds, JƗnis Rozenbergs, and OtomƗrs VilƗns 1978 Latviešu valodas mƗcƯbas pamatkurss. RƯga: Zvaigzne. Perkins, Revere D. 1989 Statistical techniques for determining language sample size. Studies in Language 13: 293–315. 2000 The view from hologeistic linguistics. Linguistic Typology 4 (3): 350–353. Pinkster, Harm 1988 Lateinische Syntax und Semantik. Tübingen: Francke. Plank, Frans 2003 There’s more than one way to make sense of one-way implications – and sense they need to made of. Linguistic Typology 7 (1): 128–140. Ponelis, Friedrich Albert 1978 Afrikaanse sintaksis. Pretoria: J.L. van Schaik. Press, Ian 1986 A grammar of modern Breton. Berlin: Mouton de Gruyter. Rasoloson, Janie Noëlle 1997 Lehrbuch der madagassischen Sprache. Mit Übungen und Lösungen. Hamburg: Buske. Rauh, Gisa 1988 Tiefenkasus, thematische Relationen und Thetarollen: die Entwicklung einer Theorie von semantischen Relationen. Tübingen: Narr. Refsing, Kirsten 1986 The Ainu language. The morphology and syntax of the Shizunai dialect. Aarhus: University Press. Rijkhoff, Jan, and Dik Bakker 1998 Language sampling. Linguistic Typology 2: 263–314.

References 527 Robinson, Charles Henry 1959 Hausa grammar. With exercises, readings, and vocabularies, and specimens of Hausa script. London: Routledge. Ruhlen, Meritt 1987 A guide to the world’s languages. London: Arnold. Saeed, John Ibrahim 1987 Somali reference grammar. Wheaton (Maryland): Dunwoody. Saint-Exupéry, Antoine de 1946 Le petit prince. Paris: Gallimard. Sakayan, Dora 2000 Modern Western Armenian for the English-speaking world. A contrastive approach. Montreal: Arod Books. Salas, Adalberto 1992 El mapuche o araucano. Madrid: MAPFRE. Saltarelli, Mario 1989 Basque. London: Routledge. Sammallahti, Pekka 1998 The Saami languages. An introduction. Karasjok: Davvi Girji. Sandra, Dominiek 1998 What linguists can and can’t tell you about the human mind: a reply to Croft. Cognitive Linguistics 9 (2): 361–378. Sapir, Edward 1922 The Takelma language of South-Western Oregon. In Handbook of American Indian languages. Part 2, Franz Boas (ed.), 1–295. New York: Smithsonian Institution. Sasse, Hans-Jürgen 1982 Subjektprominenz. In Fakten und Theorien. Beiträge zur romanischen und allgemeinen Sprachwissenschaft. Festschrift für Helmut Stimm zum 65. Geburtstag, Sieglinde Heinz, and Ulrich Wandruszka (eds.), 267–286. Tübingen: Narr. 1988 Der irokesische Sprachtyp. Zeitschrift für Sprachwissenschaft 7: 173–213. Schleicher, August 1861 Die beiden instrumentale des indogermanischen. Kuhn’sche Zeitschrift 2: 454–459. Schlesinger, Izchak M. 1979 Cognitive structures and semantic deep structures: the case of the instrumental. Journal of Linguistics 15: 330–395. 1989 Instruments as agents: on the nature of semantic relations. Journal of Linguistics 25: 189–210. 1995 Cognitive space and linguistic case. Semantic and syntactic categories in English. Cambridge: Cambridge University Press.

528 References Schmalstieg, William R. 1988 A Lithuanian historical syntax. Columbus, Ohio: Slavica Publ. Schmitt, Rüdiger 1981 Grammatik des Klassisch-Armenischen mit sprachvergleichenden Erläuterungen. Innsbruck: Institut für Sprachwissenschaft. Schröder, Jochen 1989 Mit-Phrasen im Deutschen und ihre Entsprechungen im Polnischen. Versuch einer ebenenspezifischen Analyse. In Theorie und Praxis der deutsch-polnischen Konfrontation und Translation, Andrzej Kątny (ed.), 27–40. Rzeszów: Wyd. WyĪszej Szkoáy Pedagogicznej. Schultze-Berndt, Eva 2000 Simple and complex verbs in Jaminjung. A study of event categoisation in an Australian language. Nijmegen: Max-Planck-Institute for Psycholinguistics. Schultze-Berndt, Eva, and Nikolaus Himmelmann 2004 Depictive secondary predicates in crosslinguistic perspective. Linguistic Typology 8 (1): 59–132. Schwarz, Oxana, Cornelia Stroh, and Aina Urdze 2001 BIB-KOM-TYP. Die Bibliographie zur Komitativ-Typologie. München: LINCOM Europa. Schweiger, Fritz 2000 Few languages in a short time interval? Linguistic Typology 4 (3): 353–357. Seiler, Hansjakob 1974a The principle of concomitance: instrumental, comitative and collective (with special reference to German). In Linguistic Workshop II, Hansjakob Seiler (ed.), 2–55. München: Fink. [= Seiler 1974c] 1974b The principle of concomitance in Uto-Aztecan. In Linguistic Workshop II, Hansjakob Seiler (ed.), 56–186. München: Fink. 1974c The principle of concomitance: instrumental, comitative and collective (with special reference to German). Foundations of Language 12: 215–247. [= Seiler 1974a] 1988 The dimension of participation. Guadalajara: Universidad de Guadalajara. 1993 Der UNITYP-Ansatz zur Universalienforschung und Typologie. Sprachtypologie und Universalienforschung 46 (3): 163–186. 2000 Language Universals Research: a synthesis. Tübingen: Narr. 2004 Über das Verhältnis von Sprachuniversalienforschung und Sprachtypologie. Rückblick und Ausblick. In Dimensionen und Kontinua. Beiträge zu Hansjakob Seilers Universalienforschung, Waldfried Premper (ed.), 1–16. Bochum: Brockmeyer.

References 529 Seiler, Hansjakob, and Waldfried Premper (eds.) 1991 Partizipation. Das sprachliche Erfassen von Sachverhalten. Tübingen: Narr. Serzisko, Fritz 1991 Orientierung. In Partizipation. Das sprachliche Erfassen von Sachverhalten, Seiler, Hansjakob, and Waldfried Premper (eds.), 273– 308. Tübingen: Narr. Sharma, Aryendra, and Hans J. Vermeer 1972 Einführung in die Hindi-Grammatik. Heidelberg: Julius Groos. Shopen, Timothy (ed.) 1985 Language typology and syntactic description. 3 vols. Cambridge: Cambridge University Press. Shou-Hsin, Teng 1970 Comitative versus phrasal conjunction. Papers in Linguistics 2: 315–358. Shroyer, Thomas G. 1969 An investigation of the semantics of English as a proposed basis for language curriculum materials. Unpublished PhD-Thesis. Ohio State University. Shukla, Shaligram 2001 Hindi morphology. München: LINCOM Europa. Siewierska, Anna 1984 The passive: a comparative linguistic analysis. London: Croom Helm. Skorik, Petr Jakovlevic 1961 Grammatika þukotskogo jazyka. Vol. I. Leningrad: Izdat. “Nauka”. Slobin, Dan Isaac (ed.) 1985 Crosslinguistic study of language acquisition. Volume 1: The Data. Hillsdale/N.J.: Erlbaum. Sneddon, James Neil 1996 Indonesian. A comprehensive grammar. London: Routledge. Stachowski, Marek 1994 Der Instrumental im Jakutischen und Dolganischen. Journal of Turkology 2 (2): 245–258. 1995 Der Komitativ im Jakutischen und Dolganischen. In Kuryáowicz Memorial Volume, Wojciech SmoczyĔski (ed.), 553–559. Kraków: Universitas. Stassen, Leon 2000 AND-languages and WITH-languages. Linguistic Typology 4 (1): 1– 54. Steuerwald, Karl, and Cemal Köprülü 1986 Langenscheidts Taschenwörterbuch der türkischen und deutschen Sprache. Teil II: Deutsch-Türkisch. München: Langenscheidt.

530 References Stoebke, Renate 1968 Die Verhältniswörter in den ostseefinnischen Sprachen. The Hague: Mouton. Stolz, Christel, and Thomas Stolz 1996a Funktionswortentlehnung in Mesoamerika. Spanisch-amerindischer Sprachkontakt (Hispanoindiana II). Sprachtypologie und Universalienforschung 49 (1): 86–123. 1996b ‘Transpazifische Entlehnungsisoglossen’. Hispanismen in Funktionswortinventaren beiderseits der Datumsgrenze. In Areale, Kontakte, Dialekte. Sprache und ihre Dynamik in mehrsprachigen Situationen, Norbert Boretzky, Werner Enninger, and Thomas Stolz (eds.), 262–291. (Bochum-Essener Beiträge zur Sprachwissenschaft 24.) Bochum: Brockmeyer. 1997 Universelle Hispanismen? Von Manila über Lima bis Mexiko und zurück: Muster bei der Entlehnung spanischer Funktionswörter in die indigenen Sprachen Amerikas und Austronesiens. Orbis 39: 1–77. Stolz, Thomas 1984 Para-Präpositionen, Präpositionen, Adverbien und (Pseudo-)Postpositionen im Lettischen. Indogermanische Forschungen 89: 251–269. 1991 Von der Grammatikalisierbarkeit des Körpers. I. Vorbereitung. Essen: Universität GH Essen. 1992a Von der Grammatikalisierbarkeit des Körpers. II. Einleitung. 1: Kritik der “Grammatik mit Augen und Ohren, Händen und Füßen”. Essen: Universität GH Essen. 1992b Lokalkasussysteme. Aspekte einer strukturellen Dynamik. Wilhelmsfeld: Egert. 1993 Über Komitative. Essen: Universität GH Essen. 1994 Grammatikalisierung und Metaphorisierung. Bochum: Brockmeyer. 1996 Komitativ-Typologie. MIT- und OHNE-Relationen im crosslinguistischen Überblick. Papiere zur Linguistik 54 (1): 3–65. 1997a Some instruments are really good companions – some are not. On syncretism and the typology of instrumentals and comitatives. Theoretical Linguistics 23 (1/2): 113–200. 1997b Two comitatives or more. On the degree of involvement of participants. In Semiotische Prozesse und natürliche Sprache. Festschrift für Udo L. Figge zum 60. Geburtstag, Andreas Gather, and Heinz Werner (eds.), 515–530. Stuttgart: Steiner. 1997c Companions, instruments and their absence: markedness relations. In Languages and lives: Essays in honor of Werner Enninger, James R. Dow, and Michèle Wolff (eds.), 189–204. New York: Peter Lang. 1997d Review of: Campe, Petra (1994): Case, semantic roles, and grammatical relations. A comprehensive bibliography. Amsterdam: Benjamins. Linguistics 35 (11): 183–189.

References 531 1997e

Review of: Givón, Talmy (1995): Functionalism and grammar. Amsterdam: Benjamins. Linguistics 35 (3): 422–427. 1998a UND, MIT und/oder UND/MIT? – Koordination, Instrumental und Komitativ – kymrisch, typologisch, universell. Sprachtypologie und Universalienforschung 51 (2): 107–130. 1998b Komitative sind mehr als Instrumentale sind mehr als Komitative. Merkmalhaftigkeit und Markiertheit in der Typologie der MITRelationen. In Sammelband des II. internationalen Symposions zur Natürlichkeitstheorie, Karmen Teržan-Kopecky (ed.), 83–100. Maribor: Univ. v Mariboru, Pedagoška Fakulteta. 2001a To be with X is to have X: Comitatives, instrumentals, locative and predicative possession. Linguistics 39 (2): 321–350. 2001b Comitatives vs. instrumentals vs. agents. In Aspects of typology and universals, Walter Bisang (ed.), 153–174. Berlin: Akademie Verlag. 2001c On circum-Baltic instrumentals and comitatives. To and fro coherence. In Circum-Baltic languages. Volume 2: Grammar and Typology, Östen Dahl, and Maria Koptjevskaja-Tamm (eds.), 591–612. Amsterdam: Benjamins. 2003a Comitativo e strumentale nelle lingue germaniche e romanze In Introduzione alla linguistica cognitiva, Livio Gaeta, and Silvia Luraghi (eds.), 213–231. Roma: Carocci. 2003b Possessions in the Far North. A glimpse of the alienability correlation in modern Icelandic. In Dimensionen und Kontinua: Beiträge zu Hansjakob Seilers Universalienforschung, Waldfried Premper (ed.), 73–96. Bochum: Brockmeyer. Stolz, Thomas, and Traude Gugeler 2000 Comitative typology. Nothing about the ape, but something about king-size samples, the European community, and the Little Prince. Sprachtypologie und Universalienforschung 54 (1): 53–61. Stolz, Thomas, Sonja Kettler, and Aina Urdze forthcoming European possessions. The alienability correlation in areal perspective. Stolz, Thomas, and Cornelia Stroh 2001 Wenn Komitative Instrumentale sind und umgekehrt. In Variierende Markierung von Nominalgruppen in Sprachen unterschiedlichen Typs, Winfried Boeder, and Gerd Hentschel (eds.), 387–411. Oldenburg: BIS. Stolz, Thomas, Cornelia Stroh, and Aina Urdze 2003 Solidaritäten. Lingua Posnaniensis 45: 69–92. 2005 Comitatives and Instrumentals. In The world atlas of language structures, Martin Haspelmath, Matthews S. Dryer, David Gil, and Bernard Comrie (eds.), 214–217. Oxford: Oxford University Press.

532 References forthcoming in preparation

With(out). On the markedness relation between Comitatives/ Instrumentals and Abessives. Word. Similarity and dissimilarity of European languages. A quantitative account of the comitative in areal perspective. Berlin: Mouton de Gruyter.

Stroh, Cornelia 1998 Die Geschichte der MIT-Relationen im Französischen: KomitativInstrumental-Synkretismus mit Hindernissen. Sprachtypologie und Universalienforschung 51 (2): 131–156. 1999 MIT-Relationen in der Romania: ein klarer Fall? Sprachtypologie und Universalienforschung 52 (2): 183–195. 2004 Aspekte der Arealität bei der Typologie des Komitativs/Instrumentals in den europäischen Sprachen. In Towards a dynamic theory of language, Andrea Graumann, Peter Holz, and Martina Plümacher (eds.), 257–278. Bochum: Brockmeyer. Sulkala, Helena, and Karjalainen Merja 1992 Finnish. London: Routledge. Sutton, Peter 1976 The having affix and other morphemes in fifty Australian languages. In Grammatical categories in Australian languages, Robert Malcolm Ward Dixon (ed.), 297–305. New Jersey: Humanities Press. Svane, Gunnar Olaf 1958 Grammatik der slowenischen Schriftsprache. Copenhagen: Rosenkilde & Bagger. Szabó, László 1984 The functions of the essive, comitative and abessive in KolaLappish. Nordlyd 9: 156–171. Talmy, Leonard 2000 Toward a cognitive semantics. Volume 2: Typology and process in concept structuring. Cambridge (Mass.): MIT-Press. Tauli, Valter 1983 Standard Estonian grammar. Part II: Syntax. Uppsala: University of Uppsala. Toivainen, Jorma 1980 Inflexional affixes used by Finnish-speaking children aged 1–3 years. Helsinki: Suomalaisen Kirjallisuuden Seura. Tompa, József 1972 Kleine ungarische Grammatik. Leipzig: Enzyklopädie. Traugott, Elizabeth Closs, and Bernd Heine 1991 Introduction. In Approaches to grammaticalization. Vol. 1: Focus on theoretical and methodological issues, Elizabeth Closs Traugott, and Bernd Heine (eds.), 1–14. Amsterdam/Philadelphia: John Benjamins.

References 533 Trudgill, Peter 2004 The impact of language contact and social structure on linguistic structure: Focus on the dialects of modern Greek. In Dialectology meets typology. Dialect grammar from a cross-linguistic perspective, Bernd Kortmann (ed.), 435–452. Berlin: Mouton de Gruyter. Trunte, Hartmut 1990 Ein praktisches Lehrbuch des Kirchenslavischen. Band 1: Altkirchenslavisch. München: Sagner. Ulvydas, K. (ed.) 1965 Lietuviø kalbos gramatika. I tomas. Vilnius: Mintis. 1971 Lietuviø kalbos gramatika. II tomas. Vilnius: Mintis. Valge, Jüri 1988 Komitatiiviga koos? Koos komitatiiviga? Komitatiiviga? In Lähivertailuja 3: Soome-eesti keelevigade analüüsi seminar Hailuotol 15.–16. Mail 1987, Peep Nemvalts (ed.), 34–41. Turku: Turun Yliopisto. Van den Berg, Helma 1995 A grammar of Hunzib. München: LINCOM Europa. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2 (1): 79–123. van Driem, George 1993 A synopsis of role and reference grammar. In Advances in role and reference grammar, Robert D. Van Valin (ed.), 1–164. Amsterdam: John Benjamins. Van Valin, Robert D. (ed.) 1993 Advances in role and reference grammar. Amsterdam: John Benjamins. Van Valin, Robert D., and David P. Wilkins 1996 The case for “effector”: case roles, agents and agency revisited. In Grammatical constructions: their form and meaning, Masayoshi Shibatani, and Sandra A. Thompson (eds.), 289–322. Oxford: Clarendon. Vaquero, P. Antonio 1965 Idioma Warao. Caracas: Editorial Sucre. Veenstra, Tonjes 2002 Les verbes sériels en saramaka: inventaire. Amerindia 26/27: 319–344. Veerman-Leichsenring, Annette 1992 Body-part terms occuring in Popolocan verbs. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 45 (6): 562–569. 2000 Gramática del chocho de Santa catarina Ocotlán, Oaxaca. Leiden: Research School CNWS. Vella, Joseph 1994 Learn Maltese – why not? Valletta: Valletta Publishing.

534 References Vennemann, Theo 1974 Topics, subjects and word order: from SXV to SVX via TVX. In Historical linguistics: Syntax, morphology, internal and comparative reconstruction, John M. Anderson (ed.), 339–376. Amsterdam: North-Holland. Vietze, Hans-Peter 1988 Lehrbuch der mongolischen Sprache. Leipzig: Enzyklopädie. Vincent, Alex 1973 Notes on Tairora noun morphology. In The languages of the Eastern family of the East New Guinea Highland Stock, Howard McKaughan (ed.), 530–546. Seattle: University of Washington Press. Voigtlander, Katherine, and Artemisa Echegoyen 1985 Luces contemporáneas del otomí. Gramática del otomí de la Sierra. México: El Colegio de México. Wälchli, Bernhard 2003 Co-compounds and natural coordination. University of Stockholm: Department of Linguistics. Walmsley, John B. 1971 The English comitative case and the concept of deep structure. Foundations of Language 7: 493–507. Werner, Heinz 1991 Wie komplex ist das Französische? Linguistische Berichte 132: 133–156. Whorf, Benjamin L. 1988 Grammatical categories. In Language, thought and reality. Selected writings of Benjamin Lee Whorf, John B. Carroll (ed.), Cambridge/Mass.: MIT Press. Wierzbicka, Anna 1980 The case for surface case. Ann Arbor: Karoma. Wilkins, David 1989 Mparntwe Arrernte (Aranda): Studies in the structure and semantics of grammar. Canberra: The Australian National University [PhD-Thesis]. Wolff, H. Ekkehard 1993 Referenzgrammatik des Hausa. Münster: LIT. Worth, Dean 1963 Paleosiberian. In Current trends in linguistics. Vol. 1: Soviet and East European linguistics, Thomas A. Sebeok (ed.), 345–373. The Hague: Mouton. Wurzel, Wolfgang U. 1984 Flexionsmorphologie und Natürlichkeit. Ein Beitrag zur morphologischen Theoriebildung. Berlin: Akademie-Verlag. 1987 System-dependent morphological naturalness in inflection In Leitmotifs in Natural Morphology, Wolfgang U. Dressler, (ed.), 59–96. Amsterdam: John Benjamins.

References 535 Zelinsky-Wibbelt, Cornelia (ed.) 1993 The semantics of prepositions. From mental processing to natural language processing. Berlin: Mouton de Gruyter. Zewen, François-Xavier Nicolas 1977 The Marshallese language. A study of its phonology, morphology, and syntax. Berlin: Dietrich Reimer. Zigmond, Maurice L., Curtis G. Booth, and Pamela Munro 1990 Kawaiisu. A grammar and dictionary with texts. Berkeley: University of California Press.

Index of languages

Achagua 188, 397, 422, 524 Afrikaans 397, 422, 476 Ainu (Japan) 178, 179, 397, 422, 500, 526 Akateko 177, 422 Alamblak 177, 397, 422 Albanian (Gheg) 198, 201, 204, 205, 210, 218, 219, 233, 239–242, 252, 280, 281, 422, 493, 506 Albanian (Tosk) 198, 201, 205, 210, 219, 233, 239, 240, 242, 248, 280, 281, 422, 489, 493 Aleut 101, 188, 397, 422, 427, 439, 465, 510 Alsatian 198, 205, 211, 216, 218, 222, 236, 239, 242, 422 Alyawarr 171, 397, 422 Aragonese 198, 201, 205, 211, 219, 220, 228, 236, 238, 239, 242, 422, 491 Aramaic (Old) 398, 470 Armenian 53–55, 187, 198, 210, 205, 207, 210, 216, 218, 229, 230, 232, 234–236, 239, 240, 241, 242, 248, 257, 261, 265, 267, 398, 422, 471, 482, 489, 500, 520, 527 Arrarnta (Western) 38–41, 147, 423, 452, 456 Asturian 198, 201, 206, 211, 214, 219, 220, 222, 227, 238, 242, 423 Awiyaana 58, 82, 423 Azerbaijani (North) 198, 201, 204, 205, 209, 211, 217, 219, 234, 235, 239, 242, 257, 262, 265, 267, 423, 471, 489

Badiotto198, 201, 205, 206, 211, 215, 219, 239, 242, 246, 398, 423 Baka 8–10, 12, 398, 423, 500, 519 Bamanankan 150–152, 188, 423, 475, 479, 480 Basque 168, 198, 201, 204, 205, 207–210, 214, 218, 221, 224, 236, 237, 239–242, 249, 253, 257, 262, 265, 267, 272, 352, 353, 384–386, 399, 423, 491, 527 Belarusan 198, 201, 205, 207, 210, 213, 218, 229, 234, 239, 240, 242, 264, 276, 280, 399, 423, 506 Bezhta 143, 423, 478, 519 Binumarien 399, 423, 459 Blackfoot 171, 399, 424 Boro 171, 424 Breton 36, 198, 201, 205, 209, 211, 216, 217, 219, 222, 232, 234, 238, 242, 272, 277, 369, 399, 424, 490, 506, 526 Bulgarian 198, 201, 205, 206, 209, 211, 213, 218, 229, 233, 234, 236, 238, 239, 242, 247, 248, 256, 260, 279, 354, 381, 382, 386, 399, 424 Burushaski 177, 399, 424, 478, 509 Catalan 198, 201, 205, 211, 215, 219, 238, 242, 360, 400, 424 Chemehuevi 171, 400, 424 Chinese (Mandarin) 73, 99, 400, 424, 427, 447, 454, 500 Chochotec 371, 372, 375, 376, 400, 424

Index of languages 537 Chukot 59, 60, 71, 171, 187, 424, 428 Comanche 171, 400, 424 Corsican 198, 201, 206, 212, 219, 238, 241, 242, 251, 424 Croatian 198, 201, 205, 206, 207, 211, 213, 218, 229, 239, 242, 247, 415, 425, 503 Czech 198, 201, 205, 207, 208, 211, 213, 218, 222, 226, 229, 233, 238, 239, 242, 243, 401, 425, 436, 468, 518 Dani (Lower Grand Valley) 63, 67, 401, 425, 462, 463, 511 Danish 198, 201, 205, 211, 218, 225, 226, 239, 242, 383, 402, 424, 433, 436 Dehu 139, 425, 474, 475 Djamindjung 128, 129, 144, 425 Dolgan 425, 468 Dumi 401, 425, 473 Dutch 198, 201, 205, 206, 211, 218, 225, 236, 238, 242, 250, 351, 352, 401, 425, 450, 506 English 5, 21, 22, 62, 63, 85, 86, 88, 103, 198, 201, 203, 205, 206, 211, 217, 219, 222, 225, 226, 238, 242, 245, 247, 263, 264, 268, 273–277, 280, 282, 322, 350, 351, 358, 401, 425, 435, 437, 448, 466, 468, 469, 479, 484, 485, 489, 492, 518, 525, 527, 529, 534 Estonian 30, 49, 89, 198, 201, 204, 205, 211, 216, 218, 236, 239, 241, 242, 262, 265, 267, 350, 351, 359, 360, 364–368, 383, 386, 402, 425, 453, 500, 516, 521, 532 Even 56, 402, 425, 459

Faroese 198, 201, 204–206, 210, 215, 216, 218, 221, 222, 225, 239, 240, 242, 243, 274, 275, 322, 353, 402, 426, 506 Farsi 125, 126, 413, 426, 474 Finnish 61–63, 67, 83, 141, 142, 153, 155, 198, 201, 205, 207, 210, 216, 218, 221, 222, 235, 236, 239–242, 248, 254, 257, 266, 275–278, 350, 351, 365, 368, 383–386, 402, 426, 461– 464, 466, 473, 500, 503, 504, 506, 532 Fon 132, 402, 426, 518, 521 French 64, 67, 99, 198, 200, 201, 204–206, 208, 210, 212, 214– 217, 219, 229, 234, 238–242, 249, 254, 360, 384, 385, 402, 426, 463, 464, 489 Frisian (Western) 198, 206, 211, 217, 219, 225, 226, 236, 238, 242, 402, 426 Friulian 198, 205, 206, 211, 215, 219, 238, 242, 251, 426, 489 Gaelic (Scottish) 415, 426, 490 Galician 198, 199, 201, 205, 206, 211, 214, 219, 228, 236, 238, 242, 273, 274, 426, 434, 506 Gascon 198, 206, 212, 219, 238, 242, 360, 384, 385, 426 Georgian 198, 201, 206–208, 211, 214, 218, 221, 224, 236, 239– 242, 249, 253, 254, 262, 265, 266, 268, 272, 275, 403, 426, 471, 492 German (Standard) 1–2, 5–10, 12, 18–25, 41, 48, 49, 69, 80, 88, 169, 170, 198, 202, 204–206, 211, 216, 218, 225, 227–229, 233, 234, 236, 238, 239, 242, 272, 322, 351–352, 364, 365, 367, 379, 381–383, 403, 422,

538 Index of languages 427, 430, 431, 439, 447–451, 455, 458, 463, 484, 485, 489, 495 German (Low German) 216, 383, 450 German (Old High German) 378, 379, 382, 384 German (Middle High German) 379, 382 German (Middle Low German) 383 Gilyak 427, 488 Gola 177, 403, 427 Greek 198, 202, 205, 209, 210, 218, 227, 230, 232–234, 236, 239, 242, 248, 378, 379, 381, 382, 386, 403, 427, 468, 489, 533 Greek (Byzantine Greek) 236 Greek (Middle Greek) 381 Greek (Ancient/Old) 501, 523 Greenlandic (see West Greenlandic) Guaraní (Paraguayan) 50, 55, 141, 142, 372, 403, 427, 458, 515 Hausa 150, 404, 427, 449, 479, 527, 534 Hindi 113, 114, 126, 141, 142, 145, 404, 427, 472, 509, 523, 529 Hixkaryána 167, 168, 427, 484, 487 Hokkien 427, 454 Hungarian 9–12, 102, 103, 169, 198, 202, 205, 207, 208, 210, 214, 218, 234–236, 239, 242, 249, 257, 262, 265–267, 360, 404, 427, 457, 489, 492, 502, 518, 519 Hunzib 404, 428, 478, 533 Icelandic 153–155, 198, 202, 205– 208, 211, 213, 214, 216, 219, 222, 225, 226, 240, 242, 243, 245, 250, 254, 259, 263, 266, 267, 272, 277, 282, 322–349,

353, 355, 378, 405, 428, 498, 499, 508, 509, 531 Indonesian 29, 405, 428, 452–454, 520, 529 Irish 37, 154, 155, 368, 405, 428, 456, 489, 490 Irish (Old Irish) 378, 468 Italian 198, 202, 205, 206, 212, 215, 219, 228, 236, 238, 241, 242, 382, 405, 428, 489, 491, 503 Itelmen 71, 171, 405, 428, 482 Japanese 70, 405, 428, 447, 525 Kalderash (see Romani) Kalmyk 52, 55, 153, 428, 459 Kannada 145, 359, 406, 428 Kanuri (Central) 149, 150, 152, 428 Karaim 96, 429, 472 Kawaiisu 82, 83, 429, 535 Kazakh 96, 429, 472 Ket 406, 429 Koasati 171, 407, 429 Kobon 182, 407, 429, 513 Koiarie 407, 429 Korean 407, 429, 447 Korlai Creole Portuguese 429, 454, 512 Krongo 171, 407, 429 Kurdish (Kurmanþi )198, 205, 209, 210, 216, 217, 219, 221, 232– 234, 239–242, 248, 401, 429, 471, 489 Kwami 133–134, 139, 408, 430, 475, 521 Ladin 198, 202, 205, 206, 211, 215, 219, 227, 228, 235, 238, 242, 258, 260, 276, 408, 430, 492 Lakota 73, 74, 430 Languedocien 198, 206, 212, 219, 220, 240, 242, 254, 408, 430, 489

Index of languages 539 Latin 11, 12, 37, 360, 378, 379, 381, 384, 408, 430, 470, 501, 519 Latvian 21, 137, 138, 198, 202, 205, 206, 211, 216, 218, 219, 229, 230, 232, 233, 236, 238, 242, 243, 282, 307–322, 349, 351, 353, 355, 367, 368, 383, 384, 386, 408, 430, 448, 451, 468, 477, 488, 496, 497, 499, 525 Lezgi 408, 430, 478 Limburgisch (North and South) 198, 202, 224, 430 Limburgisch (North) 205, 206, 211, 218, 225, 239, 242 Limburgisch (South) 205, 206, 210, 218, 225, 226, 238, 242 Lithuanian 171, 198, 202, 205–207, 210, 214, 216, 218, 232, 233, 238, 242, 247, 248, 257, 261, 265, 266, 275, 276, 279, 280, 307, 350, 351, 367, 368, 383– 385, 409, 430, 501, 504, 507, 513, 528 Luiseño 57, 58, 409, 431, 460, 520 Luxembourgeois 198, 202, 205, 211, 216, 218, 225, 238, 242, 250, 431 Macedonian 198, 202, 205, 206, 209, 211, 213, 219, 227, 229, 233, 236, 238–242, 247, 248, 256, 260, 279, 354, 381, 382, 386, 431 Malaccan Creole Portuguese 30, 31, 37, 431, 454, 481 Malagasy (Plateau) 142, 431, 468 Malay 409, 428, 431, 454 Malayo 37, 38, 409, 425, 428, 431, 432, 434, 435, 437, 438, 456, 481 Maltese 137, 138, 153, 155, 198, 202, 204, 205, 207, 210, 219, 234, 236, 237, 239, 242, 248,

249, 257, 262, 265, 267, 274– 277, 282–307, 349, 353, 355, 409, 431, 471, 477, 491, 495, 508, 511, 514, 533 Manchu 145, 409, 431 Maninkakan 431, 480 Maori 9–12, 410, 432, 481, 483, 509 Mapudungun 184–186, 188, 432 Mari (Eastern) 70, 410, 432 Marshallese 177, 360, 410, 432, 535 Mixe (Totontepec) 171, 411, 432 Moldavian 198, 202, 205, 209, 210, 215, 218, 227, 228, 239, 242, 246, 264, 279, 432, 492 Mongolian 51, 52, 55, 153, 406, 428, 432, 449, 458 Nahuatl (Classical) 9, 11, 135, 371, 376, 411, 432, 465, 521 Nama 72, 406, 433, 500 Ndyuka-Trio Pidgin 96, 433, 472 Ngalakan 171, 172, 412, 433 Ngaliwuru 144, 433 Norwegian (Bokmål) 198, 202, 205, 211, 219, 222, 225, 226, 239, 240, 242, 412, 433, 468 Oriya 83, 84, 140, 433, 525 Oromo 412, 433, 473, 516 Osetin 433, 478 Otomí 184–187, 371, 372, 374, 412, 487, 502, 517, 534 Palauan 171, 413, 434 Pashto 413, 434, 478, 522 Polish 198, 202, 205, 207, 210, 213, 218, 236, 239, 242, 247, 261, 264, 279, 280, 354, 368, 413, 434, 468, 501, 507, 514 Portuguese 30, 103, 154, 155, 198, 200, 202, 205, 206, 211, 214, 217, 219, 222, 238, 242, 243,

540 Index of languages 277, 414, 426, 429, 431, 434, 454, 468 Provençal 198, 205, 212, 219, 221, 229, 239, 242, 360, 434 Quechua (Ayacuchano) 126, 134, 135, 139, 414, 434, 516 Quechua (Ecuadoriano) 135, 139, 434, 474 Quechua (Imbambura) 414 Quechua (Columbiano) 434 Romani (Kalderash) 113, 114, 428 Romani (Lovari) 198, 206, 219, 234, 489 Romani (Burgudži) 414 Romani (Vlax) 434 Romanian 32, 198, 202, 205, 209, 210, 215, 217, 218, 227, 228, 230, 233, 239, 242, 246, 279, 381, 382, 414, 434, 455, 468, 489, 507 Russian 171, 172, 198, 202, 205, 207, 210, 213, 215, 218, 219, 229, 234, 239, 240, 242, 256, 257, 264, 276, 277, 279, 280, 354, 414, 435, 468, 476, 504, 507, 518 Saami (Akkala) 29, 31, 206, 209, 217, 235, 236, 239–242, 265, 267, 386, 452, 453, 489, 500, 503, 509, 527 Saami (North) 198, 202, 204, 205, 212, 219, 414, 435 Saami (Lule) 383, 503 Saami (Kola) 468 Saliba 57, 414, 435, 460, 524 Samoan 72, 141, 142, 359, 414, 435, 524 Sanskrit 359, 415, 435, 501 Saramaccan 103, 415, 435, 471, 511

Sardinian (Logudorese) 198, 205, 212, 219, 228, 239, 242, 251, 415, 435, 501 Selkup 96, 435, 472 Senoufo (Supyire) 188, 416, 435, 488, 512 Serbian 198, 202, 205–207, 211, 213, 218, 229, 233, 238, 239, 242, 256, 257, 354, 436, 489 Seri 183, 436, 524 Slovak 198, 202, 205, 207, 208, 211, 213, 217, 218, 222, 229, 233, 239, 242, 247, 351, 354, 415, 425, 436 Slovenian 198, 202, 204, 205, 209, 211, 213, 218, 222, 229, 234, 239, 242, 247, 248, 256, 260, 354, 381, 382, 386, 415, 436, 507 So (Congo) 179, 188, 415, 436, 512 Somali 71, 72, 416, 436, 527 Spanish 198, 200, 202, 205, 212, 215, 219, 228, 229, 236, 238, 239, 241, 242, 371, 372, 374– 376, 385, 416, 436, 479, 485, 488, 491, 501, 502, 507 Sursilvan 198, 202, 205, 206, 211, 215, 219, 228, 238, 242, 246, 436 Swahili 127, 135, 136, 139, 141, 142, 148, 149, 416, 436, 474, 477, 522, 524 Swedish 7–10, 12, 198, 202, 205, 210, 215, 216, 219, 222, 225, 236, 239, 242, 352, 383, 384, 416, 425, 433, 436, 437, 462, 493, 504, 507 Tagalog 59, 71, 416, 437, 460, 514 Tairora (North) 56, 57, 416, 423, 437, 459, 534 Takelma 74, 75, 416, 437, 465, 527 Tatar 96, 437, 472, 489

Index of languages 541 Tetun 96, 437, 472 Timucua 171, 417, 437 Tiri 184, 186, 187, 437 Tok Pisin 437, 481 Totonac (Papantla) 71, 371, 372, 374, 418, 437, 465, 522 Triqui (Copala) 182, 418, 437 Trumaí 177, 178, 438 Tukang Besi 96, 438, 472 Turkish 29, 36, 49, 55, 198, 202, 204, 205, 206, 209, 211, 217, 219, 235, 237, 239, 242, 262, 265, 267, 351, 359, 381, 382, 386, 418, 438, 447, 452, 453, 468, 471, 489, 500, 520 Tuvaluan 124, 125, 438, 510 Tzotzil (Chamula) 438, 468, 508 Udi 418, 438, 473, 516 Udmurt 418, 438, 489 Ukrainian 198, 202, 205, 207, 210, 213, 215, 218, 221, 234, 239, 242, 419, 438 Vallader 198, 202, 205, 210, 214, 215, 219, 239, 242, 246, 251, 438 Vod 438, 468

Warao 146, 171, 420, 439, 478, 533 Wayampi 141, 420, 439, 478, 516 Welsh 136, 153, 155, 171, 198, 202, 204, 205, 207, 209, 210, 213, 214, 216, 217, 219, 221, 222, 232–234, 239, 241, 242, 248, 252–253, 257, 261, 265–267, 277–279, 352, 368, 369, 379, 385, 386, 420, 439, 476, 489, 490, 500, 507, 519, 524 West Greenlandic 403, 427, 439, 481, 483, 515 Yakut 420, 439, 468 Yiddish (Eastern) 198, 205, 209, 211, 218, 225, 226, 237, 239, 242, 421, 439, 491 Yidiny 188–190, 439, 447 Yucatec Maya 421, 447, 465, 500, 502 Yupik (Central) 188, 189, 439 Zapotec (Yatzachi) 371–375, 421, 439, 502, 511 Zulu 96, 439, 472

Index of authors

Abaev, Vasilij Ivanoviþ 478 Abdülhayo÷lu, Suphi 440, 452 Abondolo, Daniel 503 Adrados, Francisco R. 379 Aissen, Judith 134, 468 Alhoniemi, Alho 70, 464 Altmann, Gabriel 222 Amaya, María Trillos 37, 456, 481 Anderson, John M. 77, 477 Andrews, Avery D. 346 Anttila, Raimo 164 Aquilina, Joseph 282, 303, 494 Ariste, Paul 468 Azzopardi-Alexander, Marie 282, 304 Bakker, Dik 93 Bánhidi, Zoltán 484 Barðdal, Jóhanna 323, 346, 498, 499 Barnes, Michael 274 Bartens, Hans-Hermann 29, 31, 453 Bauer, Winfried 9, 10, 481, 483 Baxter, Alan N. 30, 31, 454, 455 Bechert, Johannes 87, 243 Behagel, Otto 451 Bender, Ernst 472 Berg, Thomas 3 Berger, Hermann 478 Bergsland, Knut 465, 488 Besnier, Niko 125 Bhat, Darbhe Narayuana Shanhara 488 Biboviü, Ljiljana 457 Bisang, Walter 73, 93 Bitima, Tamene 473 Bittner, Dagmar 466 Bläsing, Uwe 53 Blake, Barry J. 144, 465, 478, 482

Blum, Martin 466 Böðvarsson, Árni 329, 497, 498 Booth, Curtis G.83 Boretzky, Norbert 114 Borg, Albert 283, 305, 496 Bornemann, Eduard 379 Bossong, Georg 2 Braune, Wilhelm 379 Brauner, Siegmund 151, 152, 481 Brincat, Joseph 237, 492 Bromley, H. Myron 63, 64, 463 Brugmann, Karl 379, 453 Buckingham, Hugh W. 77 Bukeviþinjtơ, Elvira 452, 498 Butler H., Inez M. 374–376, 503 Bybee, Joan 86, 181 Byrne, Francis 72, 103 Cadiot, Pierre 64 Campbell, Lyle 357 Campe, Petra 76 Carlin, Eithne 179, 486, 487 Carlson, Robert 488 Carpenter, Kathie L. 469 Catta Quelen, Javier 135, 474, 476 Clark, Eve 469 Claudi, Ulrike 3, 25, 100, 147, 156, 161, 193, 357, 362, 470 Clements, J. Clancy 454 Collinder, Björn 504 Comrie, Bernard 94, 279, 473, 476, 482 Coseriu, Eugenio 77, 451 Craúoveanu, Dumitru 468 Croft, William 3, 5, 25, 27–29, 31, 81, 85, 87–89, 91, 120, 129, 147, 162, 163, 448, 449, 452, 467, 468, 469, 482

Index of authors 543 Cuturi, Flavia 502 Cyffer, Norbert 149, 479 Cysouw, Michael 92 Dahl, Östen 86, 503 Dambrinjnas, Leonardas 350 Daniel, Michael 452 Davies, John 182 Delbrück, Berthold 378, 452 Derbyshire, Desmond C. 168 Dixon, Robert Malcolm Ward 189, 190, 478, 479 Dressler, Wolfgang U. 79, 80, 166, 488 Drossard, Werner 59, 460, 461 Dryer, Matthew S. 92, 94, 470 Dutton, Tom E. 460 Dyla, Stefan 134, 468 Ebbinghaus, Ernst A. 378 Echegoyen, Artemisa 185, 186, 374, 502 Eckert, Rainer 451, 497, 514 Endruschat, Annette vii, 468 EndzelƯns, JƗnis 318 Eroms, Hans-Werner 447 Ersen-Rasch, Margarete I. 29, 49 Evans, D. Simon 369 Faarlund, J. Terje 468 Fabri, Ray 494, 495, 496 Fanselow, Gisbert 448 Felix, Sascha 448 Fillmore, Charles J. 76, 469 Fliedl, Günther 488 Fonseca, Maria do Ceu 468, 482 Fortescue, Michael 481, 483 Friedländer, Marianne 480 Gabelentz, Georg von der 394 GƗters, AlfrƝds 309, 383, 501, 504 Gauto Bejarano, Miguel Ángel 50, 51, 142

Gehling, Thomas 488 Georg, Stefan 71, 482 Girke, Wolfgang 468, 476 Givón, Talmy 79, 81, 193, 358 Gnerre, Maurizio 502 Goebl, Hans 490 Grace, George W. 57, 58, 460 Greenberg, Joseph J. 192 Grenand, Françoise 141 Griefenow-Mewis, Catherine 473 Gugeler, Traude 93, 218, 440 Haase, Martin 384, 385 Habicht, Külli 364, 365, 383 Haenisch, Erich 146 Hagège, Claude 358, 359, 475 Haiman, John 451 Hakulinen, Lauri 462 Hansen, Erik 466 Hardman, Martha James 139, 477 Harris, Alice C. 473 Hartmann, Roswitha 126, 127, 135 Haspelmath, Martin 79, 80, 81, 156, 221, 240, 360, 385, 466, 469, 475, 478, 488 Hasselblatt, Cornelius 30, 49, 350, 453 Heine, Bernd 3, 6, 25, 100, 136, 147, 148, 150, 154, 156, 161, 193, 317, 357, 358, 359, 362, 377, 470, 477, 479, 499, 500, 503 Hekking, Ewald 502 Hetzer, Armin 360, 364 Hetzron, Robert 134, 466 Hill, Jane H. 502 Hill, Kenneth C. 502 Himmelmann, Nikolaus P. 20, 82, 93 Hinze, Friedhelm 451, 497 Hjelmslev, Louis 466 Hoddinott, William G. 144, 145 Höftmann, Hildegard 132 Holes, Clive 479, 493 Hollenbach, Elena 182

544 Index of authors Hollenbach, Fernando 182 Holst, Jan Henrik 497 Hovdhaugen, Even 72, 142 Hünnemeyer, Friederike 3, 25, 100, 147, 156, 161, 193, 357, 362, 470 Iordan, Iorgu 455 Ivanov, Vyacheslav 482 Jacobson, Steven A. 189 Janda, Laura A. 468 Jensen, Hans 145, 359 Johnson, Mark 92, 98, 99, 100, 104, 109, 114, 357, 470 Jókay, Zoltán 484 Jolly, Julia A. 451 Jones, Linda K. 372 Jones, Morris 136, 253 Kabelka, Jonas 351 Kämpfe, Hans-Rainer 60, 461 Karjalainen, Merja 61–63, 153, 462 Karlsson, Fred 62, 84, 142, 350, 352, 467 Kastenholz, Raimund 151, 152, 480 Kay, Paul 469, 515 Keenan, Edward L. 127 Keller, Rudi 450 Kemmer, Suzanne 469 Kempf, Zdislaw 77, 468 Kenesei, István 169 Kettler, Sonja 479 Kibrik, Aleksandr E. 143, 478 Kilian-Hatz, Christa 8, 89, 470, 479 Kircher, Chantal 11 Klein, Hans-Wilhelm 65, 67, 463 Kleineidam, Hartmut 65, 67, 463 Klimas, Antanas 350 Koch, Wolfgang 447 Kofod, F. M. 144, 145 Koptjevskaja-Tamm, Maria 503 Kornfilt, Jaklin 453

Kozintseva, Natalia 54, 482 Krause, Wolfgang 502 Kress, Bruno 322, 324, 329, 335, 336, 337, 339, 343, 344, 497, 498 Kroeber, Alfred L. 58, 460 Kross, Kristiina 462 Kühnel, Helmut 352 Kühner, Raphael 11 Kunze, Jürgen 447, 482 Kuteva, Tania 357, 358, 377, 499, 503 Kwee, John B. 29, 453 Lakoff, George 25, 77, 91, 98–100, 104, 109, 114, 160, 357, 470 Landmann, Fred 134 Lasersohn, Peter 134 Laskowski, Roman 369, 502 Lastra, Yolanda 377 Launey, Michel 11, 135 Lazard, Gilbert 79, 448 Lee, Sang-Hyup 10, 11, 102 Lefevbre, Claire 72 Leger, Rudolf 133, 475 Lehfeldt, Werner 222 Lehiste, Ilse 77, 482 Lehmann, Christian 2, 3, 6, 16, 17, 26, 33, 35, 75, 78, 147, 166, 193, 363, 365, 447, 448, 451, 465, 500, 502 Leiss, Elisabeth 79, 81, 466 Levinson, Steven C. 100 Levy, Paulette 71, 465 Lewis, Henry 378 Lötzsch, Ronald 308, 310, 321, 381, 382, 468, 497, 503 Loogman, Alfons 127, 142, 148, 474, 476 Lorenz, Manfred 478 Lüdtke, Helmut 363 Luna Alonso, Ana 200

Index of authors 545 Luraghi, Silvia 91, 120, 147, 160, 162, 468, 478 Maas, Utz 493 Mahootian, Shahrzad 125, 474 Marks, Doreen 58, 460 Martínez, Rafael 91, 464, 468 Maslova, Elena 92 Mátai, Mária D. 502 Mattissen, Johanna 488 Mayerthaler, Willi 192, 488 McKaughan, Howard 58, 460 McNally, Louise 21, 134 Meibauer, Jörg 451 Mel’þuk, Igor 181 Meléndez Lozano, Miguel Á. 188 Menges, Karl H. 468 Michaelis, Susanne 455, 469, 472, 479, 481 Miller, J. 20 Mithun, Marianne 73–74 Mitzka, Walther 378 Möhlig, Wilhelm J.G. 136, 477 Monod-Becquelin, Aurore 178 Montan, Per 7 Moravcsik, Edith 452 Moritz, Karl Philipp 457 Morris-Jones, John 379 Mosel, Ulrike 57, 72, 142, 460 Moser, Manfred 495 Moser, Mary Beck 183 Moyse-Faurie, Claire 474, 475 Müller, Nicole 468 Munro, Pamela 83 Nau, Nicole 61, 62, 384, 385, 462, 468, 497, 500, 504, 505 Neukom, Lukas 84, 140, 141 Nickel, Klaus Peter 452, 453 Nilsen, Don L.F. 77, 82–84, 466, 468, 477 Nishina, Yoko 70, 71 NƯtiƼa, Daina 496

Noreen, Alfred 76, 482 Obenland, Roland 470 Obst, Ulrich 468 O’Connor, Catherine 469 Olpp, Johann 72 Ó Siadhail, Míchéal 456 Osumi, Midori 186, 187 Pagliuca, William 181 Palancar, Enrique L. 74, 91, 127, 128, 143, 164, 362, 465, 472, 478, 500 Palmer, Frank Robert 457, 465 Patnaik, Manideepa 84, 140, 141 Paul, Waltraud 358 PauliƼš, Osvalds 308, 309 Pedersen, Holger 378 Perkins, Revere D. 93, 181, 470 Pinkster, Harm 455 Plank, Frans 92 Plungian, Vladimir A. 156, 469 Ponelis, Friedrich Albert 476 Premper, Waldfried 2 Press, Ian 36, 216 Rasoloson, Janie Noëlle 143 Rauh, Gisa 466 Refsing, Kirsten 179 Rijkhoff, Jan 93 Risch, Ernst 378 Robinson, Charles Henry 150 Robu, Vladimir 455 Rosalie, Marcel 455, 469, 472, 479, 481 Rosenqvist, Håkan 7 Rozenbergs, JƗnis 308, 309 Ruhlen, Meritt 449, 471 Saeed, John Ibrahim 72 Saint-Exupéry, Antoine de v, 198 Sakayan, Dora 53 Salas, Adalberto 184, 185

546 Index of authors Saltarelli, Mario 168, 484 Sammallahti, Pekka 503 Sandra, Dominiek 81 Sapir, Edward 74, 75, 465 Sasse, Hans-Jürgen 2, 27 Schleicher, August 466 Schlesinger, Izchak M. 85–87, 89, 91, 120, 156, 448, 467, 470 Schmalstieg, William R 350, 383, 504 Schmitt, Rüdiger 500 Schröder, Jochen 468 Schultze-Berndt, Eva 20, 128, 129 Schwarz, Oxana 76, 468 Schweiger, Fritz 470 Seiler, Hansjakob 2, 3, 16, 77, 193, 447, 450, 451, 467, 482 Serzisko, Fritz 452 Sharma, Aryendra 114, 126, 142 Shin, Yong-Min vii, 3, 16, 26, 33, 35, 75, 147, 447, 448, 451, 465, 502 Shou-Hsin, Teng 466 Shroyer, Thomas G. 82, 466 Shukla, Shaligram 113 Shyldkrot, Hava Bat-Zeev 469 Siewierska, Anna 127, 474 Skorik, Petr 60 Slobin, Jakovlevic 95 Smith, Norval 482 Sneddon, James Neil 453 Stachowski, Marek 468 Stassen, Leon 21, 91, 130, 131, 134 Stegmann, Carl 11 Stoebke, Renate 461 Stolz, Christel 450, 501 Stolz, Thomas 17, 20, 28, 29, 37, 38, 56, 59, 70, 89, 91, 92, 93, 100, 113, 114, 118, 120, 121, 122, 124, 131, 136, 139, 140, 148, 153, 154, 159, 160, 161, 166, 167, 171, 172, 173, 180, 182, 198, 209, 218, 223, 224,

253, 307, 358, 359, 360, 368, 382, 397–421, 452, 461, 465, 466, 470, 471, 472, 473, 475, 477, 478, 479, 480, 481, 485, 486, 487, 490, 491, 499, 500, 501 Stroh, Cornelia 70, 76, 91, 92, 100, 113, 171, 173, 180, 182, 198, 209, 223, 224, 360, 379, 397– 421, 468, 472, 486, 490, 491 Sulkala, Helena 61, 62, 153, 461, 462 Sutton, Peter 479 Svane, Gunnar Olaf 382 Szabó, Dénes 484 Szabó, László 468 Talmy, Leonard 450 Tauli, Valter 365, 501 Testelets, Ja.G. 143, 478 Thomas, Alan R. 136, 253 Thomas, Werner 502 Toivainen, Jorma 461 Tompa, József 360 Traugott, Elizabeth Closs 6 Trudgill, Peter 94 Trunte, Hartmut 380 Ulvydas, K. 368, 501 Urdze, Aina 76, 91, 92, 100, 171, 173, 180, 182, 198, 209, 223, 224, 397–421, 448, 468, 479, 486, 490, 491 Vago, Robert M. 169 Valge, Jüri 501 Van den Berg, Helma 478 van der Auwera, Johan 156, 469 van Driem, George 473 Van Valin, Robert D. 6, 35–37, 120, 121, 129, 147, 163 Vaquero, P. Antonio 146, 479, 480 Vásquez, Juana 139, 478

Index of authors 547 Veenstra, Tonjes 472 Veerman-Leichsenring, Annette 376, 466 Vella, Joseph 495 Vennemann, Theo 505 Verhoeven, Elisabeth vii, 16, 26, 33, 35, 149, 454 Vermeer, Hans J. 116, 128, 144 Vietze, Hans-Peter 51 VilƗns, OtomƗrs 308, 309 Vincent, Alex 57, 459 Voigtlander, Katherine 185, 186, 374, 502 Volodin, Alexander P. 60, 71, 461, 482

Wälchli, Bernhard 191, 192, 475 Walmsley, John B. 77 Werner, Heinz 467 Wierzbicka, Anna 468 Wilkins, David 35, 38–40, 129, 452, 456 Winkler, Christian 488 Wolff, H. Ekkehard 150 Worth, Dean 60 Wurzel, Wolfgang U. 69, 192 Zelinsky-Wibbelt, Cornelia 87 Zewen, François-Xavier Nicolas 360 Zigmond, Maurice L. 83

Index of subjects

This index only contains those items which occur more rarely than the key notions comitative and instrumental (which are practically mentioned on every page of this book – and thus we felt no need to include them in the index). abessive 482, 484, 532 ablative 11, 37–40, 140 absolutive 58, 244 accompanee 17–23, 25–32, 34, 40, 42–44, 48, 53, 55, 57–59, 61, 62, 69, 254, 294, 295, 305, 306, 314–316, 321, 331–333, 340, 347, 349, 350–353, 361, 374, 391, 395, 450, 451, 459, 496 accompaniment 1, 2, 5–13, 14, 17– 19, 23, 24, 26, 27, 30, 31, 36, 38, 39, 49–51, 54, 55, 61, 64, 71, 84, 88–90, 98, 99, 102, 103, 109, 113, 125–127, 129, 132, 133, 136, 137, 139, 141, 143, 146, 148–151, 154, 178– 189, 255, 262, 268, 272, 294, 306, 321, 342, 355, 356, 371, 373, 375, 380, 391, 395, 447, 450, 451, 453–456, 461, 462, 479, 484, 488, 501, 504 accusative 21, 37, 82, 124, 137, 144, 145, 154, 169, 247, 263, 307– 309, 315, 316, 322, 323, 325, 326, 329, 331–333, 335–349, 354, 367, 368, 473, 478, 497– 499, 501 actor 35, 37, 38, 59, 60, 147, 163, 170, 346, 461, 475 adessive 62, 83, 140, 142, 153, 275, 278, 365, 366, 462, 473, 504, 505 adjunct 7, 17, 20, 23–25, 28, 33, 35, 323, 345

agent 22, 23, 26, 27, 35–39, 82, 88, 89, 115, 120, 121, 125–128, 130, 143, 145, 146, 148, 151, 164, 244, 250, 276, 281, 286, 295, 298, 302, 324–326, 328, 332, 369, 392, 451, 457, 472, 475, 478, 482, 495, 522, 523, 526 agentivity 38, 43, 82, 83, 120, 121, 123, 124, 130, 164, 294, 346, 390, 457, 473 alienability 57, 152, 153, 338, 460, 480, 531 allative 140, 146 and-conjunction 133, 139, 156, 390 and-language 91, 130, 137, 138, 140–143, 145, 146, 149, 153, 154, 168, 178, 182–184, 188, 315, 383, 473, 476, 478, 481, 488, 529 animacy 11, 21, 26, 27, 30, 43, 44, 46, 51, 56, 60, 63, 75, 90, 130, 133, 135, 138, 164, 184, 294–297, 303, 305, 306, 315, 321, 331, 333, 345–348, 355, 365, 383, 451, 455, 459, 462, 482, 501 areality 204, 209, 216, 217, 233, 236, 391 argument (NP) 182, 451, 455 associative 28, 31, 50–53, 86, 149, 153, 169, 450, 452, 479, 484, 488, 513 A-type 102–107, 110–112, 117, 120, 124, 126, 127, 132, 133, 137,

Index of subjects 549 140–143, 145, 146, 148, 149, 153, 167, 168, 170, 172–184, 188, 189, 191, 196, 197, 200, 204, 207–209, 213–215, 217, 229, 233, 234, 236, 239, 240, 243, 248, 249, 253, 256–258, 260–267, 281, 282, 349, 350, 352, 355, 359, 366–368, 372, 374, 375, 377–379, 381–386, 393, 394, 473, 474, 478–481, 485–488, 490, 502 benefactive 89, 115–118, 153, 157, 158, 160, 162, 480, 481 body part instrument 43, 50, 62–66, 74, 75, 84, 179, 188, 244, 252–255, 272, 286, 311, 318, 321, 324, 373, 381, 457, 460, 462, 491, 493 B-type 102–107, 110–112, 117, 120, 124, 125, 139, 142, 150, 154, 169, 173, 187, 196, 197, 200, 201, 203, 204, 208, 209, 213– 215, 217, 226, 228, 229, 233, 234, 236, 239–241, 243, 245, 246, 248, 253, 255, 257–267, 282, 307, 310, 321, 322, 349, 350, 352, 358, 364, 366–369, 371–373, 376, 378–386, 393, 394, 472, 478, 486, 487, 489, 490, 500, 502 causative 32, 43, 115–121, 126, 127, 139, 457, 472, 473, 474, 512 combination 41, 43, 49–52, 54, 66, 90, 126, 129, 492 companion 6, 9, 11, 17–23, 25–32, 40–44, 47, 48, 51, 53, 57–59, 62, 66, 69, 91, 98, 99, 103, 107, 119, 121, 137, 148, 149, 151, 153, 154, 170, 254–258, 294, 295, 305, 314–317, 321, 332, 333, 340, 347–353, 359, 361, 364, 372–374, 389–391, 395, 450, 451, 453, 459, 461,

471, 484, 492, 495, 496, 499, 500, 503, 530 confective 24, 41, 42, 60, 61, 65, 72, 74, 90, 129, 130, 152, 169, 170, 178, 258, 272, 289, 290, 295, 306, 313, 321, 322, 329, 333, 336, 340, 348, 355, 459, 485, 500 coordination 21, 47, 48, 59, 60, 98, 117, 130–139, 150, 151, 157, 302, 306, 315, 316, 321, 458, 466, 475, 476, 477, 479, 480, 485, 487, 488, 514, 517, 523, 534 C-type 102–106, 110–112, 117, 120, 124, 126, 132–134, 136, 150, 153, 154, 169, 170, 172–174, 180, 184, 191, 196, 197, 200, 201, 204, 207, 208, 213, 214, 217, 226, 234, 236, 240, 243, 248, 249, 253, 257–259, 262, 265, 266, 268, 281, 282, 307, 322, 349, 350, 352, 355, 366– 369, 372, 373, 377–381, 383– 385, 389–391, 394, 472, 480, 485, 487, 488, 490 dative 5, 21, 48, 69, 137, 153, 154, 169, 245, 250, 257, 259, 307– 309, 315, 322–329, 331, 333, 335–349, 378, 379, 455, 457, 496, 497, 498, 499, 518 elative 140, 354, 462 ergative 89, 115–118, 120, 121, 124–126, 128, 129, 144, 145, 189, 240, 244, 250, 456, 473, 474, 477 essive 280, 503, 532 experiencer 35, 153 focus 59, 250, 314, 451, 460 genealogy 8, 204, 209, 217, 218, 225, 230, 233, 471

550 Index of subjects genitive 31, 32, 61, 153, 308, 346, 350, 354, 365, 378, 379, 457, 464, 496, 497, 498, 501, 503 grammaticalisation 3, 69, 99, 114, 160, 165, 166, 183, 187, 189– 191, 193, 194, 281, 283, 355, 357–359, 362, 363, 365, 369, 370, 377, 378, 380, 383, 386, 392, 393, 467, 470, 478, 485, 499, 500, 503 human instrument 42, 45, 46, 52–54, 63, 64, 459, 473 iconicity 48, 62, 79, 172, 451, 458 inalienability 338, 479 inessive 140, 153, 360, 462, 500 ingredient 49, 172, 244, 250, 254, 259, 263, 302, 464, 496 instrumentality 1, 50, 53, 55, 61–63, 67, 68, 72, 98, 99, 103, 109, 126, 133, 134, 139, 141, 142, 152, 153, 183, 184, 188, 244, 248, 249, 255, 259, 260, 263, 265–267, 271, 272, 274, 285, 293, 310, 313, 321, 323, 325, 355, 356, 360, 364–371, 373, 374, 376–378, 380, 381, 384– 386, 391, 393, 394, 456, 460, 464, 465, 471, 473, 477, 479, 480, 482, 483, 486, 488, 492, 499, 501, 502, 504 instructive 62, 63, 83, 84, 365, 461, 462, 464, 501, 503 language change 394, 485, 488 language contact 14, 93, 216, 236, 370–372, 376, 379, 384, 385, 393, 488, 503, 517, 533 locative 29, 37, 38, 86, 115–118, 140–147, 149, 151, 152, 157– 160, 162, 189, 305, 306, 360, 367, 390, 455–457, 478, 481, 482, 531

markedness 14, 92, 98, 166, 177, 181, 191–194, 257, 283, 390, 464, 465, 485, 530, 532 material 21, 42, 43, 49, 52, 62, 63, 66, 244, 271, 288, 306, 312, 320, 321, 327, 348, 451, 458, 462, 463, 492 means of transportation 34, 43, 50, 52, 54, 62, 64–67, 103, 152, 178, 286, 306, 311, 312, 321, 325, 344, 348, 373, 456, 463 metaphor 36, 91, 98–100, 103, 107, 119, 140, 161, 162, 249, 326, 331, 389, 390, 392, 395, 470, 471, 521, 530 object (direct, indirect) 17, 18, 26– 28, 30, 31, 64, 70, 72, 135, 136, 182, 272, 275, 287, 288, 295, 296, 326, 352, 373, 374, 453, 454, 455, 458, 468, 499 ornative 24, 34, 41, 43, 49, 54, 60, 61, 65, 130, 190, 245–249, 258–268, 272, 289, 290, 305, 306, 309, 310, 313, 318, 319, 321, 322, 333, 336, 342, 347, 348, 355, 391, 459, 485, 492, 500, 504 part-whole 33, 42, 43, 49, 51, 52, 54, 61, 302, 305, 306, 313, 319, 321, 334, 335, 348, 464, 480 passive 18, 36, 41, 42, 58, 121, 126, 127, 145, 276, 277, 369, 473, 414, 519, 529 passive agent 36, 88, 89, 115, 121, 125, 127, 128, 145, 146, 148, 151, 276, 277, 281, 369, 473, 474, 478, 480, 482, 495 patient 22–25, 35, 37, 42, 45, 50, 130, 150, 170, 244, 271, 332, 346, 392, 455, 481, 482 perlative 146, 274, 296, 306, 478, 502 polysemy 79–81, 184, 448, 449

Index of subjects 551 possession 40, 42, 43, 51–53, 57, 62, 115–117, 147–160, 162, 164, 179, 277, 278, 281, 288, 290, 291, 306, 317, 322, 338–340, 348, 369, 390, 479, 480, 481, 482, 486, 498, 517, 518, 531 possessor 32, 33, 61, 62, 75, 152– 155, 164, 168, 183, 188, 278, 340, 369, 476, 487, 500, 503 primary marker 41, 44, 47, 48, 68–70, 102, 138, 148, 152, 184, 197, 223, 253, 257, 266, 308, 321, 353–355, 357, 369, 389, 391, 448, 456, 464, 471, 481, 486 prolative 274, 296, 300, 306, 328, 329, 347, 348, 478, 499 relator 6, 11, 15, 17–20, 22, 23, 27, 29–32, 36, 40, 42, 44, 46, 48– 51, 55, 64, 67–71, 83, 84, 86, 89, 90, 92, 101–103, 106– 109, 114, 116, 125–127, 132– 135, 137–139, 141, 142, 145, 148–155, 164, 166, 167, 170– 174, 177–181, 183–185, 187– 190, 197, 200, 201, 203–205, 207–210, 213–219, 221–224, 240, 241, 244–255, 257–267, 272, 274, 275, 277, 279–282, 305, 314, 315, 321, 322, 325, 337, 338, 349–353, 355–357, 360, 364, 365, 367–370, 372, 376, 386, 388, 390, 392, 393, 395, 450, 451, 453, 454, 455, 456, 459, 462, 464, 465, 466, 472, 474, 475–480, 486, 487, 489–492, 500, 504, 521 subject 4, 16–18, 22–29, 32, 33, 35, 41, 47, 56, 62, 70, 82, 83, 88, 92, 134–138, 148, 155, 182, 186, 199, 244, 250, 254, 259, 263, 282, 283, 295, 323, 325, 330, 337, 343, 346, 369, 370, 372, 375, 376, 382, 387, 392–

395, 452, 457, 460, 466, 468, 478, 499, 508, 534 superessive 10, 140, 142, 146, 478 syncretism 12, 28, 35–38, 82, 84, 86, 88, 89, 92, 96, 98–100, 103, 104, 106, 108, 109, 114–121, 123, 126, 128, 129, 131, 132, 134, 139–141, 142–148, 157– 167, 170, 184, 189, 191, 194, 196, 338, 348, 349, 361–363, 386, 389, 390, 456, 457, 464, 465, 469, 473, 475, 478, 480, 481, 482, 497, 522, 523, 530 tool 2, 14, 21, 23–26, 31, 42, 43, 50, 51, 53, 54, 62–64, 66, 67, 71– 75, 102, 103, 113, 114, 125– 127, 129, 130, 132, 133, 152, 179, 182–186, 189, 244–250, 252, 254, 255, 267, 268, 271, 272, 285, 286, 306, 310, 312, 321, 323, 324, 326, 348, 355, 356, 371, 373, 375, 380, 391, 453, 456, 458, 460, 465, 484, 491, 504 typology 4, 15, 82, 87, 88, 93, 94, 99, 100, 204, 209, 217, 218, 229, 230, 233, 234, 388, 447, 470, 490, 510, 512, 513, 517, 518, 520, 523, 529, 530, 531, 532 undergoer 35, 37, 38, 40, 59, 60, 147, 163, 170, 346, 455, 461 universalism 2, 4, 8, 9, 13,–15, 17, 44, 77, 88, 92–94, 96, 98– 100, 104, 106, 107, 109, 114, 115, 119, 156, 161, 196, 362, 388–390, 393, 447, 470, 471, 511, 512, 513, 514, 516, 523, 528 vialis 188–189 with-language 130, 131, 138, 142, 148–150, 153, 178, 179, 315, 383, 474, 478, 479, 488, 502, 529