Relative Constructions in European Non-Standard Varieties 9783110238792, 9783110238785

Cross-linguistic studies on relative constructions in European languages are often centred on standard varieties as desc

231 66 3MB

English Pages 497 [500] Year 2011

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Acknowledgements
List of figures
List of tables
List of abbreviations
Introduction
Part 1. Theoretical and methodological premises
1. Theoretical background
1.1. Basic definitions
1.1.1. Relative relations
1.1.2. Relative constructions
1.2. Theoretical framework
1.3. ‘Standard’ and ‘non-standard’
1.3.1. Defining ‘standard’ and ‘non-standard’
1.3.2. On the usefulness of distinguishing standard from non-standard in linguistic research
1.4. Relative clauses in European languages: the state of the art
1.5. Aims of the study
2. Data sources
2.1. Looking for non-standard constructions
2.2. Language sample
2.3. A review of data sources
2.3.1. Linguistic evidence
2.3.2. Grammars
2.3.3. Linguistic studies
2.3.4. Questionnaires
2.3.5. Language corpora
2.3.6. The World Wide Web
2.3.7. Language corpora vs. the World Wide Web
2.3.8. Historical evidence
2.3.9. Summary
3. Data classification
3.1. Word order
3.2. Relative element
3.2.1 Simple relative elements
3.2.2. Combined relative elements
3.3. Syntactic positions relativized
3.4. Correlating the parameters
3.5. Classification problems
3.5.1. How ‘relative’ are relative particles?
3.5.2. The morphosyntactic status of Turkish ki
3.5.3. Zero-marker or coordinate clauses?
3.5.4. Simple or combined?
3.5.5. The status of adpositions
3.5.6. One strategy or two?
Part 2. An areal study of non-standard relative constructions in European languages
4. Typological issues
4.1. Word order
4.1.1. The postnominal strategy
4.1.2. The correlative strategy
4.1.3. The postposed strategy
4.2. Relative element
4.2.1. Inflected elements remain uninflected
4.2.2. The syntactic role of the relativized item in the RC is not encoded
4.2.3. The diffusion of relative particles
4.2.4. Other phenomena related to relative elements
4.3. Syntactic positions relativized
4.3.1. Decumulation
4.3.2. Double encoding
4.3.3. No encoding
4.3.4. Case matching
4.3.5. Avoiding the relative clause
4.4. Typological conclusions
4.5. A functional account?
5. Sociolinguistic issues
5.1. Non-standard vs. standard relative constructions: a sociolinguistic account
5.1.1. Word order
5.1.2. Relative element
5.1.3. Syntactic position relativized
5.2. A scale of standardness for relative constructions in European languages
5.2.1. The degree of standardness of individual strategies
5.2.2. The relationship between strategies
5.2.3. Factors influencing the position of languages on the scale
5.3. Standard, non-standard and speakers’ choices
6. Diachronic issues
6.1. Non-standard relative constructions: diachronic evidence
6.2. Word order
6.2.1. Turkish ki and the postnominal strategy in Basque
6.2.2. The correlative strategy
6.2.3. The postposed strategy
6.3. Relative element
6.3.1. Relative pronouns
6.3.2. Specialized relative elements
6.3.3. Relative particles
6.3.4. The zero-marker
6.3.5. Combined elements
6.4. Syntactic positions relativized
6.4.1. Decumulation
6.4.2. Double encoding
6.4.3. No encoding
6.4.4. Case matching
6.5. The development of (non-)standard relative constructions: a proposal
6.5.1. Stage 1: no endoglossic standard
6.5.2. Stage 2: the formation of an endoglossic standard
6.5.3. Stage 3: the continuum of standard and non-standard varieties
6.5.4. Further developments
6.6. Concluding remarks
Part 3. Construction-based language-specific case studies
7. Presentative constructions in German discussion forums
7.1. Aims of the study
7.2. Preliminaries to the corpus analysis
7.2.1. The constructions under investigation
7.2.2. The database
7.3. Postnominal vs. postposed relative clauses
7.4. Relative pronoun vs. relative particle
7.5. Discussion
8. Relative constructions in a spoken Russian corpus
8.1. Aims of the study
8.2. Preliminaries to the corpus analysis
8.2.1. Russian linguistic varieties
8.2.2. The corpus
8.2.3. The constructions under investigation
8.3. Quantitative analysis
8.4. Qualitative analysis
8.5. Discussion
8.6. The informants
Summary and outlook
Appendix A – The language sample
Appendix B – The questionnaires
Notes
References
Index of languages
Index of authors
Index of subjects
Recommend Papers

Relative Constructions in European Non-Standard Varieties
 9783110238792, 9783110238785

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

Relative Constructions in European Non-Standard Varieties

Empirical Approaches to Language Typology 50

Editors Georg Bossong Bernard Comrie Yaron Matras

De Gruyter Mouton

Relative Constructions in European Non-Standard Varieties by Adriano Murelli

De Gruyter Mouton

ISBN 978-3-11-023878-5 e-ISBN 978-3-11-023879-2 ISSN 0933-761X Library of Congress Cataloging-in-Publication Data Murelli, Adriano. Relative constructions in European non-standard varieties / by Adriano Murelli. p. cm. ⫺ (Empirical approaches to language typology; 50) Includes bibliographical references and index. ISBN 978-3-11-023878-5 (alk. paper) 1. Language and languages ⫺ Variation. 2. Europe ⫺ Languages ⫺ Variation 3. Grammar, Comparative and general ⫺ Relative clauses. 4. Grammar, Comparative and general ⫺ Syntax. 5. Dialectology. 6. Typology (Linguistics) I. Title. P120.V37M87 2011 417⫺dc23 2011033458

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. 쑔 2011 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany www.degruyter.com

Jede Sprache erscheint so fortwährend auf der Suche nach einem Gleichgewicht zwischen der extremen Vielfalt einer unbegrenzten Anzahl von Varianten und der kodifizierten und stabilisierten Auswahl von relativ begrenzten Untermengen dieser Varianten. (Bernard Py)

Je sais que je tombe dans l’inexplicable quand j’affirme que la réalité – cette notion si flottante –, la connaissance la plus exacte possible des êtres est notre point de contact, et notre voie d’accès aux choses qui dépassent la réalité. (Marguerite Yourcenar)

La lingüística […] debe moverse constantemente entre los dos polos opuestos de lo concreto y de lo abstracto: subir de la comprobación empírica de los fenómenos concretos a la abstracción de formas ideales y sistemáticas, y volver luego a los fenómenos concretos, enriquecida por los conocimientos generales adquiridos en la operación abstractiva. (Eugenio Coseriu)

Acknowledgements

The present work is a revised version of my Ph.D. dissertation, which I submitted and defended in summer 2009 at the University of Pavia. There is a good number of people to whom I am indebted for supporting me throughout the writing of my dissertation. First of all, I would like to thank Giuliano Bernini for following my thesis in all its phases and providing precious advice. Thank you also to Sonia Cristofaro for discussing with me parts of my dissertation, thus contributing to improve its general structure and especially the content of Part 1. Achim Rabus spurred me on throughout all phases of my work and also read a good part of it, giving me valuable feedback, in particular on Slavic languages: I want him to know that I am deeply grateful to him. I am also indebted to Christian Lehmann and Caterina Mauri, with whom I discussed single issues of this work. I first conceived the idea to write a dissertation on relative constructions in non-standard varieties of European languages during my two years’ Master in European Linguistics at the University of Freiburg from 2004 to 2006. I am indebted to Bernd Kortmann, the director of the master degree program, as well as to Stefan Pfänder, Christian Mair, Peter Auer and Christian Voß for introducing me to functional-typological linguistics, to Eurolinguistics and to the fascinating realm of language variation. Thank you also to all informants who completed the questionnaires, providing me with detailed information about the peculiarities of their mother tongues: Milena Marić (BCS), Petăr Kehajov, Maria Manova (Bulgarian), Matti Miestamo (Finnish), Ramona Baumgartner, Diana Heißler, Sandra Kohllöffel, Anna Riester (German), Sofia Lampropoulou, Nikolaos Mytilinaios (Greek), Valeria Cadau, Marianna Gionta (Italian), Jarosław Aptacy, Anna Górska, Ewa Schalley (Polish), Albertino Moreira da Silva Junior (Portuguese), Mădălina Chitez, Gabriela Diaconu (Rumanian), Natal’ja Ilina, Filipp Lupov, Ol’ga Zacharova (Russian). Finally, I owe thanks to Georg Bossong for the comments he provided on my manuscript as an editor of the EALT series, to Julie Miess and Marcia L. Schwartz from Mouton de Gruyter for their help during the publication process, to Chiara Mauri for carefully proofreading my English and to Noah Bubenhofer for his help with the indexes. Last, but not least, thanks to my friends and family for encouraging me throughout the past few years – particularly to M., to whom this book is dedicated.

Contents

Acknowledgements List of figures . . List of tables . . List of abbreviations

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

vii xiii xv xix

Introduction .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

1

Part 1. Theoretical and methodological premises

.

.

.

.

.

5

.

1. Theoretical background . . . . . . . . . . . 1.1. Basic definitions . . . . . . . . . . . . . 1.1.1. Relative relations . . . . . . . . . . 1.1.2. Relative constructions . . . . . . . . . 1.2. Theoretical framework . . . . . . . . . . . 1.3. ‘Standard’ and ‘non-standard’ . . . . . . . . . 1.3.1. Defining ‘standard’ and ‘non-standard’ . . . . 1.3.2. On the usefulness of distinguishing standard from non-standard in linguistic research . . . . . 1.4. Relative clauses in European languages: the state of the art 1.5. Aims of the study . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

6 6 7 21 26 29 29

. . .

. . .

34 37 45

2. Data sources . . . . . . . . . . . . 2.1. Looking for non-standard constructions . . . . 2.2. Language sample . . . . . . . . . . . 2.3. A review of data sources . . . . . . . . 2.3.1. Linguistic evidence . . . . . . . . 2.3.2. Grammars . . . . . . . . . . . 2.3.3. Linguistic studies . . . . . . . . . 2.3.4. Questionnaires . . . . . . . . . 2.3.5. Language corpora. . . . . . . . . 2.3.6. The World Wide Web . . . . . . . 2.3.7. Language corpora vs. the World Wide Web . 2.3.8. Historical evidence . . . . . . . . 2.3.9. Summary . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

47 47 52 55 55 57 60 62 65 67 68 70 72

. . . . . . . . . . . . .

. . . . . . . . . . . . .

x Contents 3. Data classification . . . . . . . . . . 3.1. Word order . . . . . . . . . . . . 3.2. Relative element . . . . . . . . . . 3.2.1 Simple relative elements . . . . . . 3.2.2. Combined relative elements . . . . 3.3. Syntactic positions relativized . . . . . . 3.4. Correlating the parameters . . . . . . . 3.5. Classification problems . . . . . . . . 3.5.1. How ‘relative’ are relative particles? . 3.5.2. The morphosyntactic status of Turkish ki 3.5.3. Zero-marker or coordinate clauses? . . 3.5.4. Simple or combined? . . . . . . 3.5.5. The status of adpositions . . . . . 3.5.6. One strategy or two? . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

. . . . . . . . . . . . . .

73 74 79 87 98 112 115 122 123 129 136 139 141 145

Part 2. An areal study of non-standard relative constructions in European languages . . . . . . . . . . .

149

4. Typological issues . . . . . . . . . . . . . . . 4.1. Word order . . . . . . . . . . . . . . . . . 4.1.1. The postnominal strategy . . . . . . . . . . 4.1.2. The correlative strategy . . . . . . . . . . . 4.1.3. The postposed strategy . . . . . . . . . . . 4.2. Relative element . . . . . . . . . . . . . . . 4.2.1. Inflected elements remain uninflected . . . . . . 4.2.2. The syntactic role of the relativized item in the RC is not encoded . . . . . . . . . . . . . . . . 4.2.3. The diffusion of relative particles . . . . . . . . 4.2.4. Other phenomena related to relative elements . . . . 4.3. Syntactic positions relativized . . . . . . . . . . . 4.3.1. Decumulation . . . . . . . . . . . . . . 4.3.2. Double encoding . . . . . . . . . . . . . 4.3.3. No encoding . . . . . . . . . . . . . . 4.3.4. Case matching . . . . . . . . . . . . . . 4.3.5. Avoiding the relative clause . . . . . . . . . 4.4. Typological conclusions . . . . . . . . . . . . 4.5. A functional account? . . . . . . . . . . . . .

151 151 152 158 168 172 172 182 193 207 216 218 222 228 233 237 240 251

Contents xi

5. Sociolinguistic issues . . . . . . . . . . . . . . 5.1. Non-standard vs. standard relative constructions: a sociolinguistic account . . . . . . . . . . . . 5.1.1. Word order . . . . . . . . . . . . . . . 5.1.2. Relative element . . . . . . . . . . . . . 5.1.3. Syntactic position relativized . . . . . . . . . 5.2. A scale of standardness for relative constructions in European languages . . . . . . . . . . . . . . . . . 5.2.1. The degree of standardness of individual strategies . . 5.2.2. The relationship between strategies . . . . . . . 5.2.3. Factors influencing the position of languages on the scale 5.3. Standard, non-standard and speakers’ choices . . . . . .

255

6. Diachronic issues . . . . . . . . . . . . . . 6.1. Non-standard relative constructions: diachronic evidence . 6.2. Word order . . . . . . . . . . . . . . . . 6.2.1. Turkish ki and the postnominal strategy in Basque . 6.2.2. The correlative strategy . . . . . . . . . . 6.2.3. The postposed strategy . . . . . . . . . . 6.3. Relative element . . . . . . . . . . . . . . 6.3.1. Relative pronouns . . . . . . . . . . . 6.3.2. Specialized relative elements . . . . . . . . 6.3.3. Relative particles . . . . . . . . . . . . 6.3.4. The zero-marker . . . . . . . . . . . . 6.3.5. Combined elements . . . . . . . . . . . 6.4. Syntactic positions relativized . . . . . . . . . . 6.4.1. Decumulation . . . . . . . . . . . . . 6.4.2. Double encoding . . . . . . . . . . . . 6.4.3. No encoding . . . . . . . . . . . . . 6.4.4. Case matching . . . . . . . . . . . . . 6.5. The development of (non-)standard relative constructions: a proposal . . . . . . . . . . . . . . . . 6.5.1. Stage 1: no endoglossic standard . . . . . . . 6.5.2. Stage 2: the formation of an endoglossic standard . . 6.5.3. Stage 3: the continuum of standard and non-standard varieties . . . . . . . . . . . . . . 6.5.4. Further developments . . . . . . . . . . 6.6. Concluding remarks . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

297 298 302 302 305 311 312 312 321 324 329 331 334 335 340 343 345

. . .

346 350 351

. . .

358 366 367

256 262 263 269 273 277 283 285 290

xii Contents Part 3. Construction-based language-specific case studies

.

.

371

7. Presentative constructions in German discussion forums 7.1. Aims of the study . . . . . . . . . . . . . 7.2. Preliminaries to the corpus analysis . . . . . . . 7.2.1. The constructions under investigation . . . . 7.2.2. The database . . . . . . . . . . . . 7.3. Postnominal vs. postposed relative clauses . . . . . 7.4. Relative pronoun vs. relative particle . . . . . . . 7.5. Discussion . . . . . . . . . . . . . . .

. . . . . . . .

. . . . . . . .

373 373 373 373 377 378 381 383

8. Relative constructions in a spoken Russian corpus 8.1. Aims of the study . . . . . . . . . . . 8.2. Preliminaries to the corpus analysis . . . . . 8.2.1. Russian linguistic varieties . . . . . . 8.2.2. The corpus . . . . . . . . . . . 8.2.3. The constructions under investigation . . 8.3. Quantitative analysis . . . . . . . . . . 8.4. Qualitative analysis . . . . . . . . . . 8.5. Discussion . . . . . . . . . . . . . 8.6. The informants . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

387 387 388 388 390 395 399 402 413 416

Summary and outlook .

.

.

.

.

.

.

.

.

.

.

.

.

.

417

Appendix A – The language sample Appendix B – The questionnaires . Notes . . . . . . . . . . References . . . . . . . . Index of languages . . . . . . Index of authors . . . . . . Index of subjects . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

426 428 433 440 464 468 474

List of figures

1.1. Relative constructions: terminology . . . . . . . . 1.2. The social components concerned with the standard variety .

. 7 . 31

2.1. Levels of analysis in typological studies . . . . . . . . 47 2.2. The conceptual space of scripturality and orality (according to Koch and Oesterreicher 1990) . . . . . . . . . . . 49 2.3. The position of some communicative practices on the conceptual space between scripturality and orality . . . . . . . . 50 3.1. Combinations of adpositions and simple relative elements

.

. 145

4.1. The use of [+/−case] strategies for the relativization of the positions of the AH (based on Bernini 1989: 88) . . .

.

. 216

6.1. The development of standard and non-standard relative constructions in European languages . . . . . . .

.

. 349

8.1. The Russian sociolinguistic continuum

.

. 390

.

.

.

.

.

.

List of tables

1.1. The informational structure of relative relations . . . . . 1.2. An overview of the relations between SoAs analyzed in this work . . . . . . . . . . . . . . . . . .

21

2.1. The language sample analyzed in this study . . . . 2.2. The informants and their mother tongues . . . . . 2.3. Pros and cons of different data sources: an overview .

53 64 71

. . .

. . .

3.1. The traditional classification of RCs in European languages according to word order . . . . . . . . . . . . 3.2. Lehmann’s (1984) classification of relative elements . . . 3.3. Classification of relative elements according to De Vries (2002: 176) . . . . . . . . . . . . . . . . 3.4. Relativization strategies in European languages according to Cristofaro and Giacalone Ramat (2007: 65–72) . . . . . 3.5. The features encoded by simple relative elements . . . . 3.6. Simple relative elements attested in the sample . . . . . 3.7. Combined relative elements attested in the sample . . . . 3.8. Resumptive elements attested in the sample . . . . . . 3.9. The features encoded by combined relative elements . . . 3.10. Combinations of the two parameters ‘word order’ and ‘relative element’ attested in the sample . . . . . . . 4.1. Peculiar constructions in non-standard varieties of European languages: word order . . . . . . . . . . . . 4.2. The inflection of BUL kojto and RUM care in the standard variety . . . . . . . . . . . . . . . . . 4.3. The inflection of ENG who, FRE qui and ITA che in the standard variety . . . . . . . . . . . . . . 4.4. Uses of the LOC-specialized element in some of the languages in the sample . . . . . . . . . . . . 4.5. Relative particles in East-European languages and English . 4.6. The paradigm of Dutch relative pronouns . . . . . . . 4.7. The use of relative elements in Slovenian . . . . . . . 4.8. The use of relative pronouns in Finnish and Hungarian . .

9

79 81 82 83 88 97 99 108 111 116 151 172 177 184 193 208 210 211

xvi List of tables 4.9. Relative particles in competition in non-standard varieties . 4.10. The frequency of relative elements in the languages of the sample and the features that they encode . . . . . . . 4.11. The encoding of the syntactic role of the relativized item in the RC in a sample of world’s languages (Comrie and Kuteva 2005) . . . . . . . . . . . . . . . . . . 5.1. The position of relativization strategies in individual languages on the scale of standardness . . . . . . 5.2. Relative pronouns in some West-European languages . . 5.3. Relative pronouns in Slavic languages . . . . . . . 5.4. Top-bottom distribution of European languages on the scale of standardness (with respect to the range of constructions regarded as standard) . . . . . . . . . . . . 5.5. Factors influencing speaker’s choice of linguistic variants

242 244 249

. . .

276 278 280

. .

289 291

6.1. The development of ‘which’-based relative pronouns in Romance and Slavic languages . . . . . . . . . . 6.2. Combined relative elements attested in earlier stages of European languages . . . . . . . . . . . . . 6.3. The five types of European sociolinguistic constellations according to Auer (2005b) . . . . . . . . . . . 6.4. The distribution of relative elements in the Italian section of the spoken corpus C-ORAL-ROM . . . . . . . . . 6.5. The distribution of relative elements in Northern Italian dialects . . . . . . . . . . . . . . . . . 6.6. Relativization strategies in the diasystem of Italian . . . . 6.7. Relativization strategies in the Swedish Burträsk dialect (Karlsson and Sullivan 2002) . . . . . . . . . . 6.8. Relativization strategies in the diasystem of Russian . . .

318 331 347 360 361 363 364 366

7.1. 7.2. 7.3. 7.4.

Postnominal RCs vs. IV2s . . . . . . . . Postnominal RCs vs. IV2s with N=Leute . . . . Die-RCs vs. wo-RCs . . . . . . . . . . Wo-RCs: referents and syntactic positions relativized

. . . .

. . . .

. . . .

379 380 381 382

8.1. 8.2. 8.3. 8.4.

Traditional classification of Russian linguistic varieties Occurrences of kanceljarit and prostorečie in RRR98 . Relative constructions for which RRR98 was searched The frequency of relative constructions in RRR98 . .

. . . .

. . . .

388 392 398 399

List of tables xvii

8.5. The syntactic positions relativized by each construction in RRR98 . . . . . . . . . . . . . . . . 8.6. The occurrence of C11 (asyndetic clause clusters) in RRR98 . 8.7. Occurrences of C1–C10 vs. C11 . . . . . . . . . 8.8. Correlating varieties and constructions . . . . . . . 8.9. Parameters accounting for the use of relative constructions in RRR98 . . . . . . . . . . . . . . . . . 8.10. Informants in RRR98 . . . . . . . . . . . . .

400 401 411 412 414 416

List of abbreviations

The following abbreviations are organized into two lists: the first one contains abbreviations for linguistic terms, the second one abbreviations for languages. They are used both in the flowing text and in the examples. Linguistic terms 1 2 3 A ABL ABS ACC ADV AdvP ALL AH CL CL.ADV COM COMP COND CONN CP DAT DET DEM DO DP DU ERG F FUT G G-N GEN ILL INF

first person second person third person agentive subject ablative absolutive accusative adverb adverb phrase allative Accessibility Hierarchy clitic pronoun clitic adverb comitative complementizer conditional (particle) connector complementizer phrase dative determiner (±definite) demonstrative direct object determiner phrase dual ergative feminine future gender gender and number genitive illative infinitive

INS IO L1 LOC M MC N NMLZ NOM NP NT Ø O OBL OCOMP PAR PART PL POSS PP PREP PRO PRO.ADV PRP PRS PST PTCP Q RC REFL REL RPAR

instrumental indirect object first language locative, place circumstantial masculine matrix clause noun, substantive nominalizer nominative noun phrase neuter zero-marker patient oblique object of comparison particle partitive (case) plural possessive prepositional phrase preposition pronoun, pronominal pronominal adverb prepositive (case) present tense past tense participle question particle/marker relative clause reflexive relative element relative particle

xx List of abbreviations RPRO S SG SoA SRE

relative pronoun intransitive subject singular state of affairs specialized relative element

SU TEMP V VP

subject time circumstantial verb verb phrase

ITA LAT LIT LSO LTV MAC MAL NOR OTK PER POL POR RUM RUS SLK SLN SPA SWE TUR UKR USO

Italian Latin Lituanian Lower Sorbian Latvian Macedonian Maltese Norwegian Old Turkic Persian Polish Portuguese Rumanian Russian Slovak Slovenian Spanish Swedish Turkish Ukrainian Upper Sorbian

Languages AGR ALB BAS BCS BLR BUL CAT CSL CZE DAN DUT ENG EST FIN FRE GAG GAL GER GRE HUN ICE IRI

Ancient Greek Albanian Basque Bosnian/Croatian/Serbian Belarusian Bulgarian Catalan Church Slavonic Czech Danish Dutch English Estonian Finnish French Gagauz Galician German Greek Hungarian Icelandic Irish

An ‘o’ before a language abbreviation indicates reference to earlier stages of that language.

Introduction

Non-standard varieties have been for a long time neglected in linguistics, particularly as far as European languages are concerned. Although there exists a number of studies analyzing the non-standard linguistic space within individual languages, studies investigating non-standard syntax from a cross-linguistic perspective are still a small minority. This work analyzes a specific aspect of syntax, i.e. relative constructions, in a sample of 36 European languages within an extended functional framework. Compared to their standard counterparts, non-standard varieties often display different or additional relative clause formation strategies. In this work these strategies will be the object of a cross-linguistic analysis, in order to highlight the properties shared by relative constructions in European nonstandard varieties and compare them with their standard counterparts. Since this is not a primarily sociolinguistic study, it is based on a very broad notion of ‘non-standard’. ‘Non-standard’ is defined as the linguistic space embracing all varieties that are regarded as sociolinguistically marked in reference grammars. This definition comprises both the geographical and social continuum as well as register variation within a language. In this very heterogeneous space the researcher is most likely to encounter linguistic features that have been banned from standard language. In fact, non-standard language is less constrained in the use of certain syntactic elements; moreover, in non-standard varieties some constructions persist which have developed in parallel with standard constructions, but were not accepted into standard language. Retrieving reliable data on non-standard constructions can sometimes turn into a difficult task, as reference grammars often neglect these constructions and informants tend to stigmatize them. Still, they are attested and well alive, as corpus-based studies and occurrences in language corpora and in the World Wide Web show. For the purposes of this work data were gathered from multiple sources, none of which was excluded a priori. Future studies may assess the distribution of the constructions examined within individual languages, something that has been attempted in the last two chapters of this work.

2 Introduction This study examines non-standard relative constructions from a threefold perspective: namely, a typological, a sociolinguistic and a diachronic one. They are much more interwoven than may seem at first glance. A purely typological analysis brings to light a number of syntactic structures, whose presence and use in a language can be accounted for through functional principles. Still, these principles may be invoked to account for the use of relative constructions in general, irrespective of the fact that they are pigeonholed as standard or non-standard. As a matter of fact, the distinction between standard and non-standard is grammar-external and is of a mainly sociolinguistic nature. Relative constructions can be regarded as linguistic variants and attributed a sociolinguistic status within the diasystem of each language. On this basis, one can compare the sociolinguistic status of relative constructions in languages included in the sample and compile a ‘scale of standardness’ showing which constructions are preferably regarded as standard and which as non-standard. This scale also helps uncover the actual principles underlying the distinction between standard and non-standard relative constructions – a distinction which, in turn, is closely connected with the development of a language over time. So, the diachronic perspective is justified here inasmuch as it offers valuable insight on the present-day linguistic situation and provides an explanation for the difficulties linguists are faced with when they try to classify relativization strategies in European languages. The language sample was intentionally limited to European languages. This is essentially due to three reasons: 1. This work is intended as a contribution to Eurolinguistics, a linguistic subdiscipline that despite being quite young has been noticeably expanding for the past two decades. 2. On account of their history, European languages nowadays display a very rich diasystem: still, this system is usually insufficiently mirrored in traditional typological studies, since they single out only one variety per language, and this is usually the standard one. 3. Dealing with non-standard varieties requires more than reference grammar-based knowledge of the languages under investigation. On account of my linguistic formation I am better acquainted with European languages than non-European ones – hence my decision to focus on Europe. This work is divided into three parts. Part 1 is devoted to theoretical and methodological issues. In Chapter 1 the theoretical foundations of this investigation will be illustrated: basic definitions will be introduced and the approach adopted will be described. Then, the term ‘non-standard’ will be

Introduction

3

discussed; finally, the aims of the work will be enunciated. In Chapter 2 the sources consulted in order to put together the database will be analyzed and evaluated. In Chapter 3 the parameters selected to classify relative constructions – word order, relative element and the syntactic positions relativized – will be described; then, issues of problematic classification will be covered. Part 2 represents the core of this work, as it contains the cross-linguistic investigation of relative constructions in European languages. As mentioned above, constructions are analyzed from a threefold perspective. In Chapter 4, emphasis will be placed on typological issues: the classification parameters introduced in the preceding chapter will be used to pinpoint which constructions are attested in non-standard. An attempt will be also made to account for the existence and use of these constructions on the basis of functional principles. Since a functional account of the difference between standard and non-standard constructions will not prove to be satisfactory, in Chapter 5 the relativization strategies described in Chapter 4 will be considered from a sociolinguistic perspective: each construction will be regarded as a linguistic variant occupying a position in the diasystem of a language. After establishing the sociolinguistic status of relative constructions in individual European languages, a scale of standardness will be formulated that will help to ascertain which strategies are preferably regarded as standard and which ones as non-standard. The position of relative constructions on the scale will be explained through a number of principles which served as guidelines in the formation of standard language. The latter point entails a closer look at the diachronic development of non-standard relative constructions in European languages; so, in Chapter 6 the evolution of the ‘standard vs. non-standard’ opposition will be tentatively reconstructed on the basis of Auer’s (2005b) diachronic-sociolinguistic model. The third part of this work is devoted to language-specific case studies, with the purpose to investigate the actual incidence of standard and nonstandard relative constructions, i.e. how frequent they are. The two case studies are corpus-based: the first one is concerned with presentative constructions in German and is based on data from online discussion forums (Chapter 7); the second one focuses on the use of relative constructions in a Russian corpus recorded in the 1980s in Vorkuta, a town in North-Eastern European Russia (Chapter 8). The corpus is quite interesting as social variation is fully represented in it, so that we are able to analyze a crosssection of the Russian linguistic diasystem.

4 Introduction It is my sincere wish that this study, particularly the third part, may encourage other scholars with different mother tongues to engage in the thematic spectrum covered in this work and to discuss, criticize and refine the claims that were made here. Only in this way are we likely to attain an increasingly accurate and full-fledged description of this puzzle – European languages. Remarks and conventions: 1. Non-English examples are glossed throughout the work on the basis of the Leipzig Glossing Rules (URL: http://www.eva.mpg.de/lingua/ resources/glossing-rules.php, last access: June 30th, 2011) with some slight modifications. In order to avoid excessive burdening of glosses, the metalanguage (English) was mostly relied on; morpheme glossing was added when this contributed to better comprehension of the construction under investigation. 2. For languages that do not use the Latin alphabet, the original form is given first, followed by the transliteration into the Latin alphabet. Only the line containing the transliteration is glossed. Examples from Slavic languages using the Cyrillic alphabet are transliterated according to the scientific transliteration adopted in the Enzyklopädie des Europäischen Osten (URL: http://wwwg.uni-klu.ac.at/eeo/Slawische_ Sprachen_Umschrift, last access: June 30th, 2011). As for Modern Greek, the phonologic transcription used in Joseph and PhilippakiWarburton (1987) was adopted. When examples were already transliterated in the source text, the transliteration system of the original was preserved. 3. Passages quoted from non-English references are not translated if the original language of the reference is French or German. All other quotations are translated. 4. The dissertation on which this work is based was completed in summer 2009. Literature on relative constructions appeared ever since could be considered only in part.

Part 1 Theoretical and methodological premises

In Part 1 the theoretical foundations of the whole work will be laid. Chapter 1 will have a mostly definitory character. Two notions central for the following analysis, ‘relative construction’ and ‘non-standard’, will be introduced. Then, background information on the state of the art and the theoretical approach adopted in this work will be provided. Finally, the aims of the study will be formulated. Retrieving reliable data on non-standard varieties may sometimes turn into a hard task. Linguistic research on the European area usually pays great attention to standard varieties, which, as a consequence, are best documented through reference grammars and, for quite a good number of languages, through one or more reference corpora. Instead, data on nonstandard varieties are often scarcely available or scattered among several publications which are not always easy to access. Hence, it seems reasonable to devote an entire chapter (Chapter 2) to the sources from which data on non-standard relative constructions were drawn: each source will be treated separately, evaluating its degree of accessibility and reliability as well as the pros and cons it has. After collecting the data and organizing them into a database, the next step was to classify them (Chapter 3). Classification models employed in typological literature on relative clauses will be presented; they are usually based on parameters. Three parameters – word order, relative element and syntactic positions relativized – will be picked out as they prove to be most relevant for classifying the relative constructions in the European area. Then, it will be ascertained how suitable these parameters are to catch the sometimes minimal structural differences in the constructions attested in the database. As for word order and the syntactic positions relativized, existing proposals appear to be comprehensive enough; instead, for relative elements an own proposal will be formulated. Since every model is only an attempt to grasp linguistic reality, we will discuss a number of classification problems which arise when the proposed model is adopted.

Chapter 1 Theoretical background

In this chapter basic definitions about the object of study will be first introduced (section 1.1); then the theoretical framework on which the present investigation is based will be described (section 1.2). In section 1.3 the debate on the terms ‘standard’ and ‘non-standard’ will be covered and an operational definition of ‘non-standard’ will be formulated. Finally, after reviewing literature on relative clauses in European languages (section 1.4), the aims of the work will be formulated (section 1.5). 1.1. Basic definitions This section will be devoted to the notion of RELATIVE CONSTRUCTION. Following Goldberg, a construction is defined here as “form and meaning pairing” (Goldberg 2006: 3). Constructions are learned via language acquisition – both spontaneous and conditioned – and can be set on a continuum according to their degree of idiomaticity (=schematic, recurrent, fixed character). The poles of this continuum are constituted by lexical items, which show a high degree of idiomaticity, and syntactic constructions, which display greater variability (Croft 2001: 16–18). The form of a syntactic construction is its morphosyntactic structure, which can be represented through a formula; its meaning is the function it has in speech. In order not to complicate things, the term ‘relative construction’ will be used to refer both to a construction as a whole and to the morphosyntactic structure that encodes a relative relation. Instead, the term ‘relative relation’ will refer exclusively to the ‘meaning’-side of a construction. When reference to the ‘form’-side of a relative construction is to be emphasized, the term ‘(morphosyntactic) structure of the construction’ or ‘construction scheme’ will be adopted (Figure 1.1). The ‘meaning’-side of relative constructions, i.e. relative relations, and their ‘form’-side, i.e. their morphosyntactic structure, will be at issue in 1.1.1 and 1.1.2 respectively.

Basic definitions 7

FORM: (morphosyntactic structure of the) relative construction

RELATIVE CONSTRUCTION

MEANING: relative relation

Figure 1.1. Relative constructions: terminology.

1.1.1. Relative relations In functional linguistics, a clause is syntactically defined as “any syntagm containing one predication” (Lehmann 1988: 182). The semantic-pragmatic correlate of a clause, i.e. the situation, event, process or action coded in the clause, is designated in various ways. Following, among others, Croft (2001), Cristofaro (2003: 25) and Mauri (2008: 32), the term STATE OF AFFAIRS (SoA), will be adopted: it can be regarded as a hyperonym for all previously mentioned terms and is neutral enough as to bear per se neither a static nor a dynamic connotation. Two SoAs may be put into relation with each other to form a complex SoA. The functional (i.e. “notional, cognitive, semantic/pragmatic”, Cristofaro 2003: 2) relationship between the two SoAs can be of different nature: one SoA may specify the circumstances under which the other SoA took place, or it may convey the reason bringing about the other SoA. RELATIVE RELATIONS are a subset of all possible relations between SoAs. They are defined them as follows: 1. Relative relations involve two SoAs, a main one and a secondary one. 2. The two SoAs share a participant, which has a (semantic/pragmatic) role in both SoAs. 3. The secondary SoA (which will be referred to as ‘qualifying SoA’) qualifies a participant of the main SoA. This can happen in two ways: a. the qualifying SoA identifies a participant of the main SoA within a range of possible referents (RESTRICTIVE relative relation);

8 Theoretical background b.

the qualifying SoA supplies further information on a participant of the main SoA (NON-RESTRICTIVE or appositive relative relation). The above definition of relative relation elaborates on Smits (1989: 42), Dasinger and Toupin (1994: 459) and Cristofaro (2003: 195). The two kinds of relative relation are exemplified in (1.1) and (1.2). The qualifying SoA is given in square brackets. (1.1)

The woman [we met yesterday at the movies] is Sean’s sister.

(1.2)

Sean’s sister, [whom we met yesterday at the movies,] is a beautiful woman.

In (1.1), the qualifying SoA identifies a particular woman within the range of all women: only the woman we met yesterday at the movies is Sean’s sister. In (1.2), the qualifying SoA gives additional information on Sean’s sister. Restrictive and non-restrictive relations also differ from an informational point of view. In restrictive relations, a shared piece of information is used to select one referent among others, since the referent had not been unambiguously identified beforehand. In non-restrictive relations, a piece of information – either shared or non-shared – is associated with a previously identified referent. ‘Shared’ means here that the speaker assumes that the relevant piece of information “is known or accessible to the hearer, or otherwise unlikely to be challenged as controversial new information” (Givón 1990: 646). This characterization of the shared piece of information may sound somewhat vague. Still, it is not possible to determine whether the speaker has the right to assume that the hearer has access to the piece of information without sufficient knowledge of the context in which the sentence is uttered. At best, the speaker may assume that the hearer is ready to take this piece of information for granted (Lambrecht 1994: 52) – which may not always be the case, as illustrated in (1.11) and (1.12) below. So, in (1.1) the piece of information contained in the qualifying SoA helps the hearer identify a participant of the main SoA. The speakers relies on the fact that the hearer remembers that s/he and the speaker met a woman at the movies the day before. In (1.2), the piece of information contained in the qualifying SoA does not directly contribute to the identification of a participant of the main SoA. The speaker supposes that ‘Sean’s sister’ is referential enough to identify the entity s/he is talking about. In this respect, it is irrelevant whether the piece of information in the quali-

Basic definitions 9

fying SoA is accessible to the hearer. The informational structure of relative relations can be illustrated as shown in Table 1.1 below. Table 1.1. The informational structure of relative relations.

RESTRICTIVE RELATIONS NON-RESTRICTIVE RELATIONS

PARTICIPANT OF THE MAIN SOA non-identified identified

QUALIFYING SOA shared piece of information shared/non-shared piece of information

In typological literature on relative clauses, restrictive relations are commonly seen as prototypical compared to non-restrictive ones.1 Non-restrictive relations are often dismissed as “parenthetical” (Givón 1990: 680) or “markierte Abart” (Lehmann 1984: 279). Cristofaro, who defines subordination functionally as “a cognitive asymmetry […] between two linked SoAs, such that the profile of one of the two overrides that of the other” (Cristofaro 2003: 33), proves that in non-restrictive relations there is no cognitive asymmetry between the SoAs, so they are no instances of subordination. As a consequence, Cristofaro (2003: 195–196) excludes them from her cross-linguistic analysis of subordinate constructions: in fact, she sets ‘subordination’ as a basic criterion to establish which constructions she will consider in her analysis and which ones she will disregard. Instead, in this work both kinds of relative relations will be taken into consideration since they both consist of two SoAs one of which qualifies a participant of the other one. This issue needs to be discussed further. Relative relations are traditionally regarded as instances of subordination. In functional literature on clause linkage, subordination and coordination have been defined in different ways. In coordination constructions the two clauses are seen as independent of each other, while in subordination constructions one clause is dependent on the other. Cognitively, this is explained – for instance in Croft (2001: 328–335, 349–351) – in terms of foregrounded vs. backgrounded information. Equally foregrounded SoAs are encoded by coordination constructions; if one of the two SoAs is backgrounded, it is encoded in a subordinate clause, whereas the foregrounded SoA is encoded in a main clause. Cristofaro (2003: Ch. 2) finds this proposal unsatisfactory as no sufficiently objective, context-independent criteria exist to distinguish between fore- and backgrounded information: she prefers to speak of ‘cognitive prominence’. As the function of a sentence is to designate some

10 Theoretical background SoAs, if none of the SoAs is prominent we have a coordination relation; if one designated SoA “has cognitive prominence […] over the whole sentence” (Cristofaro 2003: 30), then we have a subordination relation. A speaker grants cognitive prominence to the SoA(s) which s/he wishes to communicate to the hearer, i.e. through which s/he aims to perform a speech act. A speech act – or, more precisely, its illocutionary force – is open to be challenged by the hearer. So, if in a complex construction both SoAs exhibit illocutionary force, we have a coordination relation. If one of the two SoAs lacks illocutionary force, we have a subordination relation. In this case the SoA displaying no illocutionary force is dependent on the other one. The presence of illocutionary force in an SoA can be brought to light through a number of tests. Different tests are applied according to the speech act that the speaker performs through his/her utterance. Following König and Siemund (2007), it is assumed here that three prototypical kinds of speech acts exist across languages: assertive, interrogative and imperative ones. An assertive speech act aims to deliver some information to the hearer; an interrogative speech act aims to obtain some information from the hearer; an imperative speech act instructs the hearer to carry out an action. It must be pointed out that a correspondence between the three kinds of speech acts and declarative, interrogative and imperative sentence types cannot be assumed. For instance, the sentences “Can you pass me the bottle?”, “You’d better tell me the truth.”, are morphosyntactically interrogative and declarative respectively; still, both are used to perform imperative speech acts. In the case of interrogative and imperative speech acts, the presence of illocutionary force in the SoAs can be tested by changing the SoAs into assertions and checking whether the resulting utterances can function as appropriate answers or reactions (Mauri 2008: 42–43). Two examples are provided below. (1.3)

A: Are you going to the movies or are you staying at home tonight? B’: I am going to the movies. B”: I am staying at home.

(1.4)

A: Look for the woman we have to meet at the movies! B’: I will look for her. B”: *We have to meet her at the movies.

Basic definitions 11

In (1.3), both answers are valid: this means that both SoAs display illocutionary force. In Mauri’s (2008: 41) words, both SoAs are AT ISSUE. In (1.4), the only appropriate reaction is the one asserting the first SoA. So, the first SoA displays illocutionary force and is the only one at issue. Assertive speech acts can be tested using the so-called ‘assertiveness tests’ (Cristofaro 2003: 32): sentential negation, sentential questioning and tag-questions. An example from Cristofaro (2003) is quoted below. (1.5)

Alarms ringing, the burglar fled. (Cristofaro 2003: 32)

(1.6)

SENTENTIAL NEGATION: ‘It is not the case that, alarms ringing, the burglar fled.’ SENTENTIAL QUESTIONING: ‘Is it the case that, alarms ringing, the burglar fled?’ TAG-QUESTION: ‘Alarms ringing, the burglar fled, didn’t he?’; *‘Alarms ringing, didn’t they?, the burglar fled.’

In (1.6), the initial sentence (1.5) is first negated. What is negated is the fact that the burglar fled, not that the alarms were ringing. The other two tests – sentential questioning and tag-questions – aim to change the illocutionary force of the sentence. Again, what is questioned is the burglar’s fleeing, not the ringing of the alarms. So, only the second SoA in (1.5) displays illocutionary force. If the terms ‘coordination’ and ‘subordination’ as defined above are applied to these examples, (1.3) is an instance of coordination, while (1.4) and (1.5) are instances of subordination. In a coordination relation both SoAs are at issue; in a subordination relation, one SoA (the MAIN SOA) is at issue, the other one (the DEPENDENT SOA) is not. Applying the assertiveness tests to relative relations, Cristofaro found that restrictive relative relations are instances of subordination, whereas non-restrictive relative relations are not. This is because the qualifying SoA in non-restrictive relations actually displays some illocutionary force. In the following examples from Cristofaro (2003: 195) sentential negation is applied to the assertive speech acts in (1.7) and (1.9). (1.7)

Those CDs [you gave me] are really fabulous.

(1.8)

It is not the case that those CDs you gave me are really fabulous.

12 Theoretical background (1.9)

They went to a number of Bach concerts, [for which they had booked tickets several months in advance.]

(1.10) It is not the case that they went to a number of Bach concerts, for which they had booked tickets several months in advance. What is being negated in (1.8) is that the CDs are fabulous, not that the speaker gave them to the hearer. Instead, in (1.10) the negation may regard the fact that they went to the concerts, or that they had booked tickets in advance, or even both SoAs (Cristofaro 2003: 195). It might be useful to specify that the tests used to check the presence of illocutionary force in the SoAs are usually based on the hearer’s expected reaction from the speaker’s point of view. The speaker builds the sentence according to his/her own communicative purpose; s/he decides which SoA(s) are to be at issue and which are not. Since a speech act may be successful or not, in a real communication situation the hearer may either react according to the speaker’s expectations or ignore them and challenge, for instance, the SoA which the speaker intended as not at issue. So, possible reactions to (1.4) and (1.7) are reported in (1.11) and (1.12). (1.11) A: Look for the woman we have to meet at the movies! B: Why, do we have to meet a woman at the movies? (1.12) A: Those CDs [you gave me] are really fabulous. B: Why, did I give you any CDs? The possibility of challenging/asserting the dependent SoAs does not entail that they originally displayed illocutionary force, though. Speaker B’s reactions in (1.11) and (1.12) actually do not question the qualifying SoA, but its informational status. Speaker A chooses to qualify the woman as the one ‘we met yesterday at the movies’ and the CDs as those ‘you gave me’ because s/he supposes that these pieces of information are accessible to the hearer. Still, the hearer may not remember having met a woman at the movies or having lent any CDs. What B is challenging is the presupposed character of the piece of information contained in the dependent SoA. So, testing the illocutionary force of the SoAs involved in a relative relation, as Cristofaro does, proves to be useful to distinguish between restrictives and non-restrictives in context-independent terms. There is only one hurdle with this testing methodology, which, although not affecting its validity in principle, suggests that assertiveness tests should be integrated

Basic definitions 13

with additional semantic, pragmatic and/or contextual information when they yield ambiguous results. Cristofaro claims that assertiveness tests are independent of the structural features of any given sentence in the relevant language. […] All that is required is that a language have some means to challenge the content of the sentence. These means need not be structurally the same cross-linguistically. (Cristofaro 2003: 32–33)

Cristofaro applies assertiveness tests to sentences from many different languages. As her sample consists of some 100 languages, she is unable to carry out the tests using the specific devices that each language possesses to challenge parts of a sentence; instead, she has to rely on the English translation equivalents of the sentences and to test them through the devices available in English. This is per se methodologically acceptable, as we are to presume that an accurate and reliable translation preserves, if not the same morphosyntactic structure, at least the same conceptual structure of the original sentence (Cristofaro 2003: 40–47; Mauri 2008: 43–44). Still, it may happen that the application of language-specific tests to the original sentence and to its English translation equivalent may lead to discrepancies. For instance, when testing the English translation equivalents of the Turkish assertions in (1.13) and (1.14), the former sentence turns out to encode a restrictive relative relation and the latter a non-restrictive one, as the results of sentential negation and tag-questioning in (1.15) and (1.16) show. The sentences are based on an example quoted in Auer (1990: 279). (1.13) Onun kardeşi [on yıldan beri TUR his/her brother.POSS.3SG ten year.from since Tarabya’da kapıcılık eden] adam. Tarabya.in porter.activity do.PTCP.PRS man ‘His/her brother is the man who’s been working for ten years as a porter in Tarabya.’ (1.14) Onun kardeşi [on yıldan beri TUR his/her brother.POSS.3SG ten year.from since Tarabya’da kapıcılık eden] Ali Osman. Tarabya.in porter.activity do.PTCP.PRS Ali Osman ‘His/her brother is Ali Osman, who’s been working for ten years as a porter in Tarabya.’

14 Theoretical background (1.15) ‘It is not the case that his/her brother is the man who’s been working for ten years as a porter in Tarabya.’ ‘His/her brother is the man who’s been working for ten years as a porter in Tarabya, isn’t he?’ *‘His/her brother is the man who’s been working for ten years as a porter in Tarabya, hasn’t he?” (1.16) ‘It is not the case that his/her brother is Ali Osman, who’s been working for ten years as a porter in Tarabya.’ ‘His/her brother is Ali Osman, who’s been working for ten years as a porter in Tarabya, isn’t he?’ ‘His/her brother is Ali Osman, who’s been working for ten years as a porter in Tarabya, hasn’t he?’ However, testing the original sentences by means of the specific Turkish devices for sentential negation (doğru değil, ‘it’s not true’) and tag-questioning (değil mi, ‘innit?’) leads to different results, as shown in (1.17) and (1.18).2 (1.17) a. TUR

b.

(1.18) a. TUR

Onun kardeşinin [on yıldan his/her brother.POSS.3SG.GEN ten year.from beri Tarabya’da kapıcılık eden] since Tarabya.in porter.activity do.PTCP.PRS adam olduğu doğru değil. man be.NMLZ.POSS.3SG true not ‘It is not the case that his/her brother is the man who’s been working for ten years as a porter in Tarabya.’ Onun kardeşinin [on yıldan his/her brother.POSS.3SG.GEN ten year.from beri Tarabya’da kapıcılık eden] since Tarabya.in porter.activity do.PTCP.PRS adam, değil mi? man not Q ‘His/her brother is the man who’s been working for ten years as a porter in Tarabya, isn’t he?’ Onun kardeşinin [on yıldan his/her brother.POSS.3SG.GEN ten year.from beri Tarabya’da kapıcılık eden] since Tarabya.in porter.activity do.PTCP.PRS

Basic definitions 15

b.

Ali Osman olduğu doğru değil. Ali Osman be.NMLZ.POSS.3SG true not ‘It is not the case that his/her brother is Ali Osman, who’s been working for ten years as a porter in Tarabya.’ Onun kardeşinin [on yıldan his/her brother.POSS.3SG.GEN ten year.from beri Tarabya’da kapıcılık eden] since Tarabya.in porter.activity do.PTCP.PRS Ali Osman, değil mi? Ali Osman not Q ‘His/her brother is Ali Osman, who’s been working for ten years as a porter in Tarabya, isn’t he?’

In both (1.17) and (1.18), the only SoA to be challenged is the one that is not in square brackets (Fatma Sağır, p.c.). Nothing hints at the fact that the relative construction in (1.13) may have a different pragmatic status than the one in (1.14): in both cases, the square-bracketed SoA does not display any illocutionary force, whereas the other one does. Similarly, if the assertiveness tests are applied to the translation equivalents of the Russian sentences in (1.19) and (1.20), the former turns out to be restrictive, the latter non-restrictive: see the translations proposed in (1.21) and (1.22). (1.19) Лицо входящего в кабинет мужчины врачу показалось RUS странным. Lico [vchodjaščego v kabinet] face enter.PTCP.PRS.GEN.M.SG in surgery mužčiny vraču pokazalos’ strannym. man.of doctor.to seemed strange ‘The face of the man who was entering the surgery looked weird to the doctor.’ (1.20) Лицо входящего в кабинет Ивана Ивановича врачу показалось RUS странным. Lico [vchodjaščego v kabinet] face enter.PTCP.PRS.GEN.M.SG in surgery Ivana Ivanoviča vraču pokazalos’ strannym. Ivan.of Ivanovič.of doctor.to seemed strange. ‘The face of Ivan Ivanovič, who was entering the surgery, looked weird to the doctor.’

16 Theoretical background However, if we use the corresponding Russian syntactic devices to test the assertiveness of the SoAs involved – i.e. sentential negation through ne pravda/ne verno, čto (‘it’s not true that’), sentential questioning through pravda/verno, čto (‘is it true that’), tag-questioning through ne pravda li? (‘innit?’) – then no difference in the assertiveness status of the two sentences emerges: see the Russian original version of (1.21) and (1.22) compared to its English translation. (1.21) a. RUS b.

c.

d.

(1.22) a. RUS b.

c.

d.

Ne pravda, / Ne verno, čto lico [vchodjaščego v kabinet] mužčiny vraču pokazalos’ strannym. ‘It is not the case that the face of the man who was entering the surgery looked weird to the doctor.’ Pravda, / Verno, čto lico [vchodjaščego v kabinet] mužčiny vraču pokazalos’ strannym? ‘Is it the case that the face of the man who was entering the surgery looked weird to the doctor?’ Lico [vchodjaščego v kabinet] mužčiny vraču pokazalos’ strannym, ne pravda li? ‘The face of the man who was entering the surgery looked weird to the doctor, didn’t it?’ *Lico [vchodjaščego v kabinet, ne pravda li?,] mužčiny vraču pokazalos’ strannym. *‘The face of the man who was entering the surgery, wasn’t he?, looked weird to the doctor.’ Ne pravda, / Ne verno, čto lico [vchodjaščego v kabinet] Ivana Ivanoviča vraču pokazalos’ strannym. ‘It is not the case that the face of Ivan Ivanovič, who was entering the surgery, looked weird to the doctor.’ Pravda, / Verno, čto lico [vchodjaščego v kabinet] Ivana Ivanoviča vraču pokazalos’ strannym? ‘Is it the case that the face of Ivan Ivanovič, who was entering the surgery, looked weird to the doctor?’ Lico [vchodjaščego v kabinet] Ivana Ivanoviča vraču pokazalos’ strannym, ne pravda li? ‘The face of Ivan Ivanovič, who was entering the surgery, looked weird to the doctor, didn’t it?’ *Lico [vchodjaščego v kabinet, ne pravda li?,] Ivana Ivanoviča vraču pokazalos’ strannym.

Basic definitions 17

‘The face of Ivan Ivanovič, who was entering the surgery, wasn’t he?, looked weird to the doctor.’ In the Russian original version of both (1.21) and (1.22) the only SoA to be challenged is that the man’s / Ivan Ivanovič’s face looked weird to the doctor, not that the man / Ivan Ivanovič was entering the room. So, Cristofaro is right when she claims that every language has its own devices for testing the illocutionary force of SoAs; still, it cannot be taken for granted that these devices yield the same results when applied to the original sentences and not to their translation equivalents. This, in principle, does not affect the correctness of Cristofaro’s line of thought. However, the choice of English as a metalanguage to test the illocutionary force of the SoAs involved constitutes a questionable point. The morphosyntactic structure of English – and of every other natural language – imposes a number of constraints as to which constructions are grammatically acceptable and what their meaning is. Moreover, each language shows a different pragmatic organization of discourse. This explains why assertiveness tests yield different results when applied to translation equivalents. When such discrepancies come to light, it seems inevitable to recur to the informational-semantic status of the qualified participant in order to establish whether we are dealing with a restrictive or a non-restrictive relation. In particular, the use of a proper name in (1.14) and (1.20) should be sufficient to identify the relevant entity; as a consequence, the qualifying SoA adds further information on this entity and the relation between the two SoAs is non-restrictive. In (1.13) and (1.19) the opposite is the case: the man can only be identified through the qualifying SoA. In other cases, a context-independent decision cannot be made: (1.23) is attested in a Russian dialect. (1.23) Знаёшь ту бабушку-ту, у её сноха прошлогот умерла? RUS Znaёš’ tu babuškutu, [u её you.know that granny DET at her snocha prošlogot umerla]? daughter-in-law last.year died ‘Do you know that old lady whose daughter-in-law died last year?’ ‘Do you know that old lady, whose daughter-in-law died last year?’ ?/* ‘Do you know that old lady? Her daughter in law died last year.’(dialect, Troickij 1968: 53)

18 Theoretical background Troickij (1968) does not provide the context in which (1.23) was uttered. If the speaker is deictically referring to an old lady and his/her finger points in her direction, then one may assume that this is enough to identify her: so, the qualifying SoA adds further information and the relation between the two SoAs is a non-restrictive one. If, on the contrary, the speaker feels that the deictic reference through the demonstrative tu ‘that’ is not enough to identify the woman and s/he needs to rely on the qualifying SoA, then the relation between the SoAs is a restrictive one. In this case, neither the assertiveness tests nor the informational-semantic status of the qualified participant may help to unambiguously clarify whether the qualifying SoA possesses illocutionary force or not. The third translation equivalent of (1.23) has been marked with ‘?/*’. In fact, its acceptability is questionable, because in the original the question mark is set at the end of the sentence. Interpreting the sentence according to this translation equivalent entails regarding it as an instance of asyndetic coordination, in which both SoAs are at issue. Despite the discrepancies discussed above, Cristofaro’s procedure for establishing the assertiveness of an SoA can be in principle applied to the European languages considered in this study, provided that assertiveness tests can be adapted to the language under investigation. Being structureindependent, these tests allow distinguishing between restrictives and nonrestrictives irrespective of their morphosyntactic coding. For instance, example (1.24) consists of two juxtaposed clauses and may be structurally regarded as an instance of coordination. The assertiveness of the two SoAs constituting (1.24) is checked in (1.25) through tag-questioning. (1.24) Напротив живёт / уехала в Крым. RUS Naprotiv živët / uechala v Krym. opposite lives went.F.SG to Crimea ‘The woman living in front of us moved to Crimea.’ (Zemskaja 1987: 55) (1.25) a. RUS b.

*Naprotiv živët, ne pravda li? / uechala v Krym. *‘The woman who lives in front of us, doesn’t she?, moved to Crimea.’ Naprotiv živët / uechala v Krym, ne pravda li? ‘The woman who lives in front of us moved to Crimea, didn’t she?’

Basic definitions 19

Only the SoA uechala v Krym gives a positive answer to the assertiveness test. Thus, we are dealing with a restrictive relative relation. Adopting Cristofaro’s definition of subordination and applying the syntactic tests described above in order to keep restrictives separate from nonrestrictives somehow implies sharing her view that the former are instances of subordination, whereas the latter are not. This, though, does not entail that non-restrictives are discarded from the analysis, as Cristofaro does. In fact, this work does not primarily focus on subordination, but on relative relations. It may well be possible that relative relations do not always involve subordination. Still, this is no reason to exclude a whole subset of relative relations: in this study, subordination is not used as a criterion to establish which constructions are to be included in the analysis. On the contrary, if both restrictives and non-restrictives are considered, one becomes aware of the different ways in which two cognitively related SoAs can be morphosyntactically encoded. This is shown (1.26) to (1.28) below. (1.26) I’m very well acquainted with Lizzy: she lives next door. (1.27) I’m very well acquainted with Lizzy, who lives next door. (1.28) I’m very well acquainted with the woman who lives next door. When assertiveness tests are applied, only (1.28) turns out to be an instance of subordination. In (1.26) and (1.27), both SoAs are asserted, i.e. coordinated. This, however, does not imply that they are functionally equivalent. In (1.27), the second SoA adds further information about a participant of the first SoA. The speaker considers this piece of information to be relevant for the ongoing discourse, e.g. in order to remind the hearer who Lizzy actually is. This may also be true for the second SoA in (1.26), but does not necessarily have to be. The relation underlying the two SoAs in (1.26) may also be a causal one; however, it is not specified. From a structural point of view, the choice of a dedicated structure for expressing relative relations – in (1.27) a clause introduced by a relative pronoun, in (1.14) and (1.20) a particular word order and verbal form – helps the hearer understand that the qualifying SoA is to be interpreted as characterizing a participant of the main one, which is not the case when simply relying on asyndetic coordination as in (1.26). There, the relation between the two SoAs is implicit and the hearer has to infer it himself/herself. So, it can neither be denied that the function of non-restrictive relative relations, i.e. to qualify a referent by adding further information, can be

20 Theoretical background coded in different ways, nor that from a cognitive point of view non-restrictive relations can be instances of coordination. Their specificity resides in their narrow scope: providing further information about a participant is one of the many functions that can be coded through syndetic/asyndetic coordination, whereas it is the primary function of a non-restrictive relative relation. If a speaker chooses to code the relation between two SoAs as a non-restrictive relative one, s/he informs the hearer that one of the two SoAs is meant to qualify a participant of the other one. If a speaker chooses to rely on syndetic/asyndetic coordination, s/he does not explicitly inform the hearer of the relationship between the two SoAs. This is indirectly confirmed by the distribution of non-restrictive relative constructions in texts. Non-restrictives occur more often in conceptionally written texts, whereas in conceptionally spoken texts syndetic/asyndetic coordinate constructions are used instead. This is consistent with the difference between conceptional scripturality and orality (cf. Koch and Oesterreicher 1985, 1990), which will be discussed more thoroughly in 2.1. As a rule, the relations between SoAs are unambiguously expressed in writing; in conceptionally oral texts, speakers rely mainly on inference and the context also helps them recover implicit information. The choice of a coordinate construction over a relative one will be at issue in section 4.3.5. The considerations above may also be useful to explain why – at least in the languages investigated in this study – both restrictive and non-restrictive relations are syntactically encoded through similar constructions. Giacalone Ramat (2006: 121) suggests that this may be due to the fact that the sharing of a participant occurs both in restrictives and non-restrictives. Still, this is also the case when two clauses are simply linked through anaphora, as in (1.26). At the beginning of this section, relative relations were said to consist of two SoAs, one of which qualifies a participant of the other one. This function is common both to restrictives and non-restrictives: so, paraphrasing Giacalone Ramat (2006: 121), we may say that similarity of functions leads to relatedness of forms. This, of course, does not need to be true for each couple of relations involving functional similarity: languages might exist where the two subsets of relative relations are coded differently. In fact, they share a superordinate function, i.e. qualification, but fulfil two different subfunctions.

Basic definitions 21 Table 1.2. An overview of the relations between SoAs analyzed in this work. (A)SYNDETIC COORDINATION 1 2

3

4 5 6

NON-RESTRICTIVE RELATIVES

RESTRICTIVE RELATIVES

one SoA qualifies a participant of the other SoA one SoA provides further information on one SoA identifies a participant of the a participant of the other SoA other SoA among a range of possible referents functions 1–2 are not functions 1–2 are primary/exclusive for this kind primary/exclusive for of relation this kind of relation both SoAs are at issue only one SoA is at issue instances of coordination instances of subordination functionally related functionally and structurally related

The distinctive features of the relations between SoAs examined in this section are summarized in Table 1.2 above. Lines 1 and 2 show that all three kinds of relations share a general function, but have different specific functions. Still, in the case of relative relations, the general and specific functions are primary; in syndetic/asyndetic coordination, instead, these functions are only two of the functions that these constructions may have (line 3). As for the pragmatic status of the relations considered, in syndetic/asyndetic coordination – and, in some cases, in non-restrictive relatives – both SoAs turn out to be at issue and can be regarded as involving coordination. In restrictive relatives – as well as, in some cases, in nonrestrictive relatives – only one SoA is at issue, which makes them instances of subordination (line 4 and 5).3 Finally, syndetic/asyndetic coordination and non-restrictive relatives are functionally related; non-restrictive and restrictive relatives are functionally and structurally related, at least in the languages analyzed in this study (line 6). 1.1.2. Relative constructions Relative relations are linguistically encoded through a number of morphosyntactic structures, which constitute the ‘form’-side of relative constructions. In this study, the ‘form’-side of relative constructions is defined as follows:

22 Theoretical background 1.

Relative constructions are complex morphosyntactic structures consisting of a MATRIX CLAUSE (MC) and a RELATIVE CLAUSE (RC). Relative constructions encode relative relations: the MC encodes the main SoA, the RC the qualifying SoA. 2. The relation between the two SoAs is encoded through a link between the MC and the RC. This link may be explicitly expressed by means of morphosyntactic devices, like dedicated elements (pronouns, particles, affixes), word order, particular verb forms, or different participant coding. If the link remains implicit, it has to be recovered through inference. 3. The participant shared by the two SoAs is explicitly encoded at least in one of the two clauses; in the other clause it may be encoded or not. The elements that encode the shared participant in the MC and in the RC will be referred to as RELATIVIZED ITEMS. 4. The degree of explicitness with which the shared participant can be encoded varies according to the following scale: (inflected) nominal element > (inflected) pronominal element > uninflected element (adverb, particle) > zero. Before providing examples to illustrate this definition, some specifications are in order. The first one regards the term ‘relative clause’. In traditional school grammar, this term is used to indicate a subordinate clause introduced by a relative marker: a relative pronoun, a relative adverb or a relative particle. Since in the world’s languages relative clauses can be introduced also by a zero-marker or formed through non-finite verb forms and affixes, in this work the term is simply used to designate the part of a relative construction that encodes the qualifying SoA. Secondly, the term ‘relativized item’ was chosen because it is neutral enough to include all possible cases: the relativized item can be morphosyntactically encoded through a nominal or a prepositional phrase, a verbal phrase, a clause or a whole sentence. In typological literature on relative clauses, other terms are also used, such as ‘head noun’, ‘antecedent’ and ‘referent’. The first one is partial, because not only nouns can be relativized. It will be used only when the relativized item is actually a noun. The second term is misleading, because the prefix ‘ante-’ suggests that the relativized item is placed before the relative clause, which is cross-linguistically not always the case. The term ‘referent’ presupposes that 1. the relativized item in the MC is more explicit than the relativized item in the RC and 2. the relativized item in the RC is linked through anaphora to the relativized item in the MC. Still, the opposite is sometimes the case: for instance, in (1.32) the relativized item in the RC is more explicit than its

Basic definitions 23

counterpart in the MC. So, ‘referent’ will be used only if conditions 1. and 2. above are actually fulfilled in the construction under investigation. Third, in point 4 in the above definition ‘X>Y’ means ‘X is more explicit than Y’. The scale of explicitness is based on Lehmann’s (1984: 225) scale of anaphora; it will be used again in section 3.2.2, when dealing with resumptive elements. To illustrate the definition of relative construction, (1.1) and (1.2) will be quoted again, in (1.29) and (1.30). (1.29) The woman [we met yesterday at the movies] is Sean’s sister. (1.30) Sean’s sister, [whom we met yesterday at the movies,] is a beautiful woman. The RC is given in square brackets; the non-bracketed part of the construction constitutes the MC. In (1.30), the link between the MC and the RC is made explicit through the relative pronoun whom; Sean’s sister and whom encode the relativized item in the MC and in the RC and are COREFERENTIAL, since they designate the same entity. In (1.29), the link between MC and RC is not made explicit; the relativized item is explicitly encoded in the MC (the woman), whereas in the RC it is not. Cross-linguistically, the link between MC and RC as well as the relativized item can be encoded in various ways. In the examples below, the relativized items in the MC and the RC are marked in bold; the RC is given in square brackets. (1.31) Которы старухи не ушодци, так со старушками поговорить. RUS [Kotory staruchi ne ušodci,] tak REL old.ladies not gone PAR so staruškami pogovorit’. with old.ladies talk.INF ‘Let’s have a talk with the old ladies who haven’t gone (yet).’ lit. ‘Which old ladies have not gone, I/we shall talk with those old ladies.’ (dialect, Avanesov and Orlova 19652: 199) (1.32) Которые вот клиенты есть у меня, и тем я смотрю. RUS [Kotorye vot klienty est’ u menja,] REL PAR customers is at me i tem ja smotrju. PAR those.to I look.after

24 Theoretical background ‘Those customers that are mine, I look after them.’ (Lapteva 1976: 144) (1.33) Которые деньги останутся, положи в кошелёк. RUS [Kotorye den’gi ostanutsja] položi v košelëk. REL money remain put in wallet ‘The remaining money, put it in the wallet.’ (Zaliznjak and Padučeva 1975: 74) In (1.31) and (1.32) the relativized item is explicitly encoded both in the MC and in the RC. In (1.31) it is encoded with the same degree of explicitness by means of a nominal element; in (1.32) it is encoded more explicitly in the RC (nominal element) than in the MC (pronominal element). In (1.33), the relativized item is explicitly encoded only in the RC. The link between MC and RC is signalled in all three cases by means of a dedicated element, i.e. a relative pronoun. (1.34) La foguera [que hi dansaven al voltant] CAT DET bonfire REL CL.ADV they.danced around va cremar durant tres hores. burned for three hours ‘The bonfire they were dancing around burned for three hours.’ (López del Castillo 1999: 89) (1.35) [Beha gaude-n] lagunak erranen dauku. BAS wait we.are-REL friend.DET.ERG will.tell he.has.it.to.us ‘The friend we are waiting for will tell us.’ (Oyharçabal 1989: 68) In (1.34) and (1.35) the relativized item is encoded with the maximum degree of explicitness in the MC; in the RC, it is encoded only in (1.34), through a clitic adverb. The link between the MC and the RC is conveyed through the relative particle que in (1.34), through the affix -n in (1.35). Finally, examples (1.23) and (1.24) will be quoted again, in (1.36) and (1.37) below. (1.36) Знаёшь ту бабушку-ту, у её сноха прошлогот умерла? RUS Znaёš’ tu babuškutu, [u её you.know that granny DET at her snocha prošlogot umerla]? daughter-in-law last.year died

Basic definitions 25

‘Do you know that old lady(,) whose daughter-in-law died last year?’ (dialect, Troickij 1968: 53) (1.37) Напротив живёт / уехала в Крым. RUS [Naprotiv živët] / uechala v Krym. opposite lives went.F.SG to Crimea ‘The woman living in front of us moved to Crimea.’ (Zemskaja 1987: 55) In these two examples, the link between the clauses remains unexpressed and has to be recovered through inference. In (1.36), the relativized item is encoded both in the MC and in the RC, but with a different degree of explicitness. In (1.37) the relativized item is encoded in neither clause: the only hint that is given is that the qualified entity is feminine, as suggested by the personal ending -a of the past tense form uechala. So, (1.37) somehow constitutes a borderline construction type, as both the kind of relation between the SoAs and the entity involved remain implicit: only strong reliance on the context helps recover the missing pieces of information. Looking back at examples (1.31) to (1.37), we are faced with all possible cases: the relativized item can be morphosyntactically encoded in both the MC and the RC, only in the MC or only in the RC. The degree of explicitness with which the relativized element is encoded in the MC and in the RC may range from full explicitness – in both clauses the relativized element is encoded through a nominal element, as in (1.31) – to full implicitness, when the relativized element is encoded in neither clause, as in (1.37). The link between the MC and the RC can be signalled through dedicated elements – relative pronouns, as in (1.30), relative particles, as in (1.34), or relative affixes, as in (1.35) – or remain unexpressed, as in (1.36) and (1.37). Sometimes, the element signalling the link between the two clauses and the element encoding the relativized item in the RC may coincide, as is the case for whom in (1.30): still, this cannot be seen as crosslinguistically prototypical. The issue of relative elements is covered in detail in section 3.2. The cross-linguistic variation of relative constructions has been extensively described in a number of contributions, starting from ‘classical’ studies, like Schwartz (1971), Keenan and Comrie (1977), Downing (1978), Maxwell (1979), Comrie (1981: Ch. 7) and Lehmann (1984), up to more recent ones, like De Vries (2002) and Cristofaro (2003: Ch. 7). Languages differ particularly in the way in which relative constructions are formed. A RELATIVIZATION STRATEGY, or relativization pattern, is one of

26 Theoretical background the possible morphosyntactic patterns through which a relative construction can be formed. In typological literature devoted to RCs, relativization strategies are typically classified according to a number of parameters: 1. word order, i.e. the position of the RC with respect to the MC; 2. relative element, i.e. the element encoding the relativized item in the RC; 3. the syntactic positions that can be relativized by means of a relativization strategy; 4. TAM (=time, aspect, mood) and person changes in the verb of the RC due to relativization, in particular the use of a balanced resp. deranked verb form;4 5. changes in the coding of participants in the RC due to relativization. In this investigation, the first three parameters are adopted to classify relativization strategies in European non-standard varieties. These parameters will be presented and discussed in detail in Chapter 3. 1.2. Theoretical framework In this section, the theoretical framework adopted to analyze relative constructions will be outlined. Relative constructions will be investigated from a typological-functional perspective. Still, as a purely typological-functional approach will prove not insightful enough, this approach will be extended. As a matter of fact, traditional typology has been repeatedly criticized in the past few years. One of the main points of this argument is that typological investigations usually consider standard varieties as described in reference grammars: still, they neglect all varieties below the standard one, which often exhibit different structures (DuBois 1985: 353). This is particularly true when European languages are included in the typological sample: European languages possess rich diasystems comprising several varieties, some of which noticeably differ from their standard counterparts. In response to this criticism, new approaches were proposed, which are primarily concerned with language variation. Their aim is to develop theoretical and methodological guidelines to investigate variation within the more general framework of functional linguistics. In particular, integrative functionalism and the typological-dialectological approach are highly relevant to the present investigation. The traditional typological-functional approach is based on the assumption that variation between languages exists and is significant. So, cross-

Theoretical framework 27

linguistic comparison is a useful instrument for discovering universal patterns of variation. INTEGRATIVE FUNCTIONALISM, in turn, is based on the assumption that there exist language-internal variants. Following Ammon, a language variant can be defined as “a single unit, or form, as it can be isolated by linguistic analysis from speech or writing” (Ammon 2004 2: 274; cf. also section 5.1). Each speaker chooses a variant according to a number of sociolinguistic criteria related to the communicative situation. These criteria are grammar-external: the selection of a linguistic construction is carried out according to the context in which the speaker operates (Bisang 2004: 31). This view is in contrast with approaches such as external functionalism (Croft 1995: 512–516), which interprets the selection of grammatical structures by the speakers as the result of conflicting principles (e.g. iconicity vs. economy). For external functionalism grammar is self-contained, i.e. independent of the language-external context, which does not play any role in grammatical processes. The tenets of integrative functionalism will play a major role in the analysis of relative constructions in European languages. In fact, the areal investigation conducted in Chapter 4 will reveal that the existence of both standard and non-standard constructions can be accounted for by invoking similar functional principles, which, as mentioned above, are grammarinternal. What cannot directly be accounted for through these principles is, on the one hand, the use of one construction instead of another in a particular communicative situation – an issue that will be addressed by the case studies in Part 3 – and, on the other side, the sociolinguistic status of relative constructions. Once relative constructions are not simply regarded as syntactic structures, but as linguistic variants, the question arises of why some of them are pigeonholed as standard, whereas others are not. This question can be answered only relying on grammar-external principles. So, the main point of integrative functionalism is that each variant and its distribution within the diasystem of a language are relevant to linguistic analysis. One successful application of integrative functionalism can be seen in the TYPOLOGICAL-DIALECTOLOGICAL APPROACH, which was first described in Kortmann (2002). Kortmann states that dialectology traditionally focuses on varieties within a single language, i.e. on language-internal variation, and is mainly concerned with phonology and lexis rather than morphosyntax. Questionnaires and interviews with so-called NORM (i.e. non-mobile, old, rural, male) informants are mostly used as data sources. Additionally, dialectological findings are hardly ever integrated within a possibly more insightful and explanatory cross-linguistic framework, as

28 Theoretical background dialectology mainly pursues descriptive aims. On the other hand, traditional typology concentrates on morphosyntactic variation across languages and mainly gathers data from descriptive grammars: as a consequence, typological studies may neglect varieties that are not described in grammars. If we look for possible fields of integration, it becomes soon quite clear that both sub-disciplines can be beneficial to each other. Typology provides dialectology with a well-established reference frame to study morphosyntax: generalizations, universals and tendencies, typological hierarchies make it possible to investigate both the categorization of universal functions of language into varieties situated outside the standard one and, more specifically, a full range of cross-linguistic features on a small and large scale. Vice versa, dialectology provides typology with new research data, which can be searched to discover common dialectal features: this can lead to obtain re(de)fined typological parameters and classifications. Since the typological-dialectological approach was first formulated, there have been noticeable developments in the investigation of non-standard syntax, in particular as far as some Western European languages are concerned: cf. for instance Kortmann (2004), Kortmann, Herrmann, Pietsch, and Wagner (2005), as well as the Edisyn project (European Dialect Syntax, URL: http://www.dialectsyntax.org, last access: June 30th, 2011), led by Sjef Barbiers. This project was carried out from 2005 to 2010 and put together experts from around 20 European languages with the aim of creating instruments – especially corpora and syntactic atlases – for studying dialect syntax. The teams working on each language are composed of a very different number of people, ranging from more than 20 for Norwegian and Swedish to only one for Macedonian. For a state-of-the-art report on research on non-standard varieties in the European linguistic area see also Murelli and Kortmann (2011). In this investigation the main claim of the typological-dialectological approach serves as a starting point: dialectal data can and should be analyzed from a perspective different from the traditional dialectological one. However, at least in its original formulation, the typological-dialectological approach remains a partial application of the main tenet of integrative functionalism that language variation is cross-linguistically relevant. Traditional typological studies focus on the standard variety; typological-dialectological studies concentrate on dialects, i.e. local varieties. Still, language variation shows several dimensions and the geographical dimension is only one of them (see below, 1.3.1). A meticulous analysis of dialects in each of the languages considered cannot be carried out in this work, nor

‘Standard’ and ‘non-standard’ 29

can all other dimensions be examined in detail for each language: this would go beyond the purpose of a cross-linguistic study. Rather, after introducing a basic ‘standard vs. non-standard’ opposition, an attempt will be made to provide a full account of relativization patterns across these two groups of varieties. As mentioned in the Introduction, the ‘standard vs. non-standard’ opposition will be first considered from an areal-typological point of view; still, as will be stated in Part 2, the sociolinguistic and the diachronic components of the constructions analyzed need to be taken into account to explain the use of a construction in a particular language variety or to shed light on some controversies that may arise in the traditional typological classification of relativization strategies. 1.3. ‘Standard’ and ‘non-standard’ In the previous section, the terms ‘standard’ and ‘non-standard’ have been repeatedly mentioned, but never defined. Their definitions will be provided in this section: on the basis of existing literature on this topic, we will try to shed light on the nature of ‘standard’ and ‘non-standard’, proposing an operational criterion to establish which constructions should be focussed on (section 1.3.1); then, since the usefulness of the notion of ‘non-standard’ is far from uncontroversial in present-day linguistics, the relevance of the ‘standard vs. non-standard’ opposition in linguistic studies in general and in this investigation in particular will be briefly discussed (section 1.3.2). 1.3.1. Defining ‘standard’ and ‘non-standard’ One of the most valuable achievements of sociolinguistics is that languages are regarded as consisting of LANGUAGE VARIETIES, i.e. linguistic systems shared and used by a community of speakers (Chambers and Trudgill 19982: 5). In most languages, and in virtually all of the languages of Europe, one of these varieties is regarded as more prominent and is defined as ‘standard’. As the review in Berruto (2007) shows, the term ‘standard’ is quite controversial and is qualified in a variety of ways. One of the reasons for this is that standard languages developed at different times and under different socio-political conditions.5 The conditions under which standard French or standard Spanish began to be codified from the 17th century onwards can hardly be compared with the present-day formulation of

30 Theoretical background Galician or Rhaeto-Romance standard (cf. Auer 2005b). Surely, both cases share the common purpose of asserting the existence of a (unitary) language and of its community of speakers. As it emerges, the notion of ‘standard’ turns out to disclose quite a weighty political component, which is usually ignored in linguistic studies. These studies take the existence of standard varieties for granted and do not make the reader aware of the gap between long-tradition standard varieties and recently formed (or reformed) ones, like Macedonian, Basque or Upper Sorbian. In this regard, see Wingender (2003), who claims that languages can be classified according to their degree of standardization (Standardisierungsgrad). In this study the definition of ‘standard language’ proposed by Peter Auer will be adopted: ‘standard’ is a variety that (a) is orientated to by speakers of more than one vernacular variety […], (b) […] is looked upon as an H-variety and used for writing, and (c) […] is subject to at least some codification and elaboration […] or Ausbau (Kloss 1967). (Auer 2005b: 8)

A brief explanation of the features listed by Auer follows. An H-(=high)variety is a prestigious variety and is opposed to L-(=low)-varieties, which are not considered prestigious (cf. Ferguson 1959 and section 6.5); ‘codification’ means that there exist reference works (grammars, dictionaries, etc.) compiled with the aim of providing indications on how the standard variety should look like; as already mentioned, ‘elaboration’ (Ausbau) entails that the standard variety is or was enriched so as to be used in widely different contexts and genres, in particular in humanities and natural sciences. The first property mentioned by Auer may appear trivial. However, a standard which is not acknowledged by its speakers is simply not a standard. On the other hand, the decision of using a variety as an H-variety (property (b)) is essentially a political one – as stressed also in Wingender (2003) – whereas the task of codifying and elaborating this variety (property (c)) is taken up by some linguistic authorities. Following Chambers and Trudgill (19982: 9), it must be further pointed out that the political situation determines whether languages are heteronomous or autonomous (i.e. roofed by another language or not) in different historical periods and – in the case of dialects – by which standard language a dialect is roofed, i.e. which standard language is orientated to by speakers of a dialectal variety. For instance, Ukrainian was not granted the status of ‘language’ for a long time: in the 19th century the tsar prohibited the publications of books and journals in Ukrainian on the basis of a document from the Imperial Acad-

‘Standard’ and ‘non-standard’ 31

emy of St. Petersburg which ‘scientifically’ stated that Ukrainian was a Russian dialect (Kloss 1967: 37). So, the linguistic entity of ‘standard’ appears to be in the middle of a triangle, surrounded by three groups of people who are concerned with its rise, development and maintenance (Figure 1.2). 1. SPEAKERS’ COMMUNITY

STANDARD

2. POLITICIANS (POLITICAL EVENTS)

3. LINGUISTIC AUTHORITIES

Figure 1.2. The social components concerned with the standard variety.

Component 3, i.e. linguistic authorities, is at issue in a number of contributions written by Ammon (cf. Ammon 1995, 20042, 2005). Ammon identifies four groups of linguistic authorities, which he calls ‘social forces’: codifiers, norm authorities, model speakers and writers, and expert linguists (Ammon 20042: 276–277). They issue, respectively, linguistic codices (i.e. authoritative reference works such as grammars and dictionaries), linguistic corrections, model texts and linguistic judgements. In an ideal process of language standardization, a corpus of model texts is adopted as a source for codices, which then become a reference for linguistic corrections and can be challenged by expert linguists. However, the opposite process can also take place: once a codex is established, model speakers and writers can refer to it to compose their texts. So, according to Ammon (20042: 277) a full-fledged standard consists of graphics (an orthographical system), corpus (phonetics, grammar, lexicon and pragmatics) and status (the four social forces controlling the first two components). Ammon does not discuss component 2, i.e. politicians (and political events). This is somehow understandable, as linguistics is not primarily interested in politics. Still, it cannot be denied that the present-day linguistic panorama – both in Europe and worldwide – and, consequently, the bias leading to investigate some varieties more intensively than others actually

32 Theoretical background originates from social and political changes and events. Indeed, although the view is held in contemporary linguistics that dialects and (standard) languages should not be treated differently, this does not prevent standard languages from being much better studied and documented through reference works and corpora.6 This bears unfavourable consequences: in most typological studies, for example, the European language sample is strongly biased towards standard languages. As a consequence, European languages turn out to be typologically underdescribed, as their diasystems are not paid enough attention and structures present in the standard are implicitly assumed to be the only ones attested in the relevant language. Once a variety has been singled out in the language continuum and labelled as ‘standard’, the definition of ‘non-standard’ may simply be thought to result from the negation of its features. So, ‘non-standard’ would consist of non-acknowledged, non-prestigious, non-codified, nonelaborated varieties. However, it is not clear where these varieties are to be located on the sociolinguistic continuum. Additionally, the sociolinguistic continuum itself is modelled as bi-dimensional in the Anglo-American tradition, where scholars usually distinguish between social and geographical dialect continua (cf. Chambers and Trudgill 19982), and as three- or four-dimensional by scholars working on Romance languages (cf. Berruto 1987; Koch and Oesterreicher 1990). Here, the four dimensions, or ‘sociolinguistic axes’, are the diatopic one (representing geographical variation), the diastratic one (representing social variation), the diaphasic one (representing register variation) and the diamesic one (representing medial variation). In this work reference is made to Berruto’s (1987: 21) model of sociolinguistic variation in Italian, although it was drastically simplified so as to meet the manifold sociolinguistic situations of European languages. Additionally, following Berruto’s proposal, the diamesic dimension will not be represented by an axis like the other three dimensions, because it basically reproduces a dichotomy, i.e. ‘written vs. spoken’, and it is somehow superordinated to the other ones (Berruto 1993a: 8–9; Berruto 1993b: 37–55). So, the diamesic dimension will be covered separately in 2.1 on the basis of Koch and Oesterreicher’s (1990) distinction between two kinds of ‘orality’ and ‘scripturality’: this will serve to delimit the linguistic space to search for non-standard relative constructions. Berruto (1987: 21) sets the standard variety at the intersection of the three sociolinguistic axes described above: this highlights its unmarked, neutral character. Hence, the further from the intersection of the three axes a variety is, the more marked in a diatopic, diaphasic or diastratic sense it is. Still, what remains unclear is where the demarcation between standard

‘Standard’ and ‘non-standard’ 33

and non-standard actually lies on each of the three axes. This issue is discussed, among others, in Sgall (1992), Hinrichs (1994), Ammon (1995, 20042), Van Marle (1997) and Löffler (2005). It emerges that the difficulties connected with delimiting the boundaries of standard are related to its very rise and formation and, in particular, to the forces contributing to its creation, codification and maintenance. In spite of their reciprocal influence, it may well be that the four social forces diverge as to whether a construction should be classified as standard or non-standard. Codifiers usually mark non-standard constructions as ‘spoken’, ‘colloquial’, ‘dialectal’, ‘rural’ etc. Norm authorities reject these forms, since they are marked. Model speakers and writers behave differently according to the textual genre (fiction or non-fiction). Expert linguists, instead, behave quite neutrally towards ‘deviating’ forms as their aim is rather to ascertain the sociolinguistic context(s) where the variant occurs. In this respect, different judgements on the sociolinguistic status of constructions may well be expected (cf. the discussion in 5.1). What is certain is that the demarcation between standard and non-standard is flexible and may noticeably vary in time according to what linguistic authorities define as ‘standard’. In this respect, standard and non-standard reveal their relational character in that they both constitute a macrosystem and are to be regarded as two sides of a medal (cf. Hinrichs 1994: 97). Despite the difficulties in delimiting the non-standard linguistic space from the standard one, an operational definition of non-standard is needed here in order to delimit the constructions that will be analyzed in Part 2. It will be assumed that NON-STANDARD is the linguistic space occupying the lower portion of the sociolinguistic axes: in other words, it includes every diatopically, diastratically or diaphasically marked variety. This allows pigeonholing as ‘non-standard’ all constructions that in reference works (grammars, linguistic studies) are filed as ‘colloquial’, ‘regional’, ‘dialectal’, ‘informal’, i.e. as SOCIOLINGUISTICALLY MARKED. This has two advantages. On the one hand, it avoids getting entangled in language-specific sociolinguistic situations, which may differ considerably from language to language. For instance, the term ‘dialect’ does not have the same meaning when applied to different European varieties: English dialects have a different status and a different history from German Mundarten, Italian dialetti or Russian govory. On the other hand, constructions frequently neglected in cross-linguistic studies can be catalogued together, so as to offer an overview of the variation patterns attested outside the standard variety.

34 Theoretical background A final remark is in order. To be precise, the label ‘non-standard’ should be applied not only to the so-called ‘substandard’, i.e. to varieties situated on the lower portion of the sociolinguistic axes, but also to ‘superstandard’, i.e. to varieties situated on the upper portion of the sociolinguistic axes. For instance, poetry language is usually considered elevated and displays peculiarities which do not occur in the standard language. So, strictly speaking, also poetry language is non-standard. However, in order to delimit the database of this investigation and not to lose focus, it seemed preferable to concentrate only on substandard: correlations between substandard and the language of poetry will be cursorily mentioned in section 6.5.3. Additionally, substandard itself proves to be an extremely heterogeneous linguistic space. To overcome this hurdle, non-standard constructions were looked for within a well-defined linguistic domain, i.e. conceptional orality. This issue will be extensively covered in Chapter 2, and in particular in section 2.1. 1.3.2. On the usefulness of distinguishing standard from non-standard in linguistic research Although the opposition between standard and non-standard is commonly taken for granted in present-day linguistics, not every scholar acknowledges it as a valid instrument for linguistic research in general and for functional investigations in particular. Two different positions can be identified: some scholars (e.g. Cheshire and Stein 1997; Weiß 2004) accept the term ‘non-standard’ and actively use it in their studies; others (HennMemmesheimer 1997; Miller and Weinert 1998; Miller and FernandezVest 2006) dismiss it as unnecessarily complicating language description. Both parties produce well-founded reasons for their choice, which is often related to their theoretical approach and object of study. In their discussion on the relationship between vernacular varieties and standard language, Cheshire and Stein (1997) claim that linguistic research should first concentrate on language – on language as a whole – and then on the correlations between linguistic structures and their use in a community of speakers, i.e. on their sociolinguistic features:

‘Standard’ and ‘non-standard’ 35 if we conceptualize syntactic variation a priori in terms of discrete social, geographical or stylistic varieties […] we risk obscuring the nature of the syntactic relationships that exist within a language. […] Once the syntactic variation within a language system has been described we can attempt to identify regional or social correlations at that later stage. (Cheshire and Stein 1997: 10)

As such, the ‘standard vs. non-standard’ opposition is subsequent to the identification of linguistic constructions in a language. However, overtly addressing the issue of non-standard structures can be justified if we consider that one of the main purposes of contemporary research on European languages should be to fill a gap: corpora consist mainly of the speech of people who might be expected to use forms close to the written standard end of the continuum, and corpus linguistics has not yet produced much information about vernacular syntax. (Cheshire and Stein 1997: 3)

Cheshire and Stein (1997) call for better documentation of non-standard varieties, which, in their opinion, have not yet received the attention they deserve. The two scholars also lament the fact that there is too little clarity on the status of these varieties. They make two relevant points: 1. Non-standard is often neglected in linguistic studies, which mostly focus on standard as it is described in grammars. This prevents an allencompassing understanding of language development: constructions considered as non-standard may have a deep-rooted historical background. This issue will be covered in Chapter 6. 2. Nevertheless, too strict a distinction between ‘standard’ and ‘non-standard’ is useless as it does not correspond to linguistic reality. In fact, as mentioned above, the boundaries between standard and non-standard may be very flexible. As a solution, Cheshire and Stein propose that all syntactic constructions existing in a language should be granted equal status; correlations between constructions and sociolinguistic variation can be established later on. Weiß’s (2004) main point is that some data from standard languages should be used carefully in linguistic research since they are the result of conscious language planning and embellishment. Rationality and naturalness do not necessarily coincide: this has often led (and still leads) to exclude constructions which are regarded as illogic. As a consequence, European standard languages partially suffer from inconsistency, i.e. they may sometimes display and sometimes not display features that occur in nonstandard varieties of these languages: in this sense, Weiß does not consider

36 Theoretical background them as good candidates for typological research. One of the most striking examples is double negation: in some languages it is standard (Slavic languages, French); in other it is banned from standard (English, German, Dutch), but is attested in non-standard varieties (cf. Bernini and Ramat 1992; Auer 2004; Weiß 2004). Still, Weiß claims that many of these standard languages, which at the beginning were somewhat artificial,7 eventually underwent a process of renaturalization as they began to be acquired as L1s by children, i.e. to be actively spoken. Thus, their current status can be considered as a sort of compromise between the rationalistic guidelines residing in their formation and the modifications they have been subjected to since they became L1s. The opposite view, i.e. that the opposition between standard and nonstandard only complicates things, is held by Henn-Memmesheimer (1997), Miller and Weinert (1998) and Miller and Fernandez-Vest (2006). HennMemmesheimer (1997: 235) states that when analyzing syntactic structures, all constructions attested in a language should be catalogued as belonging to one linguistic system: in this respect, the dichotomy ‘standard vs. non-standard’ should not be used as an a priori template to categorize language structures. Miller and Weinert (1998) and Miller and FernandezVest (2006) are concerned with spontaneous spoken language. Both standard and non-standard elements are attested therein: what is more relevant, though, is that spontaneous speech exhibits intrinsic properties of its own. The dichotomy ‘written vs. spoken’ would then be prior to ‘standard vs. non-standard’ and the latter would often be shaped by the former: elements which found their way into writing also had a higher chance of being regarded as standard (also see Stein 1997 and section 5.1). The position that best suits the purposes of this study is Cheshire and Stein’s (1997): enumerating the constructions attested in a language should be prior to establishing their sociolinguistic status. While the existence of a construction in a language has little chance to be denied or challenged as objective proof of it can be provided, sociolinguistic judgements on the degree of standardness of a construction may considerably vary depending on the source consulted. In this respect, speakers may have a different opinion than codifiers or norm authorities. Nevertheless, one of the main aims of this study is to describe and classify constructions that are underrepresented in typological studies. As a consequence, a broad definition of nonstandard as ‘sociolinguistically marked’ proved to be best suitable (cf. 1.3.1 above). In first instance, i.e. in Chapter 4, these constructions will be analyzed through functional-typological parameters and principles and compared with constructions that have already been investigated in typo-

Relative clauses in European languages: the state of the art

37

logical studies on relative clauses. In a successive step, i.e. in Chapter 5 and 6, the opposition between standard and non-standard constructions will be explicitly examined, which includes considering the sociolinguistic status of relative constructions. The social force that will be referred to in order to determine the sociolinguistic status of a construction is codifiers: more specifically, codifiers’ judgement as reported in reference grammars and linguistic studies. This, of course, does not mean that different grammars may not contain different judgements and that other linguistic authorities share the judgements contained in grammars (cf. 2.3 and 5.1). Additionally, the references consulted are objectively and subjectively limited: objective limitations have to do with the existence of reference works where a judgement on the sociolinguistic status of relative constructions can be found; subjective limitations have to do with the accessibility of reference works, i.e. whether I managed to consult them, or whether they were written in a language I could understand. To summarize, the distinction between standard and non-standard serves a dual function in the present study: 1. as a parameter for linguistic comparison (which constructions are regarded as standard and/or non-standard?) and 2. as a parameter for linguistic reflection (why are constructions regarded as standard and/or non-standard?). These questions will be answered in Chapter 4 and Chapters 5 and 6 respectively. Finally, it needs to be pointed out that the position taken in this investigation does not imply that Miller and Weinert’s (1998) claims are rejected. It is true that in spontaneous spoken language standard and non-standard elements get mixed and cannot be always separated; still, reversing Miller and Weinert’s line of thought, it could be argued that once it is known which constructions are standard – i.e. which constructions are labelled as standard in reference works – all other constructions may be considered to fall into ‘non-standard’. So, spontaneous spoken language turns out to be a privileged linguistic domain where non-standard constructions can be searched for (in this regard, see 2.1). 1.4. Relative clauses in European languages: the state of the art Before enunciating the aims of the study, literature on relative clauses in European languages will be reviewed, commenting on the accuracy with which relativization strategies are described. Only cross-linguistic studies will be considered here. Language-specific studies will be quoted in the

38 Theoretical background following chapters when relative constructions in individual languages are examined. Existing cross-linguistic studies can be roughly divided into two groups, according to whether they specifically focus on European languages or analyze them within a broader language sample. The following studies are centred on the cross-linguistic analysis of European languages: Kurzová (1981), Smits (1989), Zifonun (2001a), Cristofaro and Giacalone Ramat (2007) and Fiorentino (2007). Kurzová (1981) describes the RC types found in Indo-European languages of Europe on the background of the common Indo-European RCs with inflected pronoun (*yo- and *kwi-/kwo-). After a historical part in which Kurzová describes the structure of the RC in Indo-European, in the typological part she distinguishes RC types according to the relative element introducing them: an inflected pronoun, a relative particle or a zeromarker. When reviewing relative particles, Kurzová also discusses cases of combination between a relative particle and a resumptive pronoun and distinguishes between the relativization of direct and indirect syntactic positions (“direkte/indirekte Beziehung”). Still, she does not adopt Keenan and Comrie’s (1977) Accessibility Hierarchy (see section 3.3). Kurzová also makes sociolinguistic considerations: she specifies the position of relative elements on the continuum ranging from ‘popular’ to ‘written’ (“Achse volkstümlich-schriftsprachlich”, Kurzová 1981: 83). Two drawbacks of Kurzová’s study concern the very small number of examples and the database she used. Examples are kept to a minimum and are never glossed: in more than one case, Kurzová mentions a relative element without providing any evidence for its use. As for the database, Kurzová mostly relies on older linguistic studies and, in very few cases, on grammars. As a result, the data she presents are sometimes very old and can be seen as having only historical value. Smits’s (1989) purpose is mainly a descriptive one: on the basis of the X-bar theory, he describes relative, cleft and pseudo-cleft constructions in Romance and Germanic languages. His database consists of grammars and linguistic studies, both recent and older ones. Smits’s work is named Eurogrammar: still, apart from Rumanian, only West-European languages are considered. Smits defines his sample as a “coherent and linguistically, culturally and economically prominent group of languages” (Smits 1989: v). However, this definition does not fully justify the choice of a restricted sample: among Slavic languages, one can also surely find “linguistically, culturally and economically prominent” languages. Additionally, Smits focuses only on national standard languages (apart from Catalan), because

Relative clauses in European languages: the state of the art

39

“most regional languages have hardly been subjected to systematic syntactic description yet” (Smits 1989: vi): so, he sometimes reports data from colloquial varieties and dialects, but does not pursue this issue in detail. Smits classifies relative strategies according to three parameters: the relative element, the syntactic positions that a strategy can relativize and the use in restrictive vs. non-restrictive contexts. Relative elements are further classified as relative pronouns, relative particles and relative adverbs. For each class, Smits investigates which kind of syntactic positions can be relativized: these are subjects, direct objects and prepositional objects for relative pronouns and relative particles and temporal, local and other circumstantials for relative adverbs. Under the label of LONG RELATIVIZATION Smits analyzes the behaviour of the same relative elements when the relativized item encodes a genitive or is embedded in a subordinate clause dependent on the relative clause: “We speak of long relativization when the relativized constituent is part of a domain (S, NP, PP) inside the RC, rather than part of the RC itself only” (Smits 1989: 82). In his extensive appendix, all relative elements are listed by language and several examples are provided. Smits’s (1989) study is surely quite valuable as he provides several examples of the relative constructions analyzed; still, his Eurogrammar remains a partial work, since he completely ignores Slavic languages. Zifonun (2001a) describes relativization patterns found in German from a typological-contrastive perspective on the background of patterns found in European languages. In order to pigeonhole relative constructions found in languages included in her sample, Zifonun adopts Lehmann’s (1984) cross-linguistic classification of relative clauses. Zifonun’s study is relevant for the purposes of this work because she attempts to describe relativization patterns found in a specific language by means of typological tools. The results may then be compared with contributions from other studies concentrating on different languages, where the same theoretical framework is used. Additionally, Zifonun also discusses constructions that are considered non-standard in German. The only questionable point concerns her sample. For the sake of convenience, Zifonun selects only seven languages as Kontrastsprachen (contrast languages), so that different language families are represented. However, she sometimes quotes examples from languages that are not included in the sample if these languages display structures that may be relevant for the analysis. In this respect, it is not clear which role these languages actually have. Moreover, the great majority of these additional examples are taken from Romance and Germanic languages (English, French, Spanish) as well as from Greek. Slavic lan-

40 Theoretical background guages are significantly underrepresented, the only Slavic language of the sample being Polish. It must be pointed out that Zifonun considers her work as a contribution to the didactics of German: through the contrastive comparison with other languages, learners should become aware of differences between their L1 and German. For this reason, the languages constituting Zifonun’s sample primarily include “Sprachgebiete, in denen das Deutsche intensiv gelehrt und gelernt wird” (Zifonun 2001a: 7). Still, this statement is somewhat questionable: it is not clear on which basis Zifonun assumes that German is taught more intensively in England or Spain than in Italy or Russia. Cristofaro and Giacalone Ramat’s (2007) study is extremely relevant to this work as the two scholars investigate the strategies used to code the relativized item in relative constructions in a sample of 26 European languages. On the basis of Keenan and Comrie’s (1977) [+/–case] distinction (cf. section 3.3.), they analyze how relative elements are cross-linguistically distributed and to which extent each relative element conveys the syntactic role of the relativized item, with special attention to time circumstantials and genitives, two positions that are traditionally studied to a lesser extent. Cristofaro and Giacalone Ramat (2007) summarize the relativization strategies they took into account in a table at the end of their contribution. Two major findings are to be highlighted here: 1. the coding of the relativized item in the relative clause turns out to be related with the recoverability or the relevance of the information conveyed through this item for the ongoing communicative situation; 2. the classification of relative elements may sometimes be controversial on account of diachronic changes which are still taking place in synchrony. A questionable point concerns the comprehensiveness of the table at the end of the contribution: several strategies, both standard and non-standard, do not appear in the list or appear in a somewhat incomplete way. This may be due to the fact that Cristofaro and Giacalone Ramat worked with questionnaires and reproduced in the final table only the constructions that they elicited from their informants. In general, West-European languages seem to be better covered than East-European ones. This leads to somewhat problematic results: for instance, it is claimed that “invariable elements are exclusively found in Western Europe” (Cristofaro and Giacalone Ramat 2007: 84), whereas, as will be shown in the course of this investigation, all East-European languages – with the exception of Estonian and Hungarian – also possess an invariable element (cf. section 3.2.1). Finally, Fiorentino (2007) addresses the issue of the uniqueness of the relative pronoun strategy in European written standard languages from a

Relative clauses in European languages: the state of the art

41

cross-linguistic perspective. She provides evidence from Romance, Germanic and, to a lesser extent, Slavic languages, although she does not define her sample with precision. Corpus evidence is provided for Italian, French and Spanish; evidence for other languages is taken from linguistic studies. In particular, evidence for Slavic languages is mostly taken from the cross-linguistic overview provided in Keenan and Comrie (1979a). As will be discussed below for De Vries (2002), this has some drawbacks: cross-linguistic overviews usually do not grasp the totality of the strategies attested in a language. So, Fiorentino concludes that “the greater part of Slavic languages uses the relative pronoun strategy” (Fiorentino 2007: 271): still, as Part 2 will show, Slavic languages also display several other strategies. Fiorentino’s line of thought develops e negativo: relative pronouns are not frequent in spoken varieties of European languages (typological argument); spoken varieties select a subset of grammatical tools from all those available in a language – and, significantly, do not select relative pronouns as main relativization strategies (grammatical argument); relative pronouns constitute a hurdle for online speech processing since the syntactic role of the relativized constituent must be planned in advance, before the RC is uttered (psycholinguistic argument). Moreover, even when a relative pronoun does occur in spoken varieties, its paradigm is often reduced (Fiorentino 2007: 278). On this basis, Fiorentino hypothesizes that the relative pronoun strategy is found only in European written standard varieties because it best suits the needs of (conceptionally) written language: We put some emphasis on the ‘written’ context because it is in this case that an inflected fronted relative pronoun can prove to be more useful to correctly identify, for example, a nonadjacent head. (Fiorentino 2007: 285)

Fiorentino also provides diachronic evidence for alternative relative constructions (gap strategy, pronoun retention) beside the relative pronoun. In fact, it emerges that a ‘which’-based relative pronoun arose in (written) Romance languages in the 12th century and spread through the written channel across Germanic languages, so the diffusion in Europe of the relative pronoun strategy reflects the ‘sharing’ of a common (written) cultural tradition, and its written origins explain its relative uniqueness, if cross-linguistically considered. (Fiorentino 2007: 285).

Fiorentino’s study highlights some significant points that are discussed in this study, too: in particular, instances of paradigm reduction in European

42 Theoretical background non-standard varieties will be presented in section 4.2.1 and 4.2.4; diachronic evidence for non-standard relativization strategies will be provided in Chapter 6. Additionally, she directly addresses the issue of the existence of alternative, non-standard relativization strategies in spoken varieties of European languages, which is a key theme of this work. However, Fiorentino draws her conclusion on the basis of the development of relative pronouns in West-European languages; she neglects the fact that East-European languages also display a ‘which’-based relative pronoun, which, however, did not follow the development documented for Western Europe. This pronoun is (and was) attested both in written and spoken texts and cannot directly be related to the “sharing of a common (written) cultural tradition” (Fiorentino 2007: 285). In this regard, see also section 6.6. The following studies analyze European languages within a broader language sample: Lehmann (1984), De Vries (2002), Comrie and Kuteva (2005). Lehmann (1984) conducted one of the first all-encompassing crosslinguistic investigations of relative clauses. He studies RCs from a typological, historical and theoretical point of view. In his language sample (Lehmann 1984: 12–13), European languages are mostly granted a secondary position. Lehmann explains that this is in order to try and build a language sample as balanced as possible and to avoid repeating what is already known: Dies erklärt sich durch das Bemühen um typologische Ausgewogenheit; die indogermanischen Sprachen sind ohnehin schon überrepräsentiert. Außerdem sollte dem, was allgemein bekannt ist, nicht mehr Raum als nötig gewidmet werden. (Lehmann 1984: 13 fn. 11)

European non-standard relative constructions are not directly at issue here: Lehmann sometimes specifies whether a strategy is to be found at a nonstandard level, but mostly offers no examples. De Vries (2002) aims to account for cross-linguistic variation in relativization strategies on the basis of a generative reference frame. Before formulating his analysis, De Vries tries to systematize the terminology used in typological studies on relative clauses, as regards parameters like word order or relative element (cf. the synopsis he gives in the Appendix IV, De Vries 2002: 427–431), and reviews existing literature, both functional and formal, on relative clauses (cf. his Appendix III, De Vries 2002: 413–425). In the typological part of his work, De Vries considers relativization strategies in a sample of 172 languages. The constructions attested in these languages are classified according to a number of parameters. The weak point of his work is the database: De Vries does not gather data on

Relative clauses in European languages: the state of the art

43

the 172 languages of his sample from reference grammars, but exclusively from other contributions devoted to relative clauses (which include Downing 1978; Comrie 1981; Lehmann 1984; Keenan 1985; Smits 1989). This is not a drawback per se: still, these studies do not necessarily provide comprehensive data on the languages considered; they only offer evidence supporting the claims made in each study. This bears partially unfavourable consequences. On the one hand, very few examples are provided which are mostly taken from the following languages (with decreasing frequency): Dutch, English, French, German. De Vries also declares that he will not take into account structures found in non-standard varieties (De Vries 2002: 177 fn. 29). Still, Genoese and Schwyzertütsch, two non-standard varieties, are found in the final synopsis. On the other hand, the language sample is not balanced and the synopsis in Appendix II is incomplete, at least as far as European languages are concerned. The following are three examples of this: − De Vries mentions only Finnish among the languages with “participial relatives”, but does not include Hungarian, German, Russian or other Slavic languages. − Among the languages with “correlatives” he mentions Ancient Greek and Old Russian, but leaves out Latin and other languages like Latvian and Russian. − Among the languages using resumptive clitics he mentions Genoese, Slovenian, Welsh, Greek, Czech and Schwytzerdütsch, but neglects the fact that all Romance and Slavic languages display this strategy. De Vries appears to be aware of the limited representativeness of the synopsis: he explicitly states that “statistical conclusions cannot be based on this sample without further processing.” (De Vries 2002: 384). To summarize, De Vries’s study is surely relevant from a theoretical and terminological point of view; however, his typological analysis needs to be integrated and his data should be checked against further evidence. Finally, Comrie and Kuteva’s (2005) overview of relativization strategies in the world’s languages includes 15 European languages. The sample is balanced and includes three Romance, two Slavic, two Germanic, two Celtic and two Ugro-Finnic languages, beside Greek, Latvian, Turkish and Basque. Comrie and Kuteva analyze the strategies that languages use for subject and oblique relativization and focus on the morphosyntactic coding of the relativized item in the RC. They identify four possibilities: relative pronoun, non-reduction, pronoun retention and gap. In their world maps with an overview of the world’s languages, Comrie and Kuteva report only one strategy for each language. However, it is not clear how the choice was

44 Theoretical background made. The “relativization on subjects” map (Comrie and Kuteva 2005: 498–499) shows that the majority of European languages rely on a relative pronoun, but Spanish uses the gap strategy. Actually, also Italian, English, Bulgarian and Greek can relativize subjects using a gap strategy and Spanish can also use a relative pronoun. Perhaps, more strategies could have been reported for each language, or the strategy proving to be most frequent according to language-specific corpus-based studies could have been chosen. Overviews are useful to get a general impression of a linguistic phenomenon in a particular area, but they should be regarded as a starting point for more thorough investigations. The arguments brought forward in this section can be summarized as follows. To start with, the review of cross-linguistic studies on relative clauses in European languages reveals a major drawback, i.e. an incomplete account of the relativization strategies found in individual languages – in particular of strategies that are considered as sociolinguistically marked. That individual languages often exhibit more than one relativization strategy has already been pointed out a century and a half ago by Korš (1877, quoted in Lehmann 1984: 5), who demonstrated it on the basis of Russian examples. Still, the relative pronoun strategy has often been given priority, perhaps because this strategy is typical of European standard written languages, which usually serve as databases. Moreover, relative pronouns constitute a virtually unique phenomenon in the world’s languages (cf. Lehmann 1984: 103–109; Comrie 1998; Comrie and Kuteva 2005; Fiorentino 2007) and, as such, probably attracted the scholars’ attention. This, though, led to underestimating the relevance of other relativization strategies attested in European languages. Secondly, the literature reviewed exhibits a strong West-European bias. Slavic languages are considered less often and are sometimes only partially described. Hence, it will be a major concern of this study to highlight those constructions in Slavic languages that have not been paid attention so far. Thirdly, apart from Kurzová (1981), Lehmann (1984), Zifonun (2001a) and Fiorentino (2007), who mention that some constructions are typical of non-standard varieties, other investigations do not make any distinction between standard and non-standard constructions. As stated in 1.3.2, this is per se justified: all constructions should a priori be granted equal status in each language. However, this may lead to simply excluding or not sufficiently describing the constructions attested below the standard, which are not always mentioned in reference grammars and which informants may more or less consciously censor. As a consequence, emphasis will be

Aims of the study 45

placed on non-standard constructions with the aim to offer an account as comprehensive as possible of the constructions found in individual languages. Finally, an attempt will be made not to gather data from a single source, but to rely on different sources, which may all help to put together the mosaic of relative constructions in European languages. The sources constituting the database of this work are described in section 2.3. 1.5. Aims of the study As emerged in the previous section, there is no cross-linguistic study specifically focussing on non-standard relative constructions in European languages. Even recent studies do not provide a satisfactory overview of the manifold relativization strategies which can be found in these languages. Additionally, cross-linguistic studies typically do not address the issue of the sociolinguistic status of relative constructions. So, this investigation will be devoted to two major research issues: 1. constructions that are regarded as non-standard in European languages will be surveyed with the aim to account for their non-standard status on sociolinguistic and diachronic terms. 2. possible factors explaining why a speaker chooses a relative construction instead of another will be examined. The first issue will be covered in Part 2. Each of the chapters constituting this part of the work has specific aims: a. Typological aims (Chapter 4): − to describe relative constructions which are considered nonstandard in European languages; − to understand whether and in which respect relativization patterns found in non-standard differ from the patterns found in standard; − to ascertain whether functional principles can be invoked to account for the distinction between standard and non-standard relative constructions. b. Sociolinguistic aims (Chapter 5): − to ascertain whether sociolinguistic principles are more suitable than functional ones to account for the ‘standard vs. nonstandard’ opposition as far as relativization strategies are concerned; − to attribute a sociolinguistic status to each construction on the basis of judgements contained in reference works and to set up a

46 Theoretical background scale of standardness for relative constructions in European languages. c. Diachronic aims (Chapter 6): − to look for diachronic evidence of non-standard relative constructions; − to describe the historical development of the distinction between standard and non-standard relative constructions on the basis of the model proposed in Auer (2005b). The second major research issue is covered in Part 3. Here, two languagespecific corpus-based case studies (Chapter 7 and Chapter 8) will be proposed, with the following aims: a. to establish the actual incidence of standard and non-standard relative constructions, i.e. how frequent standard and non-standard constructions are; b. to account for the distribution of standard and non-standard relative constructions within the analyzed corpora through factors like the context in which these constructions are adopted (Chapter 7), the sociolinguistic status of the construction, or the speakers’ sociolinguistic features (Chapter 8).

Chapter 2 Data sources

This chapter is concerned with data collection. It is divided into three sections. The purpose of section 2.1 will be to delimit the domain of linguistic varieties in the European sample which were searched for non-standard relative constructions. In 2.2, a few words will be spent on the choice of the language sample. Then, in 2.3 the individual sources consulted to collect data on non-standard relative constructions will be reviewed. 2.1. Looking for non-standard constructions As mentioned in 1.3, in this investigation emphasis will be placed on the non-standard linguistic space in European languages. Concretely, this research is to be located between traditional typological studies and language-specific typological-dialectological studies (Figure 2.1).

Traditional typological studies Standard varieties Nonstandard varieties

L1

L2

L3

etc.

Cross-linguistic typological-dialectological studies

Language-specific typological-dialectological studies Figure 2.1. Levels of analysis in typological studies.

In Figure 2.1, European languages (L1, L2, L3…) are represented as icebergs. The tip is constituted by standard varieties; under ‘sea level’ we find sociolinguistically marked varieties. Traditional typological studies prevailingly focus on the tip of the iceberg, i.e. European standard varieties;

48 Data sources instead, language-specific typological-dialectological studies concentrate on individual languages. Cross-linguistic typological-dialectological studies take an intermediate position: they cannot fully account for languagespecific variability, but pick data from different non-standard varieties and compare them from a typological perspective. Now, the question arises as to which types of linguistic data provide evidence for non-standard constructions. In other words, from which domains of the linguistic space can data for cross-linguistic typological-dialectological studies be concretely gathered? To answer this question, Auer (2004) identifies three levels of syntactic variation: general features of orality as such (medium-bound), strictly dialectal features (area-bound) and non-standard features that are not strictly dialectal. In his opinion, the last level should be given priority when looking at non-standard varieties since it allows a broad perspective on language as a whole, but excludes data that are directly or exclusively influenced by the oral medium or by geographical factors. In any case, data from spoken language should be preferred to written data. In this respect, Auer’s (2004) position is close to Miller and Fernandez-Vest (2006: 12), who state that standard and non-standard can be equally found in spoken language. However, Miller and Fernandez-Vest prefer not to distinguish between standard and non-standard since in their opinion spontaneous spoken language has properties of its own, whereas Auer sees the possibility of identifying features typical of non-standard varieties, but independent of the medium. Here, Auer’s position will serve as a starting point: general non-standard features can be identified; this can be best achieved if the analysis is based on data from spoken language. However, the notion of ‘spoken language’ needs to be further clarified. In this work Koch and Oesterreicher’s (1990) conceptual model will be adopted which explicitly addresses the relationship between spoken and written language, or, in Koch and Oesterreicher’s terminology, Mündlichkeit ‘orality’ and Schriftlichkeit ‘scripturality’. According to the two scholars, the relationship between ORALITY and SCRIPTURALITY can be regarded as consisting of two dimensions, a MEDIAL (medial) and a CONCEPTIONAL (konzeptionell, i.e. ‘concerning the conception’) one. The medial dimension refers to the communication medium, i.e. the written or spoken channel; the conceptional one describes how a communicative practice8 was originally conceived and relates to a number of linguistic strategies adopted by the speaker – such as syntactic structures, cohesion and coherence, linguistic variety – which are independent of the communicative channel, but can be regarded as typically oral or typically

Looking for non-standard constructions 49

written. For instance, articulated sentences, connectors and nominalizations can be seen as typical of conceptionally written language, whereas juxtaposed sentences, ellipses and reformulations characterize conceptionally spoken language. This does not prevent a medially spoken communicative practice from exhibiting articulated sentences or nominalizations: in this case, one would speak of a medially spoken but conceptionally written communicative practice. Transitions within the two dimensions are substantially different, as Figure 2.2 shows: medial scripturality and medial orality constitute a dichotomy, i.e. a communicative practice is either written or spoken or, in very few cases, both spoken and written; instead, the transition from conceptional scripturality to conceptional orality is gradual and can be represented as a continuum. The position of a communicative practice on this continuum depends on the gradient of conceptionally oral/written elements it contains. CONCEPTION

medial scripturality

distance language

conceptional orality

proximity language

medial orality

MEDIUM

conceptional scripturality

Figure 2.2. The conceptual space of scripturality and orality (according to Koch and Oesterreicher 1990).

Koch and Oesterreicher (1990: 8–9) list a number of parameters which help determine whether a text is conceptionally written or spoken: these are its public character, the familiarity between the producer of the text and its recipient, their emotional involvement, their physical proximity, the cooperation between them, the dialogical character of the text and its degree of spontaneity. In a conceptionally oral situation, the communication partners are usually acquainted with each other, are emotionally involved and produce the text in real time, with limited possibility of editing it. The two partners do not need to see each other, though: telephone calls, infor-

50 Data sources mal e-mails and chats are prototypical examples of non-face-to-face conceptionally oral communicative practices. Koch and Oesterreicher (1990: 15) label the linguistic strategies activated by speakers in such communicative practices as PROXIMITY LANGUAGE (Sprache der Nähe). At the opposite pole there are communicative practices which do not presuppose acquaintance between the partners and where the emotional involvement is low. They have a mainly monologic character and require a long editing phase: the linguistic strategies adopted in these practices are referred to as DISTANCE LANGUAGE (Sprache der Distanz). In order to better illustrate how a classification of communicative practices may look, the position of some communicative practices on the continuum between conceptional scripturality and orality is reported in Figure 2.3 below. The figure is based on Dürscheid’s (1999, 2003) proposal. The position of a communicative practice in Figure 2.3 is to be seen as relative: on the conceptional dimension, an opening discourse is ‘more written’ than a business meeting, which, in turn, is ‘less oral’ than a chat with friends. In fact, in a chat among friends the communication partners are usually well-acquainted and, consequently, emotionally involved, while this is not necessarily the case in a business meeting. An opening discourse has a monologic character and requires a long editing phase, which makes it quite close to the pole of distance language. CONCEPTION

medial scripturality medial orality

MEDIUM

conceptional scripturality

formal letter

discussion forum informal letter formal e-mail

distance language opening discourse conference

business meeting business phone call

conceptional orality

internet chat informal e-mail proximity language chat with friends informal meeting informal phone call

Figure 2.3. The position of some communicative practices on the conceptual space between scripturality and orality.

Looking for non-standard constructions 51

When Koch and Oesterreicher (1990) try to relate proximity and distance language to sociolinguistic variation, they remark that proximity language goes hand in hand with low diaphasic and diastratic varieties and strong diatopic markedness; distance language, instead, is generally associated with high diaphasic and diastratic varieties and weak diatopic markedness (Koch and Oesterreicher 1990: 14–15). Thus, it seems reasonable to consider proximity language as the best candidate for retrieving non-standard, i.e. sociolinguistically marked, constructions. This is corroborated by the fact that the features of distance language described in Koch and Oesterreicher (1990) overlap with some of the features of standard languages enunciated in Stein (1997), i.e. de-oralization, exclusion of emotional meanings and of any reference to the speech situation. So, one may assume that if distance language shows features typical of standard language, then features typical of non-standard language are more likely to be retrieved in proximity language. Support for this claim can be found in Van Marle’s (1997) observation that some constructions are present only in dialects as naive language engineering excluded them from standard. In dialects, linguistic change can take place freely as the features of real-time speech production act more unconstrainedly: “exclusively spoken dialects may incorporate the effects of speech production much more readily than the standard which need not be a spoken language at all” (Van Marle 1997: 29). Consequently, in this work data will be mainly drawn for different languages from both medially written and oral communicative practices which can be situated in the grey shaded areas in the middle and on the right side of Figures 2.2 and 2.3, i.e. in the domain of proximity language. This ensures that editing phenomena like checking what is being written or spoken against the codified standard norm are kept to a minimum, since the primary purpose is to immediately convey a message. In these communicative practices, speech elaboration and planning is not an aim in itself: hence, we are likely to encounter both standard and non-standard constructions and to be able to evaluate the incidence of both. Ideally, this investigation should be carried out using language corpora. Still, this cannot be achieved in this work, for different reasons: in the first place, conducting a corpus-based analysis for all languages investigated would be extremely time-consuming and would go beyond the scope of a cross-linguistic study; then, a good acquaintance with all of the analyzed languages would be required in order to search for non-standard constructions in a corpus; finally, even if this were possible, fully accessible, representative, reliable, up-to-date and correctly part-of-speech-tagged language

52 Data sources corpora only exist for a few European languages. Hence, data will be drawn from several sources, provided that the relevant data can be subsumed to the category ‘proximity language’. The manifold sources constituting the database are presented and discussed in section 2.3 below. 2.2. Language sample As already mentioned, this study concentrates on European languages. Still, it is not obvious which languages should be included in a European sample. In fact, while the western borders of Europe are neatly marked by the Atlantic Ocean, the eastern borders are not clearly defined: previous studies concentrating on European languages show divergences as to whether Caucasian and Altaic languages should be included, which are spoken at the eastern periphery of the European geographical space (Kortmann 1997: 38). In this study two kinds of limitations are applied, a geographical and a sociolinguistic one. Geographically, a restricted sample was selected: all Caucasian languages were excluded and only one Altaic language, i.e. Turkish, was included. This is a subjective limitation, which derives from my insufficient knowledge of Caucasian and Altaic languages. As mentioned in the Introduction, dealing with non-standard constructions implies having more than a grammar-based knowledge of the languages under investigation and, possibly, a good accessibility to original sources. Sociolinguistically, only the languages in which a distinction between standard and non-standard constructions can be applied were considered. This distinction makes sense if a codified variety referred to as standard exists in the relevant language. In Europe, this is not always the case. First of all, the codification of European languages began at different times and under different circumstances (cf. 1.3.1 and Auer 2005b). As a result, beside long-tradition standards like French, German, Russian, etc, there are more recently formed standards, like Macedonian, which was officially codified only in 1944. Additionally, there are languages, like Greek, which were recently re-standardized, and languages like Galician or Rhaeto-Romance which are currently in the process of being (re-)codified. On account of such great sociolinguistic variability, it seemed advisable to consider a convenience sample, i.e. a group of 36 languages for which “reliable information in the form of grammars and dictionaries, but especially in the form of (native speakers) specialists as well as native speakers informants could easily be accessed” (Kortmann 1997: 39). The languages

Language sample

53

included in the sample are listed in Table 2.1 below (and in Appendix A): languages are classified according to their phylum (given in upper case), family (given in small upper case) and branch (given in italic). Throughout this work, languages will be referred to by means of a three-letter code, which was reported in the List of abbreviations and is repeated here in the rightmost column. Table 2.1. The language sample analyzed in this study. INDO-EUROPEAN CELTIC Goidelic ROMANCE West

East GERMANIC

North

West

SLAVIC

West

East

South

BALTIC HELLENIC ALBANIAN

Irish Catalan French Portuguese Spanish Italian Rumanian Danish Icelandic Swedish Norwegian Dutch English German Czech Lower Sorbian Polish Slovak Upper Sorbian Belarusian Russian Ukrainian Bosnian/Croatian/Serbian Bulgarian Macedonian Slovenian Latvian Lithuanian Modern Greek Albanian

IRI CAT FRE POR SPA ITA RUM DAN ICE SWE NOR DUT ENG GER CZE LSO POL SLK USO BLR RUS UKR BCS BUL MAC SLN LTV LIT GRE ALB

54 Data sources Table 2.1. (continued) URALIC FINNO-UGRIC

Ugric

Estonian Finnish Hungarian

EST FIN HUN

South

Turkish

TUR

SEMITIC ARABIC

Maltese

MAL

ISOLATE BASQUE

Basque

BAS

ALTAIC TURKIC

Finnic

As the table shows, the sample is not typologically balanced as 30 languages belong to the Indo-European phylum. This is because the purpose of this work is to conduct an areal study, so an areal sample was relied on rather than a balanced typological one. Besides, as will become clear in the course of the study, emphasis will be placed on languages where non-standard constructions display some divergences compared to their standard counterparts. This is justified if we bear in mind that one of the aims of this work is to shed light on constructions that are traditionally not paid enough attention in typological studies. A final observation: beside the 36 languages listed above, examples from languages not included in the sample will be sometimes quoted for comparison purposes: these languages are Ancient Greek (AGR; IndoEuropean, Hellenic) Gagauz (GAG; Altaic, Turkic, South), Latin (LAT; Indo-European, Italic), Church Slavonic (CSL; Indo-European, Slavic, South), Old Turkic (OTK, Altaic, Turkic), Persian (PER; Indo-European, Indo-Iranian, Iranian). Dialects will be subsumed to their roofing language because, following Chambers and Trudgill (19982: 9), they are regarded as heteronomous varieties: for instance, Bavarian examples are quoted as German. When possible, the relevant dialect is quoted in brackets after the English translation of the example.

A review of data sources 55

2.3. A review of data sources 2.3.1. Linguistic evidence Before presenting the sources constituting the database, the issue of what counts as linguistic evidence in this work will be briefly addressed in this section. The question of evidence in linguistics is discussed in a number of recent contributions. For instance, Penke and Rosenbach (2004: 480) state that, since linguistics is an empirical science, it should work with empirical data. The same claim is made in Lehmann (2007: 16): “[S]o gilt jedenfalls, dass jegliches linguistische Datum auf eine Äußerung, also auf Performanz zurückgeht.” These two positions are key to this work, as they are based on the assumption that constructions can be theoretically described only starting from what the speakers actually utter. Later on, Lehmann (2007) deplores that, for decades, most of theoretical linguistics originating from structuralism has been only marginally concerned with the parole/performance side of language. Theories have been put forward and supported through fabricated examples and it has been claimed that these were sufficient to achieve insight into competence – i.e. speakers’ ability of mastering grammatical rules – as well as to uncover the principles of Universal Grammar, i.e. the innate linguistic common core that all people share. In fact, the study of the performance side of language has received significant input only since the development of sociolinguistics and variation linguistics, which require actual spoken data for their investigations. Without going into further detail, we may agree with Penke and Rosenbach (2004: 514) that “the question of evidence depends on the principal approach taken to linguistics”, that is, different approaches consider different kinds of data as linguistic evidence. Concretely, two attitudes can be outlined: some approaches look for what can be said in one ore more languages; others look for what is actually said. In the first case, scholars’ introspection is sufficient to speculate on the potential of the human linguistic expression; in the second, scholars find it more reasonable to look outside, since a single individual cannot claim to exhibit all possible constructions of a language in his/her idiolect. The first attitude is characteristic of formal approaches, like generativism,9 whereas the second attitude is widespread in functional and applied linguistics, which explicitly adopt speakers’ utterances in their communicative context as the starting point of linguistic analysis. In particular, functional typologists rely on data taken from reference grammars. Still,

56 Data sources the extent to which grammars depict the actual linguistic use is questionable. As Lehmann (2007: 24) argues, even in would-be descriptive grammars there is a great deal of prescriptivism: authors sometimes rely on their own competence for describing language structures and finding examples, whereas an empirically better-founded way would be to compile grammars on the basis of corpus data. Sonia Cristofaro (p.c.) points out that not every linguistic phenomenon appears in corpus data. Still, the reverse is also true: not every phenomenon appears in the idiolect of the compiler(s) of a grammar. Integrating both data sources could be a solution to obtain a good descriptive grammar. This may help avoid major discrepancies in grammars of the same language, where constructions are given more or less importance or are sometimes even left out depending on the compiler’s opinion or idiolect: see for example the review of relative constructions in Serbo-Croatian grammars in Kordić (1999: 33–37). For a comprehensive discussion of which kinds of data are regarded as evidence in formal and functional approaches the reader is referred to Penke and Rosenbach (2004). To summarize, it can be argued that it would be desirable to give preference to non-idiolectal, actually uttered data as a source for linguistic evidence. Nevertheless, it must also be acknowledged that this cannot be but an aim to pursue. The present status of language documentation does not allow us to rely only on corpus data. Great progress has been made as far as English and other major European languages are concerned; for minor languages, though, a lot still needs to be done. Since this study aims to analyze a vast number of languages, no kind of evidence can be excluded a priori. Data on non-standard varieties are often difficult to retrieve, so the most convenient way proves to be the consultation of several sources, particularly for languages that are not well-documented. This entails that in some cases, data cannot be checked against informants or corpora, so their validity has to be assumed. Still, this does not affect the validity of the data: each linguistic datum relying on actual utterances represents positive evidence of a linguistic phenomenon, i.e. proof of its existence. However, whether this phenomenon is idiosyncratic or not can be established only after a systematic collection of a large amount of data (Penke and Rosenbach 2004: 486). For our purposes, this means that unusual relative constructions like the combinations ‘specialized relative element+resumptive pronoun’ (FRE dont…la) or ‘relative particle+relative pronoun’ (POL co który) can be readily accepted as positive evidence; still, their status and frequency need to be tested against a larger database. Negative evidence, on the other hand, will not be granted scientific value: if a construction

A review of data sources 57

does not appear in a corpus, this does not necessarily mean that the relevant construction does not exist in the language considered (cf. 2.3.7). So, although the impossibility to check data against further sources may appear as a major shortcoming, it actually isn’t: one of the main purposes of this work is to draw attention to less well studied constructions or constructions whose status is not clear, and to possibly contribute to the linguistic discussion about them. Moreover, the source of the examples as well as any relevant sociolinguistic information will be, if possible, always specified, so that readers can form their own view as to their representativeness and reliability. In the following subsections each of the data sources of this work will be presented. The kind of evidence provided by each source will be weighed with respect to the aims of this work. This does not mean that some sources are per se better than others: the specific purpose of the research has to be always born in mind and the database should be chosen according to this purpose. Still, as Penke and Rosenbach (2004) point out, what actually determines the validity of a piece of evidence is not its usability within a theoretical frame, but “the way it was collected (‘systematically’) and […] how well it does account for the topic under investigation” (Penke and Rosenbach 2004: 514): data should be as neutral as possible and should be accessible to researchers of any linguistic orientation. 2.3.2. Grammars Grammars are perhaps the most valuable instruments to gather an insight into the structure of a language. With few exceptions (Noreen’s (1904) Old Swedish grammar and Endzelin’s (1923) Latvian grammar), the grammars consulted were compiled in a time span ranging from the 1960s to the present day. These grammars sometimes show considerable differences in their conception and the kind of data used as evidence. The following parameters are relevant. − THEORETICAL APPROACH. Older grammars follow the classical model of the Greco-Roman grammar-writing tradition, in that they focus on morphology and grant syntax and phonology a secondary status; more recent grammars devote a section to each of the components of a language (phonology, morphology, word formation, syntax, semantics/lexis, graphemics). Moreover, compilers dealing with all grammatical components are being gradually replaced by teams of

58 Data sources



researchers, each of whom is specialized in a different subdiscipline. As for the linguistic approach, the grammars consulted were inspired to the structural, the formal and the functional approach: for instance, Russkaja Grammatika (Švedova 1980) follows a structural approach; Grande grammatica italiana di consultazione (Renzi, Salvi, and Cardinaletti 2001) has a slightly formal approach; the German IDSGrammatik (Zifonun, Hoffmann, and Strecker 1997) adopts a functional approach. ATTITUDE. Grammars can be prescriptive, if they describe how a language should be used, or descriptive, if they simply document how a language is used. Most of the grammars consulted are descriptive. Still, as Lehmann (2007) points out, even in recent grammars there is a good deal of (crypto-)prescriptivism: [A]uch Grammatiken, die vorgeben, deskriptiv zu sein, werden de facto bis auf den heutigen Tag von Grammatikern im Vollgefühl persönlicher Sprachbeherrschung geschrieben. […] Es gibt […] keine Basis für die Annahme, der Grammatiker sei allein imstande, den herrschenden Sprachgebrauch zu beurteilen. (Lehmann 2007: 24)



This can emerge in two ways: either some phenomena or constructions are simply left out from the description, or they are marked as ‘popular’, ‘colloquial’ or sometimes as ‘incorrect’ – which is interpretable as an implicit hint that the relevant constructions should be avoided. For instance, Bartnicka et al. (2004) do not mention any other relativization strategies in their Polish grammar than the standard relative pronoun który; Zagorska Brooks (1975: 303) laconically states: “The pronoun co may, in some instances, be used for który”. The only grammar dealing more extensively with alternative relativization strategies, e.g. the particle co, is Topolińska (1984: 343–346). Topolińska also provides information about the contexts where this construction is more likely to be found (“w stylu kolokwialnym”, in colloquial style). The degree of prescriptivism present in descriptive grammars can be related to the next two parameters. COMPILERS. Grammars can be written either by language codifiers or by expert linguists. Language codifiers generally aim to describe the so-called ‘proper usage’, so it may be expected – but not automatically assumed – that grammars written by codifiers show a greater deal of prescriptivism. Expert linguists, instead, mainly aim to describe the structures found in a language, so they tend not to judge the constructions analyzed. Grammars written by linguists with the purpose of

A review of data sources 59





giving native speakers guidelines for linguistic usage or acquainting non-native speakers with the structure of a language are commonly referred to as ‘reference grammars’. For the past few decades there have been increasing efforts by functional linguists to produce reference grammars both for well-known and for less described languages. These efforts met the need of typologists for reliable, up-to-date, exhaustive and easily accessible instruments for cross-linguistic comparison. This issue is discussed in detail in Cristofaro (2006). ADDRESSEES. Grammars can either target native speakers who are looking for guidelines for ‘proper usage’ or non-native speakers who want to become acquainted with a new language, or both groups at the same time. DATA. The grammars consulted contain three different kinds of linguistic data: idiolectal data, data taken from model speakers and writers and corpus data. In some grammars, compilers rely on idiolectal data, providing fabricated examples from their own idiolect: as pointed out in Lehmann (2007), this is perhaps not the best kind of data (see point 2 above), as it is limited to the compilers’ language competence, which, though extensive, cannot be all-comprehensive. When compilers provide non-idiolectal data, they sometimes draw these data from written texts, especially fiction and poetry. These written texts do not necessarily need to be contemporary: the aim mostly being the description of a ‘pure’ language, excerpts are reported from authors regarded as model writers. This is the case with Russkaja grammatika and Gramatica limbii române (resp. Švedova 1980 and Graur 1963), which were compiled decades ago and are nowadays still regarded as reference works: one may wonder whether quotations from writers such as Fëdor Dostoevskij and Ivan Turgenev, who lived and wrote in the 19th century, actually reflect contemporary language use. Then again, it must be remembered that an old grammar is not a worthless grammar: it both provides information on earlier stages of a language and sometimes also includes constructions or explanations that cannot be found in more recent grammars focussing on standard. For example, Noreen’s (1904) Altschwedische Grammatik helps clarify how the present-day relationship between the relative particle som and the relative pronoun vilken developed. Fortunately, the tendency to rely on recent corpus data is becoming more and more widespread nowadays, so as to achieve a description of language use as up-to-date as possible: as Lehmann (2007: 24) states, “[e]s ist heute völlig klar, dass empirische Sprachbeherrschung die Auswertung von

60 Data sources Korpora erfordert.” For example, the latest release of the Duden German grammar (Dudenredaktion 20098) also includes examples from internet discussion forums; the IDS-Grammatik (Zifonun, Hoffmann, and Strecker 1997) includes passages from conversations; the English Longman grammar (Biber et al. 1999) contains statistical data on the frequency of constructions in different communicative practices. Now, the question arises as to whether data on non-standard constructions can be found in grammars. Generally, descriptive grammars address both standard and non-standard constructions and specify in which respect nonstandard constructions are sociolinguistically marked. This is not always true for grammars written for non-native speakers, which sometimes only describe the standard variety, perhaps in order not to burden language learners with too many constructions. Hence, in order to gather data on non-standard relative constructions, preference was granted to grammars written for native speakers and to reference grammars reporting clear indications on the sociolinguistic markedness of constructions. These two groups of grammars are somewhat complementary: grammars written for native speakers sometimes discuss constructions that are ignored in typologically oriented grammars, but often provide only a few examples, which do not allow us, for instance, to collect sufficient information on the syntactic positions that a relativization strategy can relativize; vice versa, typologically oriented grammars provide detailed information on the use of relativization strategies, but sometimes do not describe all of the strategies – both standard and non-standard – that are found in the relevant language. In these cases, one has to rely on grammars written for native speakers or on linguistic studies, which typically analyze a single linguistic phenomenon in depth. 2.3.3. Linguistic studies Whereas grammars are produced either by language codifiers or by expert linguists, linguistic studies are written only by the latter. As mentioned above, codifiers and linguists mainly pursue different aims. Moreover, linguists often concentrate on specific issues neglected by codifiers or try to shed light on unclear questions. Since their aim is to describe and investigate linguistic phenomena and not to codify a language, linguists also consider sociolinguistically marked constructions, especially, but not only, if they are working with corpus data. In this sense, grammars and studies complement each other.

A review of data sources 61

A major distinction between linguistic studies can be made according to the kind of evidence they are based on: some studies rely on fabricated or second-hand examples, others on data from corpora and questionnaires. The latter investigate actual language use and sometimes also contain sociolinguistic considerations on the distribution of the constructions at issue; the former reproduce the scholars’ idiolect and report data without checking them against further evidence. This cannot be condemned per se as wrong practice, since checking data against other sources may sometimes be quite difficult. However, not checking data may lead to distorted or questionable results. For instance, De Vries’s (2002) typological overview of relative clauses in the world’s languages is based only on secondhand data from other linguistic studies: these include Keenan and Comrie (1977), which actually contains nothing more than synoptic tables and a selection of examples from the 50 or so languages in the sample, and the contributions in Perenteau, Levi, and Phares (1972), which, besides being quite old, are usually based either on idiolect or on grammar data. As a result, in De Vries (2002) several constructions occurring in European languages are ignored and any imprecision or ambiguous interpretations in other contributions are simply acknowledged with no further comments. For example, in the final overview of relativization patterns in the world’s languages, Der Vries reports that Hungarian has an uncommon relative element ‘complementizer+relative pronoun’. Still, he neither comments on this nor provides any examples (De Vries 2002: 177): he simply refers to Downing (1978), who, in turn, refers to Schwartz (1971). When we read Schwartz (1971), we see that he actually proposes two interpretations for HUN aki/ami – the relative element at issue, which De Vries (2002) never explicitly mentions, though –, i.e. either as a relative pronoun with an internal structure ‘DET+wh-pronoun’ (a-ki), or as the combination of an invariable element (“an introductory element, typically associable with a pronominal root, which may be […] invariant (ENGLISH that)”, Schwartz 1971: 142) and a relative pronoun. Schwartz never uses the term ‘complementizer’: in fact, UNG a/az is a determiner or a demonstrative like ENG that. De Vries (2002) relies on the overview in Downing (1978) and probably interprets that only as a complementizer, excluding the pronominal interpretation. Still, doing so, he then attributes to Hungarian a strategy that this language does not display. What is questioned here is clearly not the reference to another source, but the fact that the reliability of second-hand data should be carefully weighed, especially if these data are found in synoptic tables and no further evidence is provided. In this respect, it is perhaps worth consulting lan-

62 Data sources guage-specific reference works, looking for occurrences of the relevant construction in a corpus or asking a native speaker, so as to check the validity of the data and/or its interpretation against further evidence. Among corpus-based studies, Pittner (2004) and Gapany (2004) may be cited. Pittner studies relative clauses in German: her data come from the spoken corpora of the Institut für Deutsche Sprache (URL: http://agd.idsmannheim.de/html/index.shtml, last access: June 30th, 2011) which contain detailed information on texts and informants, allowing her to make sociolinguistic considerations. Gapany studies relative clauses in French: his database consists of a variety of mostly oral sources (such as conversations, TV and radio programmes). What is particularly praiseworthy in both studies is that the reader has the possibility to get an insight into the data, either online, i.e. consulting the IDS-corpora, or offline, i.e. consulting the appendix of Gapany (2004), where all occurrences of the relative constructions on which the analysis was based are listed. So, it is possible to verify how the data were interpreted and/or classified and, possibly, to challenge the interpretation. Data transparency can be seen as an advantage of linguistic studies based on data from corpora and questionnaires. To summarize, studies may be very helpful to clarify issues that are not covered in grammars on account of normative pressure. Especially in corpus-based studies, the analysis is prejudice-free, even when it comes to apparently ‘ungrammatical’ constructions. In this work both idiolect-based and corpus-based studies were taken into consideration. When unusual constructions were found, they were checked against further evidence when possible; in some cases, the author of the contribution was directly asked for further explanations. 2.3.4. Questionnaires Questionnaires are usually composed of sets of sentences: data are elicited from native speakers, who are requested either to give their linguistic judgement on the sentences or to translate them into their language. The most patent advantages of questionnaires are their comparability – which is very useful when working on many languages at the same time – and their representativeness: they can be assumed to reproduce actual linguistic use. In some cases, they may provide further information as to whether different constructions are used by the same speaker and, if so, in which contexts this happens; that is to say, sociolinguistic data can be obtained, too (Cristofaro and Giacalone Ramat 2007: 64).

A review of data sources 63

Data from questionnaires can be then organized into electronic databases and made available as online resources. The search in these databases is made easier by the possibility of setting a number of search criteria; then, the output is fully accessible thanks to PoS-tagging, i.e. labelling each word of the corpus according to the part of speech it belongs to. Further morphological information can be added, too, which makes the search in the corpus easier. Examples of electronic typological databases are the database of intensifiers and reflexives, originally compiled at the Freie Universität of Berlin (URL: http://www.personal.uni-jena.de/~mu65qev/ tdir/, last access: June 30th, 2011) and the Pavia typological database (Sansò 2005), compiled at the University of Pavia. This database is composed of different modules, some of which are under construction. The modules provide data on relative clauses, noun phrases, deictic elements and coordinating structures. In order to collect data on non-standard relative constructions, two questionnaires were initially created. The first one consisted of 20 sentences to be translated from English into the informant’s mother tongue. This questionnaire was completed by eight informants and was felt to be quite long. The second version of the questionnaire only contained half of the sentences and was completed by thirteen informants. The questionnaire was designed so as to cover the syntactic positions of the Accessibility Hierarchy (Keenan and Comrie 1977, cf. section 3.3); some more sentences were added with the aim to clarify positions not represented on the Accessibility Hierarchy, like circumstantials, and cases of unclear relativization, i.e. sentences where the role of the relativized item in the relative clauses cannot be unambiguously determined (cf. 4.2.2 and 4.3.3). All informants had a degree or were graduating in linguistics and were aged between 22 and 30. Unfortunately, only a restricted range of languages was covered. This was mainly due to the fact that, after scrutinizing initial results from questionnaires, it appeared that it was perhaps not the best way of collecting data on non-standard varieties. As a matter of fact, eliciting data on standard varieties, both written and spoken, is relatively easy; instead it is much more difficult – or even impossible – to force a speaker to think only in terms of non-standard. This is clearly pointed out also by Penke and Rosenbach:

64 Data sources even if an informant judges [a] construction as bad, this does not mean that it is truly ungrammatical for her or him […]. People simply tend to evaluate the more prestigious standard forms as more ‘grammatical’, obviously mixing up ‘grammaticality’ and ‘correctness’ […]. (Penke and Rosenbach 2004: 491)

First of all, an informant may not necessarily consider as (non-)standard a construction that grammars or linguists judge as such. Additionally, in the questionnaire the informants were explicitly encouraged to translate using the most ‘non-standard’ (i.e. familiar, colloquial or vernacular) variety they could think of; still, they most often opted for the unmarked standard strategies. Only upon direct request – i.e. when presented with an alternative translation – did they concede that other solutions were also valid. Most frequent comments were “You can hear it, but I never use this structure” or “Some uncultivated people may say that, but I don’t”. All this clearly shows that the very idea of a questionnaire based on translations from English into a non-standard variety of another language was inappropriate. In fact, it was based on the assumption that it was possible to use a sort of ‘pseudo-non-standard’ English in order to hint at what kind of language variety informants were expected to adopt. Still, it was soon realized that not in every case when a relative construction is used in English, does this happen also in the target language. Moreover, as Part 2 will show, in different languages typologically similar constructions may be regarded either as standard or as non-standard, so the comparability of non-standard structures cannot be assumed a priori, but has to be demonstrated on a case-by-case basis. In this sense, the data gathered through questionnaires will be used as evidence in this work only if they have been discussed with the informants. As always happens with questionnaires, they reproduce the speaker’s idiolect, as the partial divergences in the acceptability judgements given by different informants for the same language show: hence, these data were always checked against at least one additional source. Moreover, questionnaires will not be relied on to establish the sociolinguistic status of relative constructions in Chapter 5. In Table 2.2 the number of informants for each language is reported. For the reader’s convenience, the questionnaires are reported in Appendix B. Table 2.2. The informants and their mother tongues. LANGUAGE BCS BUL FIN GER GRE ITA POL POR RUM RUS INFORMANTS 1 2 1 4 2 2 3 1 2 3

A review of data sources 65

2.3.5. Language corpora Language corpora are databases containing written and/or spoken texts in one or more languages. As Lehmann (2007: 23) and Scherer (2006: 5–10) point out, not every collection of linguistic data can be regarded as a corpus. Corpora have to fulfil at least three criteria: representativeness, quality and reliability. A corpus is REPRESENTATIVE if it consists of a significant sample of a language. Careful data collection, transcription and tagging contribute to the QUALITY of a corpus. As for RELIABILITY, it is the basic requirement to compile a corpus: the compilers commit to working professionally and to avoiding any manipulation of the collected material. Corpora can be classified with reference to the following criteria (cf. Scherer 2006: Ch. 2): 1. corpora can contain written, spoken or both kinds of data; 2. corpora contain several kinds of communicative practices, which can be differently categorized; 3. corpora can aim to reproduce sociolinguistic variation on the whole or can focus on one variety or variation axis. A corpus does not need to be homogeneous; what is relevant is that the degree of heterogeneity is made explicit by the compilers. For instance, the Russian National Corpus (Nacional’nyj korpus russkogo jazyka, URL: http://www.ruscorpora.ru/, last access: June 30th, 2011) is composed both of written and spoken texts from various periods. It also contains a dialectal and a poetry section as well as parallel subcorpora. The user can set a number of parameters, so as to create his/her own subcorpus. Representativeness also concerns the distribution of data collection points. So, the Italian LIP corpus exhibits texts recorded in Milan, Florence, Rome and Naples. North-Eastern and Southern Italy, including Sicily and Sardinia, are not represented in this sample. Moreover, Bellini (1998: 71–72) claims that the choice of the informants recorded may not be regarded as representative for the whole Italian diasystem: nel LIP gli strati socioculturali più alti [sono] fortemente sovrarappresentati rispetto alla realtà italiana odierna. […] Le testimonianze orali raccolte ci sembrano […] rendere conto in misura troppo limitata degli strati socioculturali più bassi, con un appiattimento verso l’alto del parlato documentato. (Bellini 1998: 72)

66 Data sources (in the LIP corpus the higher socio-cultural layers are strongly overrepresented with respect to present-day Italian reality. […] The collected oral texts seem to insufficiently mirror the lower socio-cultural layers, which causes an upwards levelling of the spoken language documented in the corpus.)

The Romance C-ORAL-ROM corpus (Cresti and Moneglia 2005), composed of French, Italian, Portuguese and Spanish texts, is even more strikingly biased towards individual cities or regions: most of the Italian texts were recorded in Florence (65%) or other towns in Tuscany (11%); nearly half of the French texts come from Southern France (Manosque, Marseille, Aix-en-Provence, Toulon); Spanish texts were mostly recorded in Madrid (76%) and Segovia (12,5%); only Portuguese texts show quite a regular distribution. In this sense, the geographical representativeness of these corpora may be questioned. Additional, in C-ORAL-ROM a distinction is made between formal and informal communicative practices: still, telephone calls are pigeonholed in the former group. This is somewhat perplexing: if we examine these telephone calls, we get the impression that they have quite an informal character. Corpora can be searched for type and token occurrences as well as for constructions. In order to make the search easier and more accessible to scholars who are less familiar with computer linguistics and corpus linguistics and to provide sufficient metadata to operate with corpora from languages they are not familiar with, careful PoS-tagging is required. This process is neither simple nor automatic. Great efforts are being made in computational linguistics in order to create reliable tagging software. At present, this was successful only to a limited extent. For instance, when trying to retrieve occurrence of che/que as a relative particle in the CORAL-ROM corpus, several tagging errors were noticed. The same happened when searching the Russian National Corpus for occurrences of что (čto), which can be a (relative) particle, a conjunction, an interrogative pronoun or a relative pronoun. Thanks to the richness of their data, corpora are highly valuable instruments to carry out both qualitative and quantitative analyses. Statistics can be produced on language use: the occurrences of a construction can be related to sociolinguistic parameters characterizing the informants, the communicative situation or the communicative practice. Unfortunately, not every (European) language has yet been documented through a corpus, which leads to significant asymmetries in language description. This can be easily understood if we consider that much more resources are available

A review of data sources 67

about major languages like German, English, French, Spanish, Russian, etc. than about minor languages like Slovak or Upper Sorbian. A reference list of corpora – both written and spoken – for several European languages can be found on the LARL (Laboratorio di analisi di risorse linguistiche) website at the University of Pavia (URL: http://192.167.77.47/Mambo/ index.php?option=com_weblinks&Itemid=23, last access: July 30th, 2009). Last but not least, corpora ensure free interpretability of data. If their compilation was not theory-driven, as is most often the case, each researcher could use them as databases for his/her investigations regardless of his/her linguistic theoretical background. On the other hand, corpora are samples, so they represent language use, but do not cover it fully. As Lehmann (2007: 17) states, “[m]indestens so viel muss klar sein, dass ein Textkorpus jedenfalls kein Ausschnitt aus dem Phänomenbereich, sondern eine Repräsentation davon ist.” Hence, if a construction is not attested in a corpus, it cannot be assumed not to occur at all in the relevant language; one can at most speculate on its infrequency. De facto, it may well be that in another sample the construction is widespread: resorting to the World Wide Web could help verify this against further evidence. 2.3.6. The World Wide Web A further data source must be mentioned, which, to some extent, represents a challenge for future corpus-based linguistic research: the World Wide Web. In the past decade, technical progress has led to an enormous development of new internet-based communicative practices, like e-mails, discussion forums, chats and blogs. Linguists soon became interested in these practices. From a theoretical perspective there have been attempts to pigeonhole new and traditional communicative practices, using Koch and Oesterreicher’s (1990) model as a reference frame. All of these practices are medially written: so, a distinction can be made between formal and informal practices. Formal e-mails and professional discussion forums are closer to the pole of conceptional scripturality; informal e-mails and discussion forums as well as chats and blogs tend towards the conceptionally oral pole (cf. Figure 2.3 above). In applied linguistics, the World Wide Web was enthusiastically welcomed as a continuously growing, permanently up-to-date and potentially inexhaustible database. Still, it was soon acknowledged that its being so huge and unsystematic could eventually affect the validity of linguistic studies based on it. A lively debate has been ongoing for the past few years

68 Data sources among scholars: the main point that is commonly agreed upon is that before using data from the Web, its pros and cons should be carefully weighed. For instance, Scherer (2006: 74–76) makes some considerations on the Web as an instrument for linguistic research: she states that it certainly provides a large amount of authentic linguistic material and allows us to select specific linguistic domains (economics, law, etc.); additionally, constructions apparently underrepresented in language corpora can be checked against further evidence from the Web. However, on account of its undetermined size, only qualitative analyses are possible; quantitative analyses can be carried out if a suitable subcorpus has been previously selected. Moreover, Scherer claims that we may look only for lexemes and word forms, but not for syntactic structures. Finally, frequent typing errors as well as inherent multilingualism risk to make Web searches most unreliable. Now, what Scherer says is only partially true. First, most spoken corpora are static and represent the linguistic situation in a limited period whereas the Web is permanently up-to-date, covers all linguistic domains and directly reproduces developments and tendencies in languages. Second, it is true that no concordance software satisfying linguistic needs exists for searching the World Wide Web (Scherer mentions a tool named WebConc); however, instead of looking for structures, one may look for constructions, using the advanced search mask of a search engine and typing ‘*’ as a replacement for variable elements (cf. Chapter 7, where this procedure is used to collect data from German discussion forums). Third, in order to solve the problem of multilingualism, a specific language can be selected in the same mask, although it must be born in mind that this will in no case ensure monolingualism. Fourth, the high number of typing errors cannot be possibly eliminated, but traditional corpora also contain tagging and parsing errors affecting their quality. All in all, the Web proves to be a valid source for linguistic data: however, like Scherer states, it seems more reasonable to search a subcorpus selected according to precise criteria rather than the whole WWW, at least if the purpose of the search is quantitative analysis. 2.3.7. Language corpora vs. the World Wide Web Two issues related to data sources briefly require some further discussion. The first one regards, once more, the World Wide Web. As mentioned in the previous two sections, both spoken language corpora and the Web pro-

A review of data sources 69

vide direct evidence of speakers’ utterances, especially of conceptional orality, which is the main domain of investigation in this work. So, the question arises as to whether they are to be seen as complementary or alternative sources. To begin with, language corpora may contain both medial and conceptional orality, whereas some of the Web-based communicative practices mentioned above (discussion forums, chats, blogs) are medially written, but conceptionally oral (Dürscheid 2003: 38). Thus, the criticism that texts from the internet cannot be used to investigate orality because they are medially written cannot be supported. Additionally, corpora are certainly more controlled, but also more rigid than the World Wide Web. This is because they were consciously planned and compiled for linguistic investigations and equipped with searching tools to help the researcher, like PoStagging and concordance software. These tools do not exist for search in the World Wide Web yet. The strong point of the Web rests on its dimensions. So, if a construction is not attested in a reference corpus, it does not mean that it is not attested at all in the relevant language, but only in the speech of a number of informants selected according to well-defined sociolinguistic criteria and recorded during a well-defined period of time. If the same construction is searched for in the Web, one will probably get some results, although it is nearly impossible to get sociolinguistic information about the ‘informants’. For example, when searching the LIP corpus for two typically spoken concessive constructions in Italian introduced by con tutto che and va be(ne) che (both: ‘although, even if’), no results are produced for the former and only six for the latter. If we now search discussion forums through the Google advanced search mask (URL: http://groups.google.it/advanced_ search?q=&, last access: June 30th, 2011) and limit the search to the first four months of 2008, we get 18 occurrences of con tutto che and 65 for va be(ne) che. In this sense, the results of the Web search integrate those of the corpus search and suggest that both constructions deserve further attention when investigating strategies of linking clauses in spoken Italian. A further point to be mentioned is that search results in the Web are hardly ever exactly the same, since the Web is continuously evolving: new sites are created, older ones are eliminated. These dynamics are at the same time the source of both the up-to-date, fleeting character of the Web and its presumed unreliability. Nevertheless, some recent grammars, like the German Duden-Grammatik (Dudenredaktion 20098), report quotations from the Web as well as fabricated examples and passages from language corpora as examples of contemporary linguistic usage – in particular when it

70 Data sources comes to illustrating non-standard constructions. This is a first step towards the acceptance of the World Wide Web as a source of linguistic evidence, which can be seen as complementary to all other sources traditionally used. 2.3.8. Historical evidence The second issue that needs clarification is the value of data from historical documents, which will also be used as linguistic evidence. As will emerge in Chapter 6, historical data are central to determine the development of constructions over time. These data are always medially written. Their conceptional status cannot always be established unambiguously, because most of the times sufficient information on the compilers and their intentions is not available. Moreover, the ‘standard vs. non-standard’ opposition does not directly apply to these data as they often date back to the period between the 13th and the 17th centuries: in some regions of Europe there were some initial codification efforts, although the continuum between the two poles was not as developed as it is nowadays (see section 6.5. and Auer 2005b). Since the aim of Chapter 6 is to provide diachronic evidence for constructions that are nowadays regarded as non-standard, as stated in section 1.5, all kinds of historical data available about the languages of the sample will be considered; if possible, the communicative practice in which the relevant construction was used will be reported in brackets. Doing this, it can be observed whether certain constructions generally occurred in official vs. non-official, sacred vs. lay or ‘elevated’ vs. ‘humble’ communicative practices. A further remark must be made. As this work is neither philologically oriented nor mainly concerned with collating and comparing parallel versions of historical documents, the validity of diachronic data is taken for granted. As Zifonun puts it, a typologically oriented scholar should consult historical sources and be able to draw a diachronic line, gaining insight into the development of a construction. So, it is expected that historical linguists and philologists formulate the results of their investigations as clearly as possible (Zifonun 2003: 66–67) and possibly provide information on the compilers and the sociolinguistic context in which the text was produced.

A review of data sources 71 Table 2.3. Pros and cons of different data sources: an overview. SOURCE GRAMMARS



PROS thorough description of language structures

− − −

STUDIES I (IDIOLECT-BASED)



STUDIES II (CORPUS-BASED)



− QUESTIONNAIRES

– –

may quote data neglected in grammars may quote data neglected in grammars based on actual linguistic use comparable focus on particular constructions





corpus may be biased



non-standard constructions difficult to elicit often divergences in acceptability judgements may contain tagging errors may not include all constructions under investigation may not be representative for all linguistic varieties reliability and representativeness may be questionable quantitative analyses possible only in selected subcorpora. information on their conceptional status mostly difficult to retrieve



CORPORA

– – –

WORLD WIDE WEB

– –

HISTORICAL DATA



CONS (some) prescriptivism examples sometimes not up-to-date sometimes limited to the standard variety linguist’s idiolect as a database

contain a large amount of data offer free data interpretability make both qualitative and quantitative analyses possible



contains large amounts of data may provide evidence for constructions not attested in language corpora provide valuable insight on the development of language structures











72 Data sources 2.3.9. Summary In order to summarize what has been discussed in the previous subsections, an overview of all data sources used in this work is presented in Table 2.3. The pros and cons of each source are also outlined. For two languages, German and Italian, evidence will be also provided from casual speech: these data were gathered during conversations in Italy and Germany or heard from radio or TV programmes. In these cases, the speaker’s gender and age is always indicated. Such a heterogeneous database proved to be useful in two ways. On the one hand, evidence for relative constructions could be usually gained from different sources, which made it possible to compare or integrate different pieces of information on relative constructions, their contexts of use and their sociolinguistic status. On the other, even when a construction was attested in a single source, it was included in the analysis. Each time that such constructions are quoted, it will be pointed that they need to be checked against further evidence. Some uncommon combined relative elements described in 3.2.2 are an example of this.

Chapter 3 Data classification

After describing the sources consulted in order to gather data on non-standard relative constructions, in this chapter the issue of data classification will be covered. As mentioned in section 1.1.2, in typological literature devoted to relative clauses relativization strategies are classified according to a number of parameters: 1. WORD ORDER, i.e. the position of the RC with respect to the MC; 2. RELATIVE ELEMENT, i.e. the element encoding the relativized item in the RC; 3. the SYNTACTIC POSITIONS that can be relativized by means of a relativization strategy; 4. MORPHOLOGICAL CHANGES in the verb of the RC as a result of relativization; 5. changes in the CODING OF PARTICIPANTS inside the RC as a result of relativization. In this investigation, data will be classified according to parameters 1, 2 and 3, which are separately examined in the following sections: word order will be covered in 3.1, relative element in 3.2 and the syntactic positions a strategy can relativize in 3.3. In section 3.4 it will be shown how the three parameters correlate with each other. Finally, in section 3.5 some instances of problematic classification will be discussed. Parameters 4 and 5 are not considered because they are less central when it comes to describing relative constructions in European non-standard varieties. Hence, so-called non-finite RCs, i.e. RCs in which the verb of the RC is in a non-finite form, will be disregarded. The only language in which the main relativization strategy relies on non-finite verb forms is Turkish. Still, non-standard relative constructions in Turkish can be successfully described by means of parameters 1 to 3, as will be shown in Chapter 4 (cf. in particular 4.1.1 and 4.1.2).

74 Data classification 3.1. Word order Word order is at issue in works such as Schwartz (1971), Zaliznjak and Padučeva (1975), Downing (1978), Comrie (1981: Ch. 7), Lehmann (1984: 48–49) and De Vries (2002: 22–23). The classification proposed here is based on Lehmann (1984) and De Vries (2002). In particular, De Vries (2002: 20–23) attempts to systematize previous proposals and to unify the sometimes diverging terminology. He examines possible synonyms to the terms he adopts, weighs whether they may be regarded as valid and highlights cases of terminological confusion. As mentioned above, the parameter ‘word order’ is used to describe the position of the RC with respect to the MC. An initial distinction can be made between EMBEDDED and ADJOINED RCs: embedded RCs are constituents of the NP encoding the relativized item in the MC, adjoined RCs are not. Structurally, embedded RCs can appear inside the MC, whereas adjoined RCs cannot. Adjoined RCs appear either before or after the whole MC. Within these two groups, a further distinction can be made. Embedded RCs can be classified according to their position with respect to the relativized item in the MC: if they precede it, they are PRENOMINAL; if they follow it, they are POSTNOMINAL; if the RC ‘surrounds’ the relativized item of the MC, the RC is CIRCUMNOMINAL. Examples (3.1) to (3.3) below illustrate these three types. The RC is given in square brackets and the relativized item in the MC is in bold. (3.1) BAS

[Amona hil zaio-n] mutila Granny dead she.is.to.him-REL boy.DET dantzara joan da. dance gone is ‘The boy whose Granny died went to dance.’ (Hurch 1989: 20)

(3.2) BLR

Кніга, што я прачытала, вельмі цікавая. Kniga, [što ja pračytala,] vel’mi cikavaja. book REL I read.PST.F very interesting ‘The book I read is very interesting.’ (Ramza and Hurtig 2003: 69)

(3.3) Min [bïrat-ïm at-ï Yakut I brother-POSS.1SG horse-ACC tüp-püt-ü-n] imiri-di-m. catch-PTCP-POSS.3SG-ACC beat-PST-1SG ‘I beat the horse that my brother had caught.’ (Johanson 2006: 20)

Word order 75

Adjoined RCs can be divided into two subgroups according to their position with respect to the MC: if they precede it, they are CORRELATIVE or PREPOSED, if they follow it, they are EXTRAPOSED or POSTPOSED. Both kinds are exemplified in (3.4) and (3.5). (3.4) RUS

А которые ребята были там, очень несладко им всем пришлось. A [kotorye rebjata byli tam,] očen’ and REL.NOM.PL children were there very nesladko im vsem prišlos’. unsweet to.them all came. ‘And all the children who were there didn’t have a good time at all.’ (Lapteva 1976: 144)

(3.5) HUN

Az a lány beteg volt, [aki a that DET girl sick was REL.NOM.SG DET könyvet olvassa.] book reads ‘That girl who is reading the book was sick.’ (Kenesei, Vago and Fenyvesi 1998: 39)

So-called FREE or HEADLESS RCs will be subsumed to the correlative strategy in this work. The main difference between ‘headed’ correlative RCs and headless RCs lies in the degree of explicitness with which the relativized item is encoded in the RC. Whereas in correlative RCs the relativized item is prototypically encoded with the maximum degree of explicitness, i.e. through a substantive, as in (3.4), in headless RCs the relativized item is encoded only through a pronoun or not encoded at all, as in the two examples below. (3.6) UKR

[Xto ne služyv u bahatyrja,] toj ne who not served at rich.man that not znaje horja. knows misery ‘Who has not served at a rich man’s(, that person) doesn’t know misery.’ (Gołąb and Friedman 1972: 35)

(3.7) GER

[Was what

sie/er nicht she/he not

weiß,] macht sie/ihn nicht knows makes her/him not

heiß. hot

76 Data classification ‘What a (wo)man doesn’t know cannot make her/him angry.’ (German idiom) In the Ukrainian example, the relativized item is encoded through a pronoun both in the RC (chto, transliterated as xto in the source text) and in the MC (toj); in the German example the relativized item is explicitly encoded only in the RC (was), but not in the MC. In the vast majority of languages in the sample, two different pronouns are used for animate and inanimate participants (e.g. ENG who vs. what). For a detailed analysis of headless RCs the reader is referred to Lehmann (1984: 293–325), who proposes a functional account, and Smits (1989: 140–172), who proposes a formal one. The question may arise now as to how postnominal RCs occurring in sentence-final position may be distinguished from extraposed RCs, or prenominal RCs occurring in sentence-initial position be distinguished from correlative RCs. A syntactic test can be helpful here: since embedded RCs are constituent of an NP, if the referent is shifted the RC will follow it or, vice versa, if the RC is shifted the referent will follow it. If this does not happen, i.e. if shifting either the referent or the RC does not cause the other element to shift, too, then the RC is not embedded. In (3.8) to (3.10) this test was applied to examples (3.1), (3.2) and (3.4) above. In (3.8), the determiner -a is not enclitic to the head noun but to the relative particle -n when the RC is in postnominal position (sentences b. and c.). (3.8) BAS

a. b. c. d.

[Amona hil zaion] mutila dantzara joan da. Mutil [amona hil zaiona] dantzara joan da. Dantzara joan da mutil [amona hil zaiona.] *[Amona hil zaion] dantzara joan da mutila.

(3.9) BLR

a. b. c. d.

Kniga, [što ja pračytala,] vel’mi cikavaja. Vel’mi cikavaja kniga, [što ja pračytala.] *[Što ja pračytala,] vel’mi cikavaja kniga. *Kniga vel’mi cikavaja, [što ja pračytala.]

(3.10) a. RUS b. c.

A [kotorye rebjata byli tam,] očen’ nesladko im vsem prišlos’. A [kotorye rebjata byli tam,] očen’ nesladko prišlos’ im vsem. A [kotorye rebjata byli tam,] im vsem očen’ nesladko prišlos’.

Word order 77

As (3.8) and (3.9) show, if either the referent or the RC alone are shifted and separated from each other, the results obtained are ungrammatical: this confirms that the RC in these examples is embedded. If the same is attempted with an adjoined RC, as in (3.10), no sentence turns out to be ungrammatical. However, in some cases the shifting test yields unexpected results, as shown in the examples below. The Turkish sentence in (3.11) is actually quoted in Haig (1998) as it appears in (3.12) – that is, the prenominal RC can be extraposed with no loss of grammaticality. (3.11) [Senin verdiğin] vazoyu kırdım. TUR you.GEN give.NMLZ.POSS.2SG vase.ACC I.broke ‘I broke the vase that you gave me.’ (Haig 1998: 109, modified) (3.12) Vazoyu kırdım [senin verdiğin] TUR vase.ACC I.broke you.GEN give.NMLZ.POSS.2SG ‘I broke the vase(, the one) that you gave me.’ (Haig 1998: 109) The German sentence in (3.13) behaves ambiguously: as (3.14) shows, the referent and the RC can be shifted as a whole, but the RC can also be extraposed, leaving the referent in the same position as in the original sentence. The same happens in the English sentence (3.15). (3.13) Sie würde nie mit Leuten, [die sie GER she would never with people REL.ACC.PL she nicht schätzt,] Kontakt aufnehmen. not esteems contact take. ‘She would never get in touch with people she doesn’t esteem.’ (3.14) a. GER b. c. (3.15) a. ENG b.

Mit Leuten, [die sie nicht schätzt,] würde sie nie Kontakt aufnehmen. Sie würde nie Kontakt aufnehmen mit Leuten, [die sie nicht schätzt.] Sie würde nie mit Leuten Kontakt aufnehmen, [die sie nicht schätzt.] A friend [who I hadn’t seen for ages] came round yesterday. Yesterday there came round a friend [who I hadn’t seen for ages.]

78 Data classification c.

Yesterday a friend came round [who I hadn’t seen for ages.]

Example (3.5) is quoted again in (3.16), but this time the RC is embedded: Kenesei, Vago and Fenyvesi (1998: 39) claim that both positions are acceptable, provided that the construction expresses a restrictive relative relation. (3.16) Az a lány, [aki a könyvet HUN that DET girl REL.NOM.SG DET book.ACC olvassa,] beteg volt. reads sick was ‘That girl who is reading the book was sick.’ (Kenesei, Vago and Fenyvesi 1998: 39) Now, how may these ambiguities be interpreted? There are two possibilities: these extraposable RCs are either regarded as cases of constituent right-extraposition, or we may postulate that the relevant language has two different patterns of relativization – an embedded and an adjoined one. For the purposes of this work the examples above will be regarded as instances of pre- or postnominal RCs which were subject to extraposition, so a separate type will be not postulated here. This may be useful to keep these constructions separated from so-called TRULY EXTRAPOSED or POSTPOSED RCs, i.e. RCs that cannot be moved inside the MC. These constructions are also attested in the languages of the sample. See the following example. (3.17) I have finally graduated in linguistics, [which makes me very happy.] In this example we have a truly extraposed RC because the RC cannot be moved next to a referent inside the MC. This cannot happen because the SoA encoded by the RC does not qualify a participant of the MC, but the MC as a whole. For further discussion on the issue of extraposition the reader is referred to Lehmann (1984: 203–206) and De Vries (2002: Ch. 7). Instances of the postposed strategy in the languages of the sample will be illustrated in 4.1.3. In typological literature on RCs, the European languages constituting the sample are said to exhibit following DEFAULT10 strategies: the vast majority exhibits a postnominal embedded strategy, whereas Turkish and Basque rely on a prenominal strategy (Table 3.1).

Relative element 79 Table 3.1. The traditional classification of RCs in European languages according to word order. EMBEDDED PRENOMINAL CIRCUMNOMINAL

TUR BAS

ADJOINED POSTNOMINAL

CORRELATIVE (PREPOSED)

EXTRAPOSED (POSTPOSED)

vast majority of European languages

As a matter of fact, as will be illustrated in section 4.1 non-standard varieties also display different strategies: for instance, Russian uses a correlative adjoined strategy, whereas Turkish and Basque possess a postnominal embedded strategy. 3.2. Relative element Relative elements used in RCs are at issue in works such as Zaliznjak and Padučeva (1975), Comrie (1981: Ch. 7), Lehmann (1984: Ch. III, 248– 252), De Vries (2002: Ch. 5) and, for European languages, in Kurzová (1981), Smits (1989), Zifonun (2001a), Fiorentino (2007) and Cristofaro and Giacalone Ramat (2007). Here, Lehmann’s (1984) and De Vries’s (2002) proposals will be discussed first, then the classification adopted in this work will be illustrated, which, with some modifications, is based on Cristofaro and Giacalone Ramat (2007). The necessity of proposing a different classification arose from the discrepancies found in literature on RCs: on the one hand, there are divergences as to how relative elements are defined; on the other, in European non-standard varieties there are elements whose morphosyntactic status is not easily grasped by traditional classifications. The classification of relative elements presupposes that the properties of the categories to which relative elements are subsumed are set beforehand – i.e. features of a ‘relative pronoun’ or a ‘relative particle’ are defined first, then the attested elements are pigeonholed according to these features. Relative elements can be differently classified according to the parameters that the prototypical representative of each category is assumed to have. For instance, Van der Auwera (1985) and Van der Auwera and Kučanda (1985) show that it is possible to classify resp. ENG that and BCS što both as relative pronouns and as relative particles; Berruto (1987: 123–129) reviews the contributions supporting or refuting the claim that ITA che is a relative pronoun. In

80 Data classification all of these cases, different argumentations can be presented according to the features attributed to the notion of ‘relative pronoun’ and ‘relative particle’. Hence, the classification proposed in this work is to be regarded as an attempt to create a parameter-based reference frame that helps categorize relative elements in European languages as unambiguously as possible. As the instances of problematic classification in 3.5 will show, it is sometimes impossible to univocally classify relative elements without relying on their contexts of use: this is because relative elements are essentially construction-bound (cf. 3.5.1 and 4.4). Diachronic development and grammarexternal factors led sometimes to the formation of relative elements displaying the same form, but different morphosyntactic properties in different constructions. Lehmann (1984: Ch. IV) classifies relative elements according to whether they explicitly encode the three constitutive operations of RC formation, i.e. subordination, attribution and gap construction. They can be illustrated as follows: 1. a relative element encodes SUBORDINATION if it signals that the clause where it occurs is a subordinate one; 2. a relative element encodes ATTRIBUTION if it expresses morphosyntactic agreement with the relativized element in the MC (gender, number, etc.); 3. a relative element encodes GAP CONSTRUCTION if it can be regarded as a representative of the relativized item in the RC.11 (Non-)encoding of these three operations yields different kinds of relative elements, which are reported in Table 3.2. The table is based on Lehmann (1984: 248–250). If an element encodes at least two of the three operations, it can be considered a relative pronoun. Inside this category, we may distinguish between resumptive relative pronouns, which inflect for gender, number and case and represent the relativized item in the subordinate clause (leftmost column in Table 3.2), and non-resumptive relative pronouns, which do not represent the relativized item in the RC (second column on the left in Table 3.2). Elements encoding only one function are either conjunctions, if they only encode subordination, or anaphoric personal/demonstrative pronouns, if they only encode gap construction. Elements that only encode attribution, though theoretically possible, actually cannot appear alone, since they need an additional element that expresses subordination. Lehmann (1984: 250) exemplifies this through GER der wo: he argues that wo conveys subordination, whereas der only expresses attribution.

Relative element 81 Table 3.2. Lehmann’s (1984) classification of relative elements. Subordination Attribution Gap construction Example

X X X ENG who

Relative pronoun X X X X Arabic ’allaðī

ITA che/cui

Conjunction X X X Gothic sa-ei

?

Pronoun

X X ENG that SWE som

GER der wo

Lehmann’s typology provides a good starting point for the classification of relative elements in European languages. Still, it can be further refined. Smits (1989: 60–70) divides relative elements into relative pronouns, relative adverbs and relative particles. However, he does not set any criteria or parameters to distinguish between the three categories, but draws parallels with other word classes: so, relative pronouns “behave like other pronouns”, i.e. they “inflect for some grammatical categories”; relative particles “behave like conjunctions”, i.e. they are uninflected; relative adverbs are “specific adverbs of a pronominal nature that stand for specific types of adverbial adjuncts” (Smits 1989: 60). It appears, then, that these definitions are not based on the functions that relative elements have inside the RC, but rather on their morphological properties. The category ‘relative adverb’ is relevant for the purposes of our classification and will be taken up again below, although with the different label ‘specialized relative element’. De Vries (2002: Ch. 5) considers relative pronouns, relative particles and resumptive pronouns. Looking at Table 3.3 below, where De Vries’s typology of relative elements is reproduced, we may see that some elements that Lehmann classified as relative pronouns are now considered particles (i.e. Arabic ’allaðī). This is because the two scholars define relative elements on different terms. In line with formal analyses of RCs, De Vries is concerned with the issue of the position occupied by relative elements inside the structure of an RC. He argues that fully inflected elements appear in clause-initial DP position, uninflected elements occupy the CP position and resumptive pronouns make the gap in the RC explicit, since they appear in the t-position (trace) left by the moved head NP. This explication would also account for possible combinations of relative elements. They can be used together because they occupy different positions: relative pronouns occupy the DP position and undergo wh-movement; relative particles occupy the CP position; resumptive pronouns occupy the gap in the RC and are never clause-initial.

82 Data classification Within the category ‘relative particles’, De Vries further distinguishes between relative complementizers (which include Lehmann’s category ‘conjunction’), relative markers and relative affixes. Complementizers do not display any agreement features with the relativized element in the MC, whereas markers display ‘some’ agreement (it is not further specified which kind of agreement they show); affixes do not show any agreement either, but they are bound morphemes. So, Arabic ’allaðī is a relative pronoun in Lehmann’s classification, because it encodes at least two of the three operations of RC formation; in De Vries’s classification, instead, it is a relative particle: it is not subject to wh-movement, so it occupies the CP position in the RC. Table 3.3. Classification of relative elements according to De Vries (2002: 176). RELATIVE PRONOUNS

RELATIVE COMPLEMENTIZERS

RELATIVE PARTICLES

– – – – – – – –

RELATIVE MARKERS RELATIVE AFFIXES RESUMPTIVE PRONOUNS

– – – –

with demonstrative core (DUT die) with interrogative core (BCS koji) specialized form for RCs (SLN kdor) particle identical to the complementizer (ENG that, SWE som) particle specialized for RCs (CZE co, Yiddish wo/wos) nominalizing particle (Chinese de) attributive particle used for RCs (Accadian šu) particle in non-classifier languages (Arabic ’allaðī) particle in classifier languages BAS –n resumptive affix resumptive word: demonstrative, personal pronoun

As shown in the table, De Vries introduces further subcategories within each type. Doing this, he partly considers the form, partly the function of the relative element. So, relative pronouns are classified according to their core – demonstrative or interrogative –, which implies that they are identical to either demonstrative or interrogative pronouns. As from a synchronic point of view not every relative pronoun shows this overlapping, De Vries has to introduce a third subcategory, i.e. ‘specialized forms’ for RCs. Still, he exemplifies this category through SLN kdor, which in fact has a quite transparent interrogative core, too: it originates from the combination of Proto-Slavic interrogative *kъto ‘who’ with the emphatic particle -ž(e),

Relative element 83

which phonetically modified to -r (Vondrák 19282: 482). So, this distinction cannot be viewed as insightful: on the one hand, it is not clear in which respect relative pronouns with a demonstrative or interrogative core should differ from each other; on the other, the fact that the form of a relative pronoun does not synchronically overlap with the form of a demonstrative or interrogative pronoun is not sufficient to prove that it does not have a demonstrative or interrogative core. In fact, relative pronouns in the languages of the sample originate almost exclusively from demonstrative or interrogative pronouns. To summarize, relying on the form or the etymology of a relative element seems not to be particularly insightful for cross-linguistic classification, since the synchronic and the diachronic level cannot always be kept clearly separate from each other. Table 3.4. Relativization strategies in European languages according to Cristofaro and Giacalone Ramat (2007: 65–72). CODING OF: STRATEGY: Relative pronoun Adposition+ relative pronoun Uninflected relative element+ personal pronoun/ possessive adjective Invariable relative element (1) Invariable relative element (2) Gap

Syntactic role of the head + +

Gender (M/F/N or ±animate) + +

Number

Example

+ +

RUS kotoryj ITA al quale

+

+

+

Sardinian ki dd

+











FRE où CZE co GAL que







ENG

Cristofaro and Giacalone Ramat’s (2007) proposal proves to be more helpful in order to identify further distinctions, which Lehmann’s and De Vries’s models do not grasp. As mentioned in 1.4, Cristofaro and Giacalone Ramat (2007) analyze how relative elements are cross-linguistically distributed and to which extent each relative element conveys the syntactic role of the relativized item. Their typology of relative elements relies on three parameters: the coding of the syntactic role, the gender and the number of the relativized item in the RC. The parameter ‘coding of the syntactic role of the head’ is akin to Lehmann’s ‘gap construction’, whereas the

84 Data classification features ‘coding of gender’ and ‘coding of number’ can be regarded as a refinement of Lehmann’s ‘attribution’. As will be shown in the remainder of this section, distinguishing between gender and number coding within the parameter ‘attribution’ helps to better grasp the difference between different types of relative pronouns. Cristofaro and Giacalone Ramat classify relative elements according to whether they code the parameters listed above. The strategies they identify are illustrated in Table 3.4. A basic opposition can be noticed between inflected and invariable elements. Still, it must be pointed out that not all relative pronouns code the three parameters: this is true for RUS kotoryj, but does not apply to ENG who, which only inflects for gender. Moreover, Cristofaro and Giacalone Ramat’s classification does not include all of the strategies attested in European languages. For instance, (3.18) cannot be easily classified using Table 3.4 above as a reference. (3.18) The girl [Ø I gave the book to] is a good friend of mine. Here, the relative element is constituted by a gap (Ø) and a stranded preposition (to). The preposition codes the syntactic role of the relativized item in the RC (indirect object), i.e. this information is recoverable by the hearer (Givón 1990: 650). Instead, no information is recoverable on gender and number of the relativized item in the RC. For elements displaying no agreement, Cristofaro and Giacalone Ramat adopt the label ‘invariable relative element’: this means that relative elements like FRE où or CZE co, which have a different morphosyntactic status and are clause-initial, should be considered on a par with the combination of the gap with the preposition to in (3.15), because, like to, they are invariable and convey no agreement. Still, the element to is not really invariable, because it is triggered by the argumental structure of the verb in the RC. If the verb was get, the sentence would read as The girl I got the book from… In order to avoid possible ambiguities and mismatches, a more fine-grained classification frame is called for. After reviewing the classification proposals attested in literature on RCs, the model adopted in this work will be now introduced. To start with, it is meant to equally consider both the information that the relative element provides on the relativized item and its morphosyntactic status. Moreover, the fact that relative elements can be combined with one another or with other elements is also taken into account. This issue is also covered in De Vries (2002: 177). Still, De Vries’s proposal is not fully satisfactory,

Relative element 85

as will be pointed out in 3.2.2. The classification proposal adopted in this work will be described in two steps: first, the three operations of RC formation introduced by Lehmann (1984) will be reconsidered and adapted to the theoretical background of this work; then, it will be verified how they apply to the relativization strategies observed in the European language sample. At this point, a distinction will be made between simple and combined relative elements. SIMPLE RELATIVE ELEMENTS consist of a single morphosyntactic unit; COMBINED RELATIVE ELEMENTS are composed of more units, which can be contiguous or non-contiguous. As mentioned above, Lehmann (1984) characterizes relative markers according to parameters which correspond to the the constitutive operations of RC formation, i.e. subordination, attribution, and gap construction. These parameters will be slightly modified in the light of the definition of relative construction provided in 1.1.2. That is, relative elements will be classified here according to whether they encode or not the following information: 1. the LINK between MC and RC (akin to Lehmann’s ‘subordination’), 2. the SYNTACTIC ROLE of the relativized item (akin to Lehmann’s ‘gap construction’), 3. the co-reference with the relativized item in the MC, distinguishing between GENDER and NUMBER agreement (akin to Lehmann’s ‘attribution’). As may be seen, Lehmann’s parameters were modified as follows: first, the term ‘subordination’ was given up, because relative relations do not necessarily imply subordination as defined in 1.1.1, i.e. as an asymmetry between the pragmatic status of two SoAs. Then, the term ‘gap construction’ was avoided, because it may be misleading. In fact, it relies on three assumptions: 1. the MC contains the head noun: this is not always true. For instance, in the correlative strategy the head noun appears in the RC, cf. (3.4) above or (3.19) below; 2. when the RC is linked to the MC, a (syntactic) gap is formed in the RC, which can be explicitly signalled or not: this is not always true either. As will be discussed in 4.2.2 and 4.2.3, the RC may sometimes present no syntactic gap; 3. the head noun is morphosyntactically encoded in a more explicit way than the gap in the RC: this, too, is not always the case, as exemplified in 1.1.2.

86 Data classification Assumptions 1 to 3 would imply an asymmetry between the degree of explicitness with which the relativized item is encoded in the MC and in the RC, the MC being prior to the RC. Instead, the classification frame proposed here is based on the tenet that the relativized item can be morphosyntactically encoded with a different degree of explicitness both in the MC and in RC. Neither clause is prior to the other one, as far as the coding of the relativized item is concerned. Indeed, as examples (1.31) to (1.37) in 1.1.2 show, it is not always the case that the relativized item is more explicitly encoded in the MC than in the RC: see example (3.19) below, which has been already quoted as (1.33). Here the gap, i.e. the argument not explicitly encoded, is the relativized item in the MC, not in the RC: the direct object of the verb položit’ ‘to put’ is not morphologically encoded. (3.19) Которые деньги останутся, положи в кошелёк. RUS [Kotorye den’gi ostanutsja] položi v košelëk. REL.NOM.PL money remain put in wallet ‘The remaining money, put it in the wallet’ (Zaliznjak and Padučeva 1975: 74) Finally, following Cristofaro and Giacalone Ramat (2007), ‘attribution’ was split into two subcategories, i.e. gender and number coding: this helps attain a more detailed classification of relative elements commonly referred to as ‘relative pronouns’. As pointed out in Cristofaro and Giacalone Ramat (2007: 65), the parameter ‘gender’ can refer to either grammatical gender (masculine/feminine/neuter) or animacy ([±animate]), or to other semantic features, like [±human]: in this respect, languages behave differently. Still, language-specific distinctions were disregarded and a common category of ‘gender’, including all language-specific distinctions, was posited. As mentioned above, the four parameters adopted for classifying relative elements – encoding of 1. the link between MC and RC, 2. the syntactic role of the relativized item in the RC, 3. gender and 4. number of the relativized item in the RC – were formulated on the basis of the definition of ‘relative relation’ and ‘relative construction’ given in 1.1.1 and 1.1.2 respectively. In the first place, since in a relative relation two SoAs are linked, it may be verified whether this link is explicitly encoded in the relative construction or not; also, since in a relative relation the two SoAs share a participant and one SoA qualifies a participant of the other SoA, it may be determined how this participant is expressed in the clause encoding the qualifying SoA, i.e. the RC. In particular, we may ascertain 1. whether

Relative element 87

its syntactic role inside the RC is morphosyntactically encoded and 2. whether the co-reference between the relativized items in the MC and in the RC is explicitly conveyed by means of agreement features (gender, number). The four parameters are of a Boolean nature, i.e. they only yield two possibilities each: either a feature is encoded (+) or it is not encoded (–). So, each relative element can be classified according to whether it encodes these features or not. Before illustrating how these parameters concretely apply to the relative elements attested in the sample, we need to remember the distinction mentioned above between simple and combined relative elements, which can now be further specified: the parameters encoded by simple relative elements are conveyed through a single morphosyntactic unit; in combined relative elements, different morphosyntactic units can encode different parameters and the possibility is not excluded that some of the parameters are expressed twice. 3.2.1. Simple relative elements Simple relative elements will be described first. If we examine which of the four parameters mentioned above are encoded by simple relative elements, we get the results shown in Table 3.5 below. The table is based on the data available about the languages included in the sample. It does not represent all possible combinations of the four parameters, but only those that occur in the database. The possibility is left open that other combinations could be attested in future if the sample was further extended. As the table shows, the further on the right a relative element is on the table, the fewer features it encodes. Six different kinds of relative elements can be identified. As each element in columns 1 to 6 encodes a different combination of parameters, six new labels to define them may be found. Still, this will not be done; instead, an attempt wil be made to rely on terms and notions that are well-established in literature on RCs, defining them univocally on the basis of the four parameters. This appears to be necessary because, as mentioned above, there is often no agreement on the definition of terms like ‘relative pronoun’ or ‘gap strategy’, and the same relative element can be classified differently if different parameters are set.

88 Data classification Table 3.5. The features encoded by simple relative elements. PARAMETERS link between MC and RC syntactic role of the relativized item in the RC gender agreement number agreement

1 +

2 +

3 +

+

+

+

+ +

+ – – – Relative pronoun (RPRO)

RELATIVE ELEMENT 4 5 + +

6 –

+ (!)





– – Specialized relative element (SRE)

– – Relative particle (RPAR)

– – Zeromarker (Ø)

As the table shows, it was distinguished between four kinds of simple relative elements: relative pronouns, specialized relative elements, relative particles, and zero-marker. A RELATIVE PRONOUN (abbreviated as RPRO) conveys at least the link between the MC and the RC and the syntactic role of the relativized item; it can wholly, partially or not at all convey gender and number agreement. This definition allows us to group together the relative elements in column 1, 2 and 3 in Table 3.5. All these elements are labelled as relative pronouns in literature on RCs, but their different features, i.e. the extent to which they convey gender and number agreement, are often neglected. Each kind of relative pronoun is exemplified in (3.20) to (3.25). In (3.20) and (3.21), CZE který encodes all of the four parameters. (3.20) Člověk, [který nekouří,] ušetří. CZE man RPRO.NOM.M.SG not.smokes saves ‘A man who doesn’t smoke saves money.’ (Petr 1987: 524) (3.21) Tady je ta kniha, [kterou jsi CZE there is that book RPRO.ACC.F.SG you.are mi půjčil před týdnem.] to.me lent before week. ‘Over there is the book you lent me a week ago.’ (Petr 1987: 525)

Relative element 89

Instead, ENG who/which does not encode number agreement. Its form does not change if the relativized item in the MC is singular or plural, as (3.22) and (3.23) show; still, it would change if the relativized item in the MC was [–human] or [–animate], in which case the form which would be used instead of who. (3.22) The woman [who’s coming in right now] is a good friend of mine. (3.23) The women [who are coming in right now] are good friends of mine. In (3.24) and (3.25), LTV kas encodes neither gender nor number: it is used without distinction for masculine, feminine, singular and plural referents. (3.24) Te ir grāmata [ko tu man LTV here is book RPRO.ACC you to.me devi.] gave ‘Here is the book you gave me.’ (Mathiassen 1997: 70) (3.25) Te ir grāmatas [ko tu man LTV here are books RPRO.ACC you to.me devi.] gave ‘Here are the books you gave me.’ (Mathiassen 1997: 70, modified) The three subcategories of relative pronouns can be terminologically differentiated as follows: 1. relative pronouns encoding all four parameters (column 1 in Table 3.5) can be defined as gender-number-inflecting (G-N-INFLECTING, cf. CZE který/která/které); 2. relative pronouns which do not encode number agreement (column 2 in Table 3.5) can be defined as gender-inflecting (G-INFLECTING, cf. ENG who/which); 3. relative pronouns encoding neither gender nor number agreement (column 3 in Table 3.5) can be defined as non-gender-number-inflecting (NON-G-N-INFLECTING, cf. LTV kas).

90 Data classification This subdistinction will be used in the remainder of this work only if necessary. The term ‘relative pronoun’ refers to all of the three kinds of relative elements reproduced in columns 1, 2 and 3 of Table 3.5. In some languages there are further constraints on the choice of the form of the relative pronoun. These constraints are mostly of a semantic nature. For instance, Hungarian has three forms: aki, ami, amely (Kenesei, Vago, and Fenyvesi 1998: 40). Aki is used when the relativized item is [+human], amely when the relativized item is [–human], [+specific] and [+countable], ami when the relativized item is [–human], [–specific] and [–countable]. Still, in colloquial Hungarian there exists only a basic opposition between aki, [+human], and ami, [–human], so the Hungarian relative pronoun can be classified as G-inflecting. Cf. also section 4.2.4. A SPECIALIZED RELATIVE ELEMENT (abbreviated as SRE, column 4 in Table 3.5) is a transitional element. Specialized relative elements are akin to relative particles (column 5) in that they can be invariable; at the same time they are akin to non-G-N-inflecting relative pronouns (column 3) in that they express the syntactic role of the relativized item in the RC. Like relative pronouns and relative particles, specialized relative elements explicitly encode the link between MC and RC. What makes them different both from relative particles and relative pronouns is that they can be used to relativize a single syntactic position: that is why in Table 3.5 the symbol ‘!’, ‘existing and unique’, is used. In examples (3.26) to (3.28) below, different instances of specialized relative elements are presented. (3.26) Der er menge mennesker, [der ikke har DAN there are many people SRE not have tag over hovedet.] roof over head.DET ‘There are many people who don’t have a roof over their heads.’ (Allan, Holmes, and Lundskær-Nielsen 1995: 202) (3.27) Les voisins sont atterrés par cet FRE DET neighbours are frightened by this incendie [dont on ne connaît pas encore fire SRE one not knows not yet les causes ni le montant des dégâts.] DET causes nor DET amount of.DET damages ‘The neighbours are frightened by this fire, whose causes and amount of damage are still unknown.’ (Gapany 2004: 191)

Relative element 91

(3.28) Przewodnik pokazał nam pomieszczenia, [gdzie POL guide showed us rooms SRE właśnie odbywa się remont.] just runs REFL repair ‘The guide showed us the rooms where repairs are being done just now.’ (Anna Górska, from questionnaire) DAN der is only used for the relativization of subjects, as in (3.26), FRE dont for genitives, as in (3.27) and POL gdzie for locatives, as in (3.28). Actually, FRE dont can also be used to relativize some obliques, i.e. those complements introduced by the preposition de, as in the example below with the verb hériter (de quelqu’un) ‘to inherit (from someone)’. (3.29) C’ est la vieille cousine [dont j’ ai FRE it is DET old cousin SRE I have hérité.] inherited ‘It’s the old cousin I have inherited from.’ (Riegel, Pellat, and Rioul 1994: 482) Still, the use of dont remains syntactically restricted to de-complements, so it can be further regarded as a specialized relative element. Specialized relative elements are mostly invariable, like the elements in (3.26) to (3.28), but they sometimes also inflect, like POR cujo/cuja or BCS čiji/čija/čije in the examples below. (3.30) O armário [cujas portas foram POR DET wardrobe.M SRE.F.PL door.F.PL were arranjadas] é de mogno. repaired is of mahogany ‘The wardrobe whose doors have been repaired is made of mahogany.’ (Andrade Peres and Móia 1995: 314) (3.31) to je Petar, [čija je sestra BCS this is Petar.M SRE.NOM.F.SG is sister.NOM.F.SG radnica] worker ‘This is Petar, whose sister is a worker.’ (Sussex and Cubberley 2006: 385)

92 Data classification Still, it must be pointed out that even when these elements inflect, they do not display gender and number agreement with the relativized element in the MC, but behave like adjectives in the corresponding languages: they inflect for gender, number and case and agree with the nominal that functions as a possessum in the RC. In (3.30) the feminine plural form cujas is triggered by the feminine plural portas in the RC, not by the referent o armário, which is masculine. The same can be said for (3.31): the feminine singular čija agrees with sestra in the RC and not with the masculine referent Petar. In fact, when the gender or the number of the referent is modified, the form of the specialized relative element does not change. This is shown in (3.32) and (3.33): in (3.32), the referent in the MC becomes plural, but the relative element remains unchanged; in (3.33), the referent in the MC becomes feminine but, again, the form of the relative element remains unchanged. (3.32) Os armários [cujas portas foram POR DET wardrobes.M SRE.F.PL door.F.PL were arranjadas] são de mogno. repaired are of mahogany ‘The wardrobes whose doors have been repaired are made of mahogany.’ (Andrade Peres and Móia 1995: 314, modified) (3.33) to je Ana, [čija je sestra BCS this is Ana.F SRE.NOM.F.SG is sister.NOM.F.SG radnica] worker ‘This is Ana, whose sister is a worker.’ (Sussex and Cubberley 2006: 385, modified) So, the only parameter encoded by inflected specialized relative elements is the syntactic role of the relativized item in the RC, in this case a genitive, which is expressed in the examples above through the bound morpheme cuj- and čij- respectively. What is exemplified here for POR cujo and BCS čiji also applies to SPA cuyo and to the corresponding specialized relative elements in other Slavic languages, like RUS čej, BUL čijto, etc. See Table 3.6 for a comprehensive list. To summarize, elements that in literature on RCs are referred to in various ways, like ‘relative pronoun’ (for elements like POR cujo/cuja ‘whose’, cf. Smits 1989: 62), ‘relative adverb’ or ‘locative relative’ (for elements like POL gdzie ‘where’), have been gathered here under the label

Relative element 93

‘specialized relative element’. This label is not particularly elegant: however, it successfully comprises the heterogeneous elements that belong to this category by highlighting their common feature, i.e. their being specialized to relativize only one syntactic position. Inside this category, we may (but not necessarily) further single out relative adverbs – prototypically relativizing circumstantials – and elements used to relativize genitives. Still, there are other markers which are not easily classifiable, so characterizing them as specialized relative elements is a practicable way to resolve the impasse. As previously mentioned, RELATIVE PARTICLES (abbreviated as RPAR) are invariable elements. They overtly encode only the link between MC and RC (column 5 in Table 3.5). Cristofaro and Giacalone Ramat (2007) qualify these elements as ‘truly invariable’ because they can relativize several syntactic positions, but do not change their morphological form. Examples from Portuguese are given below. (3.34) Ele era um homem [que você se dava POR he was DET man RPAR you REFL gave bem,] [que você podia rir junto,] [que well RPAR you could laugh together RPAR estava sempre bem humorado.] was always well tempered. ‘He was a man that one got along well with, that one could laugh with, that was always well-tempered.’ (Albertino Moreira, from questionnaire) (3.35) A POR DET um

criança [que você deu o dinheiro] é child RPAR you gave DET money is cigano. DET gipsy ‘The child you gave the money to is a gipsy.’ (Albertino Moreira, from questionnaire)

(3.36) O cara [que o carro foi roubado POR DET chap RPAR DET car was stolen semana passada.] week last ‘The chap whose car was stolen last week.’ (Albertino Moreira, from questionnaire)

94 Data classification In (3.34) to (3.36) the same relative element, POR que, is used to relativize different syntactic positions: oblique and subject in (3.34), indirect object in (3.35) and genitive in (3.36). In none of these cases is the syntactic case of the relativized item expressed in the RC, nor do we have any gender and number agreement; only the link between the MC and the RC is signalled through the relative particle. As mentioned above, De Vries (2002: 176) distinguishes between particles identical to the complementizer and particles used only in RCs. Still, it is actually quite difficult to find elements that are used only as relative particles. In the vast majority of European languages, relative particles also have at least another function: as complementizers (cf. ENG that, FRE/SPA que, ITA che, RUS čto, MAC da), as conjunctions (cf. SWE som, CZE jak), as specialized relative elements (cf. BUL deto, GER wo) or as inflected pronouns (cf. CZE/POL co). For instance, in (3.37) and (3.38) below FRE que is used once as a relative particle and once as a complementizer. In (3.39) and (3.40) SWE som has two different functions, as a relative particle and as a modal conjunction. (3.37) La seule chose [qu’ il faut l’ avertir.] FRE DET only thing RPAR it needs him warn ‘The only thing he needs to be warned of.’ (Gapany 2004: 190) (3.38) Nouns savons que la terre est ronde. FRE we know COMP DET earth is round ‘We know that the Earth is round.’ (Riegel, Pellat, and Rioul 1994: 491) (3.39) Min bror [som bor i Malmö]har fyra SWE my brother RPAR live in Malmö have four barn. kids ‘My brother, who lives in Malmö, has four kids.’ (Holmes and Hinchliffe 1994: 220) (3.40) Att tiga är guld, men du gör SWE to be.silent is gold but you do som du vil. as you want ‘Silence is golden, but you do as you like.’ (http://maryj.bloggagratis.se/, last access: June 30th, 2011)

Relative element 95

Diachronically, the function as a relative particle may or may not be primary. In some cases, the other functions developed out of an original relative function; in others, a particle with another function also acquired the relative one. Still, historical circumstances are not directly at issue here: the purpose of this chapter is to achieve a functional classification based on morphosyntactic features. Hence, the origin of each relative particle will be disregarded and it will be taken for granted that most particles may also have another syntactic function. Interestingly, although these particles have various functions, they share one feature: if they appear in a complex construction, they always can be said to encode the link between the matrix clause and the dependent clause, whatever its nature is (complement, modal, purpose, etc.). So, although in (3.37) and (3.38) or (3.39) and (3.40) different relations between SoAs are encoded, it can be argued that que and som make the link between the two clauses explicit. The polyvalent character of relative particles is further analyzed in section 3.5.1. Considering what has been discussed above, it seems preferable not to make any further functional differentiations between relative particles. On account of their morphological status, it will only be distinguished between affixal and non-affixal particles. The former are attached to the verb of the RC, the latter are not. The vast majority of relative particles attested in the sample are non-affixal (see examples above); instead, BAS -n is affixal, as example (3.41) shows. Here, the relative particle is attached to the auxiliary duzu. (3.41) [Karrikan ikusi duzu-n] gizonak BAS street.DET.in seen you.have.him-RPAR man.DET.ERG hemen lan egiten du. here work doing he.has.it ‘The man you saw in the street works here.’ (Oyharçabal 1989: 63) The last category of simple relative elements is the ZERO-MARKER (symbolized through ‘Ø’, column 6 in Table 3.5). This strategy is sometimes referred to as ‘gap strategy’, a term that is not univocally used, though: for instance, Comrie (1981: 144) uses it to include all [–case] strategies, i.e. all strategies where no information is provided on the syntactic role of the relativized item in the RC. So, the term ‘zero-marker’ was preferred. The link between the MC and the RC is not expressed, or, put differently, it is realized through a zero-morpheme – hence the name of the element. Syntactically, the MC and the RC are simply juxtaposed, or the RC is embedded into the MC. Neither the syntactic role of the relativized item in the

96 Data classification RC, nor gender and number agreement are overtly encoded. Instances of the zero-marker are given in the examples below. (3.42) Der ovre bor en mand [Ø hedder DAN there over lives DET man is.called Olsen.] Olsen ‘Over there lives a man called Olsen.’ (Platzack 2002: 78) (3.43) Znal jsem člověka, [Ø měl jenom půlku plic.] CZE known I.am man had only half lungs ‘I’ve known a man who had only one lung.’ (Petr 1987: 526) (3.44) А ето вот сынок в армии служиў во флоте тък. RUS A eto vot synok [Ø v armii služiŭ vo and this PAR son in army served in flote tăk.] navy PAR ‘And this here, this is the son who did his military service in the navy.’ (dialect, Lapteva 1976: 332) (3.45) The boy [Ø we met yesterday at the movies] is a good friend of mine. As the examples show, nothing overtly hints at the fact that the relativized item in the RC is a subject in (3.42), (3.43) and (3.44) and a direct object in (3.45). In literature on RCs, examples like (3.42) to (3.44) are sometimes regarded as instances of asyndetic coordination. This issue is discussed in detail in 3.5.3 below. The simple relative elements attested in the languages of the sample, both in standard and non-standard varieties, are reported in Table 3.6. Here, an account as comprehensive as possible of all relative elements attested in the languages of the sample was attempted. Still, the list can be further integrated. For instance, specialized relative elements used to relativize temporal circumstantials were not considered. In the second column, the number of languages in which the relevant relative element is attested is reported. As the table shows, the most frequent elements are relative particles, followed by relative pronouns and specialized relative elements. Only a few languages display a zero-marker. It may be noticed that some languages display more than one form for relative elements like relative

Relative element 97

pronouns or relative particles: this peculiarity will be taken up in the following (cf. for instance sections 3.5.1, 4.2.3 and 5.2.1). Table 3.6. Simple relative elements attested in the sample. RELATIVE ELEMENT

n

Relative pronoun (RPRO)

32

Specialized relative element (SRE)

31

Relative particle (RPAR)

34

Zero-marker (Ø)

7

LANGUAGES AND ELEMENTS ALB i cili, BAS zein, BCS koji, kakav, BLR jaki, katory, BUL kojto, kakăvto, CAT el qual, qui/què, CZE který, jaký, jenž, DAN hvilken, DUT die/dat, wie/wat, hetwelk, ENG who/which, EST kes/mis, FIN joka/mikä, kuka, FRE lequel, qui/que, GER der, welcher, GRE o opíos, HUN aki/ami/amelyik, ITA il quale, che/cui, LIT kurìs, LTV kas, kuŕš, LSO kotaryž, kakiž, MAC koj, kakov, NOR hvilken, POL który, jaki, POR o qual, que/quem, o que, RUM care, RUS kotoryj, kakoj, koj, SLN kateri, kakršen, SLK ktorý, aký, SPA el qual, el que, quien, SWE vilken, UKR jakyj, kotryj, USO kotryž, kajkiž, štojž. ALB ku, BAS non, BCS gdje, čiji, BLR dzej, BUL deto/kădeto, čijto, CAT on, CZE kde, DAN der, hvor, DUT waar, ENG where, EST kus, FRE où, dont, GER wo, GRE ópu, ITA dove, LIT kur, LTV kur, LSO źož, MAC kade(što), čij(što), NOR hvis, der, hvor, POL gdzie, POR onde, cujo, RUM unde, RUS gde, čej, SLN kjer, SLK kde, SPA donde, cuyo, SWE där, vars, UKR de, USO hdźež. ALB që, BAS -n, BCS što, te, gdje, da, BLR što, BUL deto, da, CAT que, CZE co, jak, že, DAN som, DUT wat, dat, ENG that, which, what, as, FIN kun, FRE que, dont, GER wo, was, we, wos, wat, GRE pu, ICE sem, IRI a, ar, nach, nar, ITA che, LIT kur, LTV kur, LSO kenž, což, ako, až, hač, haž, štož, MAC što, MAL li, NOR som, POL co, że, POR que, RUM ce, de, care, RUS što, SLN ki, SLK čo, SPA que, SWE som, TUR ki, UKR ščo, USO kiž, kož, štojž. CZE, DAN, ENG, NOR, RUS, SWE, TUR.

98 Data classification 3.2.2. Combined relative elements As mentioned above, combined relative elements consist of a number of contiguous or non-contiguous morphosyntactic units and are formed through the combination of either two simple relative elements or a simple relative element with another element – a non-relative particle or a resumptive element. Examples (3.46) and (3.47) below illustrate the combinations of two simple relative elements: in (3.46) a relative pronoun is combined with a relative particle; in (3.47), a relative particle occurs together with a specialized relative element. (3.46) D Kirch, [newe dere wu er wohnt,] GER DET church next.to RPRO.DAT.F.SG RPAR he lives isch im Griech kabütt gemacht wor. is in.DET war broken made been ‘The church next to which he lives was destroyed during the war.’ (Alemannic dialect, Balliet 1997: 214) (3.47) Kender du den mand [som der talte me DAN know you that man RPAR SRE spoke with hende]? her ‘Do you know the man who talked to her?’ (Platzack 2002: 83) The examples below illustrate the combinations of a simple relative element with a non-relative particle and a resumptive. The Dutch relative pronoun occurs with the complementizer as in (3.48); in (3.49) the resumptive element s neja is used together with the relative particle deto. As will be explained below, the relative particle encodes the link between the MC and the RC, whereas the resumptive encodes the three remaining parameters, i.e. the role of the relativized item (oblique) and gender and number agreement (feminine singular). (3.48) Dat is de man, [die as het DUT this is DET man RPRO.M/F.SG COMP DET verhaal verteld hef.] story tolf has ‘This is the man who told the story.’ (Staphorst dialect, Barbiers et al. 2005: 16)

Relative element 99

(3.49) Cега имам химикалка дето мога да пиша с нея с часове. BUL Sega imam chimikalka, [deto moga da piša now I.have pen RPAR I.can COMP I.write s neja s časove.] with her from hours ‘Now I have a pen with which I can write for hours.’ (Petăr Kehajov, from questionnaire) The combinations of relative elements occurring in European languages were reported in Table 3.7. The table is to be read as follows: the elements in the leftmost column combine with the elements in the uppermost row. The number of languages in which the relevant combination is attested is specified in each cell. Table 3.7. Combined relative elements attested in the sample. RPRO

RPRO

SRE

BAS, GER, LSO, MAC, USO (5)

GER? (1?)

Ø

OTHER PARTICLE

GER, DUT (2)

FRE, ICE, MAC, NOR, SWE (5)

SRE

RPAR

RPAR

POL, RUS (2)

DAN, RUS (2)

DAN, ICE, POL (3)

RUS (1)

RESUMPTIVE ELEMENT

ALB, BUL, CAT, ENG, FRE, GRE, ITA, MAC, POL, POR, RUM, SPA (12). CAT, FRE, GER, ITA, POL, RUM, RUS, SPA (8) ALB, BAS, BCS, BLR, BUL, CAT, CZE, DAN, DUT, ENG, FRE, GER GRE, ICE, IRI, ITA, LSO, MAC, MAL, NOR, POL POR, RUM, RUS SLN, SLK, SPA, SWE, UKR, TUR, USO (31) DAN, ENG, NOR, SWE, TUR (5)

100 Data classification De Vries (2002: 177) produces a similar table based on the 172 languages of his sample: he finds far fewer combined elements than those reported in Table 3.7, although his sample includes most of the languages considered in the present investigation. This is probably because he considers nearly exclusively standard varieties, at least as European languages are concerned. Moreover, De Vries discards a priori the combination ‘RPRO+resumptive element’ as being impossible, because it does not fit into his theoretical reference frame. In fact, this combination implies that both the clause-initial DP position and the t-position inside the RC are occupied by the relative pronoun and the resumptive element respectively. Still, from a formal perspective this is not allowed: either the t-position remains empty after fronting (‘promoting’ to the DP position, in De Vries’s terminology) the NP that occupied it – which represents the case of an RC introduced by a relative pronoun – or the NP stays in situ, leaving the clause-initial position empty – in which case we would be faced with an initial complementizer in CP position and a resumptive pronoun in the tposition. So, De Vries admits only combinations that do not violate this constraint and states that “[i]ndeed it [=the combination ‘RPRO+ resumptive element’] is not attested” (De Vries 2002: 177). However, empirical data from European standard and non-standard varieties reveal that this combination is attested in 12 languages. Each combination reported in Table 3.7 will be now briefly discussed and exemplified. Combinations of a SIMPLE RELATIVE ELEMENT WITH A RELATIVE PRONOUN are extremely rare. In the sample there is only one case of combination of two relative pronouns, which is actually a case of triple combination: the two relative elements are separated by an uninflected element, as (3.50) below shows. Since the existence of this combination is attested only by one informant, its occurrence should be definitively verified against a larger database; as a matter of fact, it may also be interpreted as a performance error. (3.50) Der Mann, [der wo welcher GER det man RPRO.NOM.M.SG RPAR RPRO.NOM.M.SG geschtern hier war,] isch heut krank. yesterday here was is today sick ‘The man that was here yesterday is sick today.’ (woman, 50 y.o.) The only other combination of a simple relative element with a relative pronoun, RPAR+RPRO, is not uncontroversial either. It is attested in two

Relative element 101

languages, Polish and Russian. The Polish example in (3.51) is taken from Szczegielniak (2004). (3.51) Marysia zna chłopców, [co których POL Marysia knows boys RPAR RPRO.M.ACC.PL Ania lubi.] Ania likes ‘Marysia knows the boys who(m) Ania likes.’ (Szczegielniak 2004: 373) This construction was criticized as being non-existent when I presented it at the 2nd meeting of the Slavic Linguistic Society in 2007 in Berlin. Adam Szczegielniak (p.c.), who was then asked for an explanation, declared that the construction is non-standard and typical of the region around Warsaw. Evidence from Russian dialects, as shown in (3.52), suggests that this combination is not to be excluded a priori, at least in the Slavic area. (3.52) Дык едить жа он вон на той-та лошади, шту у какой два RUS жьрябёнка жьрябились. Dyk edit’ ža on von na tojta PAR goes PAR he there on that- PAR lošadi, [štu u kakoj dva ž’rjabënka horse RPAR at RPRO.GEN.F.SG two foals ž’rjabilis’.] were.foaled ‘There he goes on that horse that gave birth to two foals.’ (Kursk dialect, Akimova 1964: 142) In (3.52) the relative particle štu (standard RUS što) occurs with a relative pronoun, as in the Polish example (3.51). This combination is also historically attested – once more in Russian, see (3.53). (3.53) Один анбар, что в котором была толчея, и тот анбар водою RUS подмыло, и он от подмою воды завалился. Odin anbar, [čto v kotorom byla tolčeja,] one shop RPAR in RPRO.PRP.M.SG was crowd i tot anbar vodoju podmylo, i on and that shop water.by undermined and he ot podmoju vody zavalilsja. from undermining water.of fell.down

102 Data classification ‘A shop in which there were many people was undermined by water and because of this it fell down.’ (17th century, official document, Troickij 1959: 163) Alternatively, the element što may be interpreted as a complementizer, since in Russian the complementizer and the relative particle have the same form. This may be supported by evidence from Rusyn (Duličenko 20065: 131), a minor Slavic language spoken in different central European areas, e.g. in Vojvodina (northern Serbia). In Rusyn the complementizer že is set before subordinate and indirect interrogative clauses, as in (3.54). (3.54) Вона ше пита дзивки, же хто/цо бул. Rusyn Vona še pita dzivki, že chto/co bul. she REFL asks girl COMP who/what was ‘She asks the girl who he was/what it was.’ (Duličenko 20065: 131) Combinations of a SIMPLE RELATIVE ELEMENT WITH A SPECIALIZED RELATIVE ELEMENT are also extremely rare: they are attested in Danish and in Russian dialects. In (3.55), the Danish relative particle som combines with the relative element der, specialized for the relativization of subjects. In (3.56), the Russian relative particle štu combines with jde, specialized for the relativization of locatives. Since Akimova does not provide information on the context in which (3.56) was uttered, both given interpretations are equally possible. (3.55) Kender du den mand [som der talte me DAN know you that man RPAR SRE spoke with hende]? her ‘Do you know the man who talked to her?’ (Platzack 2002: 83) (3.56) Да пашли мы па тей-та дароги, шту йде тада мяшок RUS патиряли. Da pašli my pa tejta darogi, [štu PAR went we on that- PAR street, RPAR jde tada mjašok patirjali.] SRE then sack we.lost ‘And we went on that street where we had lost the/a sack.’ ‘Let’s go on that street where we lost the/a sack!’ (Kursk dialect, Akimova 1964: 142)

Relative element 103

Combinations of a SIMPLE RELATIVE ELEMENT WITH A RELATIVE PARTICLE are attested in few languages: in Basque, German, Lower/Upper Sorbian and Macedonian a relative pronoun combines with a relative particle, as the examples below illustrate. (3.57) s Hulz, [ás deən wos dös GER DET wood from RPRO.DAT.NT.SG RPAR that gmàcht is] made is ‘The wood with which it is made.’ (North Bavarian dialect, Fleischer 2004b: 65) (3.58) Źo jo ten luź, [tom ak ja LSO where is that person RPRO.DAT.M.SG RPAR I som cora pjenjeze dała]? am yesterday money given ‘Where is the person I gave the money to yesterday?’ (dialect, Faßke 1996: 170) (3.59) Su to te holcy [z tymi kiž USO are this DET boys with RPRO.INS.PL RPAR sym ja do šule khodźił]? am I to school gone ‘Are these the boys I went to school with?’ (dialect, Faßke 1996: 172) (3.60) Човекот, когошто го сретнав, не ми е познат. MAC Čovekot, [kogo-što go sretnav] man.DET RPRO.ACC.M.SG-RPAR CL.ACC.M.3SG I.met ne mi e poznat. not to.me is known ‘I don’t know the man I met.’ (Usikova 2003: 270) In French we have the combination of the specialized relative element où ‘where’ with the relative particle que, as (3.61) exemplifies. (3.61) Ça FRE it ai have

vient justement comes of.course du travail.] DET work

le DET

jour day

[où SRE

que RPAR

j’ I

104 Data classification ‘Of course, there comes the day when I have plenty of work.’ (Gadet 1997: 118) Similarly, in Icelandic, Norwegian and Swedish the specialized relative elements þar, der, där – all of them meaning ‘where’ – combine with the relative particle sem/som, yielding ICE þar sem, NOR der som and SWE där som (Smits 1989: 341, 390, 426). In Macedonian, the SREs kade ‘where’ and čij ‘whose’ combine with the relative particle što. Here the same observation can be made as for example (3.52). In French, Icelandic, Macedonian, Norwegian, Swedish, just like in Russian, the morphological form of a relative particle overlaps with the form of a complementizer. So, this case could also be interpreted as the combination of a specialized relative element with a non-relative particle. Combinations of a SIMPLE RELATIVE ELEMENT WITH NON-RELATIVE PARTICLES are not very frequent. In (3.62) the German relative pronoun combines with the locative-determinative particle da.12 (3.62) Ich GER I [die

habe have

’ne

Telefonnummer, phone.number lautet…] RPRO.NOM.F.SG PAR sounds ‘I have a new phone number, which is...’ (man, 30 y.o.) DET

neue new da

In the following examples, a relative particle combines with a complementizer – as in (3.63) and (3.64) – or with an emphatic particle, as in (3.65). (3.63) Jens, [som at vi alle kender...] DAN Jens RPAR COMP we all know ‘Jens, whom we all know…’ (Platzack 2002: 81) (3.64) Jón, [sem að keypti husið.] ICE Jón RPAR COMP bought house.DET ‘Jón, who bought the house.’ (Platzack 2002: 82) (3.65) Duży Albin, [co to doskonale czyta.] POL big Albin, RPAR PAR wonderfully reads ‘Big Albin, who can read very well.’ (dialect, Urbańczyk 1939: 31)

Relative element 105

As the examples above show, the particles with which simple relative elements combine are of two kinds: either emphatic-determinative, like GER da and POL to, or complementizers, as DAN at and ICE að. Finally, there is also an instance of combination between the zeromarker and a particle: it is attested in Russian dialects. The particle to, which we find also in (3.65), is enclitic to the first word of the RC. (3.66) Поправился тот мужик змий-то укусил? RUS Popravilsja tot mužik [Ø zmij- to ukusil]? recovered that peasant snake PAR bit ‘That peasant who was bitten by a snake, is he well now?’ (dialect, Šapiro 1953: 58) Combinations of a SIMPLE RELATIVE ELEMENT WITH A RESUMPTIVE ELEMENT are the most frequent in the languages of the sample. Although the terms ‘resumptive pronoun’ or ‘resumptive clitic’ are often used in typological literature on relative clauses, the term ‘resumptive element’ is preferred here: as a matter of fact, resumptives can be pronouns, but also adverbs; they can be clitics, but also free morphological forms. Resumptives can be considered as representatives of the relativized item in the RC. As such, they can encode up to three of the four features according to which relative elements were classified: the syntactic role of the relativized item in the RC, gender and number agreement. The resumptive elements occurring in the sample are illustrated in the examples below. In (3.67) and (3.68) a relative particle occurs respectively with a clitic and a tonic personal pronoun; in (3.69) the resumptive consists of a preposition and a personal pronoun. (3.67) Dziecko, [co mu dałeś pieniądze,] POL child RPAR CL.DAT.NT.3SG you.gave money jest cyganem. is gipsy ‘The child you gave the money to is a gipsy.’ (Anna Górska, from questionnaire) (3.68) Così ITA so la DET

dice says DILIT DILIT

questo this ha has

tipo chap preso taken

qui, here lui him

[che RPAR

come as

poi di fatto then in.fact modello.] modello

106 Data classification ‘This is what this chap says, who in fact was taken as a model by DILIT later on.’ (woman, teacher, 45 y.o.) (3.69) Είναι πάντα πολύ δύσκολο να βρεις θέματα που να μιλάς για αυτά GRE στο μάθημα και να μη βαριούνται οι μαθητές. Íne pánta polí ðíscolo na vris is always very difficult COMP you.find θémata [pu na milás jia aftá sto topics RPAR COMP you.speak about them in.DET máθima ke na mi variúnte i lesson and COMP not get.bored DET maθités.] pupils ‘It’s always hard to find topics that you can cover during the lessons without boring the pupils.’ (Sofia Lampropoulou, from questionnaire) Also possessives can be used as resumptive elements: this is the case with joho in (3.70). (3.70) Їдемо гірским хребтом, що зветься Чорногора, а що його UKR найвищою горою є Говерля. Jidemo hirskym chrebtom, [ščo zvet’sja we.go mountainous ridge RPAR is.named Čornohora,] a [ščo joho najvyščoju horoju Čornohora and RPAR his highest mountain je Hoverlja.] is Hoverlja ‘We go along a mountain ridge named Čornohora, whose highest point is the Hoverlja Mountain.’ (Pugh and Press 1999: 296) It needs to be remarked at this point that the distinction between clitic and tonic pronouns is not equally distributed cross-linguistically. Some languages, like Romance languages, Dutch and many Slavic languages, have two sets of personal pronouns, a clitic (bound) one and a tonic (free) one; other languages, like English, German and Scandinavian languages have only one pronominal set (Zifonun 2001b). Additionally, dialects may differ from the standard variety: for instance, clitic pronouns are attested in many German dialects (Fleischer 2004a). The possibility of using both pronominal sets in RCs varies not only from language to language, but also lan-

Relative element 107

guage-internally. Since this is a cross-linguistic study, this issue will not be pursued in detail; the term ‘personal pronoun’ will be generally used to refer to both sets. Adverbs, too, are frequently used as resumptive elements. In (3.71) and (3.72) we have two instances of pronominal adverbs; (3.73) shows the occurrence of a circumstantial (local) adverb. (3.71) Nous allions nous promener dans la FRE we used.to.go REFL walk in DET forêt [où nous y construisions des cabanes.] wood SRE we CL.ADV used.to.build DET huts ‘We used to go for a walk in the wood where we used to build huts.’ (Gapany 2004: 187) (3.72) Themen [wo du GER topics RPAR you dass sich die

drüber sprichst, ohne PRO.ADV speak without Schüler langweilen.] COMP REFL DET pupils bore ‘Topics that you can cover without boring the pupils.’ (Anna Riester, from questionnaire)

(3.73) Pues nada, pasamos al claustro, [donde SPA then nothing we.passed to.DET cloister SRE allí se dio un lunch, o como se there REFL gave DET lunch or how REFL llame.] calls ‘And then we went over to the cloister, where they gave ‘lunch’, or whatever it is called.’ (Westedt 1997: 69) Here again, it must be observed that pronominal adverbs are cross-linguistically not equally distributed. Catalan, Dutch, French and Italian have clitic pronominal adverbs, resp. CAT hi, en, DUT er, FRE y, en and ITA ci, ne. Germanic languages possess pronominal elements consisting of the combination of the local adverb ‘there’ with a preposition: for instance, GER drüber in (3.72) results out of the combination of da ‘there’ and über ‘over’. The syntax of pronominal adverbs in German and German dialects is covered in Fleischer (2002).

108 Data classification One last remark regards the so-called stranded adpositions. These elements cannot be catalogued as resumptives because they are not representatives of the relativized item in the RC. However, they convey its syntactic role. For instance, in (3.74) the preposition i ‘in’ expresses the role of the relativized item in the relative clause, i.e. LOC. (3.74) Sengen [som du sover i] er hundre NOR bed.DET SRE you sleep in is hundred år gammel. year old ‘The bed where you sleep is a hundred years old.’ (Strandskogen and Strandskogen 19862: 121) In the following of this work stranded adposition will be associated with resumptive elements. Strictly speaking, they do not belong to this class. Still, the function they fulfil within the RC is close to the function of resumptives: they convey the syntactic role of the relativized item when the simple element with which they combine fails to do this. Table 3.8. Resumptive elements attested in the sample.

Personal pronoun Adposition+pers.pronoun (pronominal PP) Possessive Pronominal adverb Circumstantial adverb (Stranded adposition)

Syntactic role + +

Number agreement + +

Gender Combinable agreement with: + RPRO, RPAR + RPRO, SRE,

+ +

+/– –

+/– –

SRE, RPAR

+ +

– –

– –

SRE, RPAR

RPAR RPRO, SRE, RPAR RPRO, RPAR, Ø

The different kinds of resumptive elements attested in the sample and the features they encode are reported in Table 3.8 above. As the table shows, a major distinction can be made between personal pronouns, which encode the syntactic role of the relativized item as well as gender and number agreement, and adverbs, which only encode the syntactic role of the relativized item: pronominal elements are more explicit than adverbial elements – see the explicitness scale introduced in 1.1.2., based on Lehmann’s (1984: 225) scale of anaphora. The rightmost column of Table 3.8 reports the simple elements with which each resumptive element can combine.

Relative element 109

Possessives do not display in all languages morphologically distinct forms to express gender and number agreement: this is why the sign ‘+/–’ is used. For instance, whereas Russian distinguishes between the masculine form ego ‘his’, the feminine form eë ‘her’ and the plural form ich ‘their’, Spanish has only one form, su ‘his’/‘her’/their’, valid for both genders and numbers. So, in Spanish possessives gender and number agreement is neutralized. As a matter of fact, all simple relative elements can combine with a resumptive element: (3.67) to (3.74) show how resumptives combine with specialized relative elements and relative particles; still, resumptives also combine with relative pronouns and the zero-marker (see the rightmost column of Table 3.7). In (3.75) a relative pronoun combines with a clitic pronominal adverb; in (3.76) the zero-marker combines with a stranded preposition. (3.75) Jeffrey prépare sa valise [dans laquelle FRE Jeffrey prepares his suitcase in RPRO.F.SG il y met quelques vêtements.] he CL.ADV puts some clothes ‘Jeffrey packs his suitcase, in which he puts some clothes.’ (Gapany 2004: 187) (3.76) Här SWE here att

är en rolig historia [Ø ni kommer is DET funny history you come skratta åt.] COMP laugh.INF at ‘This is a funny story you will be amused by.’ (Holmes and Hinchliffe 1994: 221)

Resumptive elements occur also in Turkish. In the examples below, the link between the MC and the RC is encoded by means of dedicated verb forms, whereas the syntactic role of the relativized item in the MC and number agreement (Turkish makes no gender distinctions) are conveyed by resumptives: a possessive suffix in (3.77), an instance of genitive relativization; a tonic pronoun in (3.78), where an indirect object is relativized. It must be pointed out that Turkish does not possess dedicated third person pronouns, so either a demonstrative (bu, şu, o) or the reflexive pronoun kendi serve as resumptive pronouns.

110 Data classification (3.77) [Ustanın kapısını değiştireceği] TUR engineer.GEN door.POSS.3SG.ACC change.PTCP.FUT.POSS.3SG çamaşır makinası. washing mashine.POSS.3SG ‘The washing mashine of which the engineer will change the door.’ (Göksel and Kerslake 2005: 444) (3.78) [Geçen ay kendisine borç olarak beş TUR last month self.POSS.3SG.DAT loan as five yüz bin lira verdiğimiz] hundred thousand lira give.NMLZ.POSS.1PL komşuyu bir daha görmedik. neighbour.ACC one more we.not.saw ‘We haven’t seen the neighbour again(,) to whom we lent 500,000 liras last month.’ (Haig 1998: 100) We may now check which of the four parameters used to classify relative elements are encoded by combined relative elements. As discussed above, in combined relative elements the four parameters can be encoded by different morphosyntactic units. If two simple relative elements combine with each other, it can be assumed that the parameters encoded by each of the elements are added, so some parameters may be encoded twice. For instance, when a relative particle combines with a specialized relative element, as in (3.55) and (3.56) above, the link between the MC and the RC is expressed twice. This is sometimes also the case when a simple relative element combines with a resumptive element: for example, when a relative pronoun combines with a clitic adverb, as in (3.75), the syntactic role of the relativized item in the RC is encoded twice. As regards the combination of a simple relative element with a particle, if this particle is a complementizer, as in (3.63), then the link between the MC and the RC is encoded twice; other particles, as in (3.65), do not encode any of the four parameters. Table 3.9 provides an overview of the parameters encoded by combined relative elements. In two cases the symbol ‘+/–’ was used: as for resumptive elements, only personal pronouns encode gender and/or number agreement, whereas adverbs only encode the syntactic role of the relativized item in the RC. As for the category ‘other particle’, encoding of the link between the MC and the RC depends on the kind of particle: if it is a complementizer, it encodes this parameter, if not, it does not encode any parameters.

Relative element 111 Table 3.9. The features encoded by combined relative elements. Encoding of: Element: 1 RPRO+ RPRO

2 RPRO+ RPAR

3 RPRO+ other particle 4 RPRO+ resumptive elem. 5 SRE+ RPAR

6 SRE+ resumptive elem. 7 RPAR+ RPRO

8 RPAR+ SRE

9 RPAR+ other particle 10 RPAR+ resumptive elem. 11 Ø+ other particle 12 Ø+ resumptive elem.

link MC–RC

syntactic role

+ + + + + +/– + – + + + – + + + + + +/– + – – – – –

+ + + – + – + + + – + + – + – + – – – + – – – +

gender number agreement agreement + + + – + – + +/– – – – +/– – + – – – – – +/– – – – +/–

+ + + – + – + +/– – – – +/– – + – – – – – +/– – – – +/–

attested in n lgs. 1? 5 2 12 5 8 2 2 3 31 1 5

Looking at Table 3.9 one may notice that combined elements in which the link between the MC and the RC or gender and number agreement are expressed twice are not very frequent – they are attested in less than 10 languages. Combined elements in which the role of the relativized item in the RC is expressed twice are relatively more frequent: they occur in more than 10 languages. Still, in the most frequently attested combination (31 languages), i.e. ‘RPAR+resumptive element’, doubling does not occur for any of the features: the particle encodes the link between the MC and the RC and the resumptive element encodes the other parameters. This suggests that doubling of parameters tends to be avoided in combined elements: preference is given to combinations in which different morphosyntactic units encode different features. The occurrence of combined

112 Data classification elements in the languages of the sample will be further discussed in 4.2.4. A comparison of the features encoded by simple and combined relative elements is proposed and commented in section 4.4. 3.3. Syntactic positions relativized The third parameter according to which relativization strategies in European languages will be classified is the syntactic positions which a strategy can relativize. The analysis will be based on Keenan and Comrie’s (1977) ACCESSIBILITY HIERARCHY. This hierarchy is an implicational pattern which aims to describe the cross-linguistic accessibility of syntactic positions to relativization. Keenan and Comrie (1977) start from a basic distinction between case-coding and non-case-coding ([+case] and [–case]) RC formation strategies. In the first type of strategies the relativized item in the RC is encoded “AT LEAST as explicitly as is normally done in simple declarative clauses” (Keenan and Comrie 1979b: 656), in the second it isn’t. So, a relative pronoun or a relative particle combined with a resumptive element would be instances of [+case] strategies, whereas a relative particle alone or the zero-marker represent [–case] strategies. Then, Keenan and Comrie (1977) look at which syntactic positions can be relativized through [+case] or [–case] strategies. The result is the following Accessibility Hierarchy (AH): SU > DO > IO > OBL > GEN > OCOMP The hierarchy is reproduced as it appears in Keenan and Comrie (1977: 66). The abbreviations are to be read as Subject, Direct Object, Indirect Object, Oblique Case, Genitive, and Object of Comparison. The Accessibility Hierarchy was conceived to apply only to restrictive RCs. All positions are to be interpreted as argumental, not as adverbial/circumstantial. The AH predicts that if a strategy can relativize one syntactic position on the hierarchy in a particular language, then it can relativize all positions to the left of that position. So, if in a language a strategy can relativize IO, then it can also relativize DO and SU. Keenan and Comrie also formulate three Hierarchy Constraints, valid for all languages:

Syntactic positions relativized

113

1. a language must be able to relativize subjects; 2. any RC-forming strategy must apply to a continuous segment of the AH; 3. strategies that apply at one point of the AH may in principle cease to apply at any lower point. (Keenan and Comrie 1977: 67)

The AH has been an object of discussion since its very formulation. For instance, it was pointed out that some syntactic functions like time and place circumstantials are not represented and that others, like OCOMP, can often be subsumed to OBL. As for the database used to formulate the hierarchy, Keenan and Comrie relied on a strongly biased language sample: 20 of the 50 languages considered are Indo-European and 14 of these 20 languages belong to the western branch of Indo-European. Moreover, the two scholars mostly consider only codified standard varieties of European languages and neglect a number of constructions which are attested in sociolinguistically marked varieties. This leads to incomplete results: for instance, three (non-standard!) varieties – Genoese, colloquial Czech and Zurich German – exhibit the combined relative element ‘relative particle+resumptive element’, along with (standard) Slovenian and Modern Greek. Still, if all constructions attested in each European language were considered, it would appear that this pattern is more widespread than assumed in Keenan and Comrie (1977): cf. section 3.2.2 above, where this combined element turns out to be attested in 31 of the 36 European languages constituting the sample of this investigation. In later contributions, Keenan and Comrie (19872) and Keenan and Hawkins (19872) discuss the psychological validity of the AH. They try to explain why, for instance, SU proves to be more accessible to relativization than OBL. The scholars link this to ease of comprehension: for lower positions on the AH, speakers would find it more difficult to retrieve the syntactic role of the relativized item. The same would be true for the production of RCs relativizing lower positions. This would be confirmed by the presence of resumptive pronouns in several languages, which are more likely to appear the lower one proceeds down the AH: they would constitute a help for the hearer to disambiguate the position relativized. In this sense, the AH would be reversed: if a resumptive element is inserted to relativize a certain position, then it is also inserted for all positions on the right. In a number of subsequent contributions, the AH was further discussed and partially reformulated. For instance, Comrie (1981) reduces it to only four positions, subsuming IO and OBL to the same position ‘non-direct object’ and excluding OCOMP. Interestingly, this reduced version parallels

114 Data classification Herrmann’s (2005) formulation of the AH, which she obtains through a quantitative analysis of the relativized positions in the corpus FRED (Freiburg Corpus of English Dialects, URL: http://www2.anglistik.unifreiburg.de/institut/lskortmann/FRED/index.htm, last access: June 30th, 2011). Lehmann (1984: 220) suggests an alternative articulation of the AH. First, he claims that complements (verbal arguments) should be kept separate from adjuncts (circumstantials); second, he places time and place arguments on the same position as IO; third, he states that one should actually postulate at least two AHs, one for adverbal, the other for adnominal functions. Unfortunately, Lehmann provides no cross-linguistic evidence supporting his claims – he only mentions English and Portuguese structures, but gives no examples. A further point made in Lehmann (1984) is that the AH looks slightly different for nominative-accusative and ergative-absolutive languages: in the former, the two most accessible positions are SU and DO; in the latter, ABS and ERG. This issue is further raised in Keenan (1985), who acknowledges that the syntactic position SU should be actually split on semantic grounds into S (non-agentive subject) and A (agentive subject): this way, one may formulate a single hierarchy for nominative-accusative and ergative-absolutive languages. In a study on relative clauses in conversational English, Fox (1987) states that the most frequently relativized positions are intransitive subjects (S) and patients (O) rather than transitive subjects (A), so that the upper part of the AH would look as ‘S > O > A’. Cristofaro’s (2003: Ch. 7) investigation yields similar results and hints at a basic opposition between A, S and O on one side and IO and OBL on the other. As a matter of fact, the AH actually proves to be a useful generalization to study cross-linguistic variation in RCs formation strategies. Although the initial language sample was not very balanced, robust evidence supporting the cross-linguistic validity of the AH has been collected ever since. This is why it will be used as a classifying parameter in this investigation, too. In particular, the rightmost position, OCOMP, will be not considered, since there is no sufficient cross-linguistic evidence in the database. Instead, information on circumstantials will be included: emphasis will be placed on place circumstantials (LOC), leaving aside time circumstantials (TEMP). For further information on the strategies adopted in Mediterranean and European languages to encode TEMP the reader is referred to Cristofaro and Giacalone Ramat (2002, 2007). As for the leftmost part of the AH, following Cristofaro and Giacalone Ramat (2007) it will be

Correlating the parameters 115

preserved as it was first formulated. The only ergative-absolutive language in the sample, Basque, can be fitted into the general pattern: Oyharçabal (1989) states that ergatives and absolutives can be both relativized with any restrictions, “without the one having priority over the other” (Oyharçabal 1989: 66). In this sense, both positions can be placed at the top of the AH. One last remark regards the distinction [+/–case], formulated in Keenan and Comrie (1977). As mentioned above, [+case] relativization strategies are those which univocally relativize syntactic positions: this implies that a relative element should have a different morphological form for each syntactic position. Instead, in [–case] strategies the relative element does not have a specific form for each position. We may share Cristofaro and Giacalone Ramat’s (2007: 70–71) objection that, at least in European languages, [+case] strategies would be quite rare if this distinction was strictly applied: in fact, relative elements show a high degree of case syncretism. For instance, as will be discussed in 3.5.5, the Italian relative element a cui ‘to+RPRO’ can be used to relativize both IO and OBL. Keenan and Comrie make themselves a similar observation on the (non-)distinction between the form used in Spanish to relativize DO and IO (Keenan and Comrie 1979b: 656 fn. 2). So, Cristofaro and Giacalone Ramat’s (2007) proposal is adopted: in [+case] strategies the relative element expresses the syntactic role of the relativized item, and the possibility is left open that there may be identical morphosyntactic forms for different positions; consequently, in [–case] strategies the syntactic role remains unexpressed. In this sense, relative pronouns and specialized relative elements are [+case], relative particles and the zero-marker are [–case]. 3.4. Correlating the parameters In the previous sections each of the parameters used to classify relative constructions in this work has been considered individually and nothing has been said about possible correlations between parameters. In 1.1.2, a relativization strategy was defined as one of the possible morphosyntactic patterns through which an RC can be formed. Now, strategies can be described through a number of parameters. If we consider the three parameters presented above, we may state that each relativization strategy can be described in terms of 1. the word order it exhibits, 2. the relative element it contains and 3. the syntactic positions it can relativize.

116 Data classification Table 3.10. Combinations of the two parameters ‘word order’ and ‘relative element’ attested in the sample.

prenominal postnominal correlative

RPRO

SRE

RPAR

Ø

– + +

– + +

+ + +

– + +

combined element – + –

First, we can look at the correlations between word order and relative element attested in the sample, which are reported in Table 3.10. In the following, examples for each correlation are provided. To start with, in (3.79) the RC is prenominal and contains a relative particle (as mentioned in 3.2.1, BAS -n is regarded as an affixal relative particle). (3.79) [Karrikan ikusi duzu-n] gizonak BAS street.DET.in seen you.have.him-RPAR man.DET.ERG hemen lan egiten du. here work doing he.has.it ‘The man you saw in the street works here.’ (Oyharçabal 1989: 63) Postnominal RCs can contain relative pronouns, as in (3.80), specialized relative elements, as in (3.81), relative particles, as in (3.82), and the zeromarker, as in (3.83). Combined elements, too, can occur in postnominal RCs. In (3.84) a relative pronoun combines with a relative particle. (3.80) Člověk, [který nekouří,] ušetří. CZE man RPRO.NOM.M.SG not.smokes saves ‘A man who doesn’t smoke saves money.’ (Petr 1987: 524) (3.81) Der er menge mennesker, [der ikke har DAN there are many people SRE not have tag over hovedet.] roof over head.DET ‘There are many people who don’t have a roof over their heads.’ (Allan, Holmes, and Lundskær-Nielsen 1995: 202) (3.82) A POR DET um DET

criança [que você child RPAR you cigano. gipsy

deu gave

o DET

dinheiro] money

é is

Correlating the parameters 117

‘The child you gave the money to is a gipsy.’ (Albertino Moreira, from questionnaire) (3.83) The boy [Ø we met yesterday at the movies] is a good friend of mine. (3.84) To su te gelice, [z tymi ak ja LSO that are DET chaps with RPRO.INS.PL RPAR I som do šule chejźił.] am to school gone ‘Those are the chaps I went to school with.’ (dialect, Faßke 1996: 170) Also correlative RCs can occur with relative pronouns, as in (3.85), specialized relative elements, as in (3.86), relative particles, as in (3.87), and the zero-marker, as in (3.88). (3.85) И который мальчик убил волка, он дал ему три рубля. RUS [I kotoryj mal’čik ubil volka], on and RPRO.NOM.M.SG boy killed wolf he dal emu tri rublja. gave him three ruble ‘And the boy who killed the wolf, to him he gave three rubles.’ (Rjazan’ dialect, Šapiro 1953: 107) (3.86) [De borošno,] tam i porošno. UKR SRE full.of.flour there also dusty ‘Where there is flour, there is also dust.’ (Gołąb and Friedman 1972: 35) (3.87) Что у тебя есть пластинка, щас будут передавать. RUS [Čto u tebja est’ plastinka] ščas budut RPAR at you is record now they.will peredavat’. broadcast ‘The record that you have, they’ll play it now on the radio.’ (Lapteva 1976: 303)

118 Data classification (3.88) Напротив живёт / уехала в Крым. RUS [Naprotiv živët] / uechala v Krym. opposite lives went.F.SG to Crimea ‘The woman living in front of us moved to Crimea.’ (Zemskaja 1987: 55) After establishing the possible occurrences of a specific word order with a specific relative element, one can look at which syntactic positions can be relativized by means of each strategy. Here, we are confronted with different possibilities. First, there exist strategies that can relativize all positions of the AH. One of them is the prenominal strategy with a relative particle: this is illustrated in the Basque examples below. In (3.89) to (3.91) SU/DO (corresponding to ERG/ABS), DO (corresponding to ABS) and IO are relativized. The relativization of these positions does not show any constraint. As Oyharçabal (1989: 66) argues, this has probably to do with the fact that in Basque subjects, direct and indirect objects are ‘grammatical cases’ and trigger agreement in the verb. As for OBL, it can be relativized only if the relativized item has the same syntactic role in the MC and in the RC: in (3.92), both relativized items have a comitative function (cf. also 4.3.4). Relativization of GEN is possible, as shown in (3.93). So, in spite of constraints on OBL and – to some extent – on GEN, the Basque strategy can relativize in principle all positions of the AH. (3.89) [ikusi du-en] gizonarentzat ekarri dut BAS seen he.has.him-RPAR man.DET.for brought I.have.it ‘I brought it for the man who saw him/her/it.’ ‘I brought it for the man who (s)he saw.’ (Oyharçabal 1989: 66) As neither argument of the RC verb ‘see’ is expressed, the above example yields two possible interpretations: according to the first one, SU is relativized; according to the second, DO is relativized. (3.90) [Patxik aipatu dauta-n] gizona BAS Patxi.ERG mentioned he.has.it.to.me-RPAR man.DET ezagutzen duzu. knowing you.have.him ‘You know the man that Patxi mentioned to me.’ (Oyharçabal 1989: 67)

Correlating the parameters 119

(3.91) [Loreak eman dizkioda-n] neska BAS flowers given I.have.them.to.her-RPAR girl.DET hor dago. there is ‘The girl I gave the flowers to is right here.’ (Trask 1998: 320) (3.92) [bizi naize-n] lagunarekin joan niz BAS living I.am-RPAR companion.DET.COM gone I.am ‘I went with the companion I am living with.’ (Oyharçabal 1989: 70) (3.93) [izena ahantzi duda-n] gizon batek BAS name.DET forgotten I.have.it-RPAR man one.ERG erran dau told he.has.it.to.me ‘A man whose name I forgot told this to me.’ (Oyharçabal 2003: 780) As Oyharçabal (2003) points out, GEN-relativization by means of the prenominal strategy in Basque is possible if the relationship between the possessor and the possessum is immediately recoverable. So, a possessive relationship between a person and his/her name, as in (3.93), can be regarded as more semantically transparent than the relationship between a person and his/her house (Oyharçabal 2003: 780). Other strategies that can relativize all positions of the AH are the postnominal strategies introduced by a relative particle or a relative pronoun: the former was shown in the Portuguese examples (3.34) to (3.36) above; the latter is illustrated by the Estonian examples below. In (3.94) to (3.98) the relative pronoun kes is used to relativize the following positions: SU, DO, IO, OBL and GEN. All examples are taken from Erelt (2003: 124). In (3.96), there is no number agreement between the referent in the MC and the relative pronoun: this is because Estonian uses the singular form of the relative pronoun unless plurality is to be explicitly highlighted (Erelt 2003: 124). (3.94) mees, [kes juhib EST man RPRO.NOM.SG is.driving ‘The man who’s driving the car.’

autot.] car

120 Data classification (3.95) mees, [keda ma EST man RPRO.PART.SG I ‘The man that I met.’

kohtasin.] met

(3.96) mehed, [kellele te maja müüsite.] EST men RPRO.ALL.SG you house sold ‘The men to whom you sold the house.’ (3.97) mees, [kellega ma EST man RPRO.COM.SG I ‘The man that I spoke with.’

rääkisin.] spoke

(3.98) mees, [kelle autoga ma EST man RPRO.GEN.SG car I ‘The man whose car I was driving.’

sõitsin.] was.driving

Second, there exist strategies that can relativize only a segment of the AH. Two examples are the correlative strategy introduced by a relative pronoun in Russian and the postnominal strategy introduced by the relative particle kenž in Lower Sorbian. As examples (3.99) to (3.102) show, the correlative strategy in Russian can relativize only the positions from SU to OBL. (3.99) Которые деньги останутся, положи в кошелек. RUS [Kotorye den’gi ostanutsja,] položi v RPRO.NOM.PL money.NOM will.remain put in košelëk. wallet ‘Put the remaining money in the wallet.’ (Zaliznjak and Padučeva 1975: 74) (3.100) Которых ребят увидишь, тем передай привет. RUS [Kotorych rebjat uvidiš’,] tem RPRO.ACC.PL boys.ACC you.will.see those.to peredaj privet. give greeting ‘Say hello to the boys that you will see.’ (Zaliznjak and Padučeva 1975: 73)

Correlating the parameters 121

(3.101) Которым дети (sic!) седьмой год – сейчас они в эту субботу RUS уезжают уже в Евпаторию. [Kotorym deti (sic!) sed’moj god] – sejčas RPRO.DAT.PL children.NOM seventh year now oni v ėtu subbotu uezžajut uže v they in this Saturday go already in Evpatoriju. Evpatorija ‘The seven-year-old children are already going to Evpatorija this Saturday.’ (Kručinina 1968: 85) In the RC in the example above the referent noun is in the nominative form, deti, instead of the expected detjam, the dative form. Still, the construction pattern is clearly correlative. (3.102) У которых комсомольцев была лопата, они рыли ямы. RUS [U kotorych komsomol’cev byla lopata,] at RPRO.GEN.PL komsomolers.GEN was shovel oni ryli jamy. they dug holes ‘The komsomolers that had a shovel dug holes.’ (Kručinina 1968: 84) In the above example, the genitive form kotorych is required by the preposition u: so, this is an instance of OBL-relativization. As for GEN-relativization, it seems to be only possible if the specialized relative element čej is used, as shown in the example below: however, this is a different strategy, because čej is a specialized relative element, not a relative pronoun. (3.103) Чья свеча заплывёт дальше, та девушка будет счастливее RUS всех. [Č’ja sveča zaplyvët dal’še,] ta devuška SRE.F.SG candle will.float furthest that girl budet sčastlivee vsech. will.be luckiest all.of ‘The girl whose candle will float furthest (from the shore) will be the luckiest of all.’ (Beličová and Sedláček 1990: 183)

122 Data classification Also the postnominal strategy introduced by the relative particle kenž in Lower Sorbian can relativize only a segment of the AH, i.e. SU and DO, as illustrated in (3.104) and (3.105). (3.104) Wiźiš muske, [kenž tam stoje]? LSO you.see man RPAR there stands ‘Can you see the man over there?’ (Janaš 1976: 187) (3.105) Pśedaj mě zešywka, [kenž njetrjebaš.] LSO give to.me copybook RPAR not.you.need ‘Give me the copybook you don’t need.’ (Janaš 1976: 187) Third, there are strategies than can relativize only a single position of the AH. This is for instance the case with postnominal strategies introduced by a specialized relative element – see examples (3.23) to (3.29) above. Finally, it must be pointed out that the positions relativized by a strategy may vary cross-linguistically: for instance, whereas in all the languages exhibiting the postnominal strategy introduced by a relative pronoun this strategy can relativize all of the positions of the AH, in languages exhibiting the postnominal strategy introduced by a relative particle this strategy tends to be used to relativize SU and DO (cf. the Lower Sorbian examples above). Only in a restricted number of languages can it relativize also IO, OBL and GEN: this is for instance the case with Portuguese. So, the correlation of the three parameters should first be attempted on a languagespecific level; then, different languages could be compared, so as to verify whether they possess similar strategies and whether these strategies can relativize the same syntactic positions. Additionally, the same strategy may behave differently language-specifically according to which variety is considered. So, while in standard Russian the relative particle čto is used only to relativize SU and DO, in non-standard varieties – dialects, prostorečie (a low urban sociolect) – it is also used to relativize lower positions of the AH (Švedova 1980: 524). See 3.5.6 for examples. 3.5. Classification problems The classification parameters outlined in the previous three sections help us find common patterns of variation within the relative constructions attested in European languages. Still, as it happens in any classification attempt, there are a number of issues that remain controversial. Also Gapany

Classification problems 123

notices the difficulty of unambigously classifying relative constructions, even within an individual language: [c]haque tentative de typologie débouche sur un nombre plus ou moins important d’ambigüités, c’est-à-dire de situations dans lesquelles un même énoncé peut être rangé dans plus d’une catégorie, sans qu’il soit possible d’opter de façon définitive en faveur de l’une ou de l’autre. (Gapany 2004: 169)

In order to partially overcome this impasse, in this section some cases of ambiguous classification will be discussed. In 3.5.1 to 3.5.3 issues related to simple relative elements, i.e. relative particles and the zero-marker, will be covered. Instead, 3.5.4 to 3.5.6 are devoted to questions of problematic classification of combined relative elements: the combination of a simple relative element with a particle, the status of adpositions and the combination of a relative particle with a resumptive element. 3.5.1. How ‘relative’ are relative particles? In 3.2.1, relative particles were defined as relative elements that convey only the link between the MC and the RC. Expressing the link between two clauses is not exclusive to relative particles, but is a common feature of conjunctions. In the database a number of conjunctions are attested which are not primarily used as relative elements, but which in particular contexts do have a relative function: in these contexts the relation between two SoAs patterns with the definition of relative relation given in 1.1.1. This raises the issue of the polyvalency of relative particles. A distinction can be made between three groups of particles. First, in several languages of the sample relative particles have the same morphosyntactic form as conjunctions or complementizers. These elements are attested in Romance languages (ITA che, FRE/POR/SPA que), Germanic languages (ENG that, SWE/DAN/NOR som), Slavic languages (BLR/BCS što, RUS čto, UKR ščo) and even Basque: as shown in Oyharçabal (1989: 64–65), -n also acts as a complementizer. Additionally, these elements also combined with prepositions, adverbs or pronouns and grammaticalized to new conjunctions over time: cf. POR/SPA porque ‘because’, where the complementizer combines with por ‘for’; SWE eftersom ‘as, since’, where the complementizer combined with efter ‘after’, or UKR tomu ščo ‘because’, where the complementizer combines with a demonstrative pronoun in the

124 Data classification dative. This issue was covered by Bernd Kortmann in his work on adverbial subordinators in European languages (Kortmann 1997). In another group of languages, a relative element extended its usage to a range of non-relative constructions (causal, temporal, concessive, etc.): this is for instance the case of ALB që (Lambertz 1959: 173), GER wo (Günthner 2002) and RUM de (Nilsson 1969: 27–28; Vulpe 1980: 136; Rusu 1984: 184). Examples (3.106) and (3.107) illustrate the relative and non-relative use of RUM de. (3.106) Era RUM was [de-

odată la noi în sat unul then at us in village one i ziceau Ionu Bodii.] RPAR CL.DAT.3SG they.said Ionu Bodii ‘At that time there was a man in our village called Ionu Bodii.’ (Nilsson 1969: 27)

(3.107) Mi se pare că eşti amorezat, de RUM to.me REFL seems COMPL you.are in.love because le faci toate pe dos. them you.make all on back ‘I guess you’re in love, since you’re doing everything the wrong way.’ (Nilsson 1969: 28) Third, in some languages (Bosnian/Croatian/Serbian, Bulgarian, Czech, Macedonian, Polish, Rumanian) particles that are not primarily relative and that usually code different relations may be used in relative constructions. These particles occur for instance in consecutive and final constructions and sometimes also act as complementizers with declarative verbs (e.g. CZE že, POL że, RUM că) or with modal verbs (e.g. MAC/BUL da, RUM să). This phenomenon is also historically attested: for instance, in Early New High German und ‘and’ and so ‘so’ could be used as relative particles (cf. section 6.3.3). As regards POL że and CZE že, in 6.3.1 it will be shown that they do have a relative origin, but later on grammaticalized to complementizers. So, their present-day use in relative constructions cannot be directly related to their relative origin: their use is more likely to have developed from consecutive constructions, as may be seen in the following examples. In (3.108) and (3.109) the SoA in brackets identifies a participant of the main SoA – ‘winter’ and ‘man’ – among a range of possible referents.

Classification problems 125

(3.108) Zimu, [že by řeka zamrzla až na CZE winter RPAR would.be river frozen even to dno,] nepamatuji. bottom not.I.remember ‘I don’t remember a winter that the river froze to the very bottom.’ (Petr 1987: 526) (3.109) człowiek, [że lepszego nie znajdziesz] POL man RPAR better not you.will.find ‘A man that you won’t find any other better.’ (Topolińska 1998: 221) This construction originates from a consecutive construction, in which a demonstrative meaning ‘such’ (POL taki, CZE taký) or an adverb meaning ‘so’ (POL, CZE tak) occurs in the MC (Topolińska 1998: 221; Petr 1987: 526). In fact, (3.108) may be also attested as in (3.110). (3.110) Takú zimu, [že by řeka zamrzla CZE such winter COMP would.be river frozen až na dno,] nepamatuji. even to bottom not.I.remember ‘I don’t remember such a winter that the river froze to the very bottom.’ (Petr 1987: 526, modified) Interestingly, in (3.111) the complementizer could be replaced by a relative pronoun, which would change the construction into a (primarily) relative one. (3.111) Takú zimu, [v které by řeka CZE such winter in RPRO.PRP.F.SG would.be river zamrzla až na dno,] nepamatuji. frozen even to bottom not.I.remember ‘I don’t remember such a winter in which the river froze to the very bottom.’ (Petr 1987: 526, modified) So, it seems reasonable to consider constructions like (3.108) and (3.109) as instances of relative relations, although they structurally originate from a different kind of construction. Similar considerations apply to Rumanian constructions in (3.112) and (3.113): să and că typically appear in complement clauses; in these examples, they introduce relative clauses.

126 Data classification (3.112) da om [să doarmă cu ochii RUM but man RPAR sleeps with eyes.DET deschişi] nam văzut open not I.have seen ‘But a man who sleeps with open eyes, well, I have never seen this.’ (Vulpe 1980: 136) (3.113) acceleratu de Brad cu vagoane [că nu RUM local.train from Brad with waggons RPAR not poţi trece dintr- unu în altu] // you.can pass from one in other ‘The local train from Brad with waggons that one cannot go from one to the other.’ (Vulpe 1980: 137) The particle da in Bosnian/Croatian/Serbian, Bulgarian and Macedonian seems to leave several interpretations open, as multiple translations of (3.114) and (3.115) suggest. Mišeska-Tomić (2006: 435) specifies that constructions like (3.114) are attested in Macedonian only when the verb of the MC is ima ‘to have’ or bara ‘to look for’. (3.114) Имам браќа да ми помогнат. MAC Imam brak’a [da mi pomognat.] I.have brothers RPAR me help ‘I have brothers who help me.’ ‘I have brothers to help me.’ (Mišeska-Tomić 2006: 435) (3.115) Uvijek je tako teško naći teme BCS always is so difficult find.INF topics [da razgovaraš o njima na času] i RPAR you.speak about them in lesson and [da učenicima ne bude dosadno.] RPAR pupils.to not will.be boring ‘It’s always so difficult to find (such) topics that you can cover during the lesson without boring the pupils.’ ‘It’s always so difficult to find topics in order to cover them during the lesson and not to bore the pupils.’ (Milena Marić, from questionnaire) The relation between the two SoAs in (3.114) and (3.115) can be seen as relative and final at the same time. In this case, multiple interpretations are

Classification problems 127

essentially connected with the polyvalency of da in the above mentioned languages (a peculiarity that they share with other Balkan languages: cf. the elements ALB të, GRE na, RUM să). For the purposes of this work, constructions containing these elements are regarded as cases of mixed relations, i.e. as instances of relative and of another relation. For a thorough analysis of polyvalent connectors in Balkan languages the reader is referred to Bužarkova (2004) and Mišeska-Tomić (2006: Ch. 6). The diachronic development of these connectors makes it quite difficult to pigeonhole them only as relative particles: most of them also have several other functions. The label ‘relative particle’ may be abandoned and a more generic one may be adopted, such as ‘complementizer’ or ‘connector’. Still, this is not satisfactory, because these elements mostly have different functions in different contexts. Hence, an element can be differently defined according to the construction in which it occurs. For instance, RUM de is a relative particle, i.e. it has a relative function, in (3.106), and it is a causal conjunction, i.e. it has a causal function, in (3.107). To summarize, relative particles can be considered as a subgroup of the category ‘connectors’. All connectors share a general function, i.e. signalling the link between two clauses. The nature of the link and, consequently, the nature of the relation between the two linked SoAs allows classifying connectors into different subgroups: for instance, the causal subgroup, the temporal subgroup, etc. The possibility exists that the same connector may belong to different subgroups, if it can be used in constructions encoding different kinds of relations between SoAs. So, the polyfunctionality of connectors can be also interpreted in terms of ‘construction-boundedness’: the specific function(s) of a connector can be established only by considering the construction(s) in which it occurs. In this sense, relative particles can be seen as connectors that can be used to encode relative relations. This does not impede that the same connectors – both coordinating and subordinating – may also appear in other constructions and encode other kinds of relations. This is due to the fact that the opposition between coordination and subordination is of a cognitive-pragmatic nature, as mentioned in 1.1.1: it is connected with the illocutionary force of the SoAs involved in a relation, i.e. with their being or not being at issue. Instead, the relation between the SoAs is of a semantic nature. A typical case of polyfunctional connector is BCS te, which is is attested in standard Bosnian/Croatian/Serbian and in southern dialects, particularly in Montenegro (Gołąb and Friedman 1972: 46). In (3.116) and (3.117) it has a consecutive function irrespectively of the fact that the relation between the two SoAs can be one of either coordination or subordina-

128 Data classification tion, as the multiple translations reveal. In both cases the semantic meaning of the relation is ‘the second SoA is a consequence of the first SoA’. (3.116) Prvi obišao morem Afriku te otvorio BCS first he.went.around over.see Africa so he.opened novi put u Indiju. new route in India. ‘He was the first to circumnavigate Africa and so he opened a new route to India.’ ‘He was the first to circumnavigate Africa, so that he opened a new route to India.’ (Katičić 1986: 162) (3.117) Sama se bacila na pobožnost i BCS herself REFL was.keen on piety and milosrdnost te su je u gradu charity so they.are her in town već smatrali sveticom. already considered as.saint ‘She was keen on piety and charity, so in the town she was already considered as a saint.’ ‘She was keen on piety and charity, so that in the town she was already considered as a saint.’ (Katičić 1986: 162) In (3.118) and (3.119) the same connector is used with a non-restrictive relative function: the SoA in brackets supplies further information on a participant of the non-bracketed SoA, ‘tsar’ and ‘angel’ respectively. As the two translations suggest, this relation can be interpreted as one of coordination or of subordination. (3.118) Био jе jедан цар те се звао цар Дукљан. BCS Bio je jedan car [te se zvao been is one tsar RPAR/and REFL named car Dukl’an.] tsar Dukl’an ‘Once upon a time there was a tsar whose name was tsar Dukl’an.’ ‘Once upon a time there was a tsar and his name was tsar Dukl’an.’ (Gallis 1958: 145)

Classification problems 129

(3.119) Ja ugledam ispred sebe anđela [te me BCS I see before REFL angel RPAR/and me gleda nestašno i veselo.] he.looks impishly and merrily ‘I see before me an angel looking at me impishly and merrily.’ ‘I see before me an angel, and he looks at me impishly and merrily.’ (Katičić 1986: 162) Finally, in (3.120) the same connector is used in a restrictive relative construction. In this case, only one interpretation is possible. (3.120) Kao onaj [te za majkom tuguje.] BCS like that RPAR for mother mourns ‘Like he who mourns for his mother.’ (Katičić 1986: 162) So, looking back at (3.116) to (3.120), it emerges that the connector te in Bosnian/Croatian/Serbian has two main functions: a consecutive and a relative one. From a pragmatic point of view, the relation between the two SoAs can be a coordinative or a subordinative one when the connector has a consecutive or non-restrictive relative function; when te has a restrictive relative function, only one of the SoAs is at issue. 3.5.2. The morphosyntactic status of Turkish ki Special attention needs to be devoted to the Turkish particle ki. Its morphosyntactic status is quite controversial and cannot be unambiguously established sometimes. All grammars and studies concerned with ki (Auer 1990; Kornfilt 1997: 321; Lewis 2000: 210; Johanson 2006: 19) underline that it was borrowed from Persian, but was fully integrated in the morphosyntactic structure of Turkish. Actually, as will be shown in section 6.2.1, a relative particle (kim) had already developed in Old Turkic before the contact with Persian. Ki has two main functions: as an emphatic particle and as a connector. The first use is exemplified in (3.121) to (3.123). (3.121) Anlayamıyorum ki! TUR I.can’t.understand PAR ‘I just can’t understand!’ (Göksel and Kerslake 2005: 112)

130 Data classification (3.122) Londra sonbaharda o kadar güzel TUR London autumn.in that much beautiful olur ki! is PAR ‘In autumn London is sooo beautiful!’ (Göksel and Kerslake 2005: 112) (3.123) Siz ki bir Rastignac olduğunuzu TUR you PAR DET Rastignac be.NMLZ.POSS.2PL.ACC söylemekten hoşlanıyorsunuz. say.INF.from you.take.pleasure ‘And you, you take pleasure in saying you are a Rastignac!’, ‘And you, who take pleasure in saying you are a Rastignac!’ (Haig 1998: 124) In (3.124) and (3.125) ki is used respectively as a complementizer and a consecutive connector. (3.124) İnanıyorum ki herşey düzelecek TUR I.believe CONN everything will.come.right ‘I believe that everything is going to be all right.’ (Göksel and Kerslake 2005: 112) (3.125) Sana bunu söylüyorum ki sonradan TUR to.you this I.say CONN later.on şaşırmayasın you.shall.not.be.surprised ‘I’m telling you this now, so that you won’t be surprised later.’ (Göksel and Kerslake 2005: 112) In (3.126) ki is used to introduce a parenthetical statement – parallel to ENG which/what. (3.126) Bugün hava açarsa, [ki TUR today weather if.opens RPAR açacağını pek sanmıyorum,] open.NMLZ.POSS.3SG.ACC really I.not.believe bahçede mangal yakacaklarmış. garden.in barbecue they.are.said.to.be.going.to.put.on

Classification problems 131

‘If the weather brightens up today, which I don’t really think it will, they’re going to have a barbecue in the garden, as I heard.’ (Göksel and Kerslake 2005: 112) Finally, in (3.127) and (3.129) ki is used as a relative particle: it introduces a non-restrictive relative clause in (3.127), an embedded non-restrictive relative clause in (3.128) and a restrictive relative clause in (3.129). In (3.127) and (3.129) resumptive elements are inserted to encode the syntactic role of the relativized item as well as number agreement – gender is not morphologically encoded in Turkish. Although Kornfilt calls ki a “relative pronoun” (Kornfilt 1997: 321), according to the classification used in this work it rather falls under the category ‘relative particle’, because it only encodes the link between the MC and the RC. (3.127) Sınıflarında Gülten diye çok şımarık TUR class.POSS.3PL.in Gülten saying very self-indulgent bir kız vardı [ki Hülya ondan DET girl there.was RPAR Hülya that.from nefret ederdi.] hate did ‘In their class there was a very self-indulgent girl called Gülten, whom Hülya hated.’ (Göksel and Kerslake 2005: 458) (3.128) O kız, [ki hiç gülmezdi,] sonunda TUR that girl RPAR at.all not.laughed at.the.end güldü. laughed ‘That girl, who hadn’t laughed at all, eventually laughed.’ (Erkman-Akerson and Ozıl 1998: 323) (3.129) Bir çocuk [ki babası kendisini TUR DET child RPAR father.POSS.3SG self.POSS.3SG.ACC sevmez] katil olur. not.loves murderer becomes ‘A child whom his father doesn’t love becomes a murderer.’ (Kornfilt 1997: 62) Additionally, the form ki also functions as an attributive suffix: it changes circumstantials into adjectives that can be attributed to nouns or be substantivized. For instance, buradaki inslanlar means ‘the people who are

132 Data classification here’ (from burada, ‘here’), bügünkü toplantı means ‘today’s meeting’ (from bügün, ‘today’); okuldakiler means ‘those (people) who are in the school’ (okul-da-ki-ler, school-LOC-ki-PL). In this case ki has an interphrasal, not an interclausal function, so this use will not be further discussed. Before examining the morphosyntactic status of ki in the examples above, a parallel may be drawn between TUR ki and PER ke/ki (‫)ﻛﻪ‬. In Persian, the form ke is standard, the form ki is colloquial (Lambton [1953] 1974: 10). If the contexts of use of the particles in the two languages are observed, it may be noticed that they mostly overlap. In (3.130) and (3.131) PER ke has an emphatic function, as TUR ki in (3.121) to (3.123). (3.130) ‫ﻧﺮﻓﺘﻢ ﻛﻪ‬ PER naraftam ke not.I.went PAR ‘I didn’t go!’ (Lambton 1974: 155) (3.131) ‫ﺍﻭ ﻛﻪ ﻣﺮﺩﻩ‬ PER u ke morde he PAR dead ‘He, he’s dead.’ (Lambton 1974: 155) Moreover, ke is also used as a connector to introduce different kinds of subordinate clauses (Lambton 1974: 57, 155; Lehmann 1984: 86; Haig 1998: 121). In (3.132) ke has a consecutive-final function, similar to (3.125) above. (3.132) ‫ﺍﻳﻨﺮﺍ ﺑﺎﻭ ﺩﺍﺩﻡ ﻛﻪ ﺑﻤﻨﺰﻝ ﺑﺒﺮﺩ‬ PER inrā be=u dādam ke be=manzel this.ACC to=him I.gave CONN to=house bebarad he.may.bring ‘I gave this to him so that he may bring it at home.’ (Lambton 1974: 57) Finally, ke is also used in relative clauses – both restrictive and non-restrictive (Lambton 1974: 73–77). It may occur alone, as in (3.133), where SU is relativized, or with a resumptive element, as in (3.134), where IO is relativized.

Classification problems 133

(3.133) ‫ﻣﺮﺩﻳﻜﻪ ﺁﻧﺠﺎ ﺑﻮﺩ ﻛﺘﺎﺑﺮﺍ ﺑﻤﻦ ﺩﺍﺩ‬ PER mardi=[ke ānjā bud] ketābrā be=man man=RPAR there was book.ACC to=me dād gave ‘The man that was there gave me the book.’ (Lambton 1974: 73) (3.134) ‫ﻣﺮﺩﺋﻴﻜﻪ ﻛﺘﺎﺑﻬﺎﺭﺍ ﺑﺂﻧﻬﺎ ﺩﺍﺩﻩ ﺑﻮﺩﻳﺪ ﺭﻓﺘﻨﺪ‬ PER mardhāi=[ke ketābhārā be=ānhā dāde budid] men=RPAR books.ACC to=them given you.were raftand went ‘The men to whom you gave the books went.’ (Lambton 1974: 74) So, it seems that Turkish borrowed the whole spectrum of functions that ki has in Persian. Still, as Auer (1990) and Haig (1998: 125) point out, Turkish adapted it to its structure. In fact, both languages are OV-languages, but Turkish is left-branching (modifier-head), whereas Persian is rightbranching (head-modifier). This entails that a clause-initial subordinator should not exist in Turkish: indeed, Turkish subordinators are typically clause-final, as shown in (3.135), where the subordinate clause is given in square brackets. (3.135) Adam [hayalet görmüş gibi] sapsarı kesildi. TUR man ghost saw as.if very.pale turned ‘The man turned very pale, as if he had seen a ghost.’ (Göksel and Kerslake 2005: 466) If this is true, how should TUR ki be classified? First, it needs to be remarked that in some of the constructions in which ki occurs, it is not clear at all whether it formally belongs to the MC or to the non-MC. In (3.121) and (3.122) ki is undoubtedly clause-final and in (3.126) it is undoubtedly clause-initial, but in other cases this is not so clear. For instance, in (3.124) ki may also have an emphatic function and the sentence could be interpreted as ‘I do believe (it)! Everything is going to be all right’. Similarly, (3.127) could also be interpreted as ‘In their class there was a very selfindulgent girl called Gülten, and Hülya hated her’ or ‘In their class there was a very self-indulgent girl called Gülten – Hülya hated her’. Now, as Auer (1990: 281) points out, even Turkish native speakers cannot always

134 Data classification distinguish between the two functions of ki. Auer reports the following passage from a conversation. (3.136) sen şimdi dedin ki e // elektrikler kesiliyo (sic!) TUR you now said ? electricity is.cut.off ‘You’ve said now that they are going to cut off the electricity.’, ‘You’ve said it now! They’re going to cut off the electricity.’ (Auer 1990: 281) The speaker makes a pause after ki. Is it because in this case ki has an emphatic function? Still, the pause is preceded by a hesitation: if so, it could well be that ki acts here as a connector and that the speaker simply hesitates before producing the second clause. Auer proposes that ki should be regarded as a cataphoric particle (Auer 1990: 282): the speaker instructs the hearer that another clause may follow, although this does not necessarily need to be the case. Similarly, Haig, quoting Schroeder (1997: 359), sees clauses containing ki as “a textual exploitation of the postpredicative slot in Turkish” (Haig 1998: 125). Since the issue of the nature of ki is quite intricate, only cases in which ki is said to introduce an RC will be covered. For a discussion of ki in other contexts, the reader is referred to Auer (1990), Schroeder (1997), Haig (1998: 121–128) and references therein. To start with, Haig (1998: 127) states that RCs introduced by ki can be always changed into prenominal RCs – the default relativization strategy of Turkish. So, (3.127) to (3.129) above can be changed into (3.137) to (3.139). (3.137) Sınıflarında [Hülya’nın (kendisinden) TUR class.POSS.3PL.in Hülya.GEN self.POSS.3SG.from nefret ettiği] Gülten diye çok hate do.NMLZ.POSS.3SG Gülten saying very şımarık bir kız vardı. self-indulgent DET girl there.was ‘In their class there was a very self-indulgent girl called Gülten, whom Hülya hated.’ (Göksel and Kerslake 2005: 458, modified)

Classification problems 135

(3.138) [Hiç gülmeyen] o kız sonunda TUR at.all laugh.not.PTCP.PRS that girl at.the.end güldü. laughed ‘That girl, who hadn’t laughed at all, eventually laughed.’ (Erkman-Akerson and Ozıl 1998: 323) (3.139) [Babasının kendisini TUR father.POSS.3SG.GEN self.POSS.3SG.ACC sevmediği] bir çocuk katil not.love.NMLZ.POSS.3SG DET child murderer olur. becomes ‘A child whom his father doesn’t love becomes a murderer.’ (Kornfilt 1997: 62, modified) Haig (1998: 127) argues that the main difference between sentences like (3.127) to (3.129) and (3.137) to (3.139) is to be found in their pragmatic status. The first group of sentences “contains two almost equal assertions”; in the second group, “one assertion is lost, or backgrounded” (Haig 1998: 127). In other words, in the first group both SoAs would be at issue, whereas in the second group only one SoA would be at issue. This is certainly true for (3.127) and (3.128) and for their counterparts (3.137) and (3.138). Instead, in (3.129) only one SoA – ‘a child becomes a murderer’ – seems to be at issue. This suggests that (3.127) and (3.128) are instances of coordination, whereas (3.129) – as well as sentences (3.137) to (3.139) – is an instance of subordination. Semantically, (3.127) and (3.128) are nonrestrictive, as their pragmatic status suggests, whereas (3.129) is restrictive. So, we are faced with two interpretations. If we admit that ki is essentially an emphatic, clause-final particle, then (3.127) to (3.129) are not instances of relative relation, but of sequential coordination: in (3.127) ki would be an emphatic-cataphoric particle announcing a subsequent clause, as proposed in Auer (1990); in (3.128) and (3.129) ki would be an emphatic-topicalizing particle, similar to ki/ke in (3.123) and (3.131). However, if we accept the analysis according to which ki acts as a clause-initial connector, we have to admit that Turkish also possesses extraposed RCs – as in (3.127) – and postnominal RCs – as in (3.128) and (3.129). In this work the latter analysis is adopted, agreeing with those scholars who admit the possibility that Turkish has clause-initial connectors. This is not surprising per se. As Slobin notices, “[t]hroughout the history of the Turkic

136 Data classification languages, subordinating particles are repeatedly borrowed from neighboring languages or invented on neighbouring models, and verb forms constantly shift between participles and tensed forms” (Slobin 1986: 275). Beside ki, at least two other clause-initial connectors that derived from Persian are attested in present-day Turkish: eğer ‘if’, from PER agar (‫‘ )ﺍﮔﺮ‬if’, which may be inserted at the beginning of conditional clauses, and çünkü ‘because, for’, from PER cun ke (‫‘ )ﭼﻮﻥ ﻛﻪ‬because’.13 So, it may well be that ki was borrowed as a clause-initial connector along with the other two. As shown above, it was not only borrowed as a relative element: in present-day Turkish, ki appears in the manifold contexts in which it appears in Persian, and these include also restrictive and non-restrictive relative relations. Postnominal and extraposed RCs introduced by ki, like in (3.127) to (3.129), constitute an alternative to the prenominal strategy shown in (3.137) to (3.139): still, as Haig (1998: 20) stresses, the former strategies are not undermining the latter one, which remains the most used in present-day Turkish. Moreover, RCs introduced by ki are more frequently used in non-restrictive relative relations than in restrictive ones (Kornfilt 1997: 60; Göksel and Kerslake 2005: 457–458). In particular, they often introduce parenthetical remarks (Haig 1998: 125), as exemplified in (3.126). 3.5.3. Zero-marker or coordinate clauses? As mentioned in 3.2.1, the postnominal strategy introduced by the zeromarker is attested in few of the languages in the sample (Czech, Danish, English, Norwegian, Russian, Swedish). Now, in some cases it can be questioned whether we are in front of a zero-marker strategy or two coordinate clauses – i.e. whether we are faced with instances of relative relations or sequential coordination. This issue cannot possibly be given a definitive answer – at least, not at a cross-linguistic level. In standard English and Scandinavian languages, the zero-marker is used for all positions on the AH except for SU. Still, in non-standard varieties of the same languages the zero-marker is also attested for SU-relativization, as shown in (3.140). (3.140) Der ovre DAN there over Olsen.] Olsen

bor lives

en a

mand [Ø man

hedder is.called

Classification problems 137

‘Over there lives a man who is called Olsen.’ ‘Over there lives a man, (and) he’s called Olsen.’ (Platzack 2002: 78) From a pragmatic point of view, both of the SoAs constituting the sentence are asserted: so, we are faced either with a non-restrictive relative relation or with a coordination relation (see Table 1.2). From a morphosyntactic point of view, if (3.136) is interpreted as an instance of sequential coordination, then the second clause is to be regarded as subjectless. Still, Danish, like English and other Scandinavian languages, is a non-pro-drop-language: typically, in each clause the subject needs to be overtly expressed. So, one may either agree that non-standard varieties of these languages allow subjectless sentences or that in the same varieties, the zero-marker strategy extends to the SU-position. In accordance with the second of these options, such sentences are regarded here as cases of ‘paradigm regularization’ (cf. 4.2.4): if in a (standard) language a strategy can be used to relativize only a segment of the AH, it is possible that in non-standard varieties this limitation does not apply, so the relevant strategy can be used to relativize the whole AH. Instances of zero-marker in Slavic languages need to be dealt with separately. The structure of these sentences in English, Scandinavian and Slavic languages is quite similar, as may be noticed comparing (3.140) with (3.141) and (3.142) below. (3.141) Znal jsem člověka, [Ø měl jenom půlku plic.] CZE known I.am man had only half lungs ‘I’ve known a man who had only one lung.’ ‘I’ve known a man, (and) he had only one lung.’ (Petr 1987: 526) (3.142) Приехал товарищ работал в посольстве я его спрашиваю. RUS Priechal tovarišč [Ø rabotal v posol’stve] arrived comrade he.worked in embassy ja ego sprašivaju. I him ask ‘There arrived a/the comrade who worked in the embassy – I’ll ask him.’ ‘There arrived a/the comrade – he worked in the embassy: I’ll ask him.’ (Lapteva 1976: 294)

138 Data classification The main difference between sentences like (3.140) and (3.141) or (3.142) is that Slavic languages – at least in their non-standard varieties – are generally pro-drop-languages. In this sense, (3.141) and (3.142) could also be interpreted as instances of sequential coordination. Russian scholars have different opinions about this: Lapteva regards constructions like (3.142) as (spoken) functional equivalents both of RCs (Lapteva 1976: 301) and of coordinate clauses (Lapteva 1976: 303). The zero-marker can be substituted by the relevant form of the relative pronoun kotoryj in the former case and by the relevant form of the personal pronoun on in the latter. If we consider (3.142) again, these changes would look like (3.143a) and (3.143b) respectively. (3.143) a. RUS

b.

Priechal tovarišč, [kotoryj arrived comrade RPRO.NOM.M.SG rabotal v posol’stve,] ja ego sprašivaju. worked in embassy I him ask ‘There arrived a/the comrade who worked in the embassy – I’ll ask him.’ Priechal tovarišč, on rabotal v arrived comrade he worked in posol’stve, ja ego sprašivaju. embassy I him ask ‘There arrived a/the comrade, he worked in the embassy – I’ll ask him.’

Other scholars – like Zemskaja (1973: 327–329) and Širjaev (1986: 116– 123) – suggest that constructions like (3.142) should be treated as instances of so-called asyndetic clause clusters (bessojuznye složnye predloženija, lit. ‘conjunctionless complex sentences’, or bessojuznye polipredikativnye vyskazyvanija, lit. ‘conjunctionless polypredicative utterances’, cf. Širjaev 1986). These clusters occur more frequently in spoken than in written Russian. According to Širjaev (1986: 123), they are semantically ambiguous: the link between the SoAs is not signalled through a connector, so multiple interpretations are possible – temporal, causal, relative, consecutive, etc. This is for example the case in (3.144).

Classification problems 139

(3.144) Плохо я спала / зайду к этой сестре / вот давление меряет // RUS Plocho ja spala / zajdu k ėtoj sestre / badly I slept I.will.go to this nurse vot / davlenie merjaet / PAR pressure measures ‘I didn’t sleep well – I’ll go to this nurse who measures blood pressure.’ ‘I didn’t sleep well – I’ll go to this nurse, because she measures blood pressure.’ ‘I didn’t sleep well – I’ll go to this nurse, so that she measures my blood pressure.’ (Širjaev 1986: 123) Širjaev (1986: 123) explains that the relation between the second and the third cluster could either be a causal or a relative one. In fact, it may also be a consecutive one. Such cases can be regarded as instances of multiple relations, in which the interpretation remains ambiguous: if necessary, the context helps to disambiguate the assertion (Širjaev 1986: 124). Similarly, Miller and Weinert (1998) propose two different analyses. First, they analyze constructions like (3.142) and (3.144) as “three main clauses” (Miller and Weinert 1998: 119–120); later on, they regard them as instances “of a relative clause construction with an even lower degree of cohesion” (Miller and Weinert 1998: 352–353). Jim Miller (p.c.), who was asked for an explanation, confirmed that in such cases both analyses are to be considered valid: it has to be accepted that this ambiguity cannot be eliminated, unless one knows the exact context in which the relevant sentence was uttered. So, in this work these constructions will be regarded as instances of the zero-marker strategy – although we have to bear in mind that alternative interpretations are possible and in some case co-exist with the interpretation of these constructions as relative relations. 3.5.4. Simple or combined? In section 3.2.2, combined relative elements were introduced. These include combinations of a simple relative element with an emphatic-determinative particle, as in the Polish example below. (3.145) Duży Albin, [co to doskonale czyta.] POL big Albin, RPAR PAR wonderfully reads ‘Big Albin, who can read very well.’ (dialect, Urbańczyk 1939: 31)

140 Data classification In (3.145) the relative particle and the emphatic-determinative particle are two independent morphosyntactic units; the RC could also be introduced by the relative particle alone. Instead, in other languages the particle has grammaticalized and is nowadays regarded as an integral part of the relative element: this is why relative elements in (3.146) to (3.148) have been classified as simple relative elements, although they share a common origin with the combined element illustrated in (3.145). (3.146) Той бе човек който винаги бе в добро настроение. BUL Toj be čovek [kojto vinagi be v he was man RPRO.NOM.M.SG always was in dobro nastroenie.] good mood ‘He was a man who was always well-tempered.’ (Maria Manova, from questionnaire) (3.147) To so ludźo, [ktorymž móžeš dowěrjeć.] USO it are people RPRO.DAT.PL you.can trust.INF ‘These are people that you can trust.’ (Faßke 1996: 169) (3.148) Wiźiš muske, [kenž tam stoje]? LSO you.see man RPAR there stands ‘Do you see the man over there?’ (Janaš 1976: 187) Whereas in Polish and in other languages in which simple relative elements combine with non-relative particles (cf. Table 3.7 above) the simple relative element can also occur alone, i.e. not combined with the relevant particle, in Bulgarian and Lower/Upper Sorbian this is not possible: these elements always occur together with a particle, if they have a relative function. So, in (3.146) to (3.148) the relative elements can be broken down into koj-to, ktorym-ž and ken-ž respectively. Still, it can neither be argued that BUL koj or USO ktorym alone are relative pronouns, nor that LSO ken alone is a relative particle, because none of these elements occurs in this form with this function. The case of Macedonian in (3.60) above is different. Although kogošto is written as one word, it can be split into kogo-što; moreover, the pronoun koj, of which kogo is the accusative form, can occur alone as a relative pronoun (cf. Friedman 2002: 44; Usikova 2003: 269). In this respect MAC koj behaves differently than BUL koj, which cannot occur as a relative pronoun without -to – although some Bulgarian dialects exhibit the simple

Classification problems 141

from koj (Bauer 1967: 303; Radeva 2003: 288). In general, Bulgarian relative elements always occur in combination with -to; in Macedonian, the combination with što is optional. Cramer (2003: 152) points out that in the standard variety the relative pronoun koj and the specialized relative element čij ‘whose’ are more frequently used alone, whereas the specialized relative element kade ‘where’ mostly combines with što. To summarize, in nearly all Slavic languages combinations of simple relative elements with the emphatic particles ž(e) and to are diachronically attested (Vondrák 19282: 478–79, 485; Urbańcyzk 1939: 4, 31; Bauer 1962: 104; see also section 6.3.5). They originally served to distinguish between the interrogative and the non-interrogative (i.e. indefinite and relative) function of simple relative elements. Language-specific developments led to different synchronic situations: in some languages, such as Russian, there is no longer a trace of the emphatic particles and the relative and interrogative pronouns are nowadays morphologically identical (e.g. kotoryj ‘who/which’ and ‘which?’). In other languages the particles retained their function and are attached to each relative element: so, for instance, in Bulgarian we find the opposition between koj ‘who?’ as an interrogative and kojto ‘who’ as a relative pronoun. Slovenian retained this distinction only partially (Herrity 2001: 102–106): the relative pronoun kateri and the relative particle ki do not combine with any particle. Still, all other relative elements, both inflected and non-inflected, differ from their interrogative counterparts through the affix -r (deriving itself from the particle že through rhotacism): kdo vs. kdor ‘who’, kaj vs. kar ‘what’ (both used in free relatives), kakšen vs. kakršen ‘which’ (a relative pronoun with a more restricted use, cf. 5.2.1), čigav vs. čigar ‘whose’, kje vs. kjer ‘where’ (both specialized relative elements). 3.5.5. The status of adpositions Stranded adpositions were dealt with together with resumptive elements in 3.2.2: there, it was said that they encode the syntactic role relativized, although they do not constitute representatives of the relativized item within the RC. In the language sample analyzed in this work, stranded adpositions are attested in Germanic languages and in French.

142 Data classification (3.149) Här är en rolig historia [som / Ø ni SWE here is DET funny history RPAR you kommer att skratta åt.] come COMP laugh.INF at ‘This is a funny story you will be amused by.’ (Holmes and Hinchliffe 1994: 221) (3.150) Elle avait son manteau [qu’ elle allait FRE she had her coat RPAR she went régulièrement au marché avec.] regularly to.DET market with ‘She had her coat, which she used to wear when she went to the market.’ (Gadet 1997: 118) In the vast majority of the languages in the sample, adpositions are also used in combination with relative pronouns to relativize some positions of the AH: in particular, in synthetic languages adpositions are used only for OBL and some circumstantials, whereas in analytic languages they are used for several positions of the AH starting with IO. Additionally, in some analytic languages like Spanish, Portuguese and Rumanian, a preposition (SPA/POR a, RUM pe) is used also for DO-relativization in cases of differential object marking (see Bossong 1991). Still, this preposition is never stranded. Prepositions usually precede the relative pronoun, as (3.151) and (3.152) show: this method of coding the relativized item in the RC mirrors the way of coding syntactic positions in simple sentences. As Haig puts it, “[m]any languages essentially use the same means for case recovery in RCs as they do in simple sentences, for example via case marking morphology or adpositions” (Haig 1998: 129). (3.151) Zanimam se samo za probleme, [za SLN I.am.interested REFL only for problems for katere hitro najdejo rešitev.] RPRO.ACC.F.PL easy they.find solution ‘I’m only interested in the problems people quickly find a solution for.’ (Sansò 2005, SLV_072_a) (3.152) Il ITA DET i DET

bambino child soldi] è money is

[a to uno DET

cui

hai you.have zingaro. gipsy RPRO

dato given

Classification problems 143

‘The child you gave the money to is a gipsy.’ (Marianna Gionta, from questionnaire) Cristofaro and Giacalone Ramat (2007) regard this combination as a separate relativization strategy. In fact, whether this is to be seen as a simple or combined element depends very much on the function attributed to the adposition. If the adposition is claimed to encode the parameter ‘syntactic role of the relativized item’, whereas the relative pronoun expresses the link between the MC and the RC as well as gender and number agreement, then we are faced with a combined element. However, for the sake of cross-linguistic comparability, it seems reasonable to regard the combination ‘adposition+relative pronoun’ as a simple strategy. On the one hand, adpositions in analytic languages have the same role of case endings in synthetic languages: if we compare (3.152) with (3.153), we may notice that in both cases IO is relativized. This is expressed by a preposition in (3.152) and by the ending of the relative pronoun in (3.153). (3.153) Ребёнок, которому ты дал деньги, – цыган. RUS Rebënok, [kotoromu ty dal den’gi,] – child RPRO.DAT.M.SG you gave money cygan. gipsy ‘The child you gave the money to is a gipsy.’ (Filipp Lupov, from questionnaire) On the other hand, even synthetic languages use adpositions to relativize some positions of the AH. If we compare (3.154) and (3.155), in both cases OBL is relativized: this is expressed by the preposition a ‘at’ in (3.154) and by the preposition v ‘in’ and the case ending -oj in (3.155). (3.154) La conferenza [a cui abbiamo ITA DET conference at RPRO we.have partecipato] è stata proprio interessante. taken.part is been really interesting ‘The conference in which we took part was really interesting.’ (Marianna Gionta, from questionnaire)

144 Data classification (3.155) Конференция, в которой мы приняли участие, была очень RUS интересная. Konferencija, [v kotoroj my prinjali conference in RPRO.PRP.F.SG we took učastie,] byla očen’ interesnaja. part was very interesting ‘The conference in which we took part was very interesting.’ (Filipp Lupov, from questionnaire) As a matter of fact, clause-initial and clause-final adpositions show a complementary distribution. The comparison of (3.149)–(3.150) with (3.151)– (3.155) shows that stranded adpositions combine with relative particles and with the zero-marker, i.e. with simple relative elements which only encode the link between the MC and the RC: in this case, stranded adpositions may be seen as encoding the syntactic role of the relativized item. In turn, non-stranded adpositions combine with relative pronouns, i.e. with elements that encode at least two of the four parameters reported in Table 3.5; when pronouns are inflected for case, prepositions act as heads, i.e. they trigger case agreement in the relative pronoun. So, the link between nonstranded adpositions and the relative pronoun with which they occur is tighter than the link between stranded prepositions and relative particles or the zero-marker. This is particularly evident in English, the only language that exhibits both the structure ‘non-stranded adposition+relative pronoun’, as in (3.156), and ‘relative pronoun+stranded adposition’, as in (3.157). Whereas in (3.156) the preposition triggers the non-subject form whom, in (3.157) the stranded adposition has no influence on the case of the relative pronoun. (3.156) Janet posted a letter for me last week for a friend [with whom I ENG worked at Ipswich.] (Herrmann 2005: 47, modified) (3.157) Janet posted a letter for me last week for a friend [who I worked ENG with at Ipswich.] (Herrmann 2005: 47, original) If we try to systematize what has been told so far, we may represent the use of non-stranded and stranded adpositions on a scale, ranging from synthetic-simple to analytic-combined relative elements (Figure 3.1).

Classification problems 145 synthetic-simple 1. case-inflecting relative pronoun (no adposition), cf. (3.153) 2. (case-agreement-triggering) non-stranded adposition + relative pronoun, cf. (3.154)–(3.156) 3. (non-case-inflected) relative pronoun + stranded adposition, cf. (3.157) 4. relative particle + stranded adposition, cf. (3.149)–(3.150) analytic-combined 5. zero-marker + stranded adposition, cf. (3.149) Figure 3.1. Combinations of adpositions and simple relative elements.

With reference to Figure 3.1, relative elements with non-stranded adpositions (points 1 and 2) will be considered as simple and relative elements with stranded prepositions (points 3 to 5) as combined. The demarcation is constituted by the absence vs. presence of case-agreement-triggering. 3.5.6. One strategy or two? Comparing Table 3.6 and Table 3.7, it emerges that the vast majority of languages possessing a relative particle also exhibit the combined relative element ‘relative particle+resumptive element’. The simple relative particle and the relative particle with resumptive element are usually seen as complementary: the particle is used alone for the higher segment of the AH (SU and DO) and is combined with a resumptive for the lower segment of the AH (DO to GEN). This is stated by scholars including Bernini (1989) for Romance languages, Křížková (1970) for Slavic languages and Fleischer (2004a) for German dialects. The Slovenian examples below illustrate this case. In (3.158) the particle ki alone relativizes SU. In (3.159) and (3.160) DO and IO are relativized through the combination of ki with resumptive clitics. In (3.161) ki combines with a pronominal PP to relativize OBL. Finally, ki combines with a possessive in (3.162), where GEN is relativized. (3.158) Rešil je otroka, [ki se je utapljal.] SLN saved is boy RPAR REFL is drowned ‘He saved the boy who was drowning.’ (Herrity 2001: 107) (3.159) Ta sta SLN this are.3DU iskali v looked.for in

punci, [ki smo ju girl.DU RPAR we.are CL.ACC.3DU hotelu.] hotel

146 Data classification ‘These are the two girls we have looked for in the hotel.’ (Herrity 2001: 107) (3.160) Učenci, [ki ste jin dali knjige,] so SLN students RPAR you.are CL.DAT.3PL given books are naredili izpit. passed exam ‘The students you (pl.) gave the books to have passed their exam.’ (Herrity 2001: 107) (3.161) Človek, [ki sva moja sestra in jaz SLN man RPAR are.1DU my sister and I z njim potovala,] je zelo pomembna with him travelled.DU is very important oseba. person ‘The man with whom my sister and I travelled is a very important person.’ (Sansò 2005, SLV_078) (3.162) Človek, [ki moja sestra pozna njegovega brata,] je SLN man RPAR my sister knows his brother is zelo pomembna oseba. very important person ‘The man whose brother my sister knows is a very important person.’ (Sansò 2005, SLV_056) Still, the fact that ki appears in all examples does not mean that we are faced with the same strategy in (3.158) to (3.162). In (3.158) the syntactic role of the relativized item is not expressed, whereas in (3.159) to (3.162) it is. Since relative elements are classified according to the features they encode, if two relative elements encode a different number of features, they cannot be regarded as representatives of the same strategy. Additionally, as will be discussed in section 4.2.3, in various languages two parallel strategies are attested: the first one uses a relative particle to relativize all of the positions of the AH; the second one uses a relative particle combined with a resumptive element to relativize the same positions. So, it seems reasonable to distinguish between the simple and the combined relative elements at issue. It may well be that the first strategy (relative particle alone) is used more frequently to relativize the upper segment of the AH, whereas the second one (relative particle and resumptive element) is used more

Classification problems 147

frequently to relativize the lower segment of the AH. Still, both strategies apply in principle to the whole AH. A relative particle can be used to relativize lower positions of the AH, both in Romance languages, as the Portuguese examples (3.34) to (3.36) above show, and in Slavic languages, as can be seen in (3.163) to (3.165). See also section 4.3.3. (3.163) Lopata [što se žito veje.] BCS shovel RPAR REFL wheat winnows ‘A shovel that wheat is winnowed with.’ (Maček 1986/1987: 109) (3.164) А ой! Не знаешь / возьмешься в ведро / вода / замерзла / таз / RUS что умывались / все замерзло // A oj! Ne znaeš’ / voz’meš’sja v and oh not you.know you.take in vedro / voda / zamerzla / taz / [čto umyvalis’] / bucket water froze basin RPAR we.washed vse/ zamerzlo / all froze ‘Oh, you don’t know! When I wanted to wash myself in the bucket, the water had frozen, the basin we used to wash ourselves, everything had frozen.’ (Zemskaja and Šmelëv 1984: 183) (3.165) А дзе тая крынічанка, што голуб купаўся? BLR A dze taja kryničanka, [što golub kupaŭsja]? and where that fountain RPAR dove bathed ‘And where is that fountain where the/a dove bathed?’ (Atrachovič 1966: 609) A relative particle is used to relativize OBL in (3.163) and (3.164) and LOC in (3.165). In turn, a relative particle with a resumptive pronoun can be used also for higher positions, as shown in (3.166) and (3.167), where a clitic is inserted in SU-relativization (see also 4.2.1). Gärtner (1998: 219) specifies that (3.166) is rather typical of Brazilian than European Portuguese. (3.166) Aqui nesta aldeia tem um caboclo [que ele é POR here in.this village there.is DET caboclo RPAR he is ideal.] ideal ‘In this village there is a caboclo that is ideal.’ (Gärtner 1998: 219)

148 Data classification (3.167) Знаці чашніка маладзенькага, што ён жонкі не мае? BLR Znaci čašnika maladzen’kaha, [što ёn žonki ne you.know butler young RPAR he wife not mae]? has ‘Do you know the young butler who has no wife?’ (Atrachovič 1966: 609) So, the two strategies mainly differ in their degree of explicitness: whereas the combined element encodes all four parameters, the relative particle only encodes the link between MC and RC. The fact that the second strategy is preferably used for the upper segment of the AH and that the first one is found particularly when low positions of the AH are relativized is not surprising: it conforms to the statement made in Keenan and Comrie (1977) that lower positions require explicit coding of the syntactic role of the relativized item because they are more difficult to retrieve. Still, it was demonstrated that there are also cases in which the two strategies are used respectively for the relativization of lower and higher positions of the AH. As a consequence, the recoverability of the syntactic role of the relativized item alone cannot account for the choice of one strategy instead of the other one. In this regard, cf. the discussion in 4.3.3 and 5.3 and Ariel (1999).

Part 2 An areal study of non-standard relative constructions in European languages

In the first part we discussed the theoretical framework in which this investigation is to be situated and the parameters according to which relative constructions are classified. The second part will focus on the cross-linguistic analysis of non-standard relative constructions in European languages. The analysis will be conducted from a threefold perspective. In Chapter 4, relative constructions will be analyzed from a purely typological point of view. That is, the peculiarities displayed by non-standard relative constructions will be described on the basis of the parameters introduced in Chapter 3. Emphasis will be placed on the structural differences between these constructions and their standard counterparts. As will be shown in section 4.5, these differences cannot be accounted for on purely functional grounds, since the same functional principles can be invoked to explain the existence and use of both standard and non-standard constructions. We need to resort to other principles instead, of a sociolinguistic nature. These sociolinguistic principles will be at issue in Chapter 5, where relativization strategies described in Chapter 4 will be considered from a sociolinguistic perspective. This means that each construction will be considered as a linguistic variant occupying a position in the diasystem of a language. On the basis of the sociolinguistic status of relative constructions in European languages, a ‘scale of standardness’ will be formulated. This scale mirrors which strategies are preferably considered standard and which ones are viewed as non-standard in the specific socio-historical context of European languages as they have developed for at least the past 1,000 years. The possibility of comparing the sociolinguistic status of relativization strategies is due to the relatively strong cultural unity that has characterized Europe for the past millennium (cf. Haspelmath 1998, 2001; Auer 2005b; Stolz 2006). Auer (2005b: 7) also speaks of “Europe’s sociolinguistic unity”, thus implying that the sociolinguistic constellations characterizing Europe nowadays and in the past can be grasped by means of a common reference frame (cf. sections 5.2 and 6.5.) The position of constructions on the scale of standardness can be explained through the sociolinguistic principles introduced at the beginning of Chapter 5. As these

150

An areal study of non-standard relative constructions

principles often served as guidelines for the codification of European standard varieties, in Chapter 6 we will take a closer look at the diachronic development of non-standard strategies. First, evidence will be provided for the existence of relative constructions that are nowadays considered as non-standard in past stages of the languages in the sample; then, on the basis of Auer’s (2005b) diachronic-sociolinguistic model, it will be ascertained how the opposition between standard and non-standard strategies came into being. The analysis will reveal that the formation of standard language was a crucial moment in the development of this opposition: during this process, a number of principles and guidelines were followed leading to the inclusion of some constructions and the exclusion of others from the standard language – which eventually created the present-day situation. Since language is continuously evolving, it may well be that constructions that are nowadays considered non-standard will be accepted into the standard norm in the long run or that standard constructions that are rarely used will eventually disappear.

Chapter 4 Typological issues

In this chapter the relative constructions attested in European non-standard varieties will be examined. To this purpose, the three parameters introduced in Chapter 3 will be considered, identifying the distinctive features which non-standard constructions show compared with their standard counterparts. Section 4.1 will discuss word order, section 4.2 relative elements and section 4.3 the syntactic positions relativized. In each section, emphasis will be placed on cross-linguistically significant phenomena: that is, in this chapter a language-by-language analysis is not aimed to; instead, general tendencies observed in non-standard varieties of different languages will be outlined. 4.1. Word order As mentioned in section 3.1, in non-standard varieties of European languages we find relative constructions that have a different word order from constructions attested in standard varieties. The languages that display different constructions are reported in Table 4.1 below. Table 4.1. Peculiar constructions in non-standard varieties of European languages: word order. EMBEDDED

ADJOINED

PRENOMINAL CIRCUMNOMINAL POSTNOMINAL

BAS, TUR

CORRELATIVE (PREPOSED)

EXTRAPOSED (POSTPOSED)

GRE, HUN, LTV, RUS, (TUR), UKR

DUT, GER, TUR

As the table shows, there are three main relevant phenomena: 1. languages with a default (=prototypical, most frequent) prenominal strategy – Basque and Turkish – also exhibit a postnominal one; 2. in a group of languages with default embedded strategies, a preposed (correlative) adjoined strategy also occurs; 3. in three languages that have default embedded strate-

152 Typological issues gies, a postposed adjoined strategy is attested. The three groups of constructions are separately discussed in 4.1.1 to 4.1.3. 4.1.1. The postnominal strategy As Table 4.1 shows, two languages that usually form RCs through a prenominal strategy, i.e. Basque and Turkish, also display a postnominal strategy. This strategy can have two different structures: 1. the internal structure of the prenominal RC is preserved, but the RC appears after the referent instead of before it; 2. the postnominal RC is introduced by a relative element. The first case is attested in Basque, as shown in (4.1). (4.1) BAS

Gizon [karrikan ikusi duzu-n]-a-k man street.DET.in seen you.have.him-RPAR-DET-ERG hemen lan egiten du. here work doing he.has.it ‘The man you saw in the street works here.’ (Oyharçabal 1989: 64)

(4.2) BAS

[Karrikan ikusi duzu-n] gizonak street.DET.in seen you.have.him-RPAR man.DET.ERG hemen lan egiten du. here work doing he.has.it ‘The man you saw in the street works here.’ (Oyharçabal 1989: 63)

If we compare (4.1) and (4.2), we will notice that in the former the RC is placed after the referent, but before the determiner -a and the ergative suffix -k. That is, the RC still belongs to the same NP, but it appears after the head noun. The internal structure of the RC remains unchanged in both positions: the clause is verb-final, the link between the MC and the RC is expressed through the affixal clause-final particle -n and the syntactic role of the relativized item is not explicitly encoded. Turkish also displays a shift of the RC from its original prenominal position, as shown in (4.3) to (4.6). (4.3) TUR

Daha yirmi otuz more twenty thirty [tutuklanan.] be.arrested.PTCP.PRS

kişi varmış person there.was

Word order 153

‘There were another 20 or 30 people who were arrested.’ (Haig 1998: 108) (4.4) TUR

Vazoyu kırdım [senin verdiğin] vase.ACC I.broke you.GEN give.NMLZ.POSS.2SG ‘I broke the vase – (the one) you gave me.’ (Haig 1998: 109)

(4.5) TUR

Kitabı okudun mu [getirdiğim]? book.ACC you.read Q bring.NMLZ.POSS.1SG ‘Have you read the book – (the one) I brought?’ (Haig 1998: 109)

(4.6) TUR

İki mektup var [atacağın.] two letter there.is throw.PTCP.FUT.POSS.2SG ‘There are two letters – (those that) you shall post.’ (Haig 1998: 109)

These constructions are parallel to the prenominal ones reported below. As (4.3) to (4.6) and (4.7) to (4.10) show, the internal structure of the RC remains unchanged: the clause is verb-final, the verb in the RC has a deranked form with respect to the verb in the MC and the syntactic role of the relativized item is not encoded in the RC. (4.7) TUR

[Tutuklanan] daha yirmi otuz kişi varmış. be.arrested.PTCP.PRS more twenty thirty person there.was ‘There were another 20 or 30 people who were arrested.’ (Haig 1998: 108, modified)

(4.8) TUR

[Senin verdiğin] vazoyu kırdım. you.GEN give.NMLZ.POSS.2SG vase.ACC I.broke ‘I broke the vase you gave me.’ (Haig 1998: 109, modified)

(4.9) TUR

[Getirdiğim] kitabı okudun mu? bring.NMLZ.POSS.1SG book.ACC you.read Q ‘Have you read the book I brought?’ (Haig 1998: 109, modified)

(4.10) [Atacağın] iki mektup var. TUR throw.PTCP.FUT.POSS.2SG two letter there.is ‘There are two letters you shall post.’ (Haig 1998: 109, modified)

154 Typological issues However, other than in Basque the Turkish RCs in (4.3) to (4.6) should be perhaps better classified as extraposed, because the RC appears after the whole MC, not directly after the referent. As their internal structure remains unchanged with respect to their prenominal counterparts, they may be regarded as cases of context-bound right-extraposition: in 3.1 it was argued that such instances would not be regarded as a separate relativization strategy in the relevant language (examples were quoted for German, English and Hungarian). Additionally, Haig (1998), from whom these examples are quoted, stresses that they should be regarded as afterthoughts: the speaker further specifies the referent as s/he feels that it was not univocally identified – cf. also the partially diverging English translation of (4.3)–(4.6) vs. (4.7)–(4.10). Such constructions are most common “in casual speech and in contemporary fiction” (Haig 1998: 108). As Haig argues, prenominal RCs are not the only nominal (sub)constituents that can be right-extraposed in Turkish: also genitives can be right-extraposed (Haig 1998: 109–113). The second structure that postnominal RCs may have in Basque and Turkish is illustrated in the examples below. (4.11) Neska [zeini loreak eman BAS girl.DET RPRO.DAT flowers.DET given dizkioda-n] hor dago. I.have.them.to.her-RPAR there is ‘The girl I gave the flowers to is right here.’ (Trask 1998: 320) (4.12) O kız, [ki hiç gülmezdi,] sonunda TUR that girl RPAR at.all not.laughed at.the.end güldü. laughed ‘That girl, who hadn’t laughed at all, eventually laughed.’ (Erkman-Akerson and Ozıl 1998: 323) (4.13) Şu kız, [hani dün rastlamıştık,] bugün TUR that girl RPAR yesterday we.met today geldi. came ‘That girl, whom we met yesterday, came today.’ (ErkmanAkerson and Ozıl 1998: 321)

Word order 155

As the examples show, not only is the RC placed after the referent, but the internal structure of the RC differs from the prenominal strategy: in (4.12) and (4.13) the verbs in the RC and the MC are balanced; in all three examples the RC contains a clause-initial relative element. In particular, in (4.11) the postnominal RC is introduced by zein ‘which’, which behaves here as a relative pronoun: it encodes the link between the MC and the RC and the role of the relativized item in the RC. Trask (1998: 320) suggests that this construction was probably borrowed from French or Spanish, both of which possess a postnominal construction with a relative pronoun. Heine and Kuteva (2005: 2–3) call this process REPLICATION: grammatical meanings (or functions) are transferred from a model language into a replica language. However, it must be pointed out that the structure ‘(clause-initial) wh-pronoun + (clause final) -n’ is not extraneous to Basque syntax. It is used for instance in embedded questions, as the example below shows. (4.14) zer egin duzu-n galdatu daut BAS what done I.have.it-COMP asked he.has.it.to.me ‘He has asked me what I have done.’ (Oyharçabal 1989: 64) In this sense, the postnominal strategy could be regarded as the application of an existing morphosyntactic structure to a new function – to qualify an entity – through replication, i.e. under the influence of ‘model languages’ French and Spanish. As will be discussed in 6.2.1, this contact is not recent, but dates back to some centuries ago. So, in (4.11) we see a clause-initial relative element similar to the relative pronoun in French and Spanish occurring with the clause-final relative particle -n, which appears also in (4.1) and (4.2). In this respect, it remains questionable whether the clause-initial relative element can be said to encode the link between the MC and the RC, since this function is already taken over by the clause-final relative particle. Still, this is not an unexpected construction: as we saw in 3.2.2, combinations of a relative pronoun with a relative particle are attested also in German, Lower/Upper Sorbian and Macedonian. The difference between (4.11) and the constructions in examples (3.57) to (3.60) lies in the position of the particle, which is clause-final in Basque and clause-initial in the other languages. What may be typologically more relevant is that Basque is a left-branching language, but the relative pronoun is clause-initial. However, clause-initial connectors are also attested in the other left-branching language of the sample: cf. the discussion of TUR ki in 3.5.2.

156 Typological issues Like prenominal RCs, also postnominal clauses introduced by ki can be extraposed in Turkish, as may be seen in (4.15) below: the RC appears after the whole MC. (4.15) Sınıflarında Gülten diye çok şımarık TUR class.POSS.3PL.in Gülten saying very self-indulgent bir kız vardı [ki Hülya ondan DET girl there.was RPAR Hülya that.from nefret ederdi.] hate did ‘In their class there was a very self-indulgent girl called Gülten, whom Hülya hated.’ (Göksel and Kerslake 2005: 458) In (4.13) we saw a postnominal RC containing a balanced verb form introduced by hani. In that example hani can be regarded as a relative particle, since it conveys the link between the MC and the RC. However, the same construction is also attested as it appears in (4.16) and (4.17). (4.16) Şu kız, [hani dün rastladığımız,] TUR that girl PAR yesterday meet.NMLZ.POSS.1PL bugün geldi. today came ‘That girl, (the one) whom we met yesterday, came today.’ (Erkman-Akerson and Ozıl 1998: 322) (4.17) [Dün rastladığımız] şu kız bugün TUR yesterday meet.NMLZ.POSS.1PL that girl today geldi. came ‘That girl, whom we met yesterday, came today.’ (ErkmanAkerson and Ozıl 1998: 322) In (4.17) we have a prenominal RC containing a deranked verb form. In (4.16) we see an intermediate construction: the RC is postnominal and is introduced by hani, but the verb form is deranked. This may suggest that hani cannot possibly be considered as a relative particle in Turkish – although in (4.13) it was glossed as such. In fact, hani is a discourse particle signalling that what is said refers to a piece of information that the speaker assumes to be shared with the hearer. Haig (1998: 119) glosses it as ‘you know’, as in (4.18) below.

Word order 157

(4.18) Benim de hakkıma zil düşmüştü, TUR my PAR right.POSS.1SG.DAT cymbal fell hani iki tencere kapağı gibi şey PAR two saucepan lid.POSS.3SG like thing olur ya […] is PAR ‘And me, I got cymbals (lit. to my right fell cymbals), you know, (it’s) something like two saucepan lids.’ (Haig 1998: 119) Similarly, (4.13) and (4.16) could be rendered as ‘That girl, you know, we met (her) yesterday, (she) came today’ and ‘That girl, you know, the one we met yesterday, (she) came today’. Vice versa, (4.18) could be rendered by means of an RC: ‘And me, I got cymbals, that are something like two saucepan lids, (you know).’ This suggests that hani, like ki, allows for multiple interpretations: formally, it is not a relative particle (yet), but it is used in contexts in which an SoA qualifies a participant of another SoA. As for the nature of the relation between the two SoAs linked by hani, these relations tend to be non-restrictive, as in (4.18), but can also be restrictive. Erkman-Akerson and Ozıl (1998: 322) explain the difference between (4.16) and (4.17) in terms of restrictiveness: in (4.17) the RC has a descriptive function, i.e. it is non-restrictive; in (4.16) it helps identify the referent, i.e. it is restrictive. The ambiguous status of hani may have been the source for a grammaticalization path leading to the development of postnominal RCs in Gagauz, a Turkish dialect spoken on the Black Sea coast. In Gagauz, postnominal RCs contain finite verb forms and are introduced by the particle ani, deriving from hani, as the examples below show. (4.19) adam [ani geler] GAG man RPAR will.come ‘The man that will come.’ (Menz 2006: 143) (4.20) adam [ani para verdim] GAG man RPAR money I.gave ‘The man that I gave the money to.’ (Menz 2006: 143) So, we may conclude that non-standard Turkish has postnominal RCs which contain a finite verb form and are introduced by a relative particle. Some of these constructions – namely those introduced by hani – have an

158 Typological issues ambiguous status in standard Turkish, whereas in Turkish dialects they can be regarded as instances of relative constructions. How can the existence of these constructions in Basque and Turkish be accounted for? To start with, one may notice that they share a common feature: they shift the RC after its referent in the MC and qualify it ex post. This may be due to the organizational needs of spoken discourse, which tends to expand rightwards. In particular, in (4.3) to (4.5), (4.15) and (4.18) we are presented with afterthoughts or additional information; in (4.17) the speaker specifies the identity of the referent, which s/he believes not to be clear enough. In several of the Turkish constructions – (4.13), (4.15), (4.17) and (4.18) – the RC shows an MC-like structure, where the verb has a finite form: speakers may find it easier to build chains of MCs instead of embedding clauses through deranked verb forms (cf. also Auer 1990). As for the Basque construction in (4.11), the only significant difference from the prenominal strategy is constituted by the pronoun-like relative element, which serves (at least) to encode the syntactic position of the relativized item in the RC. 4.1.2. The correlative strategy In a group of European languages a preposed (correlative) adjoined strategy introduced by a relative pronoun is attested, similar to the correlative strategy of Hindi (Lehmann 1984: 133). The structure of this construction is the following: [RPRO N ...]RC [(DEM)(N) ...]MC. Typically, the relativized item is encoded in the RC with an equal or higher degree of explicitness than in the MC (cf. the scale of explicitness introduced in 1.1.2). In the formula representing the morphosyntactic structure of the construction this peculiarity is signalled through brackets around DEM and N: that is, a demonstrative and a nominal element encoding the relativized item can – but do not have to – appear in the MC, as shown in examples (1.31) to (1.33). The correlative construction is attested in languages such as Greek, Hungarian, Latvian and Ukrainian. (4.21) [Óses jinékes máθane já to GRE RPRO.NOM.PL women heard about DET stílane ðóra. sent gifts

moró] tu baby to.him

Word order 159

‘All the women who heard about the baby sent him gifts.’ lit. ‘How many women heard about the baby, (they) sent him gifts.’ (Joseph and Philippaki-Warburton 1987: 166) (4.22) [Amelyik kutya ugat,] az nem harap. HUN RPRO.NOM.SG dog barks that not bites ‘A barking dog does not bite.’ lit. ‘Which dog barks, that does not bite.’ (Tompa 1968: 66) (4.23) [Kuŕš puĩsītis čakli gãja,] tas LTV RPRO.NOM.M.SG boy quickly was.walking that bij liẽls dziêdâtâjs. was great singer ‘That boy who was walking quickly was a great singer.’ lit. ‘Which boy was walking quickly, that was a great singer.’ (Endzelin 1923: 401) (4.24) В яку сторону вiнок попливе, туди дiвка замiж пiде. UKR [V jaku storonu vinok poplyve,] tudy in RPRO.ACC.F.SG side garland will.flow there divka zamiž pide. girl in.marriage will.go ‘The direction that the garland flows on water, in that direction the girl will get married.’ lit. ‘In which direction the garland flows, there the girl will go in marriage.’ (Dnipropetrovsk dialect, Mel’nyčuk 1962: 114) In the examples above, the qualified participant is encoded in the RC through a substantive – respectively ‘women’, ‘dog’, ‘boy’ and ‘direction’; in the MC, it is encoded less explicitly, by a Ø-element in (4.21), by a demonstrative pronoun in (4.22) and (4.23) and by an adverb (tudy) in (4.24). As examples (4.21) to (4.23) are taken from grammars or linguistic studies none of which is corpus-based, we may only state that the correlative strategy is attested in non-standard varieties of Hungarian and Latvian, but no claims can be made about its frequency ot the syntactic positions it can relativize. As for Greek, further examples are reported in Holton, Mackridge, and Philippaki-Warburton (1997: 446–447): the correlative construction with ósos seems to be used in contexts in which the SoA encoded in the MC regards all of the participants qualified by the preposed RC. This explains the translation of (4.20) as ‘All the women who heard about

160 Typological issues the baby…’. Finally, in Ukrainian dialects this construction can relativize segment SU–OBL of the AH, as the examples in Mel’nyčuk (1962: 114) demonstrate. The correlative construction is attested also in Russian, both in colloquial language, as in (4.25) and in dialects, as in (4.26). (4.25) Которые вот клиенты есть у меня, и тем я смотрю. RUS [Kotorye vot klienty est’ u menja,] RPRO.NOM.PL PAR customers is at me i tem ja smotrju. PAR those.to I look.after ‘Those customers that are mine, I look after them.’ (Lapteva 1976: 144) (4.26) Которы старухи не ушодци, так со старушками поговорить. RUS [Kotory staruchi ne ušodci,] tak RPRO.NOM.PL old.ladies not gone PAR so staruškami pogovorit’. with old.ladies talk.INF ‘Let’s have a talk with the old ladies who haven’t gone (yet).’ lit. ‘Which old ladies are not gone, I/we shall talk with them.’ (dialect, Avanesov and Orlova 19652: 199) Once more, the relativized item is encoded through a substantive in the RC and through a demonstrative in the MC. The correlative construction in Russian is far better documented than in the languages mentioned before: it is widespread both in colloquial varieties (russkaja razgovornaja reč’, cf. 8.2.1), as shown in Kručinina (1968: 83–85), Zaliznjak and Padučeva (1975: 73–75) and Lapteva (1976: 144, 302), and in dialects, as documented in Šapiro (1953: 107–110) and Avanesov and Orlova (19652: 199). As examples (3.99) to (3.102) illustrated, this strategy can relativize segment SU–OBL of the AH. Still, it is more frequently used to relativize subjects, as we may conclude if we examine the occurrences of this construction in the references mentioned above. In Russian the correlative strategy can exhibit a number of variants slightly deviating from the formula provided at the beginning of this subsection. Deviations regard the word order inside the RC and the relative element. They are separately discussed here below.

Word order 161

VARIANT 1: the relative pronoun in the RC can be preceded by a demonstrative (ėtot ‘this’, tot ‘that’) – as in (4.27) – or an indefinite pronoun (vse ‘all’) – as in (4.28). (4.27) Вот тем, которые ребята не подготовлены, тем приходилось RUS туго. Vot tem, [kotorye rebjata ne podgotovleny,] PAR those.to RPRO.NOM.PL guys not prepared tem prichodilos’ tugo. those.to came hard ‘Now, those guys who were not prepared, they were in a difficult position.’ (Lapteva 1976: 144) (4.28) В общем-то все, которые коми находятся или ненцы RUS находятся в тундре, они все и члены колхозов. V obščem-to vse, [kotorye komi in general PAR all RPRO.NOM.PL Komi nachodjatsja ili nency nachodjatsja v tundre,] are.found or Nenets are.found in tundra oni vse i členy kolchozov. they all PAR members kolkhozes.of ‘Generally speaking, all Komi and Nenets who live in the tundra are members of kolkhozes.’ (Sergieva and Gerd 1998: 129) The demonstrative or indefinite pronoun can be repeated in the MC, as in the two examples above, or it can be replaced by a personal pronoun, as in the example below, where the accusative form eë ‘her’ is used. (4.29) Та, которая здесь стояла лампа, я ее не брал. RUS Ta, [kotoraja zdes’ stojala lampa,] ja eë that RPRO.NOM.F.SG here stood lamp I her ne bral. not took ‘The lamp that stood here, I didn’t take it.’ (Lapteva 1976: 144) The main function of this variant seems to be a preliminary determination or quantification of the referent introduced in the RC: since the MC follows the RC, a determination or quantification of the referent would be only possible in the MC, when the referent itself has been already introduced. As will be discussed below, this construction originally encoded a

162 Typological issues conditional relation and the pronoun in the RC had an indefinite meaning, ‘who/what/whichever’ (see also Lehmann 1984: 370–372). The presence of an element before the relative pronoun itself may have arisen out of the necessity of determining or quantifying an entity that as such was not definite. This variant of the correlative construction was attested also in Latin: see Lehmann (1979: 9), in particular example (6). VARIANT 2: the relativized item in the RC may not directly follow the relative pronoun. As in (4.29), also in (4.30) and (4.31) the relative element is clause-initial, whereas the relativized item is clause-final. (4.30) Которые у нас жили дачники, они этот год не приехали. RUS [Kotorye u nas žili dačniki,] oni RPRO.NOM.PL at us lived residents they ėtot god ne priechali. this year not came ‘The summer residents who used to stay at ours didn’t show up this year.’ (Kručinina 1968: 84) (4.31) Мам, вон которая наши везла электричка обратно идет. RUS Mam, von [kotoraja naši vezla Mum there RPRO.NOM.F.SG ours brought ėlektrička] obratno idet. local.train back goes ‘Mum, look, the local train with which our friends arrived is going back.’ (Lapteva 1976: 302) This construction can be explained through the relatively free word order of Russian non-standard varieties (cf. Zemskaja 1973; Lapteva 1976). In fact, examples like (4.32) are also attested in which both the relative pronoun and the referent – or, more specifically, part of it, since the relative pronoun splits the referent nervnaja klinika ‘nerve clinic’ into its two components – are RC-final. (4.32) У нас нервная которая клиника, так она лечит с параличами. RUS [U nas nervnaja kotoraja klinika,] tak at us nerve RPRO.NOM.F.SG clinic so ona lečit s paraličami. she cures with paralyses ‘At the nerve clinic that we have here, they cure patients who are paralysed.’ (Lapteva 1976: 144)

Word order 163

One remark on the translation of the example above: as will be discussed below for Variant 5, in spoken Russian a PP is considered sufficient to identify an entity. So s paraličami ‘with paralyses’ means ‘Xs that are paralysed’, which in the given context can be interpreted as ‘people/patients that are paralysed’. VARIANT 3: the relativized item in the RC may display case agreement with the relativized item in the MC, not with the relative pronoun. (4.33) Во-первых, которую она купила побелка – она вся будет RUS отлетать. Vo-pervych, [kotoruju ona kupila pobelka] – firstly RPRO.ACC.F.SG she bought whitewash.NOM ona vsja budet otletat’. she.NOM all will vanish ‘First of all, the whitewash that she bought will completely vanish.’ (Kručinina 1968: 85) (4.34) Которым дети седьмой год – сейчас они в эту субботу RUS уезжают уже в Евпаторию. [Kotorym deti sed’moj god] – sejčas RPRO.DAT.PL children.NOM seventh year now oni v ėtu subbotu uezžajut uže v they in this Saturday go already in Evpatoriju. Evpatorija ‘The seven-year-old children are already going to Evpatorija this Saturday.’ (Kručinina 1968: 85) In (4.33) DO is relativized; however, only the relative pronoun conveys the syntactic position relativized, since the substantive pobelka is in the nominative case. Similarly, in (4.34) the substantive deti is in the nominative case whereas the syntactic position relativized, IO, is expressed through the relative pronoun. There seems to be a conflict between the role of the relativized item in the RC and in the MC: the referent is introduced in the RC, where a non-SU position is relativized, but its morphological form agrees with the syntactic role it has in the MC, i.e. SU – a sort of attractio relativi,14 which, however, does not affect the relative element, but the referent in the RC. This peculiarity would require further investigation.

164 Typological issues VARIANT 4: instead of a relative pronoun, a relative particle may be used: in (4.35) čto is used instead of kotoraja. (4.35) Что у тебя есть пластинка, щас будут передавать. RUS [Čto u tebja est’ plastinka] ščas budut RPAR at you is record now they.will peredavat’. broadcast ‘The record that you have, they’ll play it now on the radio.’ (Lapteva 1976: 303) This variant is quite rare: only Lapteva (1976) reports it. However, as will be shown in 6.2.2, it is also historically attested. VARIANT 5: the NP [RPRO N]NP in the RC can be left out. This is a somewhat extreme case, where hearers have to infer themselves the relation between the two SoAs, as in (4.36). (4.36) Напротив живёт / уехала в Крым. RUS [Naprotiv živët] / uechala v Krym. opposite lives went.F.SG to Crimea ‘The woman living in front of us moved to Crimea.’ (Zemskaja 1987: 55) It may be questioned whether this variant should be regarded as a correlative construction at all: the RC is adjoined to the MC, but nothing signals the correlation between the two clauses. We may postulate a zero-marker in the RC: both the relativized item in the RC and in the MC would be morphologically zero-marked. That is, the participant of the main SoA qualified by the other SoA would not receive any morphosyntactic encoding, except for the verbal ending -a, which tells the hearer that the entity at issue is a woman. As Weiss (1993) points out, spoken Russian exhibits a “Liebe zur Null”, i.e. a strong tendency to leave unexpressed not only verbal arguments, but also connectors and predicates. This is possible because speakers strongly rely on hearers’ inferential skills. Russian scholars who analyzed this kind of construction proposed a slightly different interpretation (cf. Zemskaja 1973: 239; Zemskaja 1987: 53–59), i.e. that it may be related to the ‘nominative’ discourse function (nominacija). Nominating a referent, i.e. introducing the topic of an assertion, can be achieved in different ways in Russian: through a NP, through a

Word order 165

PP or through a whole clause. In examples (4.37) to (4.39) the topic is introduced through a NP: ‘student’, ‘our neighbour’, ‘that child’. (4.37) Студент опять приходил / просил почитать что-нибудь // RUS Student opjat’ prichodil / prosil počitat’ student again came.round he.asked read.INF čto-nibud’ // something ‘The student came round again and asked if I/you/we could read something.’ (Zemskaja 1987: 53, modified) (4.38) У нашего соседа недавно ногу ампутировали // RUS U našego soseda nedavno nogu at our neighbour recently leg amputirovali // they.amputated ‘Our neighbour recently had his leg amputated.’ (Zemskaja 1987: 53, modified) (4.39) Тот мальчик / ты его знаешь? RUS Tot mal’čik / ty ego znaeš’? that child you him know? ‘That child, do you know him?’ (Zemskaja 1973: 239, modified) Examples (4.37) to (4.39) have been modified: in the original source the topics are not introduced by means of NPs, but by means of a PP, as in (4.40) and (4.41), or by means of a whole clause, as in (4.42). Instead of identifying topics through a substantive, the speaker chooses to qualify them through a distinctive feature referring to their appearance, their location or something they are doing. (4.40) В очках опять приходил / просил почитать что-нибудь // RUS V očkach opjat’ prichodil / prosil počitat’ in glasses again came he.asked read.INF čto-nibud’ // something ‘The X with glasses came round and asked if I/you/we could read something.’ (Zemskaja 1987: 53)

166 Typological issues (4.41) Из соседней квартиры недавно ногу ампутировали // RUS Iz sosednej kvartiry nedavno nogu from neighbouring flat recently leg amputirovali // they.amputated ‘The X from the neighbouring flat recently had his/her leg amputated.’ (Zemskaja 1987: 53) (4.42) Вон дерётся / ты его знаешь? RUS Von derëtsja / ty ego znaeš’? there fights you him know ‘The X who is fighting over there, do you know him?’ (Zemskaja 1973: 239) In the above examples it is not possible to define the entity at issue with precision: this is signalled through ‘X’ in the translation. In all these constructions the speaker introduces a topic not by means of a full NP, but by means of a distinctive feature that identifies it. This way of introducing a topic is akin to what happens in correlative constructions: Lehmann (1979: 9) discusses correlative constructions in Latin and argues that the RC has the function of introducing a topic which is commented on in the following MC. In this sense, constructions like (4.36) and (4.42) can be assimilated to correlative constructions: a topic is introduced by means of a predication, although the RC shows neither a relative element nor an explicitly encoded referent. So, (4.36) and (4.42) will be regarded as instances of ‘zeromarked’ correlative constructions. Moreover, both correlative constructions and zero-marked constructions like (4.36) and (4.42) can be included within the constructions that Maslova and Bernini (2006: 83) define as “htopic”, i.e. hanging topic: through the preposed RC the speaker creates a reference point which allows “anchoring the message in the shared knowledge” (Maslova and Bernini 2006: 90); in the following MC an SoA is coded in which the entity introduced in the RC usually acts as a participant. As the examples from different languages reported in this section have shown, the entity can be encoded with either a high or a low degree of explicitness. Finally, the correlative strategy is attested also in Turkish, as can be seen here below. Only Kornfilt (1997: 60) and Erkman-Akerson and Ozıl (1998: 325) mention this structure among Turkish relative constructions.

Word order 167

(4.43) [Hangı resmi beğeniyorsan,] onu TUR REL? picture.ACC if.you.are.loving that.ACC alalım. we.shall.take ‘Which(ever) picture you like best, we’ll buy it.’ (Erkman-Akerson and Ozıl 1998: 325) In (4.43), hangı ‘which’ was glossed as ‘REL?’: although hangı is said to have only an interrogative function in present-day Turkish, in (4.43) it can be thought to encode the link between the MC and the RC, on a par with all the relative elements marked in bold in (4.21) to (4.24). So, in (4.43) hangı acts as a relative element. Here it behaves like an adjective and does not inflect, so it encodes neither the syntactic role of the relativized item nor number agreement (Turkish does not encode gender distinctions). The syntactic position relativized, DO, is encoded by means of the referent resim ‘image’, which is in the accusative case resmi. Additionally, the construction in (4.43) can be regarded as displaying also a (primary) conditional value, as the RC-verb in the potential mood suggests: consequently, hangı may be better interpreted as ‘whichever’. This is true not only for Turkish, though: in all languages mentioned in this section the correlative strategy developed out of a conditional construction, which later on grammaticalized to a relative one: originally, the pronoun had an indefinite function. See Lehmann (1979, 1984: 368–375, 386–389) for the development of this construction in Latin, Troickij (1968: 66–80) for old Russian, Vondrák (19282: 484–485) and Beličová and Sedláček (1990: 181) and for Slavic languages. In particular, Beličová and Sedláček (1990: 181) claim that the correlative strategy is attested in all Slavic languages: it is still currently used in Slovenian (although no evidence is provided for this), whereas it is archaic in Bulgarian, Czech, Slovak and Ukrainian. The fact that this construction is attested also in Turkish – a non-Indo-European language – suggests that the development from conditional to (cor)relative may be a cross-linguistically widespread grammaticalization pattern for RC-formation and not a specifically IndoEuropean one. This issue will be further discussed in 6.1 and 6.2. Like the postnominal strategy introduced by hani, the Turkish correlative strategy further developed in Gagauz: this language exhibits postnominal RCs introduced by angï(sï), deriving from hangı (combined with the third person possessive -sı). That is, we may argue that hangı has fully grammaticalized into a relative pronoun, as the examples below demon-

168 Typological issues strate. Language contact with Russian or Bulgarian probably played a central role in this process. Further investigation would be needed here. (4.44) Adam [angïsïnï gördüm.] GAG man RPRO.ACC I.saw ‘The man I saw.’ (Menz 2006: 143) (4.45) Adam [angïsïna para verdim.] GAG man RPRO.DAT money I.gave ‘The man I gave the money to.’ (Menz 2006: 143) 4.1.3. The postposed strategy Instances of the postposed strategy occur in Dutch and German, as may be seen in (4.46) and (4.47). The RC follows the MC and is introduced by a relative pronoun. These RCs are different from postnominal RCs in Dutch and German as they are not verb-final, but verb-second. In literature regarding this type of RCs – including Gärtner (2001), Zifonun (2001a: 79–83) and Weinert (forthc.) – they are called integrated verb-second clauses (IV2s). (4.46) In het publiek zaten mensen, [die DUT in DET audience sat people RPRO.PL hadden geen idee van syntaxis.] had no idea of syntax ‘There sat people in the audience who had no idea of syntax.’ (Gärtner 2001: 136) (4.47) und GER and [die

dort there

hab have

ich ’ne Frau kenn’gelernt, I DET woman got.acquainted erzählte mir ’ne ganze RPRO.NOM.F.SG told to.me DET whole Menge interessante Sachen] amount interesting things ‘There I got acquainted with a woman who told me a lot of interesting things.’ (radio program, woman, 30 y.o.)

Instances of the postposed strategy are attested also in Turkish, as mentioned in 4.1.1, and in other languages of the sample, including English and

Word order 169

Hungarian, as discussed in 3.1: however, in all these languages postposed RCs can be interpreted as embedded (pre- or postnominal) RCs subject to right-extraposition, so they will not be covered in this section. Instead, the postposed RCs in the two examples above have a different structure from their embedded counterparts: if we try to embed the two RCs in (4.49a) and (4.50a), the resulting sentences are unacceptable. To make them acceptable, we need to make them verb-final, i.e. to turn them into proper postnominal RCs, as in (4.49b) and (4.50b). (4.48) a. DUT b.

*Mensen, [die hadden geen idee van syntaxis,] zaten in het publiek. Mensen, [die geen idee van syntaxis hadden,] zaten in het publiek.

(4.49) a. GER b.

*und dort hab ich ’ne Frau, [die erzählte mir ’ne ganze Menge interessante Sachen,] kenn’gelernt und dort hab ich ’ne Frau, [die mir ’ne ganze Menge interessante Sachen erzählte,] kenn’gelernt

Another feature characterizing this kind of constructions is that the referent in the MC is often indefinite: it is preceded by an indefinite article, as in (4.47), by no article, as in (4.46), or by indefinite quantifiers (GER viele ‘many’, einige ‘some’, welche ‘some’). This is also linked to the context in which these constructions often appear, “i.e. with certain existential/presentative constructions which perform particular discourse functions” (Miller and Weinert 1998: 113). In fact, as will be more thoroughly discussed in Chapter 7, this strategy is particularly frequent in presentative constructions in which the N acting as a referent in the MC exhibits a hyperonymous semantic character: typical referents would be substantives like ‘person’ or ‘thing’. In those constructions, the SoA encoded through the IV2 is pragmatically more relevant than the SoA encoded through the MC: the latter has minimal semantic content, whereas the former encodes the property characterizing the new referent. The issue of whether constructions containing IV2s encode restrictive or non-restrictive relative relations is often debated in literature on IV2s. Gärtner argues that IV2s preferably have a restrictive meaning and quotes the following example.

170 Typological issues (4.50) a. GER

b.

Apfeldorf hat viele Häuser, (/) Apfeldorf has many houses [die stehen leer.] RPRO.NOM.PL stay empty ‘Apfeldorf has lots of houses which are uninhabited.’ Apfeldorf hat viele Häuser, (/) Apfeldorf has many houses [die leer stehen.] RPRO.NOM.PL stay empty ‘Apfeldorf has lots of houses(,) which are uninhabited.’ (Gärtner 2001: 113)

Both (4.50a) and (4.50b) display rising intonation, signalled through ‘(/)’. Whereas (4.50b) allows both for a restrictive and a non-restrictive interpretation, as the bracketed comma in the translation suggests, the sentence in (4.50a) can only receive a restrictive interpretation. Gärtner argues that “definite descriptions […] are not accessible antecedents for IV2” (Gärtner 2001: 115). That is, constructions containing definite referents in the MC, like (4.51), would be excluded. (4.51) Sie GER they die

organisieren heute ’ne Geburtstagsparty für organize today DET birthday.party for Tochter [die is’ fünf.] DET daughter RPRO.NOM.F.SG is five ‘Today they are organizing a birthday party for their daughter, who is five.’ (woman, 35 y.o.)

Sentence (4.51) was uttered with no break between the first and the second clause. The RC appears to be an IV2; however, a restrictive interpretation is excluded as the couple mentioned by the speaker does not have other kids. So, in this case the IV2 should be assigned a non-restrictive function. Since IV2s are verb-second, constructions containing IV2s are sometimes claimed to represent coordination relations rather than relative relations (Miller and Weinert 1998: 113). This alternative interpretation is possible DUT die/dat and GER der/die/das can be both demonstrative and relative pronoun, with only slight differences in their paradigm in German. Indeed, IV2s could in principle be regarded as instances of asyndetic coordination. Still, if we apply a syntactic test that turns them into clear instances of coordination, we get a different meaning. Clauses in (4.46) and (4.50) are now syndetically coordinated through the connector ‘and’.

Word order 171

(4.52) In het publiek zaten mensen, en DUT in DET audience sat people and die hadden geen idee van syntaxis. those had no idea of syntax ‘There sat people in the audience and they had no idea of syntax.’ (Gärtner 2001: 136, modified) (4.53) Apfeldorf hat viele Häuser und die stehen GER Apfeldorf has many houses and those stay leer. empty ‘Apfeldorf has lots of houses and they are uninhabited.’ (Gärtner 2001: 113) As the examples show, turning the IV2-constructions into clear instances of coordination gives them a different meaning: the second SoA does not narrow down the range of possible referents of a participant in the first SoA. So, in (4.52) it is implied that the totality of the people who were at the conference had no idea of linguistics, whereas in (4.46) only a group of people who had no idea of syntax is meant. Similarly, in (4.53) it is stated that in Apfeldorf there are a lot of houses and that they all are uninhabited, while in (4.50) being uninhabited refers to many, but not all of the houses in Apfeldorf. To summarize, the postposed strategy attested in non-standard varieties of Dutch and German has the function of highlighting the information provided about a participant of the main SoA: the use of a dedicated word order “signal[s] the importance of the postmodifying clause explicitly” (Weinert forthc.). As shown above, an IV2 can encode different relative relations according to the contexts in which it is used: if the referent in the MC is indefinite, as in (4.46), then we have a restrictive relation; if the referent in the MC is definite, as in (4.51), then we have a non-restrictive relation. That is, IV2-constructions can be considered to be more specialized than their standard postnominal counterparts (as in (4.50.b)), whose interpretation requires a more in-depth knowledge of the communicative situation (cf. also Weinert forthc.).

172 Typological issues 4.2. Relative element The most noticeable divergences between standard and non-standard varieties are found in the use of relative elements. The following phenomena are attested: 1. In non-standard varieties the same relative elements occur as in standard varieties, but they sometimes display different morphosyntactic behaviour: in standard varieties they are inflected, in non-standard varieties they remain uninflected. This peculiarity is discussed in section 4.2.1. 2. Elements that in standard varieties encode the syntactic role of the relativized item in the RC may happen not to encode it in non-standard varieties. This case is presented in section 4.2.2. 3. In non-standard varieties a full range of particles is attested, which often do not find their way into the standard variety. This issue will be discussed in 4.2.3. 4. In non-standard varieties some cases of paradigm regularization occur. This phenomenon is at issue in 4.2.4. 4.2.1. Inflected elements remain uninflected In some of the languages in the sample, a relative element that in the standard variety is inflected remains uninflected in non-standard varieties. This is the case with BUL kojto and RUM care. Both are G-N-inflecting relative pronouns, as Table 4.2 shows. For Bulgarian it needs to be remarked that the accusative form kogoto is used with [+human] masculine referents, whereas for [–human] ones the form kojto is used. Table 4.2. The inflection of BUL kojto and RUM care in the standard variety. BUL M.SG NOM kojto DAT komuto

F.SG kojato –

NT.SG koeto –

PL koito –

RUM NOM GEN/ DAT ACC kogoto kojato koeto koito ACC PP na kogoto na kojato na koeto na koito PP

SG care M: căruia F: căreia (pe) care din care

PL care cărora (pe) care din care

Relative element 173

In non-standard varieties, the nominative form is used for all syntactic positions. The pronoun remains uninflected: either a preposition or a resumptive element is used to encode the syntactic position relativized and gender and number agreement, as the examples below show. (4.54) čovekăt, [kojto vidjaxme...] BUL man.DET RPRO.M.SG we.saw ‘The man we saw.’ (Mišeska-Tomić 2006: 271) (4.55) čovekăt, [kojto go vidjaxme...] BUL man.DET RPRO.M.SG CL.ACC.M.3SG we.saw ‘The man we saw.’ (Mišeska-Tomić 2006: 271) (4.56) čovekăt, [kogoto (*go) BUL man.DET RPRO.ACC.M.SG CL.ACC.M.3SG vidjaxme...] we.saw ‘The man we saw.’ (Mišeska-Tomić 2006: 270) If we compare the standard construction in (4.56) with the non-standard constructions in (4.54) and (4.55), we will notice that the relative pronoun kojto does not inflect. That is, it still conveys gender and number agreement, but it does not explicitly encode the syntactic position relativized, i.e. DO. In (4.54) the role of the relativized item in the RC has to be inferred by the hearer; in (4.55), a resumptive pronoun is inserted, which, as well as encoding gender and number agreement, also encodes the syntactic position relativized. In the following examples IO is relativized. (4.57) čovekăt, [na kojto dadoxme BUL man.DET to RPRO.M.SG we.gave pismoto] letter.DET ‘The man we gave the letter to.’ (Mišeska-Tomić 2006: 271) (4.58) čovekăt, [na kojto mu dadoxme BUL man.DET to RPRO.M.SG CL.DAT.M.3SG we.gave pismoto] letter.DET ‘The man we gave the letter to.’ (Mišeska-Tomić 2006: 271)

174 Typological issues (4.59) čovekăt, [na kogoto (mu) BUL man.DET to RPRO.ACC.M.SG CL.DAT.M.3SG dadoxme pismoto] we.gave letter.DET ‘The man we gave the letter to.’ (Mišeska-Tomić 2006: 271) (4.60) čovekăt, [komuto (*mu) dadoxme BUL man.DET RPRO.DAT.M.SG CL.DAT.M.3SG we.gave pismoto] letter.DET ‘The man we gave the letter to.’ (Maria Manova, p.c.) (4.59) and (4.60) are considered as standard: here, the relative element inflects according to the position relativized (IO). Instead, in (4.57) and (4.58) the relative pronoun remains uninflected: the syntactic role of the relativized item in the RC is expressed through the combination with the preposition na in (4.57) and through the combination with the preposition na and the resumptive clitic mu in (4.58). The examples above can be explained if we take another look at the inflection of BUL kojto in Table 4.2: only the masculine singular displays specific forms for dative, accusative and when combined with a preposition. In non-standard varieties this paradigm is regularized and the nominative form is used in all cases, as already happens for feminine, neuter and plural forms. If the speaker wants to explicitly encode the syntactic role of the relativized item in the RC, s/he can insert a resumptive pronoun. The same phenomenon is attested also in Macedonian (Cramer 2003: 152). In Rumanian we have a similar situation. As Table 4.2 above shows, the form care is used in the standard variety for the nominative and accusative form of all genders and numbers and in combination with prepositions. In non-standard varieties, the following peculiarities occur. PECULIARITY 1: the form care combines with a preposition and replaces GEN/DAT case forms. (4.61) Conferinţa de RUM conference.DET of un eveniment DET event sǎi COMP CL.DAT.3SG

sǎptǎmâna viitoare e sigur week.DET next is surely [cǎruia trebuie RPRO.DAT.M.SG is.necessary dedicǎm toate eforturile.] we.devote all efforts.DET

Relative element 175

‘The conference we have next week is surely an event that we need to devote all our efforts to.’ (Mădălina Chitez, from questionnaire) (4.62) Copilul [la care iai RUM boy.DET to RPRO CL.DAT.3SG you.have dat bani] e un ţigan. given money is DET gipsy ‘The boy you gave the money to is a gipsy.’ (Mădălina Chitez, from questionnaire) (4.63) oamenii [la care le faci bine] RUM men.DET to RPRO CL.DAT.3PL you.do good ‘People to whom you do something good.’ (Muntenian dialect, Rusu 1984: 177) In all three examples above IO is relativized. Whereas in (4.61) the dative form of the relative pronoun is used, in (4.62) and (4.63) this form is replaced by the combination of the preposition la with the nominative form of the relative pronoun. Additionally, in both examples a resumptive element is inserted, which encodes both the syntactic role of the relativized item in the RC and number agreement. Gender agreement is not encoded since the Rumanian third person clitic shows no gender opposition in the dative form: (î)i is the singular form, le is the plural form.) The resumptive has a doubling function. Still, it doubles different features in the examples: in (4.61) the relative pronoun care behaves like a G-N-inflecting pronoun, so the resumptive element doubles both the syntactic role and number agreement, which are already encoded through the relative pronoun; instead, in (4.62) and (4.63), care behaves like a non-G-N-inflecting relative pronoun: the form la care is used in both examples to refer to a singular and a plural referent respectively. So, in (4.61) and (4.62) the resumptive element doubles the syntactic role of the relativized item in the RC, but is the only element encoding number agreement – singular in (4.62) and plural in (4.63) – since the relative pronoun fails to do this. According to Rusu, constructions like (4.61) and (4.62) are widespread “pe aproape tot teritoriul dacoromân” (“in nearly the whole of the Daco-Romanian territory”, Rusu 1984: 205 fn. 87). An alternative version of (4.62) is reported in (4.64). Here, the preposition la ‘to’ is replaced by the apocoped form of the article lui. This preposed article is usually inserted before masculine proper names to build the GEN/DAT form, as in (4.65)

176 Typological issues (4.64) Copilul [lu’ care iai RUM boy.DET DET.DAT RPRO CL.DAT.3SG you.have dat bani] e un ţigan. given money is DET gipsy ‘The boy you gave money to is a gipsy.’ (Mădălina Chitez, from questionnaire) (4.65) Iam dat lui Alexandru RUM CL.DAT.3SG I.have given DET.DAT Alexandru toate cărţile mele de română. all books.DET my of Rumanian ‘I gave Alexandru all my Rumanian textbooks.’ (Cojocaru 2003: 38) In non-standard varieties, the same apocoped form can be used also with feminine nouns: for instance, we have cartea lui Maria instead of cartea Mariei ‘Maria’s book’ (Iliescu 2002: 150). So, in Rumanian non-standard varieties proper names are invariable and do not encode any case distinction. The same is true for care. In both cases, the speakers resort to another morphosyntactic means to signal case-marking: as an alternative to the combination of the preposition la with an inflected (pro)noun, they use an inflected determiner with an uninflected (pro)noun. In particular, the form lu’ care – and also lu cari, a lu cari – is generally attested in supraregional non-standard varieties of Rumanian (Nilsson 1969: 11) and is found in several dialects (Rusu 1984: 177, 205, 222). PECULIARITY 2: the form care behaves like a relative particle. That is, care neither inflects nor combines with prepositions; it only combines with resumptive elements, which encode the syntactic role of the relativized item as well as gender and number agreement. (4.66) Cartea [pe care am terminatRUM book.DET RPRO.ACC I.have finished o de citit azinoapte.] CL.ACC.F.3SG of reading today night ‘The book that I finished reading tonight.’ (Mădălina Chitez, from questionnaire)

Relative element 177

(4.67) Urmele [care le lasă în RUM traces.DET RPAR CL.ACC.F.3PL leaves in limba.] language.DET ‘The traces it leaves in language.’ (Nilsson 1969: 12) (4.68) oamenii [care le faci bine] RUM men.DET RPAR CL.DAT.3PL you.do good ‘People to whom you do something good.’ (Muntenian dialect, Rusu 1984: 177) (4.69) Fata [care mam plimbat cu ea.] RUM girl.DET RPAR REFL I.have walked with her ‘The girl who I went for a walk with.’ (Nilsson 1969: 12) In (4.66) DO is relativized through the standard form pe care combined with a resumptive pronoun. In (4.67), the resumptive pronoun is still present, but the accusative marking preposition pe is left out. This may suggest that care is no longer used as a relative pronoun, but as a relative particle – which is confirmed in (4.68) and (4.69), where care is used in combination with a resumptive clitic and a resumptive pronominal PP respectively. In these two examples, care only encodes the link between the MC and the RC, whereas the syntactic role of the relativized item in the MC, gender and number agreement are encoded by the resumptive element. This use, too, is attested both in supraregional non-standard varieties and in several Rumanian dialects (Rusu 1984: 222). Table 4.3. The inflection of ENG who, FRE qui and ITA che in the standard variety. ENG NOM ACC GEN PP

[+human] who who(m) whose with whom who… with

[–human] which which whose with which which…with

FRE NOM ACC PP

[+anim] qui que à qui

[–anim] qui que à quoi

ITA NOM ACC DAT PP

che che (a) cui con cui

Analogous phenomenona can be observed in English, French and Italian. In the standard variety of these languages, relative pronouns exhibit fewer morphological forms than in Bulgarian and Rumanian. In fact, whereas Bulgarian and Rumanian relative pronouns can be classified as G-N-in-

178 Typological issues flecting, at least partially, ENG who is G-inflecting, FRE qui is partially Ginflecting and ITA che is non-G-N-inflecting. Still, ENG who, FRE qui and ITA che show some case distinctions, as Table 4.3 shows. In standard English, a major distinction is made between [+human] and [–human] referents. The former are relativized using who, as in (4.70), the latter using which, as in (4.71). (4.70) Janet posted a letter for me last week for a friend [who I worked ENG with at Ipswich] (Herrmann 2005: 47) (4.71) these are the walls [which Leo scribbled his theories on] after he ENG had been imprisoned for heresy (Smits 1989: 291) In standard French there exists a three-partite opposition between NOM, ACC and prepositional cases: see the examples below. In (4.72) and in the first two RCs in (4.74), qui is used to relativize SU; in the third RC in (4.74) qui is preceded by a preposition and relativizes OBL. The form que in (4.73) relativizes DO. (4.72) tu vois / c’est deux pièces de puzzle / [qui FRE you see it’s two pieces of puzzle RPRO vont s’ assembler] // will REFL assemble ‘You see, it’s like two pieces of a jigsaw, which will get together.’ (Cresti and Moneglia 2005, C-ORAL-ROM, ffamcv01) (4.73) par conséquent / les gestes [que vous FRE for consequent DET gestures RPRO you faisiez / ils sont [/] ils étaient [/] ils étaient made they are they were they were très bien / mais / very good but ‘As a consequence, the gestures that you were making, they are, er, they were very good, but…’ (Cresti and Moneglia 2005, C-ORALROM, fpubmn05) (4.74) c’est le FRE it’s DET me l’ to.me it

président de president of avait apprise / had taught

Lionel _P1 Lionel-P1 qui est RPRO is

[qui RPRO

médecin] / doctor

Relative element 179

[avec qui j’ ai rendez-vous d’ailleurs euh with RPRO I have appointment besides er demain] / tomorrow ‘It’s the president of Lionel-P1 who had taught me this – he’s doctor and, by the way, I have to meet him, er, tomorrow.’ (Cresti and Moneglia 2005, C-ORAL-ROM, dffamdl22) In standard Italian there is a basic opposition between the use with preposition, where cui is used, and the use without preposition, where che is used – the only exception being the use of cui alone to encode the dative form. For example, in (4.75) che is used to relativize DO, in (4.76) a cui is used to relativize IO. (4.75) Sono solo pregiudizi stupidi [che non ITA they.are only prejudices silly RPRO not dovresti avere.] you.should have ‘It’s only silly prejudices that you shouldn’t have.’ (Marianna Gionta, from questionnaire) (4.76) Il ITA DET i

bambino [a cui hai dato child to RPRO you.have given soldi] è uno zingaro. DET money is DET gipsy ‘The child you gave the money to is a gipsy.’ (Marianna Gionta, from questionnaire)

Now, in non-standard varieties of these three languages these case oppositions are neutralized. Pronouns do not inflect any more: instead, ENG which, FRE que and ITA che are the only forms used and can be combined with resumptive pronouns. In this sense, they behave as relative particles. For instance, in (4.77) which is used with a [+human] referent to relativize SU; in (4.78), it is combined with a possessive to relativize GEN. (4.77) and then there was C. [which caught his hand in the machinery up ENG here] and he had his hand off, being severed at the wrist (Herrmann 2005: 42)

180 Typological issues (4.78) But you’d got to watch, there again, that er you didn’t exceed the ENG width of er of your waggon, [which its maximum limit was er would be er eight foot three, or er eleven foot six, high] (Herrmann 2005: 71) As for FRE que, in (4.79) and (4.80) it combines with a resumptive pronoun and a stranded preposition to relativize DO and OBL respectively. (4.79) Il FRE he type guy vu seen le

était was [que

grand et costaud mais c’est un big and sturdy but it.is DET je l’ ai toujours RPAR I CL.ACC.M.3SG have always avec une bouteille de whisky dans with DET bottle of whisky in sac.] DET bag ‘He was big and sturdy, but he’s a guy who I have always seen with a bottle of whisky in his bag.’ (Gadet 1997: 118)

(4.80) Elle avait son manteau [qu’ elle allait FRE she had her coat RPAR she went régulièrement au marché avec.] regularly to.DET market with ‘She had her coat, which she used to wear when she went to the market.’ (Gadet 1997: 118) In Italian, DO and IO can be relativized through the combination of the relative particle che with resumptive clitics, as in (4.81) and (4.82). (4.81) Abbiamo avuto le porte chiuse da tutti // ITA we.have had DET doors closed by all gente [che l’ abbiamo aiutata.] people RPAR CL.ACC.F.3SG we.have helped ‘Nobody gave us assistance, even people who we had helped.’ (TV program, woman, 50 y.o.) (4.82) E ITA and bene] good

le

persone [che non gli people RPAR not CL.DAT.3PL vogliono abbassar-li, questi want turn.down.INF-them these DET

va goes decibel. decibels

Relative element 181

‘And the people that don’t find it good want these decibels to be turned down.’ (radio program, student, 25 y.o.) Something of particular interest is the reanalysis operated on the nominative form of FRE qui: on account of its pronunciation, [ki], it is interpreted as qu’i, deriving from the contraction of the combination que+il (relative particle and masculine subject clitic pronoun). This gives rise to the forms qu’elle, qu’ils – phonetically realized as [ki], if the following verb begins with a consonant, or [kiz], if the following verb begins with a vowel – and qu’elles: cf. Blanche-Benveniste (1990: 322) and Miller and Weinert (1998: 111–113). Two examples are reported below. In (4.83), the relative elements are given in phonetic transcription. This is why they appear in square brackets. (4.83) On a même des clients FRE one has even DET customers [[k-iz] arrivent] et là RPAR-CL.NOM.M.3PL arrive and PAR [[k-i] font pas d’ échauffement.] RPAR-CL.NOM.M.3PL do not of warming-up ‘We also have some customers who come (to the gym) and do not do any warming-up.’ (Gapany 2004: 182) (4.84) J’ ai téléphoné à Martine / [qu’ FRE I have phoned to Martine RPAR elle était pas là.] she was not there ‘I have phoned Martine, who was not at home.’ (Gadet 1997: 118) So, the first relevant phenomenon regarding the parameter ‘relative element’ in non-standard varieties of European languages seems to be a reduction of the paradigm of relative pronouns. In particular, non-standard varieties show a tendency towards the use of a single, invariable form, which can combine with resumptive elements to relativize different positions of the AH. One concluding remark: the issue of the nature of Romance que/che, i.e. whether it should be viewed as a relative pronoun, a relative particle or a general connector (complementizer), is not further pursued here. As mentioned in 3.2, in this work the position is taken that elements like Romance que/che, ENG that, BCS što can be differently pigeonholed according to

182 Typological issues the parameters selected. Moreover, in 3.5.1 it was claimed that the syntactic status of particles can be disambiguated only if they are located inside a construction. As for the specific question of whether que/che is to be treated as a relative particle or a relative pronoun, in this work it is assumed that que/che can be both a relative pronoun, because it occupies a position inside a morphological paradigm, and a relative particle, because there are contexts in which this element only encodes the link between the MC and the RC. In a number of cases it is certainly impossible to establish whether we have to do with a relative pronoun or a relative particle, for example when ITA che is used alone to relativize SU or DO. According to the theoretical framework adopted in this work, such instances should be regarded as cases of formal coincidence between che as a pronoun and che as a particle. 4.2.2. The syntactic role of the relativized item in the RC is not encoded The second peculiarity that emerges when examining the relative elements attested in the sample is that some elements that in standard varieties encode the syntactic role of the relativized item in the RC may happen not to encode it in non-standard varieties. This applies to both relative pronouns and specialized relative elements: by not encoding of the position relativized, they act as relative particles (cf. Table 3.5). As far as relative pronouns are concerned, this property has already been described in the previous section. For instance, when BUL kojto or RUM care are not inflected, but occur with a preposition to encode some syntactic positions, they are still relative pronouns because they encode the syntactic role of the relativized item in the MC: see (4.59)–(4.60) vs. (4.57)–(4.58) or (4.61) vs. (4.62). Instead, when these elements are invariable and combine with a resumptive element, they could be regarded as encoding only the link between the MC and the RC: in other words, they behave as relative particles, cf. (4.67) to (4.69). Similar considerations can apply to also the English, French and Italian examples reported in 4.2.1. As for specialized relative elements, this phenomenon seems to be quite widespread, although it has not reached the same stage in all the languages in which it is attested. The specialized relative elements that prototypically cease to encode the syntactic role of the relativized item in the RC and only express the link between the MC and the RC are those used to relativize LOC and, in only one case, GEN. This case is represented by FRE dont: as discussed in 3.2.1, dont is used to relativize GEN and, in some specific

Relative element 183

cases, OBL. However, this specialized relative element derives from LAT de unde, which had a locative meaning. So, from a diachronic point of view it may be assimilated to the other LOC-specialized elements. The languages in which this phenomenon is attested can be placed onto a scale consisting of five stages: 1. the specialized relative element is used to relativize LOC; 2. the specialized relative element is used as an unspecific connector to link MC and RC; 3. the specialized relative element is used to relativize [–animate] OBL; 4. the specialized relative element is used to relativize [+/–animate] OBL; 5. the specialized relative element is used to relativize all positions of the AH. Some remarks are in order. To start with, one should actually speak of specialized relative elements only in Stage 1, where the (simple) relative element encodes both the link between MC and RC and the syntactic role of the relativized item in the RC. From Stage 2 onwards, the specialized relative element behaves like a relative particle in that it only encodes the link between the MC and the RC; the other parameters are encoded by means of a resumptive element, if they are encoded at all. So, in the examples below specialized relative elements will be glossed as ‘SRE’ only in Stage 1, otherwise they will be glossed as ‘RPAR’. Secondly, both in Stage 2 and in Stage 5, the relative element does not encode any syntactic role. Still, in Stage 5 the syntactic role (SU, DO, IO…) can be pragmatically inferred. Instead, in Stage 2 the syntactic role remains unspecified and the role of the participant shared by the two SoAs, first encoded in the MC, turns into a “reference point” (Maslova and Bernini 2006: 90) for the SoA encoded in the RC: this will be discussed below and, later on, in section 4.2.3. Thirdly, the five-stage sequence described above can be seen as having implicational value: for instance, if in a language a specialized relative element can be used as described in Stage 3, then it can be used also as described in Stages 1 and 2. In Table 4.4 the group of languages for which information on this phenomenon was available is reported: languages appear on different rows according to the uses that the LOC-specialized element has. GER wo appears twice in Table 4.4: in supraregional non-standard varieties the use of wo ranges from Stage 1 to Stage 3; instead, in (Alemannic) dialects it ranges from Stage 1 to Stage 5.

184 Typological issues Table 4.4. Uses of the LOC-specialized element in some of the languages in the sample. 1. LOC

2. CONN

3. [–anim] OBL

4. OBL

5. SU, DO, IO

+









+

+







+

+

+





+

+

+

+



+

+

+

+

+

EST kus POL gdzie RUM unde RUS gde BCS gdje ENG where ITA dove GER wo (supraregional) DUT waar BUL deto FRE dont GER wo (dialects) GRE pu LIT/LTV kur

Examples for each stage are now provided. STAGE 1. The Estonian specialized relative element kus ‘where’ can be used only to relativize LOC, as (4.85) shows. (4.85) maja, [kus ma elan] EST house SRE I live ‘The house where I live.’ (Erelt 2003: 124) Although referents relativized by means of a LOC-specialized relative element are prototypically [–animate], in some languages the referent can be [+animate]: see for instance the Rumanian example (4.86), where a resumptive element is added to specify that the referent is animate. (4.86) şţiţi ROM you.know Femeii woman.to.DET la ea] // at her

cui ito.who CL.DAT.3SG [unde- aţi SRE you.have

a has fost been

spus? told dumneavoastră you

Relative element 185

‘Do you know who he told it to? To the woman that you have visited.’ (Vulpe 1980: 129) Also in (4.87) the resumptive element has a specifying function: the specialized relative element encodes the syntactic role LOC (‘in’), the resumptive adverb specifies the nature of the spatial relation (‘under’) (4.87) Ils voulaient le torturer sur une FRE they wanted him torture.INF on DET grille [où dessous il y avait des flammes.] grill SRE under there.was DET flames ‘They wanted to torture him on a grill under which there was fire.’ (Gapany 2004: 188) STAGE 2. BCS gdje can be used both to relativize LOC, as in the second RC in (4.88), and as an unspecific connection marker, as in (4.89). (4.88) Vodič nam je pokazao [...] veliku salu [gdje BCS guide to.us is shown big hall SRE će uskoro biti umjetnička galerija.] will soon be artistic gallery ‘The guide showed us […] the big hall where there’ll be an art gallery soon.’ (Milen Marić, from questionnaire) (4.89) Znam neke žene [gdje se uvijek pitam kako BCS I.know some women RPAR REFL always I.ask how sve stižu.] everything they.manage ‘I know some women that I’m always wondering how they get to do everything.’ (Milen Marić, from questionnaire) Structures similar to (4.89) are attested in other languages. Consider the examples below. (4.90) Познавам жени дето винаги съм се чудела как успяват с BUL всичко. Poznavam ženi [deto vinagi săm se čudela I.know women RPAR always I.am REFL wondered kak uspjavat s vsičko.] how they.come.to.terms with everything

186 Typological issues ‘I know women that I’ve always wondered how they get to do everything.’ (Maria Manova, from questionnaire) (4.91) Ich spüre Schmerz an Muskeln, [wo ich GER I feel pain in muscles RPAR I gar nicht wusste, dass sie da sind.] at.all not knew COMP they there are ‘I’m feeling pain in muscles whose existence I ignored’ (man, 50 y.o.) (4.92) Dickens is one of the few authors [where I prefer to watch the ENG video.] (Comrie 2003: 34) (4.93) Con ITA with aprire open [dove

certi adolescenti si riesce ad certain adolescents REFL manages to un canale, ma ci sono anche ragazzi DET channel but there.are also teenagers gli insegnanti si trovano in RPAR DET teachers REFL find in seria difficoltà.] serious difficulty ‘With some adolescents one can open a communication channel, but there are also teenagers with whom teachers have serious trouble.’ (TV program, woman, 45 y.o.)

In (4.89) to (4.93), the specialized relative element encodes the link between the MC and RC, but cannot be claimed to encode the syntactic position LOC – actually, it cannot be claimed to explicitly relativize any syntactic positions at all. Taking a closer look, a distinction can be made between two types of constructions. a. (4.89) to (4.91) represent instances of long relativization (cf. Smits 1989: 82): the relativized item does not have a syntactic role in the RC, but in a clause dependent on the RC. For instance, in (4.89) to (4.91) the relativized item is the subject of a complement clause dependent on the RC. It appears that the relativized item does not need to be explicitly codified: it is codified through the resumptive pronoun sie in (4.91) because in German subordinate clauses the subject cannot be left out, whereas it is not expressed in (4.89) and (4.90) because in Bosnian/Croatian/Serbian and Bulgarian subject pronouns can be dropped.

Relative element 187

b.

In (4.92) and (4.93) the relativized item does not have any syntactic function inside the RC: the RCs can be regarded as semantically and syntactically autonomous. So, the relative element introducing the RC simply signals that the MC and the RC are in relation with each other and that the SoA encoded in the RC regards a participant of the MC. In order to account for both types of constructions, we may first observe that (4.89) to (4.93) can be paraphrased as shown in (4.94). (4.94) a. b. c. d.

‘I’ve known some women ← as far as they are concerned, I’m always wondering how they get to do everything.’ (4.89 and 4.90) ‘I’m feeling pain in muscles ← as far they are concerned, I didn’t know that they existed.’ (4.91) ‘Dickens is one of the authors ← as far as they (or: their works) are concerned, I prefer to watch the video (instead of reading the book).’ (4.92) ‘There are teenagers ← as far as they are (or: their behaviour is) concerned, teachers have serious trouble.’ (4.93)

In the above paraphrases the link between the MCs and the RCs is clarified through the phrase ‘as far as X is concerned’: so, one could advance the hypothesis that in these examples the relativized item in the RC has the role of ‘topic’. This role can either be expressed syntactically, as in (4.90), or it may remain unexpressed, as in (4.92) and (4.93). In these two examples the RCs are syntactically complete, i.e. no syntactic gap needs to be filled: so, the relative element can be seen as fulfilling the function of indicating that the topic of the RC is to be found in the MC, or, vice versa, that the RC encodes an SoA whose topic is a participant of the main SoA. With reference to Maslova and Bernini (2006), the shared participant may be regarded as a sort of ‘reference point’: first, a participant is introduced in the main SoA; then, either the secondary SoA makes direct reference to this participant – as in (4.89) to (4.91), where the same participant is shared in the MC and in the RC – or reference is made to “an entity closely related to one of [the] participants” of the main SoA (Maslova and Bernini 2006: 83) – like in (4.92), where ‘the authors’ works’ rather than the referent ‘authors’ constitute a participant of the qualifying SoA, or in (4.93), where ‘the teenagers’ behaviour’ rather than ‘the teenagers’ is referred to in the RC. Resorting to this kinds of constructions, speakers have “the possibility to introduce a reference to the topic [=the shared participant, A.M.] […] without integrating this reference into the propositional structure of

188 Typological issues the sentence” (Maslova and Bernini 2006: 88). In these cases the RC is reduced to its primary function, summarized in Kuno’s “thematic constraint on relative clauses [:] a relative clause must be a statement about its head noun” (Kuno 1976: 418) – or, as we saw, to an entity closely related to the head noun. When the role of the shared participant in the RC is not expressed through morphosyntactic devices, the hearer relies on a number of semantic and pragmatic factors, which are discussed in detail in section 4.3.3. In his study on the distribution of the relative element wo in Swiss German, Lötscher reaches a similar conclusion: Wenn kein konkret bestimmbares Satzglied vorhanden ist, bleibt die Relativpartikel ein allgemeiner “Relator”, dessen Bedeutung nichts mehr besagt als “in Bezug auf den/die/das”. Natürlich muss eine pragmatische Verknüpfung zwischen dem Kopf des Relativsatzes und dem Relativsatz selbst vorhanden sein […]. Näher zu bestimmen, worin die sachliche Relation zwischen der Aussage des Relativsatzes und dem übergeordneten Nomen oder der Proposition besteht, ist jedoch nicht Sache der Grammatik, sondern der pragmatischen Interpretation auf der Basis der Kontextbedingungen. (Lötscher 2004: 176)

STAGE 3. The specialized relative element is used to relativize inanimate obliques. The syntactic role of the relativized item is usually encoded through a resumptive element. In (4.95) to (4.96) below, the syntactic role is encoded respectively through a pronominal adverb and a stranded preposition. (4.95) Es sind Szenen, [wo man immer gerne GER it are scenes RPAR one always willingly drüber nachdenkt.] PRO.ADV think ‘These are scenes that one always likes to think about.’ (TV moderator, 35 y.o.) (4.96) De grammofoonplaten [waar ik gisteravond DUT DET longplays RPAR I yesterday.evening met jou naar heb geluisterd,] waren mooi. with you to have listened were nice ‘The longplays I’ve listened to with you yesterday evening were nice.’ (Fontein and Pescher-ter-Meer 19963: 151)

Relative element 189

STAGE 4. The specialized relative element is used for both inanimate and animate obliques. A resumptive is prevalently used: in (4.97) it is a stranded preposition, in (4.98) it is a pronominal PP. Instead, in (4.99) no resumptive is set. (4.97) De man [waar ik in de winkel mee DUT DET man RPAR I in DET shop with stond te praten,] is mijn oom. stayed to talk.INF is my uncle ‘The man who I was talking with in the shop is my uncle.’ (Donaldson 1997: 72) (4.98) Kséro tin kopéla [pu eksartáse apó aftín.] GRE I.know DET girl RPAR you.depend on her ‘I know the girl you depend on.’ (Joseph and Philippaki-Warburton 1987: 166) (4.99) Elle est liée avec Anne [dont elle n’ FRE she is close with Anne RPAR she not rien en commun.] nothing in common ‘She is very close to Anne, with whom she has nothing in common.’ (Gapany 2004: 189)

a has

As mentioned above, FRE dont can be used in the standard variety to relativize some obliques, i.e. those introduced by the preposition de. Still, in the example above the relativized oblique is introduced by the preposition avec (avoir quelque chose en commun avec X ‘to have something in common with X’). So, the use of dont in this example is to be regarded as nonstandard (see also Guiraud 1966: 42). STAGE 5. At this stage, the specialized relative element can be used to relativize higher positions of the AH: SU, as in (4.100) and (4.101), DO, as in (4.102) and (4.103), and IO, as in (4.104) and (4.105). As the examples show, a resumptive element can be inserted, but not necessarily. (4.100) и хората дето видяха катастрофата бяха като онемели BUL i chorata [deto vidjacha katastrofata] and people.DET RPAR saw accident.DET bjacha kato onemeli were like unable.to.speak

190 Typological issues ‘And the people who saw the accident could not utter a single word.’ (Petăr Kehajov, from questionnaire) (4.101) Ich han welle erfaare, wie sìn d GER I have wanted know.INF how are DET Gedankegäng vome Mänsch, [wo sich in trains.of.thought of.DET man RPAR REFL in dere Uusnaamesituazion vor emene Gricht emal this exceptional.situation before DET court PAR mues verantwòrte.] must take.responsibility ‘I wanted to know which thoughts a man has who in an exceptional situation is forced to stand trial and to take responsibility.’ (Alemannic dialect, Lötscher 2004: 165) (4.102) Dans cette ville, il y a dix règles [dont toute FRE in this town there.is ten rules RPAR every personne doit respecter et savoir.] person must follow and know ‘In this town there are ten rules that must be known and followed by everybody.’ (Gapany 2004: 189) (4.103) Il s’ est vendu une armoire FRE he REFL is sold DET wardrobe fribourgeoise [dont un de mes amis a from.Fribourg RPAR one of my friends has été la voir.] been CL.ACC.F.3SG see ‘He sold a Fribourg-style wardrobe, which a friend of mine went to see.’ (Gapany 2004: 189) (4.104) Το παιδί που έδωσες τα λεφτά είναι γυφτάκι. GRE To peðí [pu éðoses ta leftá] íne DET child RPAR you.gave DET money is jiftáki. gipsy ‘The boy you gave the money to is a gipsy.’ (Nikolaos Mytilinaios, from questionnaire)

Relative element 191

(4.105) John, tipul [de i sa RUM John chap.DET RPAR CL.DAT.3SG REFL has furat maşina săptămâna trecută.] robbed car.DET week.DET last John, the chap whose car was stolen last week.’ (Gabriela Diaconu, from questionnaire) Kurzová (1981: 83) and Endzelin (1923: 401) claim that also LIT/LTV kur can be used to relativize higher positions of the AH. This would be restricted to dialectal varieties, though: “wie nämlich in litauischen Mundarten […], so wird auch in lettischen Mundarten kur ‘wo’ als Pronomen relativum gebraucht: Tas bij jūsu virsinieks, kur... [That was your chief RPAR, A.M.]” (Endzelin 1923: 401). Unfortunately, Endzelin provides only this example, which is taken from a folk song. This phenomenon should be further investigated. If we look at Table 4.4 again, we will notice that Stage 1 to Stage 5 constitute a sort of reverse version of the AH: that is, in the languages of the sample the LOC-specialized relative element seems to climb up the AH, from the bottom to the top. So, we may formulate the hypothesis that the LOC-specialized relative element first becomes a general connector, then extends its use to obliques and finally to higher positions of the AH. In this progression, which can be seen as a grammaticalization path, the specialized relative element undergoes semantic bleaching in that it becomes a relative particle, i.e. it does not encode the syntactic role of the relativized item in the MC any longer. Sometimes it also undergoes phonetic erosion. This is for instance the case in Balkan languages: BUL deto derives from the kădeto ‘where’, RUM de from unde. In these languages, the grammaticalized specialized relative element – which is now a relative particle – can be used to relativize all syntactic positions, including LOC, whereas the contrary is not possible: for instance, RUM unde can relativize only LOC, whereas RUM de can relativize all syntactic positions, as reported in Table 4.4. For further details on specialized relative elements in Balkan languages see Murelli (2009). Additionally, in some languages, the specialized relative element extends its use to temporal circumstantials: this is the case with ITA dove (cf. Bernini 1989: 91 and Alfonzetti 2002: 99), FRE où (Blanche-Benveniste 1990: 327–328), GER wo (Pittner 2004: 363–365). Still, this use does not seem to be cross-linguistically widespread: for this reason it was not included in the hierarchy of the five stages listed above.

192 Typological issues Finally, it must be pointed out that the classification of the specialized relative elements on the five-stage scale described above is based on the default uses of these elements: still, in some languages non-default occurrences can be attested. This is the case, for instance, with BCS gdje and ITA dove, which were both placed in Stage 2. Dmitriev (1972: 245) and Gołąb and Friedman (1972: 43) quote examples in which gdje relativizes SU. Dmitriev claims that this mostly happens with perception verbs and that this use is restricted to some Serbian dialects. An example is given below. (4.106) Aли он, кад види маjку гдjе плаче, навали jош већма и стане jе BCS заклињати, да му каже, кто jе. Ali on, kad vidi majku [gdje plače,] navali but he when he.sees mother RPAR cries assails još većma i stane je zaklinjati, da still more and starts her beg COMP mu kaže kto je. to.him tells who is ‘But when he sees his mother crying, he becomes even more insistent and begs her to tell him who s/he is.’ (Dmitriev 1972: 245) Similarly, Bernini (1989: 91) and Alfonzetti (2002: 96–100) quote examples in which ITA dove is used for higher positions of the AH. Still, they admit that this use is quite marginal: on the one hand, some of these occurrences are cases of construction reformulation; on the other, dove is particularly used when the antecedent displays a locative or temporal semantic content, as in the example below. Here, dove is combined with a resumptive clitic to relativize DO and the referent has a clear locative semantic content (‘room’). These additional, though marginal occurrences once more underline the difficulty of pigeonholing relative elements unambiguously. (4.107) m’ han ITA to.me they.have [dove la RPAR CL.ACC.F.3SG di danza] of dance

detto che said COMP useranno they.will.use

è is per to

una DET

fare do

stanzetta little.room corsi di / courses of

Relative element 193

‘I was told that this is a little room that’s going to be used for dance courses.’ (Bernini 1989: 91) 4.2.3. The diffusion of relative particles As already observed by Cuendet (1939), a striking property of European non-standard varieties is the great number of relative particles that they exhibit. In Table 3.6 it was shown that relative particles are attested in nearly all of the languages in the sample. Table 4.5. Relative particles in East-European languages and English. LANGUAGE BCS BUL CZE ENG LSO MAC POL RUM USO

RELATIVE PARTICLES što, te, da deto, što(to), da co, jak, že that, as, what, which kenž, což, ako, hač, (h)až što, da co, że ce, de, că, să kiž, kož, štojž

In Table 4.5 above languages that possess more than one relative particle are reported: this is the case with some Slavic languages, Rumanian and English; especially West and South-Slavic languages exhibit more than one particle. Examples for their use were given in 3.5.1. Here, it may be added that relative particles are not randomly used, but show a specific distribution. On the one hand, some particles are variety-bound: they are either widespread in some dialectal areas but not in others or they are used in local varieties, but not in supraregional varieties. This is for instance the case with English and Sorbian dialects. The frequency of relative particles in English dialects is investigated in Herrmann (2005), who finds that that is most common in Northern Ireland and Northern/Central England; as can be found prevalently in Central England; what is widespread in Southern/ Eastern England; which is equally distributed throughout England, but is used very rarely in Northern Ireland (Herrmann 2005: 25–26). The distribution of relative particles in Sorbian dialects is at issue in Faßke (1996: 170–176): in the standard varieties of Lower and Upper Sorbian the particles kenž and kiž are used respectively. In Lower Sorbian dialects, the par-

194 Typological issues ticle ak(o) is gaining importance, so that it has begun to be accepted in the standard variety. In Upper Sorbian dialects, the particles kož and štojž are used beside kiž. In the transitional area between Lower and Upper Sorbian dialects, the particles což, štož, hač, (h)až occur. Examples for different particles are given below. (4.108) Źo jo ta droga, [ak my smy pe LSO where is DET street RPAR we are on njej šli]? her gone ‘Where is the street on which we walked?’ (dialect, Faßke 1996: 170) (4.109) Źo jo ten clejak, [což jo mě te LSO where is DET person RPAR is to.me DET pjenjeze dal]? money given ‘Where is the person who gave me the money?’ (dialect, Faßke 1996: 171) (4.110) Dźe ten člowjek jo, [hač som cora USO where DET person is RPAR am yesterday jom te pjenjeze dał]? to.him DET money given ‘Where is the person who I gave the money to yesterday?’ (dialect, Faßke 1996: 176) (4.111) Dźe je tón člejek, [kož sym jom čera USO where is DET person RPAR am to.him yesterday te pjenjezy dal]? DET money given ‘Where is the person who I gave the money to yesterday?’ (dialect, Faßke 1996: 172) As the examples above show, the Upper and Lower Sorbian particles can be used for all positions of the AH – mostly in combination with a resumptive element when a non-SU position is relativized. Also in Bosnian/Croatian/Serbian, Bulgarian and Czech particles show a particular distribution. According to Gołąb and Friedman (1972: 46) BCS te is more common in Montenegro. BUL što/štoto has a strong dialectal-

Relative element 195

archaic character, as argued in Genadieva-Mutafčieva et al. (19983: 84): “В народното творчество и у по-стари автори се среща и формата щото.” (“In folk production and older authors the form štoto can also be found”). Finally, CZE jak is claimed to be widespread only in dialects, but not in colloquial varieties (Gołąb and Friedman 1972: 37), although Petr (1987: 528) and Sgall et al. (1992: 138) claim that jak has found its way into the supraregional variety obecná čeština (Common Czech) together with the particle co. So, in (4.112) below co may be found instead of jak. (4.112) To je ten chlapec, [jak jsem ti CZE that is that boy RPAR I.am to.you o něm říkal.] about him spoken ‘That’s the boy I told you about.’ (Gołąb and Friedman 1972: 37) On the other hand, some particles are semantics-bound in that they are used in contexts where the relative relation also exhibits a nuance of consecutivity or finality. This is for instance the case with BCS/BUL/MAC da, CZE že, POL że, RUM că, să: examples were given in 3.5.1. A partial exception to this is mentioned by Laskowski (1991: 275): in South-Eastern Polish dialects the relative particle że is used not only in constructions with a consecutive nuance, but has generalized to an all-purpuse (relative) particle. (4.113) Ten chłopak, [że-smy go wczoraj POL that lad RPAR-we.are CL.ACC.M.3SG yesterday spotkali.] met ‘The lad we met yesterday.’ (Laskowski 1991: 275) Additionally, elements that in the standard variety do not act as relative particles may serve as such in non-standard varieties. The great majority of specialized relative elements occupying a position from Stage 2 to Stage 5 in Table 4.4 are examples of this, along with FIN kun and GER was. FIN kun ‘when’ is used in dialects as a relative particle (Sulkala and Karjalainen 1992: 42; Miller and Weinert 1998: 349; Buchholz 2004: 196). Some more words will be spent on GER was. Was is accepted as a relativizer in standard German in the following cases (cf. Dudenredaktion 20098: 1031–1033, 1037):

196 Typological issues a. b. c. d.

in free relatives, as in (4.114); when the referent is the whole MC, as in (4.115); when the referent is a [–animate] indefinite pronoun, as in (4.116); when the referent is a substantivized neuter adjective, as in (4.117).

(4.114) [Was sie/er nicht weiß,] macht sie/ihn GER RPRO.NOM.SG she/he not knows makes her/him nicht heiß. not hot ‘What a (wo)man doesn’t know cannot make her/him angry.’ (German idiom) (4.115) Ich musste lachen, [was die anderen GER I had.to laugh RPRO.NOM.SG DET others ziemlich irritierte] quite annoyed ‘I could not but laugh, which quite annoyed the others.’ (Dudenredaktion 20098: 1030) (4.116) Ist Zeit etwas [was man besitzen GER is time something RPRO.NOM.SG one own kann]? can ‘Is time something you can own?’ (Dudenredaktion 20098: 1032) (4.117) Das Zweite [was ich dir erzählen GER DET second RPRO.NOM.SG I to.you tell möchte ist sehr vertraulich. would.like is very confidential ‘The second thing I would like to tell you is very confidential.’ (Dudenredaktion 20098: 1032) In supraregional non-standard varieties, the use of was is increasing (cf. also Romaine 1984: 450). First, it can be used with all neuter referents. In (4.118) and (4.119) the referents Geld ‘money’ and Gut ‘good’ are neuter. (4.118) Mit dem GER with DET haben,] können have can

Geld, money Sie you

[was

Sie you schon zufrieden already content RPRO.NOM.SG

gewonnen won sein. be

Relative element 197

‘You can already be content with the money you’ve won.’ (TV moderator, about 40 y.o.) (4.119) Denn die Gesundheit ist ein langfristiges GER because DET health is DET long-term Gut, [was erhalten werden muss.] good RPRO.NOM.SG preserved be must ‘In fact, health is a long-term good that must be preserved.’ (TV program, lawyer, about 50 y.o.) A further step consists of the extension of was to masculine or feminine referents, as documented in the two examples below. While (4.120) can be considered transitional, since eine ganze Menge could be considered as a synonym of the neuter adjective Vieles ‘much’, which could explain the use of was as a relativizer, in (4.121) the antecedent is a masculine substantive. Hence, in these examples was is glossed as ‘RPAR’: it doesn’t explicitly encode the position relativized or gender agreement, but actually conveys number agreement, as it is not used with plural referents. (4.120) Es gibt eine ganze Menge, [was man da vermissen GER there.is DET whole amount RPAR one there miss könnte.] could ‘There are a lot [of things] you could miss there.’ (TV program, man, 40 y.o.) (4.121) Dazu GER additionally [was sich

kommt noch Nebel oder Hochnebel, comes still fog or low.stratus inzwischen breit gemacht hat.] RPAR REFL in.meantime broad made has ‘Additionally there’s also fog or low stratus, which has spread in the meantime.’(radio-program, man, 35 y.o.)

So, was in supraregional German varieties should be ascribed a transitional status. For corpus evidence and further details on was see also Murelli (2011). In other Germanic languages a similar development of relative/interrogative ‘what’ is attested: cf. for instance Van der Wal (2002: 31–32) for DUT wat. In Afrikaans, North Frisian and German dialects the particles wat and was (wås, wos, we) are used for all kinds of referents: cf. Hoekstra (2002), Patocka (2000) and Fleischer (2004a, 2004b, 2006).

198 Typological issues The degree of standardness of constructions where specialized relative elements or other elements are used as relative particles will be discussed in section 5.1.2 and 5.2: there, it will emerge that they are sometimes considered standard, as is the case with GRE pu; in others cases they are regarded as non-standard, as it happens with GER wo or was. Probably a unique phenomenon in the realm of relative particles, in South-Eastern Serbian dialects the relative particle što “occurs in association with a definite article” (Mišeska-Tomić 2006: 279), yielding the forms štoto, štovo and štono. These forms are used according to the (intended) degree of proximity of the referent: the t-form is neutral, the v-form indicates proximity to the speaker, the n-form indicates distance from the speaker. (4.122) onija čovek [štono mu tvoja majka BCS that man RPAR.DET CL.DAT.M.3SG your mother dala pismo] gave letter ‘The man that your mother gave a letter to.’ (South-Eastern Serbian dialect, Mišeska-Tomić 2006: 279) (4.123) prsten [štovo ti ga majka kupila] BCS ring RPAR.DET to.you CL.ACC.M.3SG mother bought ‘The ring that your mother bought you.’ (South-Eastern Serbian dialect, Mišeska-Tomić 2006: 279) In (4.122) the referent is distant, so the form štono is used, which is parallel to the demonstrative onija in the MC; in (4.123) the speaker may be thought to point at the ring that the hearer has on his/her finger, so the form štovo occurs. As Mišeska-Tomić remarks, this phenomenon probably correlates with the fact that “definite articles in these dialects make triple special distinctions” (Mišeska-Tomić 2006: 279 fn. 101) – a peculiarity that these dialects share with Bulgarian dialects (Stojkov 19933: 232) and Macedonian. As discussed in 3.2.1 and 3.2.2, relative particles can either be used alone or combine with resumptive elements to relativize all positions of the AH. Here, a difference between standard and non-standard varieties emerges: if relative particles are attested in standard varieties, they are prevalently used alone and can relativize only higher positions of the AH, i.e. SU and DO; instead, in non-standard varieties relative particles occur both alone and in combination with resumptive elements to relativize all

Relative element 199

positions of the AH. This use is exemplified below for POR que and CZE co. In (4.124) to (4.130) POR que is used alone to relativize respectively SU, DO, IO, OBL, GEN as well as LOC and TEMP (cf. also Andrade Peres and Móia 1995: 269–317; Arim, Ramilo, and Freitas 2005). This construction is attested in several other languages: French (Gapany 2004: 190–191), German dialects (van Riemsdijk 1989), Greek (Nikolaos Mytilinaios, p.c.), Italian (Alfonzetti 2002: Ch. 2) and Spanish (Westedt 1997: 59–70). Again, it must be pointed out that the languages in the sample behave differently in terms of their acceptance of relative particles and combinations of relative particles with resumptive elements in the standard variety. This issue will be discussed in sections 5.1.2 and 5.2.1. (4.124) Ele era um homem [...] [que estava sempre POR he was DET man RPAR was always bem humorado.] well tempered ‘He was a man who was always good-tempered.’ (Albertino Moreira, from questionnaire) (4.125) É o livro [que eu terminei de POR is DET book RPAR I finished to ler semana passada.] read week last ‘It is the book that I finished reading last week.’ (Albertino Moreira, from questionnaire) (4.126) A POR DET um

criança [que você deu o dinheiro] é child RPAR you gave DET money is cigano. DET gipsy ‘The child you gave the money to is a gipsy.’ (Albertino Moreira, from questionnaire)

(4.127) Eles produzem música [que o povo gosta.] POR they produce music RPAR DET people like ‘They make music that people like.’ (Gärtner 1998: 219)

200 Typological issues (4.128) O cara [que o carro foi roubado] POR DET chap RPAR DET car was stolen semana passada. week last ‘The chap whose car was stolen last week.’ (Albertino Moreira, from questionnaire) (4.129) A rua [que eu moro.] POR DET street RPAR I live ‘The street where I live.’ (Perini 2002: 513) (4.130) A noite [que nós andamos em redor POR DET night RPAR we went around desse lago] estava chovendo! of.this lake it.was raining ‘The night that we walked around this lake it was raining!’ (Albertino Moreira, from questionnaire) In (4.124) and (4.125) que was glossed as ‘RPAR’. As mentioned in 4.2.1, in these cases it is not possible to establish whether we have to do with a (non-G-N-inflecting) relative pronoun or a relative particle, because they are formally identical. In (4.131) to (4.136) the combination of the Czech relative particle co with a resumptive element serves to relativize respectively SU, DO, IO, OBL, GEN and LOC. As reported in Table 3.7, this strategy is attested in 31 of the 36 languages of the sample. (4.131) To je ten chlap, [co ani von CZE that is that guy RPAR not.even he nevěděl co a jak.] not.knew what and how ‘That is the guy who didn’t know himself what he’d better do.’ (Lešnerová and Oliva 2003: 243) (4.132) To je ten chlap, [co ho viděli v CZE that is that guy RPAR CL.ACC.M.3SG saw in tramvaji.] tramway ‘This is the guy who they saw on the tram.’ (Toman 1998: 310)

Relative element 201

(4.133) To je ten chlap, [co mu každej CZE that is that guy RPAR CL.DAT.M.3SG everyone pomáhá.] helps ‘This is the guy who everyone helps.’ (Lešnerová and Oliva 2003: 243) (4.134) To je ten člověk, [co jsem ti o CZE that is that man RPAR I.am to.you about něm mluvil.] him spoken ‘This is the man I told you about.’ (Gołąb and Friedman 1972: 37) (4.135) To je ten pán [co jeho pes CZE that is that gentleman RPAR his dog pokousal nedávno Vaška.] bit recently Vašek ‘This is the gentleman whose dog bit Vašek not too long ago.’ (Beličová and Sedláček 1990: 185) (4.136) ten dům, [co sem tam bydlel] CZE that house RPAR I.am there lived ‘The house where I lived.’ (Sgall et al. 1992: 138) As (4.137) shows, co can be used also to relativize TEMP. Still, in this case it occurs with no resumptive element. The alternation of strategies introduced by a relative particle and by the combination of a relative particle with a resumptive element will be at issue in 4.3.1 and 4.3.3. (4.137) To léto, [co bylo tolik CZE that summer RPAR was so.much komárů.] mosquitoes.of ‘That summer in which there were so many mosquitoes.’ (Toman 1998: 312) Finally, in many languages relative particles can also be used as unspecific connection markers, like the specialized relative elements discussed for Stage 2 in 4.2.2. As mentioned above, in these cases a participant of the main SoA serves as a reference point for the SoA encoded in the RC, with-

202 Typological issues out further determining the syntactic link between the two clauses; the RC does not have to show any syntactic gap to be filled, either. See (4.138) and (4.139). (4.138) J’ FRE I [que

ai un bouquin en allemand have DET book in German j’ aimerais bien savoir ce RPAR I would.like well know.INF that [que ça veut dire.]] RPAR it means ‘I’ve got a book written in German and I would like to know what it’s about.’ (Gadet 1997: 118)

(4.139) the one down Taunton left his pottery to his son, [which all he ENG wanted was drink] (Herrmann 2005: 87) In both examples, the RC shows no syntactic gap: as discussed in 4.2.2, one may argue that the role of the relativized item in the RC is ‘topic’. In fact, the two sentences can be paraphrased as follows. (4.140) ‘I have a book written in German ← as far as it is concerned, I would like to know what it is about.’ (4.138) ‘The one down Taunton left his pottery to his son ← as far as he is concerned, all he wanted was drink.’ (4.139) Relative particles also prove to be quite useful in cases of multiple embedding, as shown in (4.141). (4.141) To je ta ženská, [co sem ti dal CZE that is that hussy RPAR I.am to.you given ten časopis, [co v něm byla její that magazine RPAR in it was her fotka.] RC2] RC1 photo ‘That is the hussy whose photo is in the magazine that I gave you.’ lit. ‘This is the hussy that I gave you the magazine that there’s her photo in it.’ (Toman 1998: 307)

Relative element 203

Here, the participant shared by the MC and the RC1 is ‘hussy’, whereas the participant shared by the RC1 and the RC2 is ‘magazine’: still, both relativized items are encoded inside the RC2 through a resumptive element: její encodes the role of the relativized item ‘hussy’ (GEN); v něm encodes the role of the relativized item ‘magazine’ (LOC). As these examples demonstrate, relative particles can be thought to increase syntactic flexibility: as they basically encode the link between the MC and the RC, the speaker can choose to encode the syntactic role of the relativized item in the RC or to leave it unexpressed. This is not possible with relative pronouns: they are clause-initial and encode the syntactic role, so that the speaker has to establish the syntactic role of the relativized item in the RC at the beginning of the clause (cf. also Fiorentino 2007: 276, 279). Adopting the terminology of incremental syntax (Auer 2000, 2005a), we may say that RCs introduced by relative particles are projective structures, since they inform the hearer that the following SoA will have a participant of the main SoA as its topic; the syntactic role of this participant inside the RC can be then either explicitly encoded or pragmatically inferred, according to the structure chosen by the speaker. A similar interpretation is proposed also in Sornicola (1981: 73) and Alfonzetti (2002: 118–119) for ITA che and Miller and Weinert (1998: 108–109) for ENG that. Additionally, these structures are recursive: they allow multiple embedding, as in (4.141) above, but they can also build chains of RCs introduced by the same element. This requires minimal planning effort and the speaker is free to encode the syntactic role of the relativized item in the RC or not. This is illustrated in the examples below. In the first two RCs of (4.142) OBL is relativized, in the third one SU: still, the speaker relies on the same relative element, i.e. the relative particle che, and encodes the syntactic role of the relativized item in none of the RCs. (4.142) Be’ ITA well [che

comunque era una persona anyway s/he.was DET person c’ intendevamo,] [che si rideva,] RPAR REFL we.got.along.well RPAR REFL laughed [che era ironica...] RPAR was humorous ‘Anyway, s/he was a person with whom I got along well, with whom you could laugh, who had a sense of humour…’ (woman, 55 y.o.)

204 Typological issues Fiorentino (1999: 169–170) makes similar considerations: the use of che as an unspecific connection marker in Italian would constitute a further resource of the linguistic system (“risorse ulteriori del sistema”, Fiorentino 1999: 169), since it allows speakers to link in a relative relation clauses that would probably be hardly ever connected if the standard strategy (relative pronoun) was used. Fiorentino provides the examples below. (4.143) l’ unica persona [che l’ incarico di ITA DET only person RPAR DET assignment of telefonar-gli mi emoziona] ring.up.INF-CL.DAT.3SG me excites ‘The only person that I’m excited about having the assignment of ringing him/her up.’ (Fiorentino 1999: 170) (4.144) ?l’ unica persona [l’ incarico di ITA DET only person DET assignment of telefonare alla quale mi emoziona] ring.up.INF to.RPRO.F.SG me excites ‘The only person that I’m excited about having the assignment of ringing him/her up.’ (Fiorentino 1999: 170) As Fiorentino herself remarks, (4.144) would hardly be uttered – or even written – as such. Instead, (4.143), despite being considered non-standard, is much more likely to be heard. Here, too, the role of the relativized item in the RC can be seen as topic: ‘The only person that, as far as she is concerned, I’m excited about having the assignment of ringing her up.’ In (4.145), the particle da is used three times: in the first and in the third RC it is used alone, whereas in the second one it is combined with a resumptive pronominal PP. Although in the first two RCs OBL is relativized, the syntactic role of the relativized item is explicitly encoded only in the second RC. (4.145) Uvijek je BCS always is [da se

tako teško naći teme so difficult find.INF topics učenici zainteresuju,] znaš, RPAR REFL students get.interested you.know teme [da razgovaraš o njima na času] i topics RPAR you.speak about them in lesson and [da učenicima ne bude dosadno.] RPAR pupils.to not will.be boring

Relative element 205

‘It’s always so difficult to find topics that pupils are interested in, you know, topics you can cover during the lesson without boring the pupils.’ ‘It’s always so difficult to find topics in order to get pupils interested in them, you know, topic to cover during the lesson and not to bore the pupils.’ (Milena Marić, from questionnaire) In (4.146) below the syntactic role of the relativized item is explicitly expressed in neither of the RCs. This implies that the first RC can receive multiple interpretations as far as the syntactic position relativized is concerned: it could be both LOC (‘a film in which the characters are credible’) or GEN (‘a film whose characters are credible’); instead, in the second RC the role of the relativized item is ‘topic’ (‘a film that, as far as it is concerned / when you watch it, you think…’). (4.146) Исках да снимам филм дето героите да бъдат достоверни, BUL като в ежедневния живот, филм дето си казваш: «Ох, това ми се е случвало и на мен!» Iskach da snimam film [deto I.wanted COMP I.take movie RPAR geroite da bădat dostoverni, kato v characters.DET COMP will.be credible like in ežednevnija život,] film [deto si kazvaš]: everyday.DET life movie RPAR REFL you.tell “Och, tova mi se e slučvalo i na oh this to.me REFL is happened also to men.” me ‘I wanted to do a movie that the characters would be credible, like in everyday life, a movie that you think: “Oh, that happened to me, too!”.’ (Maria Manova, from questionnaire) Finally, in the two RCs in (4.147) SU and GEN are relativized: still, the syntactic role of the relativized item is encoded only in the second RC.

206 Typological issues (4.147) Spotkała POL she.met [co to

mię ta pani, [co tam mieszka,] me that lady RPAR there lives jej mąż umarł.] RPAR PAR her husband died ‘I met that woman who lives over there, whose husband died.’ (dialect, Urbańczyk 1939: 60)

What all these examples have in common is the use of the same relative particle as a means for speech progression: speakers can build chains of RCs relying on the same relativization strategy (same word order, same relative element) to relativize different syntactic positions. Opposite examples are also attested, i.e. examples where a speaker uses different relative elements to encode the same syntactic position: still, these occurrences are rarer. Two examples are reported below. (4.148) Firmaet søger en en person, [der er DAN company.DET looks in DET person SRE is velkvalificeret,] og [som kan klare well-qualified and RPAR can cope.with arbejdspresset.] job.pressure.DET ‘The company is looking for a person who is well-qualified and who can cope with the pressure of the job.’ (Allan, Holmes, and Lundskær-Nielsen 1995: 203) (4.149) А буровые и бурят, и тут сваи до грунта, который там и что RUS уже вечный он, пробивают и ставят. A burovye i burjat, i tut svai do and drills PAR drill and there piles to grunta, [kotoryj tam] i [čto uže ground RPRO.NOM.M.SG there and RPAR already večnyj on,] probivajut i stavjat. firm he excavate and set ‘And the drills drill and excavate and set the piles to the ground that is there and is already firm.’ (Sergieva and Gerd 1998: 79) In both (4.148) and (4.149) SU is relativized: in (4.148) a specialized relative element is used in the first RC and a relative particle in the second RC; in (4.149) a relative pronoun is used in the first RC, in the second RC the speaker relies on the combination of a relative particle with a resumptive.

Relative element 207

4.2.4. Other phenomena related to relative elements The peculiarities described in the previous sections are cross-linguistically attested in several languages of the sample. Phenomena that are limited to one language or to a small group of languages will be dealt with in this section. The first phenomenon can be regarded as an instance of paradigm regularization. As described in 3.2.1, in a number of languages a zeromarker is attested. When the zero-marker occurs in the standard variety, it usually cannot relativize SU. It can be used alone to relativize DO, as in (4.150), or it can be combined with a stranded preposition to relativize lower positions of the AH, as in (4.151), where OBL is relativized. (4.150) The boy [Ø we met yesterday at the movies] is a good friend of ENG mine. (4.151) Här SWE here att

är en rolig historia [Ø ni kommer is DET funny history you come skratta åt.] COMP laugh at ‘This is a funny story you will be amused by.’ (Holmes and Hinchliffe 1994: 221)

In dialectal varieties, this ‘irregularity’ is no longer present: the zeromarker can also be used to relativize SU, as illustrated in the examples below. (4.152) I seen a chap at Broughton Moor, [Ø got his leg took off] ENG (Herrmann 2005: 69) (4.153) Vi ha en ganska häfti kyrkoherde [Ø spela SWE we have DET pretty cool vicar plays golf.] golf ‘We have a pretty cool vicar who plays golf.’ (Burträsk dialect, Karlsson and Sullivan 2002: 104)

208 Typological issues (4.154) Der ovre bor en mand [Ø hedder DAN there over lives DET man is.called Olsen.] Olsen ‘Over there lives a man called Olsen.’ (Platzack 2002: 78) Additional data on the use of the zero-marker to relativize SU can be found in Herrmann (2005) for English dialects, Platzack (2002: 78) for Danish dialects and Rendahl (2001: 145), Karlsson and Sullivan (2001) and Platzack (2002: 78) for Swedish dialects. Table 4.6. The paradigm of Dutch relative pronouns. Common gender Standard NOM ACC/DAT GEN PP Non-standard NOM ACC/DAT GEN PP

Neuter

Plural (both genders)

die die wiens/wier van wie op wie

dat dat waar…van

die die van wie

waar…op

op wie (non-NT) waar…op (NT)

wie wie wie z’n/d’r waar…van waar…op

wat wat (wat z’n) waar…van waar…op

wie wie wie d’r waar…van waar…op

The second phenomenon regards relativization in Dutch. According to Van der Wal (2002), the standard Dutch relative pronoun paradigm is going through a number of changes due to pressure from non-standard varieties. The paradigm of Dutch relative pronouns is shown in Table 4.6. The standard paradigm is reported on the left (following Donaldson 1997: 71) the non-standard on the right (following Van der Wal 2002). The non-standard paradigm shows the following properties: 1. as already mentioned in 4.2.2, the relative particle waar combines with a stranded adposition to relativize OBL not only when the referent is neuter or [–inanimate], but also when it is [+animate]; 2. the pronouns die and dat, which formally coincide with the demonstrative pronoun ‘that’, are replaced by the pronouns wie and wat, which formally coincide with the interrogative pronouns ‘who’

Relative element 209

and ‘what’ – see the two examples below, where they relativize SU and DO respectively; (4.155) dat is nou die jongen [wie gisteren DUT that is now that boy RPRO yesterday aan de deur kwam] at DET door came ‘So, this is the boy that stood at the door yesterday.’ (Van der Wal 2002: 34) (4.156) dat zijn nou ideeën [wie ik heel interessant DUT that are now ideas RPRO I very interesting vind] find ‘Those are ideas that I find very interesting.’ (Van der Wal 2002: 34) 3.

for the relativization of GEN, an analytic solution is preferred over a synthetic one: that is, either the GEN is assimilated to OBL – this would be the waar…van variant – or a combination of the relative pronoun with the possessives z’n ‘his/its’ and d’r ‘her/their’ is adopted. Observing the non-standard system, we may notice that it is composed only of w-elements: that is, the mixed demonstrative-interrogative (d-/w-) system is replaced by a solely interrogative one. Also in this case we may speak of paradigm regularization. According to the Syntactic atlas of the Dutch dialects, this happens in particular in the southern Dutch dialects; in the rest of the Dutch-speaking area wie is used for low positions of the AH, whereas SU and DO are most often relativized through die (Barbiers et al. 2005: Map 82a–b). Hoekstra (2002: 72) describes the relative elements used in Afrikaans: here, we may notice that the paradigm is even more reduced and corresponds to the neuter column of the Dutch non-standard paradigm in Table 4.6. Whereas the Dutch non-standard relative pronoun remains G-N-inflecting at least in the singular form, the Afrikaans relative pronoun is nonG-N-inflecting and in the process of becoming a relative particle. In standard Slovenian the use of the relative particle ki and of the relative pronoun kateri is regulated as shown in Table 4.7: ki is used for SU and is combined with resumptive clitic pronouns to relativize DO and IO, but cannot be used to relativize OBL. This may be related to the fact that

210 Typological issues the resumptive elements for DO and IO-relativization are clitic pronouns, whereas the resumptive to relativize OBL would be a ‘heavier’ pronominal PP. Instead, kateri is preferred to relativize OBL, but can also be used for DO and IO (Křížková 1970: 26; Gołąb and Friedman 1972: 41; Herrity 2001: 106). Table 4.7. The use of relative elements in Slovenian. ki kateri

SU standard

DO IO OBL GEN standard standard non-standard non-standard (+resumptive) (+resumptive) (+resumptive) (+resumptive) (standard) standard standard standard standard

Now, in non-standard varieties ki can be combined with a resumptive element also to relativize OBL and GEN, as illustrated in section 3.5.6. So, on a non-standard level both relative elements can be used to relativize all positions of the AH. This is confirmed if searching the Slovenian questionnaire of the paviatyp database, which is based on translations: all Slovenian sentences present a double translation, one with ki and one with kateri (Sansò 2005, URL: http://193.206.78.130:8080/jsp-examples/rel_cl/ jsp/executer.jsp?lang=slo&head=unmodified&rel=unmodified&role=unmo dified, last access: June 30th, 2011). Turning now to (standard) Macedonian, the relative pronoun koj(što) and the relative particle što behave pretty similarly to relative elements in Slovenian (Křížková 1970: 26; Gołąb and Friedman 1972: 43–44): koj(što) relativize preferably lower positions of the AH, što higher ones. Still, in non-standard varieties, both relative elements are used to relativize all positions of the AH. See for instance (4.157), where što is combined with a pronominal PP to relativize OBL. (4.157) Kako se vika čovekot [što se MAC how REFL calls man.DET RPAR REFL šetaše so nego včera]? s/he.walked with him yesterday ‘What’s the name of the man with whom s/he walked yesterday?’ (Gołąb and Friedman 1972: 44) As a general tendency, it can be remarked that when relative elements in the standard variety exhibit mixed paradigms, discrepancies or inconsistent use – as could be the case with the non-use of the zero-marker for SU or the mixed paradigm of the Dutch relative pronoun – in non-standard

Relative element 211

varieties these irregularities tend to be eliminated (in this regard see also section 5.1). This is also true for languages like Finnish and Hungarian which are usually thought to display little or no variation in the realm of relative elements. The use of relative elements in these two languages is illustrated in table 4.8. Table 4.8. The use of relative pronouns in Finnish and Hungarian. FIN with substantives not with substantives LOC with superlative

joka mikä missä mitä

HUN [+spec] [+count]

[+human] aki

[–human] amely

[–spec] [–count]

aki

ami

In Finnish the pronoun joka is used when the referent is a substantive, otherwise mikä is used, which is identical to the interrogative pronoun ‘what’ (Sulkala 1992: 44, 286). Still, in two cases mikä is used when the referent is a substantive: to relativize LOC – in this case, the locative form missä is used – and when the referent is a superlative, relying on the partitive form mitä, as in (4.158). (4.158) Hän on kaunein [mitä tiedän.] FIN s/he is the.most.beautiful RPRO.PART I.know ‘S/he is the most beautiful (person) that I know.’ (Sulkala 1992: 44) In colloquial varieties, the distinction [+/–substantive] is replaced by the distinction [+/–animate]: joka disappears, mikä is used for [–animate] referents and kuka, which is identical to the interrogative pronoun ‘who’, is used for [+animate] referents. Comparing (4.159) with (4.160), in both cases the referent is [+animate], but in (4.159) joka is used, whereas in (4.160) the speaker relies on kuka. (4.159) ja se ihmiset, [jotka näki sen FIN and DET people RPRO.NOM.PL saw DET onnettomuuden,] ei saanu sanaa accident not.3SG got word suustaa mouth.from.POSS3 ‘And the people who saw the accident could not utter a single word.’ (Matti Miestamo, from questionnaire)

212 Typological issues (4.160) ne on ihmisii [kehen sä et voi FIN they is people RPRO.ILL.PL you not.2SG can luottaa] trust.INF ‘These are people who you cannot trust.’ (Matti Miestamo, from questionnaire) In Hungarian there are even stricter semantic constraints on the use of relative elements: aki is used when the referent is [+human]; when the referent is [–human], amely(ik) is used if the referent is [+specific] and/or [+countable], otherwise ami is used. In colloquial varieties this subdistinction is obliterated and ami generally substitutes amely (Forgács 2001: 217– 218; Kenesei, Vago, and Fenyvesi 1998: 40–41), as in (4.161). (4.161) Elolvastam a könyvet, [amit HUN I.finished.to.read DET book RPRO.ACC.SG küldtél nekem.] you.sent to.me ‘I finished the book that you sent me.’ (Rounds 2001: 136) A final remark concerns combined elements. These elements were illustrated in detail in 3.2.2; an overview of the elements attested in the languages of the sample was provided in Table 3.7. Here, two considerations must be added. First, combined elements are mostly attested in non-standard varieties. Only the following also occur in standard varieties: – the combination of a relative pronoun with a relative particle in Macedonian (kogo+što in the example below); (4.162) Човекот, когошто го сретнав, не ми е познат. MAC Čovekot, [kogo-što go man.DET RPRO.ACC.M.SG-RPAR CL.ACC.M.3SG sretnav, ne mi e poznat.] I.met not to.me is known ‘I don’t know the man I met.’ (Usikova 2003: 270) – the combination of a relative pronoun with a resumptive clitic in Albanian, Bulgarian, Catalan, Greek, Macedonian, Rumanian and Spanish (căruia+i in the example below);

Relative element 213

(4.163) Conferinţa de sǎptǎmâna viitoare e sigur RUM conference.DET of week.DET next is surely un eveniment [cǎruia trebuie DET event RPRO.DAT.M.SG is.necessary sǎi dedicǎm toate eforturile.] COMP CL.DAT.3SG we.devote all efforts.DET ‘The conference we have next week is surely an event that we need to devote all our efforts to.’ (Mădălina Chitez, from questionnaire) – the combination of a relative pronoun with a stranded preposition in English (who+with in the example below); (4.164) Janet posted a letter for me last week for a friend [who I worked ENG with at Ipswich.] (Herrmann 2005: 47) – the combination of a relative particle with a resumptive element in Greek, Irish, Maltese (respectively pu+jia aftá, a+dó and li+magħa in the examples below); (4.169) Είναι πάντα πολύ δύσκολο να βρεις θέματα που να μιλάς για αυτά GRE στο μάθημα και να μη βαριούνται οι μαθητές. Íne pánta polí ðíscolo na vris is always very difficult COMP you.find θémata [pu na milás jia aftá sto topics RPAR COMP you.speak about them in.DET máθima ke na mi variúnte i lesson and COMP not get.bored DET maθités.] pupils ‘It’s always hard to find topics that you can cover during the lessons without boring the pupils.’ (Sofia Lampropoulou, from questionnaire) (4.166) an fear [a dtugaim an t-arán dó] IRI DET man RPAR I.give DET bread to.him ‘The man I give the bread to.’ (McGonagle 1988: 52)

214 Typological issues (4.167) Il-bieraħ iltqajt mattfajla [li ħja MAL yesterday I.met with.DET girl RPAR brother my ltaqa’ magħ- ha filparty.] met with her in.DET party ‘Yesterday I met the girl my brother met at the party.’ (Sansò 2005, MLS_027) – the combination of a relative particle with a stranded preposition in Danish, Durch, English, Norwegian and Swedish (som+i in the example below); (4.168) Sengen [som du sover i] er hundre NOR bed.DET SRE you sleep in is hundred år gammel. year old ‘The bed where you sleep is a hundred years old.’ (Strandskogen and Strandskogen 19862: 121) – the combination of the zero-marker with a stranded preposition in Danish, English, Norwegian and Swedish (Ø+åt in the example below). (4.169) Här SWE here att

är en rolig historia [Ø ni kommer is DET funny history you come skratta åt.] COMP laugh at ‘This is a funny story you will be amused by.’ (Holmes and Hinchliffe 1994: 221)

Possible reasons for the inclusion/exclusion of these combined relative elements in standard varieties of different European languages will be discussed in section 5.1.2 and 5.2. The second consideration is that triple combinations are attested in a small group of languages. They are illustrated in the examples below. (4.170) Der Mann, [der wo welcher GER det man RPRO.NOM.M.SG RPAR RPRO.NOM.M.SG geschtern hier war,] isch heut krank. yesterday here was is today sick ‘The man that was here yesterday is sick today.’ (woman, 50 y.o.)

Relative element 215

(4.171) Човекот, когошто го сретнав, не ми е познат. MAC Čovekot, [kogo-što go man.DET RPRO.ACC.M.SG-RPAR CL.ACC.M.3SG sretnav, ne mi e poznat.] I.met not to.me is known ‘I don’t know the man I met.’ (Usikova 2003: 270) In (4.170) we are presented with the combination ‘RPRO+RPAR+RPRO’, which is uniquely attested in Alemannic, a German dialect. Other German informants completely rejected this example, so further evidence is definitively needed. In (4.171) we have a combination attested only in Macedonian, ‘RPRO+RPAR+resumptive element’. A third triple combination, ‘RPAR+other particle+resumptive element’, is attested only in Polish dialects, where it is used to relativize lower positions of the AH. In (4.172) it relativizes IO, in (4.173) OBL and in (4.174) GEN. (4.172) O tym panu, [co to mu POL about that gentleman RPAR PAR CL.DAT.M.3SG lekarz zrobił operację.] doctor did operation ‘About the man that on whom the doctor operated.’ (dialect, Urbańczyk 1939: 60) (4.173) Zwolon, ten pan, [co to oni POL Zwolon, that gentleman RPAR PAR they rzuczali nań.] relied on.him ‘Zwolon, the man who they relied on.’ (dialect, Urbańczyk 1939: 31) (4.174) Spotkała POL she.met [co to

mię ta pani, [co tam mieszka,] me that lady RPAR there lives jej mąż umarł.] RPAR PAR her husband died ‘I met that woman who lives over there, whose husband died.’ (dialect, Urbańczyk 1939: 60)

In French, a triple combination ‘SRE+RPAR+resumptive element’ is attested, as shown in (4.175)

216 Typological issues (4.175) Une poussette [où qu’ ya plein d’ FRE DET pushchair SRE RPAR there.is a.lot of habits dedans.] clothes inside ‘A pushchair in which there are a lot of clothes.’ (Gapany 2004: 188) So, although triple combinations are quite rare, they are not excluded a priori: as the examples above demonstrate, all simple relative elements as well as particles and resumptive elements can appear in triple combinations. 4.3. Syntactic positions relativized In section 4.2 relative constructions attested in non-standard varieties of European languages have been analysed with an emphasis on relative elements; the syntactic positions which could be relativized by means of these elements were considered only cursorily. Now the same constructions will be scrutinized once more, paying special attention to the encoding of the syntactic role of the relativized item inside the RC. As mentioned in 3.3, the AH will serve as a reference frame to test how non-standard strategies relativize different syntactic positions. The starting point is constituted by Keenan and Comrie’s (1977: 92) statement that higher positions on the AH are preferably relativized through [–case] relativization strategies; instead, the higher we move down the AH, the higher the probability that a [+case] strategy is used. Basing on this Bernini (1989: 86–88) formulates a slightly different proposal, which he claims to be valid for Italian and Romance languages. It is reported in Figure 4.1. SU

DO

IO

OBL/GEN

[–case] [+case] (–Rel ?) Figure 4.1. The use of [+/–case] strategies for the relativization of the positions of the AH (based on Bernini 1989: 88).

Syntactic positions relativized

217

According to Figure 4.1, higher positions on the AH are mostly relativized through a [–case] strategy, whereas lower positions prefer a [+case] one. Bernini also points out that for two positions, OBL and GEN, the RC may be introduced through elements that are not strictly relative. Here, Bernini refers in particular to ITA che, which can be used as a general connector to introduce most different kinds of subordinate clauses (Bernini 1989: 88). Additionally, DO appears to have a particular status, being positioned at the intersection point between [+case] and [–case] strategies. That is, it can be equally relativized through both [+case] and [–case] strategies (cf. also Ariel 1999: 250–254). It may be recalled here that in this work Cristofaro and Giacalone Ramat’s (2007) definition of [+/–case] strategies is adopted: [+case] strategies are those in which the syntactic role of the relativized item in the RC is explicitly encoded, [–case] strategies are those in which it is not explicitly encoded (cf. section 3.3). Now, the aim of this section is to verify whether non-standard relative constructions in European languages pattern with the tendencies reported in Figure 4.1. We will examine how the syntactic role of the relativized item is encoded in the RC. In particular, we will concentrate on five crosslinguistically significant phenomena: 1. the syntactic role of the relativized item is not encoded through a clause-initial element (decumulation); 2. the syntactic role of the relativized item is encoded by means of two different morphosyntactic elements in the same RC (double encoding); 3. the syntactic role of the relativized item is not encoded at all in the RC (no encoding); 4. the syntactic role of the relativized item is not encoded in the RC only if the relativized item has the same syntactic function (or morphosyntactic form) in the RC and in the MC (equi type, equi-case or case matching);15 5. non-use of the RC. These phenomena will be discussed separately in the next subsections. It must be pointed out a priori that the description of these syntactic phenomena cannot go beyond a bird’s eye view of cross-linguistically most significant tendencies. In fact, the languages in the sample often display great language-internal variation as far as the syntactic positions that each strategy can relativize are concerned – a variation that, unfortunately, cannot be fully grasped in a cross-linguistic study. In this sense, more language-specific studies would be needed: they would provide the basis for a more solid cross-linguistic comparison.

218 Typological issues 4.3.1. Decumulation In the languages of the sample, the syntactic role of the relativized item in the RC can be encoded through the same clause-initial morphosyntactic element that also encodes the link between the MC and the RC: this is the case with simple relative elements like relative pronouns and specialized relative elements (see Table 3.5). In other cases, a simple element that does not encode the syntactic positions relativized (i.e. a relative particle or the zero-marker) combines with a resumptive element, which encodes at least the syntactic position relativized. This phenomenon is usually labelled ‘decumulation’: the four parameters encoded by the relative element are not conveyed through a simple relative element, but through a combined one – that is, they are ‘decumulated’, i.e. distributed on more morphosyntactic units.16 We may speak of decumulation in two cases: 1. when a relative particle combines with a resumptive element, the relative particle only encodes the link between the MC and the RC, whereas the other parameters (syntactic role and possibly gender and number agreement) are encoded through the resumptive element; 2. when the zero-marker combines with a resumptive element, the link between the MC and the RC is not encoded; the resumptive element encodes the syntactic role and, in some cases, gender and number agreement. As shown in Table 3.7, these two combinations are not equally widespread in the languages of the sample: in particular, the combination ‘RPAR+resumptive element’ is attested in 31 languages, the combination ‘zero-marker+resumptive element’ only in 5. In this sense, languages seem to prefer relative elements that explicitly encode the link between the MC and the RC – cf. also section 4.4. The great diffusion of the former strategy in (Indo-)European languages had been already noticed by Cuendet (1939) and Tesnière (1959: 570–573). Both kinds of decumulation mentioned above are exemplified below: in (4.176) the particle qi encodes the link between the MC and the RC; the position relativized is encoded through the pronominal PP për até. The form qi is the Geg dialectal variant; standard ALB që derives from the Tosk variant (Friedman 20062: 1876). Instead, in (4.177) the only parameter encoded is the syntactic positions relativized, expressed through the stranded preposition på.

Syntactic positions relativized

219

(4.176) Njû heqimi [qi për até folëm.] ALB there doctor.DET RPAR about him we.spoke ‘That is the doctor who we were speaking about.’ (Lambertz 1959: 173) (4.177) Grasset [Ø vi sad på] var endnu vådt. DAN grass.DET we sat on was still wet ‘The grass we were sitting on was still wet.’ (Allan, Holmes, and Lundskær-Nielsen 1995: 204) Traditionally, only examples like (4.176) are referred to as cases of decumulation: all of the four parameters are encoded, but they are distributed onto different morphosyntactic units. Instead, in (4.177) only one parameter is encoded. In this work, all RCs in which the syntactic role is encoded through a morphosyntactic unit distinct from the unit encoding the link between the MC and the RC will be considered as instances of decumulation. This definition includes examples both like (4.176) and (4.177). It implies that the morphosyntactic unit encoding the link between MC and RC can also be silent, as is the case with the zero-marker. According to the predictions reported in Figure 4.1, one should find cases of decumulation when lower positions of the AH are relativized, since they are thought to be more difficult to process (Keenan and Comrie 1977: 92): that is, one may expect the segment DO–GEN to be relativized through a combined relative element – a relative particle or the zero-marker plus a resumptive element – whereas a relative particle or the zero-marker alone should be preferred to relativize SU and DO. (As discussed in 3.5.6, in this work the uses of a relative particle or the zero-marker with and without a resumptive are considered as two different relativization strategies.) As a matter of fact, there are several cases in which this prediction is not fulfilled. In a group of languages, a resumptive element is inserted also to relativize SU, as shown in the examples below. (4.178) To je ten chlap, [co ani von CZE that is that guy RPAR not.even he nevěděl co a jak.] not.knew what and how ‘That is the guy who didn’t know himself what he’d better do.’ (Lešnerová and Oliva 2003: 243)

220 Typological issues (4.179) C’era anche una hölderlinista americana ITA there.was also DET Hölderlin.specialist American [che lei è quella [che ha RPAR she is that RPRO/RPAR has parlato sabato.]] spoken Saturday ‘There was also an American Hölderlin specialist, who was the one that presented on Saturday.’ (man, 30 y.o.) (4.180) Aqui nesta aldeia tem um caboclo [que ele é POR here in.this village there.is DET caboclo RPAR he is ideal.] ideal ‘In this village there is a caboclo that is ideal.’ (Gärtner 1998: 219) (4.181) ten, [co on potem uciekł ze szkoły] POL that RPAR he then escaped from school ‘The one that escaped from school then.’ (Topolińska 1984: 345) (4.182) Знаці чашніка маладзенькага, што ён жонкі не мае? BLR Znaci čašnika maladzen’kaha, [što ёn žonki ne you.know butler young RPAR he wife not mae]? has ‘Do you know the young butler who has no wife?’ (Atrachovič 1966: 609) (4.183) Well, it’s what they fed, you used to put it [i.e. treacle] on hay that ENG it was mouldy, you know, bad hay. (Herrmann 2005: 71) We can distinguish between two different uses, an emphatic and a nonemphatic one. In (4.178) and (4.179) the focus is on the resumptive pronoun: this is confirmed by the presence of the focalizing particle ani ‘not even’ in the Czech example and by the high pitch on lei in the Italian example. Instead, in (4.180) to (4.183) the resumptive element has no emphatic function. Its presence may be due to the necessity of explicitly encoding the syntactic position relativized. As predicted in Figure 4.1, in the majority of the languages in the sample DO has an intermediate position: it can be relativized either by means of a relative particle alone or by means of a relative particle combined with

Syntactic positions relativized

221

a resumptive element. Sometimes grammars try to set rules in order to discipline this use. For instance, in Bosnian/Croatian/Serbian grammars the distinction [+/–animate] is advocated to distinguish between cases in which a resumptive has to be set (i.e. with [+animate] referents) and cases in which it must be left out (i.e. with [–animate] referents). However, exceptions can be found (Gołąb and Friedman 1972: 42; Kordić 1999: 145–146): in (4.184) a resumptive element is used with a [–animate] referent, while in (4.185) no resumptive is used with a [+animate] one. (4.184) to je grad, [što sam ga posetio BCS that is town RPAR I.am CL.ACC.M.3SG visited prošle godine.] last year ‘That is the town that I visited last year.’ (Gołąb and Friedman 1972: 42) (4.185) to je najbolji konj [što sam video] BCS that is biggest horse RPAR I.am seen ‘That’s the biggest horse I’ve ever seen.’ (Gołąb and Friedman 1972: 42) Other scholars try to formulate general tendencies on the basis of corpus analyses. For example, Berruto (1987: 123–129) investigates RCs in Italian: he finds that resumptive clitics are more often used to relativize the segment of the AH between DO and GEN when the relative construction is non-restrictive and when the referent is [+animate]. However, exceptions can be found also in this case: in (4.186) the referent is [+animate] and the construction is non-restrictive, but no clitic is used; in (4.187) a resumptive clitic is inserted even if the referent is [–animate] and the construction is restrictive. (4.186) Come mai alle capre, [che piace il sale,] ITA how.come to.DET goats RPAR likes DET salt non viene mai la pressione alta? not comes never DET pressure high ‘How come goats, which love salt, never have high blood pressure?’ (TV program, woman, 40 y.o.)

222 Typological issues (4.187) Gli scatoloni [che non li hai ITA DET boxes RPAR not CL.ACC.M.3PL have ancora aperti] li prendi e li yet opened them you.take and them porti al mercatino delle pulci. you.bring to.DET little.market of.DET fleas ‘The boxes you haven’t opened yet, you can take them directly to the flea market.’ (woman, 45 y.o.) To summarize, the prediction made in Figure 4.1 does not seem to apply to decumulation: other factors beside the relativized position probably play a role in the choice of the strategy (cf. for instance Ariel 1999; Dasinger and Toupin 1994: 475–481). In some – extremely rare – cases speakers may encode the link between the RC and the MC and the syntactic role through a simple or combined relative element even inside the same sentence, as in (4.188): in the first RC kotoryj encodes all of the four parameters; in the second RC čto encodes the link between MC and RC, whereas on encodes the syntactic role – SU – as well as gender and number agreement. (4.188) А буровые и бурят, и тут сваи до грунта, который там и что RUS уже вечный он, пробивают и ставят. A burovye i burjat, i tut svai do and drills PAR drill and there piles to grunta, [kotoryj tam i [čto uže ground RPRO.NOM.M.SG there and RPAR already večnyj on,] probivajut i stavjat. firm he excavate and set ‘And the drills drill and excavate and set the piles to the ground that is there and is already firm.’ (Sergieva and Gerd 1998: 79) 4.3.2. Double encoding In some of the languages in the sample, the syntactic role of the relativized item in the RC can be encoded twice: this is the case with combined elements composed either of a relative pronoun and a resumptive element or of a specialized relative element and a resumptive element. As indicated in Table 3.7, the former combination is attested in 12 languages, the latter in 8 languages. These figures are based on the available data: both combinations may be attested in more languages; further research is needed on this

Syntactic positions relativized

223

point. Examples of double encoding are reported below: in (4.189) the syntactic role IO is expressed both through the relative pronoun tis opías and through the resumptive clitic tis; in (4.190), the syntactic position relativized, LOC, is encoded both through the specialized relative element donde and through the resumptive adverb allí. (4.189) Kséro tin kopéla [tis opías tis GRE I.know DET girl RPRO.GEN/DAT.F.SG CL.GEN/DAT.F.3SG éðoses leftá.] you.gave money ‘I know the girl you gave money.’ (Joseph and PhilippakiWarburton 1987: 27) (4.190) Pues nada, pasamos al claustro, [donde SPA then nothing we.passed to.DET cloister SRE allí se dio un lunch, o como se there REFL gave DET lunch or how REFL llame.] calls ‘And then we went over to the cloister, where they gave ‘lunch’, or whatever it is called.’ (Westedt 1997: 69) As in the case of decumulation, this phenomenon is not limited to particular syntactic positions, but is attested throughout the AH, as the following examples illustrate. In (4.191) and (4.192) SU is relativized by means of the combination of a relative pronoun with a resumptive pronoun. (4.191) But anyway there was so many people and one chap [who he he ENG was, as a matter of fact, he was organizer with Communist Party,] for whom I’ve got the very greatest respect, the very greatest respect (Herrmann 2005: 47) (4.192) şi el so dus la ceilalţi / la RUM and he REFL has gone to others to C.V. / la B. / [care ei ... bea vin la B.] C.V. to B. RPRO they drank wine at B. ‘So he went to the others, to C.V, to B., who... were drinking wine at B’s.’ (Vulpe 1980: 129)

224 Typological issues In the example above, the use of a resumptive pronoun could be viewed as further evidence for the incapacity of the relative pronoun care to mark the position relativized (cf. 4.2.1 above). So, care could also be glossed as ‘RPAR’. In (4.193) to (4.195), DO and IO are relativized. In all three cases the relative pronoun combines with a clitic pronoun. (4.193) Lexova një libër [të cilin e ALB I.read DET book RPRO.ACC.F.SG CL.ACC.F.3SG mora në bibliotekë.] I.got in library ‘I read a book that I got from the library.’ (Mišeska-Tomić 2006: 314) (4.194) weź mój kij, [który ja POL take my stick RPRO.ACC.M.SG I go nosze ze sobą] CL.ACC.M.3SG I.take with REFL ‘Give me my stick, which I usually take with me.’ (dialect, Urbańczyk 19622: 57) (4.195) uns marrecs [a qui la Maria CAT some kids to RPRO DET Maria els dóna galetes] CL.DAT.M.3PL gives cookies ‘Some kids whom Maria gives cookies.’ (Hualde 1992: 58) In (4.196) to (4.198) OBL is relativized through different relative elements. In (4.196) a relative pronoun is combined with a resumptive pronominal PP, in (4.197) a relative pronoun is combined with a clitic pronominal adverb and in (4.198) a specialized relative element is combined with a clitic pronominal adverb. (4.196) am crezut că ia ... lucruri de-a lu ... RUM I.have believed COMP he.took things from DET.DAT femeia / [cu care- o trăit cu ea] // woman.DET with RPRO he.has lived with her ‘I thought he took... things from... the woman he has lived with.’ (Vulpe 1980: 129)

Syntactic positions relativized

225

(4.197) Ci sono cose del passato [alle quali ci ITA there.are things of.DET past to.RPRO.F.PL CL.ADV tengo moltissimo.] I.care.about a.lot ‘There are some things in the past which I really care about.’ (TV program, man, 60 y.o.) (4.198) Ça c’est effectivement un problème [dont on FRE this is indeed DET problem SRE one en a discuté hier.] CL.ADV has discussed yesterday ‘In fact, this is a problem which was discussed yesterday.’ (Gapany 2004: 185) In (4.199) GEN is relativized through the combination of a specialized relative element with a possessive. (4.199) Une réalité [dont sa compréhension n’ est pas FRE DET reality SRE her comprehension not is not à la portée de tout le monde.] suitable of everybody ‘A reality that cannot be comprehended by everybody.’ (Gapany 2004: 186) Finally, in (4.200) and (4.201) LOC is relativized by means of the combination of a specialized relative element with a locative adverb and a clitic pronominal adverb respectively. (4.200) Przyjachał na jednę zieloną łąkę, [gdzie POL he.came on one green meadow SRE tam nigdy drógi nie było.] there never street not was ‘He arrived at a green meadow where there had never been any streets.’ (dialect, Urbańczyk 1939: 30) (4.201) la casa [on els teus cosins hi viuen] CAT DET house SRE DET your cousins CL.ADV live ‘The house where your cousins live.’ (Hualde 1992: 59)

226 Typological issues As the examples above show, all types of resumptive elements listed in Table 3.8 occur in these combinations. Sometimes the two units constituting the relative element somehow integrate each other. In the example below, the specialized relative element encodes the syntactic role LOC and the resumptive adverb specifies the nature of the spatial relation (‘under’). (4.202) Ils voulaient le torturer sur une FRE they wanted him torture.INF on DET grille [où dessous il y avait des flammes.] grill SRE under there.was DET flames ‘They wanted to torture him on a grill under which there was fire.’ (Gapany 2004: 188) If we are to look for explanations for the existence of these constructions, we may produce the following ones. They do not cover all attested cases, but can be applied to a good number of them. To start with, in Balkan languages, Catalan, Portuguese and Spanish clitic doubling is found also in MCs. That is, verbal arguments encoded as NPs or PPs – usually direct and indirect objects – are doubled through clitics pronouns, as in the Macedonian examples below, where clitics mu and go double the PP na edno dete ‘to a child’ and the NP pismoto ‘the letter’. (4.203) Jana MAC Jana edno

mu go dade pismoto na CL.DAT.NT.3SG CL.ACC.NT.3SG gave letter.DET to dete. DET child ‘Jana gave the letter to a child (that I know).’ (Mišeska-Tomić 2006: 255)

Clitic doubling – both in MCs and in RCs – is generally connected with semantic traits of the object, like [+/–specific] and [+/–definite]. Crosslinguistic generalization cannot easily be made, though, as the above mentioned languages show different constraints on clitic doubling and variation sometimes seems to be inherently idiolectal (see for instance the discussion on Balkan languages in Mišeska-Tomić 2006: Ch. 4). For our purposes, we can maintain that the presence of a clitic doubling some features of the relative pronoun in Albanian, Bulgarian, Catalan, Greek, Macedonian, Rumanian and Spanish is most probably related to constraints governing clitic doubling in these languages. Clitic doubling in MCs is attested also

Syntactic positions relativized

227

in Italian and French: still, it is mainly restricted to instances of constituent extraposition. Additionally, clitic doubling in Italian and French is not codified in the standard variety, whereas in the language mentioned above it is. This fact will be taken up again in 5.1.3. Secondly, in other languages the combination ‘pronominal clitic+verb’ may be regarded as having gained a high degree of idiomaticity: that is, although the syntactic role is encoded through the relative pronoun, a resumptive element is inserted in the RC because the group ‘pronominal clitic+verb’ is felt as a single unity in terms of meaning. This is for instance the case with the Italian example (4.197): the combination ‘pronominal clitic+verb’ ci tengo has lexicalized, acquiring the meaning ‘I care about’, so it is no longer regarded as the syntactic combination of the two elements ci ‘to it’ and tengo ‘I hold’. Hence, the pronominal adverb ci is no longer felt to occupy the argumental position IO, which is therefore encoded by means of the relative pronoun in the dative case: cf. also Alfonzetti (2002: 60). Thirdly, in French, the resumptive may be regarded as disambiguating the role of the relativized item in the RC, which is felt to be not sufficiently determined by the specialized relative element: as mentioned in 3.2.1, FRE dont can be used to relativize both GEN and OBL, namely when it is introduced by the preposition de. So, the clitic may be inserted to distinguish between the two positions: a pronominal adverb is inserted when OBL is relativized, as in (4.198), and a possessive is inserted when GEN is relativized, as in (4.199). Additionally, we need to bear in mind that FRE dont is also used as a relative particle: in this case, it doesn’t encode the syntactic position relativized (cf. 4.2.2). If we interpret dont as a relative particle, (4.198) and (4.199) could also be seen as instances of decumulation. Similar considerations could also apply to (4.200) and (4.201), although POL gdzie and CAT on are used only for LOC. In general, one may assume that a resumptive element is inserted when the speaker feels that the syntactic role of the relativized item is not encoded with a sufficient degree of explicitness through the clause-initial relative element. One the other hand, it may also be argued that a speaker uses a relative pronoun with a resumptive element instead of a relative particle with a resumptive element because s/he feels that the latter is more colloquial than the relative pronoun. Still, in choosing the strategy of the relative pronoun, s/he simply replaces the particle with the pronoun, but does not eliminate the resumptive, which s/he does not consider a colloquialism (cf. the data reported in Alfonzetti 2002: 133). This hypothesis should be further tested, though.

228 Typological issues Finally, it needs to be pointed out that the acceptance of double encoding of the syntactic role of the relativized item in the RC varies from language to language The sociolinguistic status of this construction will be discussed in detail in section 5.1.3. 4.3.3. No encoding This phenomenon can be considered as somewhat complementary to those described in the previous sections: whereas in the case of decumulation the syntactic role is encoded once and in the case of double encoding it is expressed twice, in this case the syntactic role is not expressed at all. That is, simple and combined relative elements are used which do not encode the syntactic role of the relativized item in the RC: relative particles, the zeromarker and the combinations of these two elements with non-relative particles. Following the prediction made in Figure 4.1, we may expect these strategies to be mostly used when SU and DO are relativized. Instead, in non-standard varieties of European languages no encoding of the syntactic position relativized is attested for any of the other positions of the AH. The Portuguese examples (4.124) to (4.130) above illustrate this: the relative particle que, which only conveys the link between the MC and the RC, is used in all sentences. So, the task of recovering the syntactic position relativized is left to the hearer. Semantic and pragmatic (i.e. co-textual and contextual) factors play a central role in retrieving the syntactic role if it is not explicitly expressed (cf. Alfonzetti 2002: 94–95). This issue is discussed in detail in Haig (1998: 128–146) and Fiorentino (1999: 36–41). Haig examined case recoverability17 in Turkish: in the Turkish prenominal strategy the syntactic role of the relativized item is usually not explicitly encoded unless it is to be emphasized, as in and (3.78), or GEN is relativized, as in (3.77). Fiorentino analyzes case recoverability in Italian non-standard RCs introduced by the relative particle che alone, where the syntactic position relativized is not encoded either. Although the two scholars conduct language-specific investigations, they reach similar conclusions, which may be tested here for cross-linguistically validity. Haig and Fiorentino identify a number of parameters that help recover the syntactic role of the relativized item in the RC when it is not explicitly encoded. In particular, Haig examines four of them: 1. argument filling, 2. preferred interpretation, 3. semantic predict-

Syntactic positions relativized

229

ability and 4. contextual information. Each parameter is considered separately below. 1. ARGUMENT FILLING. If in the RC all of the verbal arguments are filled but one, this one is likely to be the syntactic position relativized. Haig exemplifies this through the following example. (4.204) [bir şey sorduğumuz] komşu TUR DET thing ask.NMLZ.POSS.1PL neighbour ‘The neighbour that we have asked something.’ (Haig 1998: 141) In Turkish, the verb sormak ‘to ask’ can have two arguments beside the SU-argument, DO and IO. Since in the example above the SU-slot is morphologically encoded on the verb and the DO-slot is filled, the role of the relativized item in the RC is necessarily IO. If neither argumental slot is filled, hearers will rely on the next parameter, ‘preferred interpretation’. Similar examples can be quoted from other languages. (4.205) Come mai alle capre, [che piace il sale,] ITA how.come to.DET goats RPAR likes DET salt non viene mai la pressione alta? not comes never DET pressure high ‘How come goats, which love salt, never have high blood pressure?’ (TV program, woman, 40 y.o.) (4.206) Kséro tin kopéla [pú eksartáse (apó aftín).] GRE I.know DET girl RPAR you.depend on her ‘I know the girl you depend on.’ (Joseph and Philippaki-Warburton 1987: 166) In (4.205) the IO-slot of the verb piacere ‘to like’ is not filled, so the hearer can assume that the syntactic position relativized is IO. In (4.206), the resumptive apó aftín can be left out as it is the only object of the verb eksartóme ‘to depend’ and the speaker can easily recover the position relativized, i.e. OBL. Arim, Ramilo, and Freitas (2005) state that in Portuguese conceptionally oral texts there is a strong tendency not to encode the syntactic position relativized when verbs like gostar ‘to like’, precisar ‘need’ and falar ‘to speak’ occur in the RC. This could be explained through argument filling: gostar and precisar have a single object argument (introduced by the preposition de), which, consequently, is the only one that can be relativ-

230 Typological issues ized. Instead, falar has two object arguments, the addressee and the topic of the conversation. Here, one relies on the factor ‘preferred interpretation’: when neither of the object arguments is filled, hearers tend to interpret the sentence so that the relativized item is the the topic of the conversation. 2. PREFERRED INTERPRETATION. If in the RC more than one verbal argument is not filled, a syntactic position could be preferred to another depending either on the semantics of the verb in the RC or on the referent. Haig quotes a modified version of (4.206). (4.207) [sorduğumuz] komşu TUR ask.NMLZ.POSS.1PL neighbour ‘The neighbour about whom we have asked.’ (Haig 1998: 141) As Haig (1998: 140) states, if both non-subject arguments of the verb sormak are missing and no further contextual information is present, the hearer tends to interpret the DO as the position relativized. Haig provides an example of preferred interpretation with verbs of saying in Turkish (Haig 1998: 133–135). He also introduces a hierarchy for the preferred interpretation of the syntactic positions relativized, which – excluding SU – would read as “DO > others” (Haig 1998: 136).18 Still, as Haig himself states, the hierarchy is not immutable: in many cases, the semantics of the referent concurs with the semantics of the verb (Haig 1998: 139). (4.208) a. TUR b. c.

[beklediğimiz] arkadaş wait.NMLZ.POSS.1PL friend ‘The friend we’ve waited for.’ [beklediğimiz] oda wait.NMLZ.POSS.1PL room ‘The room in which we’ve waited.’ [beklediğimiz] tren wait.NMLZ.POSS.1PL train ‘The train we’ve waited for.’ (Haig 1998: 139–140)

As (4.208) shows, the preferred interpretation for an RC containing the verb beklemek ‘to wait’ changes according to the referent: if the referent is [+human], as in (4.208a), the preferred syntactic role is DO (beklemek is a transitive verb in Turkish); if the referent describes a place, as in (4.208b), the syntactic role would rather be interpreted as LOC. Still, when both

Syntactic positions relativized

231

interpretations are possible, one is always preferred to the other, as (4.208c) suggests: one can both wait for a train and wait in a train, but native speakers would rather choose the first interpretation. Haig specifies that if the position relativized was LOC, a resumptive element would be inserted, which would result in the example below. (4.209) [içinde beklediğimiz] tren TUR interior.POSS.3SG.in wait.NMLZ.POSS.1PL train ‘The train in which we waited.’ (Haig 1998: 140, modified) 3. SEMANTIC PREDICTABILITY. If in the RC all verbal arguments are filled, the syntactic position relativized could be a circumstantial. Here, too, the semantics of the referent plays a central role. Givón argues that the referent can specify through its lexical meaning which syntactic position is relativized: this happens in particular with “adverbial […] arguments [=circumstantials, A.M.], such as time, manner, reason, purpose and, occasionally, even place” (Givón 1990: 679 [emphasis in the original, A.M.]). This is shown in the following examples. In (4.210) and (4.211) the referents are two places, respectively ‘little market’ and ‘sauna’. As in the RC there is no argumental gap to be filled, the most likely syntactic position relativized is LOC. In (4.212) the referent is a substantive describing a time slot, ‘summer’. Consequently, one tends to interpret the syntactic position relativized as TEMP. Finally, in (4.213), the referent is a receptacle: so, the relativized position may be either OBL (‘with which’) or LOC (‘in/from which’). (4.210) ummercatino [kevvendono molto] ITA DET little.market RPAR they.sell a.lot ‘A little market where they sell very much.’ (Sornicola 1981: 69) (4.211) Наш дом у байны вы вцера купались. RUS Naš dom u bajny [Ø vy vcera our house at sauna you yesterday kupalis’.] had.a.bath ‘Our house (is) near the sauna where you had a bath yesterday.’ (dialect, Šapiro 1953: 58)

232 Typological issues (4.212) To léto, [co bylo tolik CZE that summer RPAR was so.much komárů.] mosquitoes.of ‘That summer in which there were so many mosquitoes.’ (Toman 1998: 312) (4.213) Am avut un ol în mînă [care beam RUM I.have had DET pitcher in hand RPAR I.drank apă.] water ‘I had a pitcher in my hand, from which I drank water.’ (Vulpe 1980: 127) 4. CONTEXTUAL INFORMATION. On the basis of their encyclopaedic knowledge and of their knowledge of the communicative situation, hearers can retrieve the role of the relativized item. This factor helps particularly when the other factors are not successfully applied or if their application yields nonsense. Haig (1998: 141) quotes the following example, for which the original translation is reported. (4.214) A aldın mı parayı? TUR B tamam canım aldım A peki sonra ne yaptın? B ne olacak, bir arkadaşima verdim A [verdiğin] arkadaş kim? B sen tanımazsın

‘did you get the money?’ ‘sure I got (it)’ ‘good – and what did you do (with it) then?’ ‘what was I supposed to do? – I gave (it) to one of my friends’ ‘who’s the friend [you gave (it) (to)]?’ ‘you wouldn’t know (him)’

In the RC [verdiğin] neither of the arguments – DO and IO – is filled. So, the hearer could rely on the preferred interpretation, which, according to the hierarchy “DO > others” (Haig 1998: 136) would be DO. Still, this yields the interpretation ‘the friend you gave’, which is nonsense. So, the hearer relies on the context: the topic of the conversation being para ‘money’, it may well be that the DO in the RC is occupied by para, although this is not expressed explicitly. This is not uncommon: as it emerges in (4.214), (pronominal) arguments may remain unexpressed also in MCs (this is hinted at by means of ‘(it)’ in the English translation). So,

Syntactic positions relativized

233

the latter reading proves to be better: the sentence Verdiğin arkadaş kim? is correctly interpreted as ‘Who is the friend [you gave it to]?’. In other cases, only hearers’ knowledge of the communicative situation helps to retrieve the syntactic role. (4.215) l’ uomo [che ho parlato] non è ITA DET man RPAR I.have spoken not is più tornato more come.back ‘The man I spoke with/about has never come back.’ (Fiorentino 1999: 38) If (4.215) is taken in isolation, one can by no means establish the role of the relativized item in the RC. Instead, if the sentence is inserted in a proper context, the hearer can rely on his/her knowledge of the situation and retrieve the role of the relativized item. To summarize, as the examples above demonstrate, syntax alone not always helps to retrieve the syntactic role of the relativized item in the RC. Semantic and pragmatic factors also play a central role in case recovery. This is true for both high and low positions of the AH. 4.3.4. Case matching Case matching is akin to the phenomenon described in the previous section: the syntactic role of the relativized item in the RC is not expressed. Still, whereas in the examples discussed in 4.3.3 the syntactic position was not encoded irrespective of the role of the relativized item in the MC and in the RC, in languages in which case matching is attested this is only possible under particular conditions, i.e. when the relativized item has the same syntactic role or the same case form in the MC and in the RC. This identity allows the speaker to recover the syntactic role of the relativized item in the RC from the MC. As the examples below demonstrate, case matching is often attested in dialects. The positions relativized are OBL in (4.216) to (4.219), IO in (4.220) and DO in (4.221).

234 Typological issues (4.216) Won jezo z tym awtom, [ako cora LSO he goes with DET car RPAR yesterday jo jeł.] he.is gone ‘He goes with the car he went with yesterday.’ (Janaš 1976: 187) (4.217) Amb els amics [que sortie abans,] CAT with DET friends RPAR I.went.out before vaig arribar a avorrirm’ hi I.arrived to get.bored.INF REFL CL.ADV molt. a.lot ‘I began to get very bored with the friends I used to go out with.’ (Wheeler, Yates, and Dols 1999: 538) (4.218) a. ITA

b.

(4.219) a. BAS

b.

El

ga fatto he.has made [che ti RPAR CL.NOM.2SG

CL.NOM.M.3SG

baruffa col quarrel with.DET ga fatto baruffa have made quarrel

fio boy ti.] you ‘He has quarrelled with the boy you (also) quarrelled with.’ *El ga conossuo el fio CL.NOM.M.3SG he.has known DET boy [che ti ga fatto baruffa ti.] RPAR CL.NOM.2SG have made quarrel you ‘He became acquainted with the boy you quarrelled with.’ (Venetian dialect, Fiorentino 1999: 30) [bizi naiz-en] lagunarekin joan living I.am-RPAR companion.DET.with gone naiz I.am ‘I went with the companion I’m living with.’ ?? [bizi naiz-en] laguna living I.am-RPAR companion.DET ‘The companion I’m living with.’ (Oyharçabal 1989: 69–70)

Syntactic positions relativized

(4.220) a. GER

b.

(4.221) a. MAL

b.

sie gem ’s dem Mo they give it DET.DAT.M.SG man [(dem) wo mir g’hoifa RPRO.DAT.M.SG RPAR we helped ‘They give it to the man that we helped.’ der Mantl [*(den) wo DET coat RPRO.ACC.M.SG RPAR kaffd hob] bought have ‘The coat I bought.’ (Bavarian dialect, Fleischer 2006: 230–231)

235

hom] have i I

Rajt ilqattus [li ttfal I.saw DET cat RPAR DET children xtraw il-bieraħ.] bought yesterday ‘I saw the cat that the children bought yesterday.’ Qattus [li ttfal jitimgħu-h cat RPAR DET children feed-CL.M.3SG kuljum.] every.day ‘A cat that the children feed every day.’ (Borg and Azzoppardi 1997: 35)

In the sources from which (4.218) to (4.221) are taken, the authors provide additional contrasting examples (the ‘b.’-examples above) in order to show that relativization through a relative particle would not be possible if the relativized item in the MC and in the RC did not have the same case form. In (4.218) and (4.219) case matching is possible only when the relativized item in the MC is a comitative, like in the RC. Additionally, Oyharçabal (1989: 70) points out that case matching is the only instance where the Basque prenominal strategy can be used to relativize non-argumental positions. In (4.220a) the relative pronoun dem can be left out because it has the same case form as the referent in the MC: instead, in the contrasting example the relative pronoun cannot be left out because the relativized item is SU in the MC and DO in the RC (Fleischer 2006: 230–231). Finally, in (4.221a) the resumptive -h is not inserted because the relativized item is DO both in the MC and in the RC.

236 Typological issues When some of these examples were checked with native speakers, there were quite different reactions: as regards (4.218), both sentences actually sound acceptable to a Venetian informant (Maria Busetto, p.c.), so that the ‘*’ sign set by Fiorentino may be questionable. The same can be said for (4.220): a Bavarian-speaking informant (Robert Nockermann, p.c.) stated that the relative pronoun can be left out also in the contrasting example. Finally, checking the validity of case matching in Maltese through the corresponding questionnaire in the database paviatyp (Sansò 2005, URL: http://193.206.78.130:8080/jsp-examples/rel_cl/jsp/executer.jsp?lang=mal&head=unmodified&rel=un modified &role=unmodified, last access: June 30th, 2011), it emerges that sometimes no resumptive element is set even if the relativized item has a different syntactic role in the MC and in the RC – e.g. SU and DO as in (4.222) – and that vice versa a resumptive element can be set even if the relativized item has the same syntactic role in the MC and in the RC, as in (4.223), where OBL is relativized. (4.222) Irraġel [li qed iffittex ħja] MAL DET man RPAR is.looking.for brother POSS.1SG huwa ħabib tiegħ-i. he friend of.me ‘The man my brother is looking for is a friend of mine.’ (Sansò 2005, MLS_026) (4.223) Il-bieraħ iltqajt mattfajla [li ħja MAL yesterday I.met with.DET girl RPAR brother my ltaqa’ magħ- ha filparty.] met with her in.DET party ‘Yesterday I met the girl my brother met at the party.’ (Sansò 2005, MLS_027) So, it seems quite difficult to establish whether case matching is really attested in a language or if the syntactic role of the relativized item in the RC is not encoded in general, irrespective of the syntactic role of the relativized element in the MC. In the database case matching seems to be attested with certainty in Basque, Catalan and Lower Sorbian. Still, further language-specific investigations on this issue are needed, for instance in the spirit of Salzmann and Seiler (2010), who offer an Optimality Theory based account of clitic resumption and matching effect in Swiss dialects.

Syntactic positions relativized 237

4.3.5. Avoiding the relative clause Bernini (1989: 88) claims that when lower positions of the AH are relativized the RC may be introduced by elements that are not exclusively relative. In Figure 4.1 above this is signalled through the symbol ‘–Rel’ under the positions OBL and GEN. Still, this also applies to higher positions of the AH: general connectors, like ALB që, GER wo or RUM de, can be used, at least in some varieties, as relative elements to relativize the whole AH and as connectors encoding most different relations (temporal, causal, etc.; see 3.5.1). What is perhaps more curious is that when a low position of the AH is to be relativized or when the nature of the link between MC and RC is not clearly identifiable, speakers may decide not to rely on a relative construction, but on an alternative one, which is usually an asyndetic coordinate construction. This point is certainly questionable: how can we know whether a speaker consciously decides to use an asyndetic coordinate construction instead of a relative one? A partial answer comes from the questionnaires collected for this investigation. As discussed in 2.3.4, questionnaires are not the best instruments to investigate non-standard relative constructions; still, they contain useful suggestions that can be checked against further materials. One of the advantages of questionnaires is that they enable the researcher to see how a construction is rendered in another language. Sometimes, the informants avoided relative constructions and relied on an asyndetic coordinate construction. Interestingly, this was the case with the relativization of GEN, as shown in (4.224), and with instances of ‘long relativization’, as in (4.225). (4.224) i BCS and da

ona druga cura, ona nova, ne mogu that other girl that new not I.can joj se sjetim imena COMP CL.DAT.F.3SG REFL remember name ‘And the other girl, the new one, whose name I can’t remember.’ lit. ‘And the other girl, the new one, I cannot remember her name.’ (Milena Marić, from questionnaire)

(4.225) Znam kobiety / podziwiam POL I.know women I.wonder wszystko robią. all do

jak how

one they

to this

238 Typological issues ‘I know women that I wonder how they can do all this.’ lit. ‘I know women, I wonder how they do all this.’ (Ewa Schalley, from questionnaire) So, it seems that when it comes to relativizing positions that are less accessible, speakers are likely not only to rely on elements that are not exclusively relative, as suggested by Bernini, but also to give up the relative construction in favour of asyndetic coordination. Another case is represented by the use of asyndetic clause clusters in non-standard Russian. As illustrated in 3.5.3, asyndetic clause clusters (bessojuznye složnye predloženija) allow for a plurality of interpretations. Syntactically, they look like asyndetic coordinate clauses; functionally, they can express most different relations – that is, they do not specifically encode relative constructions, as shown in the example below. (4.226) Плохо я спала / зайду к этой сестре / вот давление меряет // RUS Plocho ja spala / zajdu k ėtoj sestre / badly I slept I.will.go to this nurse vot / davlenie merjaet / PAR pressure measures ‘I didn’t sleep well – I’ll go to this nurse who measures blood pressure.’ ‘I didn’t sleep well – I’ll go to this nurse, because she measures blood pressure.’ ‘I didn’t sleep well – I’ll go to this nurse, so that she measures my blood pressure.’ (Širjaev 1986: 123) Even when the relation encoded can be interpreted as a relative one, it is difficult to determine whether it is a restrictive or a non-restrictive one. As will be discussed in section 8.3, although asyndetic clause clusters can encode both restrictive and non-restrictive relative relations they seem to be preferred for encoding the latter type – irrespective of the position relativized. In principle, these Russian constructions can be equally viewed as instances of relative constructions and of coordinative constructions. When they are interpreted as relative constructions, they may be considered as applying to all positions of the AH: SU is relativized in (4.227), DO in (4.228), OBL in (4.229), GEN in (4.230) and LOC in (4.231).

Syntactic positions relativized

239

(4.227) Возьми газету / у меня на столе лежит. RUS Voz’mi gazetu / [Ø u menja na stole ležit.] take newspaper at me on table lies ‘Take the newspaper, which is in my room on the table.’ ‘Take the newspaper, it’s in my room on the table.’ (Zemskaja 1973: 240) (4.228) Там у меня висят чулки я постирала вчера вечером принеси их RUS пожалуйста. Tam u menja visjat čulki [Ø ja postirala there at me hang socks I washed včera večerom] prinesi ich požalujsta. yesterday evening bring them please ‘The socks I washed yesterday evening are in my room, bring them to me, please.’ ‘In my room there are socks, I washed them yesterday evening, bring them to me, please.’ (Zemskaja 1973: 336) (4.229) Эта книжка – вы говорили – она появилась. RUS Ėta knižka – [Ø vy govorili] – ona pojavilas’. this book you told she appeared ‘This book, which you told me about, has been published.’ ‘This book, you told me about it, now it has been published.’ (Kručinina 1968: 86) (4.230) вот стул-от нога-то изломана RUS vot stul- ot [Ø noga- to izlomana] here chair DET leg PAR broken ‘Here is the chair which has a broken leg.’ ‘See this chair, it has a broken leg.’ (dialect, Lapteva 1976: 332) (4.231) Увидела тут будочка-то с Андреем все ходили. RUS Uvidela tut budočkato [Ø s I.saw here pub PAR with Andreem vse chodili.] Andrej always we.went ‘I saw (it), over there is a/the pub where we used to go with Andrej.’ ‘I saw (it), over there is a/the pub, we used to go there with Andrej.’ (Lapteva 1976: 294)

240 Typological issues As the multiple translations of the above examples suggest, it seems that it is always possible to interpret these constructions also as instances of coordination relations. Jim Miller (p.c.) argues that the interpretation of asyndetic clause clusters must be definitively left open: first, without sufficient knowledge of the communicative context it is impossible to establish whether they encode restrictive or non-restrictive relations; second, even if the context is known, it is difficult to decide whether they are instances of relative relations or of asyndetic coordination. In both cases, the use of these constructions partially contradicts the predictions made in Figure 4.1: in Russian, non-specifically relative constructions are not only used for lower positions of the AH, but also for higher ones. Vice versa, if we interpret them as asyndetic coordinate clauses, a relative construction is avoided not only to relativize lower positions of the AH, but also to relativize higher ones. 4.4. Typological conclusions As the analysis conducted in sections 4.1 to 4.3 showed, in non-standard varieties a number of relative constructions appear that are not attested in standard varieties. The peculiarities described in the previous sections are now briefly summarized and commented on. As regards the first parameter, WORD ORDER, in non-standard varieties of European languages all possible strategies are attested except for the circumnominal one. In some languages, a strategy can specialize and acquire a specific function. For instance, the postnominal strategy in Turkish and Basque is used as a device for encoding afterthoughts (cf. 4.1.1); the correlative strategy is often used when speakers want to introduce a topic and immediately qualify it, therefore turning it into a participant of the main SoA encoded through the MC (cf. 4.1.2); the postposed strategy in Dutch and German is used to encode restrictive or non-restrictive relative relations depending on the definiteness of the referent (cf. 4.1.3). So, the presence of additional word orders in non-standard varieties seems not to be random, but can be connected to functional specialization. As for the second parameter, RELATIVE ELEMENT, the cross-linguistic analysis brought to light great variability, which can be looked upon as a hallmark of relativization strategies in European varieties. Comparing relative elements in standard and non-standard varieties, a number of peculiarities were identified, which are listed below.

Typological conclusions

241

First, as it emerged in the previous sections, in many languages relative elements are polyvalent. In particular, simple elements that in the standard variety are used as relative pronouns or specialized relative elements act as relative particles in non-standard varieties (cf. 4.2.1 and 4.2.2) – that is, they do not encode the syntactic role of the relativized item in the RC or gender and number agreement. As a consequence, relative elements need to be differently pigeonholed according to the construction in which they appear and can be regarded as inherently construction-bound. Only in some cases does the classification of a relative element remain ambiguous, as was discussed in 4.2.1: we may think of FRE/POR/SPA que, ITA che, RUM care and, to some extent, ENG which: when they are used alone to relativize SU and DO they may be regarded both as relative pronouns and as relative particles. A second consideration regards relative particles, which are quite widespread in non-standard varieties (cf. 4.2.3). In some languages there seems to be a competition between relative particles for the relativization of the positions of the AH, whereas other languages rely on a single particle. The particles attested in the majority of languages show a common origin, according to which they can be subsumed to three main groups: the ‘what’ (or ‘which’) group, the ‘where’ group and the ‘how’ (or ‘as’) group. The classification of particles according to the group they belong to is represented in Table 4.9 below: if no comment is added next to the particle, this means that the particle is generally attested in non-standard varieties of the relevant language; if the use of a particle is restricted, a comment is added in brackets. It must be also pointed out that the table is compiled on the basis of the available data. If the database was expanded, even more particles may be found in a language. As the table shows, a distinction can be made between languages exhibiting a single relative particle, languages in which two or more particles are in competition and languages in which two or more particles show equal distribution. The majority of languages exhibit a single relative particle: in Greek, Lithuanian, Latvian this particle belongs to the ‘where’ group; in Albanian, Belarusian, Catalan, Macedonian, Polish, Russian, Slovak, Spanish, Ukrainian and Upper Sorbian it belongs to the ‘what’ (or ‘which’) group; in Danish, Icelandic, Norwegian and Swedish it belongs to the ‘how’ (or ‘as’) group. In some languages there seems to be a competition between particles in the ‘what’ and in the ‘where’ group. In Bulgarian deto is gaining ground at the expense of što(to) (Radeva 2003: 289). In French and Rumanian relative particles are in competition, but, apparently, no clear winner has emerged yet (cf. Gapany 2004; Vulpe 1980).

242 Typological issues Table 4.9. Relative particles in competition in non-standard varieties.

FRE RUM BCS

‘what’ / ‘which’ – – – što(to) (rare; archaic) que ce, care što

ITA

che

DUT

wat (for higher AHpositions) was (for higher AHpositions)

GRE LIT LTV BUL

GER

ENG

was (in dialects: for all AH-positions) which, what

ALB BLR CAT MAC POL POR RUS SLK SPA UKR USO LSO CZE DAN ICE NOR SWE

që što que što co que čto čo que ščo štojž což co – – – –

‘where’ pu kur kur deto

‘how’ / ‘as’ – – – –

don’t de gdje (restricted to some dialectal areas) dove (rarely used for higher AH-positions) waar (for lower AHpositions) wo (for lower AHpositions; generic connector) wo (in dialects: for all AH-positions) where (used only as a generic connector) – – – – – – – – – – – – – – – – –

– – – – – –

as (restricted to some dialectal areas) – – – – – – – – – – – ak(o) jak som sem som som

Typological conclusions

243

English is the only language in which particles from the three groups are attested. Still, where has limited applicability (cf. 4.2.2) and as is used in some dialects and its use is decreasing; only which and what show a more or less strong regional character. The most widespread particle is that, which has a different origin (Herrmann 2005: 25). In Bosnian/Croatian/ Serbian the competition between particles in the ‘what’ group and those in the ‘where’ group seems to be restricted to areally limited varieties. In supraregional varieties, the particle from the ‘what’ group clearly dominates. In Lower Sorbian and Czech there is a competition between particles in the ‘what’ group and those in the ‘how’ group: CZE jak and LSO ak(o), which were originally limited to local dialects, are spreading also in supraregional non-standard varieties (Petr 1987: 528; Faßke 1996: 169). Finally, in German and Dutch particles in the ‘what’ group and those in the ‘where’ group show a complementary distribution: the former relativize higher positions of the AH (SU and DO), the latter lower ones (OBL and GEN). This equal distribution disappears in German dialects: as described in Fleischer (2004a, 2004b), in some dialects (e.g. Alemannic) the particle from the ‘where’ group is spread throughout the AH, in others (e.g. Bavarian) it was the particle of the ‘what’ group which supplanted the other one. A further point regards the relationship between simple and combined elements. As discussed in 4.2.4, combined elements are much more frequent in non-standard varieties than in standard ones: in this respect, nonstandard varieties display greater typological diversity than their standard counterparts. In order to get an overview of relative elements attested in European varieties, the content of Tables 3.5, 3.6 and 3.7 were reported in Table 4.10. Relative elements are ordered starting from the most frequent ones; the triple combinations discussed in 4.2.4 were been reported, too. As in Table 3.9, the symbol ‘+/–’ is used to highlight that an element may or may not convey the relevant feature. As for relative pronouns, they convey gender and number agreement if they are G-N-inflecting (e.g. CZE který), only gender agreement if they are G-inflecting (e.g. ENG who) and neither gender nor number agreement if they are non-G-N-inflecting (e.g. ITA che/cui): cf. the discussion in 3.2.1. Additionally, as was shown in Table 3.8, there are different kinds of resumptive elements: some of them convey gender and number agreement, others do not. So, also for resumptive elements the symbol ‘+/–’ is used. Finally, other particles may or may not convey the link between the MC and the RC depending on whether they are complementizers or emphatic particles: again, this was represented through ‘+/–’ in the table.

244 Typological issues Table 4.10. The frequency of relative elements in the languages of the sample and the features that they encode.

1 2 3

4 5 6 7 8

9 10

11

ENCODING OF: ELEMENT: RPAR RPRO SRE RPAR + resumptive RPRO + resumptive SRE + resumptive Zero-marker Zero-marker + resumptive RPRO + RPAR SRE + RPAR RPAR + other particle RPAR + RPRO RPAR + SRE RPRO + other particle Zero-marker + other particle RPRO + RPAR + resumptive RPAR + other particle + resumptive SRE + RPAR + resumptive

LINK

MC–RC + + + + – + – + – – – – + + + + + +/– + + + + + +/– – – + + – + +/– – + + –

SYNT. ROLE

GENDER AGR.

NUMBER AGR.

OCCURS IN N LGS.

– + + – + + + + + – – + + – + – – – – + – + + – – – + – + – – + + – +

– +/– – – +/– +/– +/– – +/– – – +/– +/– – – – – – – +/– – – +/– – – – +/– – +/– – – +/– – – +/–

– +/– – – +/– +/– +/– – +/– – – +/– +/– – – – – – – +/– – – +/– – – – +/– – +/– – – +/– – – +/–

34 32 31 31 12 8 7 5 5 5 3 2 2 2 1 1

1

1

Typological conclusions

245

The following figures are significant: − There are no constraints on the parameters: a relative element can encode each of the four parameters from zero to three times. This suggests that the morphosyntactic units forming a combined element are syntactically independent of one another. − Simple elements are generally more frequent than combined elements. The only combined element that is as frequent as simple elements is ‘RPAR+resumptive element’. − Elements that encode the link between the MC and the RC more than once are dispreferred: among the most frequent combined elements no element is found in which the link is encoded more than once. − Elements that encode the syntactic role of the relativized item more than once are not dispreferred, although they are not among the most frequent ones (positions 4 and 5). − Elements that encode the syntactic role of the relativized item at least once are preferred: considering the 11 most frequent elements (positions 1 to 9 in Table 4.10), the syntactic role is encoded by 8 out of 11 elements. Still, the most frequent element, the relative particle, does not encode the syntactic role. − Elements that encode gender and number agreement are as preferred as elements that do not encode these two parameters: considering the 11 most frequent elements (positions 1 to 9), gender and number agreement are encoded in 6 cases and not encoded in 5 cases. − Elements that encode each of the four parameters only once are most preferred: in positions 1 to 3 we find no elements that encode a parameter more than once. − Elements that encode none of the four parameters are dispreferred: they are attested only in 7 out of 36 languages. − Combinations of simple elements with elements that do not encode any parameters (usally other particles) are dispreferred. On the basis of the above results, a number of tendencies may be outlined. First, the four parameters encoded by relative elements seem to rank differently as far as their encoding is concerned. This can be expressed through the following hierarchy: link between MC and RC > syntactic role > gender and number agreement The hierarchy is to be read as follows: if a relative element encodes one parameter on the hierarchy, then it encodes all parameters to its left. This applies to all relative elements reported in Table 4.10 but one, i.e. the com-

246 Typological issues bination of a zero-marker with a resumptive element. Still, this combination is not very frequent, as it is attested only in 5 out of 36 languages. How can this hierarchy be explained? One explanation may be in terms of ease of recoverability. Gender and number can be easily retrieved when they are encoded in the relativized item in the MC. The syntactic role of the relativized item in the RC cannot be directly recovered from the head noun (except in the case of case matching): still, as illustrated in 4.3.3, there are a number of semantic and pragmatic factors on which the hearer can rely in order to retrieve it. Instead, the link between the MC and the RC seems to be the most central parameter. In fact, it is encoded in all relative elements but two. That is, signalling that there is a link between two clauses must be one of speakers’ main concerns. Interestingly, Antinucci, Duranti, and Gebert (1979) reach the same conclusion after conducting a perceptual experiment.19 They claim that [t]he effect of this marker [i.e. a relative element, A.M.] on perceptual strategies is obvious: it blocks the operation of BS [Basic Strategy, i.e. the unmarked MC word order A.M.] immediately after the head noun and tells the hearer that what comes next is a new, subordinate clause. (Antinucci, Duranti, and Gebert 1979: 162)

So, if a language has a relative element, its main function consists in increasing the ease of discourse processing. Since spoken language is processed in real time and speakers “tend to segment and interpret the language input as soon as possible” (Antinucci, Duranti, and Gebert 1979: 164), signalling the link between clauses facilitates the segmentation and interpretation of the speech flow. This connects to what has been discussed in 4.2.3 on relative particles: they can be seen as projective structures, because they inform the hearer that the following SoA will have a participant of the main SoA as its topic, but do not commit to establishing which syntactic role this participant will have. So, the basic function of relative elements in the languages of the sample is to convey the link between the MC and the RC. Hence, it is not surprising that all elements attested more frequently encode this parameter. This matches also Dasinger and Toupin’s observation that [w]hen commitment to the expression of meaning is at stake, an unvarying relativizer allows for a delay in this commitment, since it leaves the speaker with numerous options as to what can follow, while preserving the ‘speed, coherence, and well-formedness’ of the utterance. (Dasinger and Toupin 1994: 510)

Typological conclusions 247

Languages in which the zero-marker or the combination of the zero-marker with a resumptive element are attested rely on different factors to signal that the MC and the RC are linked. As for the Turkish prenominal strategy, this is usually done by means of a non-finite verb form. So, in the construction [… Vnon-finite]RC [NP … V]MC the sequence Vnon-finite–NP instructs the hearer that the clause ending with V non-finite is linked to the clause beginning with NP. Moreover, the RC occupies the prenominal position, which is usually reserved for determiners and qualifying elements like adjectives, genitive attributes or first elements of nominal compounds (Haig 1998: 111–115). As for the zero-marker in English, Russian and Scandinavian languages, it can be accounted for in terms of word order, an explanation which was put forward in Antinucci, Duranti, and Gebert (1979: 158). In these languages the basic word order is SVO, that is, sequences like [NP V NP]clause [NP V NP]clause are most common. When a hearer is confronted with a sequence like [NP [NP V (NP)]RC V (NP)]MC – as it is the case with the structure of (4.232a), analysed in (4.232b) – s/he notices a deviation from the basic word order, which leads to a re-segmentation of the sequence as being constituted by an MC and an embedded RC. (4.232) Рука-то перелом был дает себя знать иногда. RUS a. Ruka- to [Ø perelom byl] arm PAR fracture was daet sebja znat’ inogda. gives REFL know.INF sometimes b. [Ruka-toNP [perelomNP bylV]RC daet sebja znat’V inogdaAdvP]MC ‘The arm I fractured sometimes still aches.’ (Zemskaja 1973: 336) This issue would surely need further investigation, which, however, would go beyond the purposes of this work. The point that can be made is that strategies in which the link between the MC and the RC is not explicitly conveyed are attested in the languages of the sample, but constitute a small minority. Among the less frequently attested relative elements there are combinations of simple elements with non-relative particles. They are never attested in more than three languages. The most likely explanation for the uncommonness of these relative elements relies on the fact that non-relative particles encode none of the four parameters. As was discussed in 3.5.4, these particles – at least in Slavic languages – constitute historical

248 Typological issues remnants. They had a functional value in the past (Urbańczyk 1939: 4), but lost it gradually and were eliminated from most of the languages, unless they could be attributed functional distinctions (see also 5.1.2 and 6.3.5). A different account can be given when a simple element is combined with a complementizer or a relative particle that can also act as a complementizer, as is the case in Danish, French, Icelandic, Macedonia, Norwegian and Swedish (cf. 3.2.2). Here, the particle may be seen as encoding the link between the MC and the RC: it is probably added when the speaker feels that this parameter is not sufficiently expressed through the preceding relative element, or it may be used to disambiguate the function of the preceding element, which is typically polyvalent. (4.233) här är huset [där som jag SWE here is house.DET SRE RPAR I ‘Here is the house where I live.’ (Smits 1989: 426)

bor] live

In (4.233) the particle som is added so as to specify that där is used as a relative element – or, more generally, as a connector – and not as a locative adverb. This phenomenon is also diachronically attested, as will be shown in 6.3.5. We may now ascertain whether the tendencies highlighted in European languages correspond to variation patterns in the languages of the world. To this purpose, the relative elements attested in the sample will be compared with Comrie and Kuteva’s (2005) cross-linguistic analysis. Comrie and Kuteva (2005) classify relativization strategies in the languages of the world according to a single parameter, i.e. the encoding of the syntactic role of the relativized item in the RC. They distinguish between relativization on subjects and relativization on obliques. Their results are reported in Table 4.11. The strategies are to be interpreted as follows: ‘gap’ means that the syntactic role of the relativized item is not encoded in the RC; ‘nonreduction’ means that the syntactic role is expressed by means of a full NP; ‘relative pronoun’ means that the syntactic role is encoded through a relative pronoun; ‘pronoun retention’ means that the syntactic role is encoded through a resumptive element.

Typological conclusions

249

Table 4.11. The encoding of the syntactic role of the relativized item in the RC in a sample of world’s languages (Comrie and Kuteva 2005). RELATIVIZATION ON SU gap nonreduction relative pronoun pronoun retention

ATTESTED IN N LANGUAGES

125 24 12 5

RELATIVIZATION ON OBL gap pronoun retention nonreduction relative pronoun not possible

ATTESTED IN N LANGUAGES

55 20 14 13 5

As the table shows, strategies that do not express the syntactic role of the relativized item predominate both in SU and in OBL-relativization. Strategies in which the syntactic role is expressed are far less frequent: considered altogether, they still are not as frequent as the ‘gap’ strategy. If we compare Comrie and Kuteva’s (2005) results with the results of the present analysis, we will notice that European relativization strategies pattern ambiguously with the world’s relativization strategies. First, both in the languages of the European sample and in the world’s languages, encoding the syntactic role of the relativized item does not seem to be a primary function of relativization strategies. In fact, the parameter ‘syntactic role’ is in the middle rather than on the left of the hierarchy enunciated above. Moreover, the relative element most frequently attested in European varieties, the relative particle, does not encode the syntactic role of the relativized item. On the other hand, if we observe positions 1 to 3 in Table 4.10, we may remark that apart from the first position, all other relative elements actually encode the syntactic role. So we may conclude that in the languages of the European sample the ‘gap’ strategy, the ‘relative pronoun’ strategy and the ‘pronoun retention’ strategy are attested with a nearly identical frequency: in this respect, they differ from the cross-linguistic tendencies described in Comrie and Kuteva (2005). Additionally, Comrie and Kuteva’s (2005: 498) statement that “in Europe, the relative pronoun strategy dominates” must be corrected: relative pronouns are as frequent as specialized relative elements, relative particles and combinations of a relative particle with a resumptive element. Similarly, data from non-standard varieties do not confirm Cristofaro and Giacalone Ramat’s claim that “invariable relative elements are exclusively found in Western Europe” (Cristofaro and Giacalone Ramat 2007: 84) – on the contrary, invariable elements are attested in 34 out of 36 languages of the European sample. As it appears, if both standard and non-standard relativization strategies are considered, the

250 Typological issues European typological panorama looks different than it is commonly described. Finally, the main findings on the third parameter, the SYNTACTIC POSITIONS RELATIVIZED, will be briefly summarized. As already observed by Miller and Weinert (1998: 108–112), the relativization strategies used in non-standard varieties seem not always to be sensitive to the Accessibility Hierarchy. In standard varieties we may notice the following tendency, which patterns with the predictions made in Figure 4.1: the further one moves to the right on the AH, the higher the chance that a [+case] strategy is used, i.e. that distinct morphosyntactic forms exist to encode the individual positions of the AH. Instead, as illustrated in 4.3.1 to 4.3.5, in nonstandard varieties [+case] strategies may be used to relativize all positions of the AH along with [++case] strategies (i.e. strategies in which the syntactic role of the relativized item is encoded twice) on the one hand and with [–case] strategies on the other. As far as the syntactic positions relativized are concerned, the generalization that “the pronoun retention strategy is preferred at the lower end of the hierarchy [the AH, A.M.]” (Comrie and Kuteva 2005: 499; cf. also Keenan and Comrie 1977: 92) is confirmed in Comrie and Kuteva’s study, but not in the present analysis. As illustrated in 4.3, there are many languages in which resumptive elements are used also for higher positions of the AH, i.e. SU and DO; on the other hand, there are languages which can relativize lower positions of the AH without inserting a resumptive element. Still, we are aware that more (quantitative) data would be needed in order to establish how frequent each strategy is and to make more solid language-specific and cross-linguistic generalizations. The result of Cristofaro and Giacalone Ramat’s (2007) analysis is not confirmed, either. Cristofaro and Giacalone Ramat identify two geographical areas in Europe. [The first area] roughly cover[s] Eastern Europe, and includ[es] Slavic languages, Finnish, Hungarian, and Rumanian. German and Greek also follow this pattern. Languages in this area typically present an inflected relative element that can be used for all syntactic roles. [The second area] roughly corresponds to Western Europe, and includes Romance and Germanic languages (except German), Greek and Irish. In this area, a variety of strategies are found that provide an overt indication about the syntactic role and possibly the gender and number of the relativized item […]. These strategies alternate with strategies that provide no indication about the syntactic role, gender and number of the relativized item. (Cristofaro and Giacalone Ramat 2007: 83–84)

A functional account?

251

As the analysis conducted in this chapter demonstrated, strategies that make the syntactic role of the relativized item in the RC explicit alternate with strategies that do not make it explicit not only in Western European languages, but also in Eastern European ones. A distinction can be made not between languages, but between language varieties: when it comes to retrieving the syntactic role of the relativized item in the RC, non-standard varieties rely both on syntactic and on semantic-pragmatic factors (cf. 4.3.3), whereas standard varieties mostly rely only on syntactic factors. 20 Adopting Givón’s (1979: 97–98) terminology, we may say that standard varieties preferably rely on a syntactic mode whereas non-standard ones rely both on a syntactic and on a pragmatic mode. In this sense, the main difference between relativization strategies in standard and non-standard varieties lies in the greater variability of the latter with respect to the former. This variability is based on the fact that both strategies relying on a syntactic mode and strategies relying on a pragmatic mode are attested; instead, in the vast majority of European standard varieties, strategies relying on a syntactic mode strongly predominate. This divergence will be taken up again in 5.1. To sum up, the following two phenomena can be cross-linguistically regarded as the hallmarks of European non-standard varieties as far as relativization strategies are concerned: 1. the great variability in the realm of relative elements, and particularly of relative particles; 2. the co-existence of strategies that make the syntactic role of the relativized item in the RC explicit and of strategies that do not make it explicit within individual languages. 4.5. A functional account? The occurrence of the relativization strategies described in sections 4.1 to 4.3 can be accounted for by a number of functional principles. They are described below. 1. ECONOMY. This principle can account for several phenomena regarding relative constructions in non-standard varieties. First, the formal reduction of the paradigms of some relative pronouns – as described in 4.2.1 for Bulgarian, English, French, Italian and Rumanian – as well as the cross-linguistically widespread use of invariable relative elements – as documented in 4.2.2 and 4.2.3 – can be seen as economic. Instead of relying on multiple morphosyntactic forms, a single

252 Typological issues

2.

3.

form is adopted as a general relative element. Second, the non-encoding of the syntactic role of the relativized item in the RC can be seen as economic, too, be it generalized to the whole AH or only attested in instances of case matching. The speaker encodes the minimum information needed – i.e., as stated above, the link between the MC and the RC – relying on the fact that the hearer will recover the missing information through co-textual and contextual inference. Finally, also the preference for relative elements that encode the four parameters no more than once can be seen as economic, since redundant information is avoided. In fact, looking at Table 4.10, the most frequent elements (positions 1 to 3) do not encode the four parameters more than once. ANALOGY. This principle seems to be at work in particular in cases of paradigm regularization, like those described in 4.2.4 for Dutch, Macedonian and Slovenian: if different morphosyntactic elements are used to encode different syntactic positions, there may be a tendency to extend the use of one or the other element to the whole AH. Similarly, the regularizations attested in non-standard varieties of Finnish and Hungarian can be related to the creation of a basic distinction [+animate] vs. [–animate], obtained by obliterating other semantic constraints. Finally, the use of strategies that allow speakers to rely on unmarked MC word order – like those described in 4.1.1 and 4.1.3 – can also be accounted for by the principle of analogy: a speaker may find it easier to build chains of MCs than complex sentences requiring different word orders, as is the case, for instance, with German or Turkish (Auer 1990: 279–284; Auer 2002). ICONICITY. This principle predicts a 1:1-correspondence between form and function: distinct functions are encoded by means of distinct morphosyntactic units. A prototypical case of iconicity is represented by decumulation (Giacalone Ramat 1982). Here, each of the morphosyntactic units constituting the combined relative element has a distinct function: the relative particle encodes the link between MC and RC, whereas the resumptive element encodes the syntactic role of the relativized item in the RC as well as gender and number agreement. Iconicity could be also invoked to account for the additional word orders found in Basque and Turkish (cf. 4.1.1) as well as in Dutch and German (cf. 4.1.3): they seem to have specialized for encoding particular functions, i.e. supplying additional information or highlighting the SoA encoded through the RC. In this sense, a distinct form (word order) conveys a distinct function, and this can be regarded as iconic. Finally, the correlative strategy reproduces a 1:1-topic-comment-ar-

A functional account?

253

ticulation: the topic is introduced and qualified in the preposed RC and commented on in the following MC. Cf. also Lehmann: “Die Funktion dieser Struktur [ist], im Vordersatz [RC, A.M.] einen Referenten auf der Basis eines komplexen Begriffs einzuführen, der dann im Eventum [main SoA, A.M.] eine Rolle spielt.” (Lehmann 1995: 1206). On the other side, it may be remarked that not all relativization strategies match functional principles. Two noticeable exceptions are represented by the [++case] strategy (double encoding, see 4.3.2) and by combinations of simple relative elements with non-relative particles (see 3.2.2, 3.5.4 and 4.2.4). Both can be considered as instances of non-economy, for different reasons. In the former case, the syntactic role of the relativized item and sometimes gender and number agreement are encoded twice, once in the clause-initial element, once in the resumptive element. This causes redundancy of information. In the latter case, if the particle is not identical to the complementizer, it usually constitutes a historical remnant that lost its functional value, so it does not contribute to the encoding of any parameter. The question may arise now as to whether the strategies attested in nonstandard varieties follow different principles than the strategies attested in standard varieties. The answer is negative: the same functional principles can be invoked to account for both groups of relativization strategies. Some examples are: − the use of a relative pronoun, which encodes all four parameters within one morphosyntactic form, can be seen as more economic than spreading the parameters onto distinct morphosyntactic units, as is the case with the combination of a relative particle with a resumptive element, which in this sense is non-economic. Still, one could also define the relative pronoun as non-iconic, since more functions are condensed in one form, whereas the combination of a relative particle with a resumptive would be iconic, as discussed above. − as was mentioned in 3.5.1, relative particles often have additional functions. The use of polyvalent elements in non-standard varieties is economic: minimal information is expressed and the hearer retrieves the missing information through inference. Still, this feature can also be regarded as non-iconic, because several functions are encoded by means of the same form. On the other hand, the use of dedicated forms to encode distinct functional relations can be seen both as iconic and as non-economic at the same time.

254 Typological issues If we try to account for the difference between standard and non-standard relativization strategies on purely functional grounds, we will not achieve convincing results: both groups of varieties exhibit strategies that follow one functional principle, but contradict another one. This means that the same principles are equally in competition in both groups, but do not constitute a valid criterion to distinguish the strategies attested in the one group from those attested in the other one. So, if we are to formulate general functional principles governing the use of relative constructions in European standard and non-standard varieties, we may tentatively express them as follows: − NON-STANDARD: “Make at least the link between the MC and the RC explicit.” − STANDARD: “Make the relation between MC and RC as unambiguous as possible and as explicit as needed.” Through these principles we may account for many, but not for all of the strategies attested in standard and non-standard varieties. For instance, the use of a zero-marker contradicts the overall principle for non-standard varieties; the use of a relative pronoun with a resumptive element, which, as discussed in 4.2.4, is regarded as standard in some languages, contradicts the overall principle for standard varieties. So, it seems that the ‘standard vs. non-standard’ opposition needs to be described in different terms. This will be done in Chapter 5.

Chapter 5 Sociolinguistic issues

In this chapter the relative constructions described in Chapter 4 will be considered from a sociolinguistic perspective: that is, they will be regarded as linguistic variants. As mentioned in 1.2, a LINGUISTIC VARIANT is regarded here as a single unit, or form, as it can be isolated by linguistic analysis from speech or writing […]. For a linguistically identifiable form to be a variant, it has to be an element of a variable, i.e. to be exchangeable paradigmatically (in de Saussure’s sense) for at least one other variant. (Ammon 20042: 274)

Variants exist on all linguistic levels – phonological, morphological, semantic, syntactic, and so on. So, if in a language a number of relativization strategies occur that are partially or totally exchangeable, they can be regarded as linguistic variants. Linguistic variants can be variety-independent or variety-bound: in the former case they occur in all varieties of a language, in the latter they only occur in one or a few varieties. Identifying linguistic variants and determining the variety or varieties in which they are attested is not always an easy task, especially as far as syntactic structures are concerned. To do this some criteria are needed according to which constructions can be pigeonholed. The criterion adopted in this work in order to distinguish between standard and non-standard linguistic variants is the judgement contained in grammars and linguistic studies. In reference grammars and in linguistic studies indications are sometimes given on the variety or varieties in which relativization strategies can be found. In 1.3 ‘non-standard’ was defined as a synonym for ‘sociolinguistically marked’. So, if in a reference work a relative construction is filed as ‘colloquial’, ‘regional’, ‘dialectal’ or ‘informal’, it will be catalogued as nonstandard. This makes it possible for each relative construction to be assigned a sociolinguistic status, i.e. to establish in which varieties that construction is used. A methodological remark. To gather information on the sociolinguistic status of relative constructions, grammars were mainly relied on. When grammars did not provide sufficient information, linguistic studies were consulted. Informants’ judgements were excluded a priori as they are often subjective and can noticeably diverge from one another. For sure, if several

256 Sociolinguistic issues grammars are consulted for the same language they may display different judgements on the sociolinguistic status of (relative) constructions, too, as the review of Bosnian/Croatian/Serbian grammars in Kordić (1999: 33–37) shows. This, however, does not prevent the formulation of general tendencies: a distinction can be made between constructions that are unanimously considered standard and constructions that are unanimously deemed as non-standard; a third group of constructions can be identified, too, for which judgement varies depending on the source consulted. In Table 5.1 below these constructions will be labelled as ‘standard/non-standard’. On the basis of the theoretical framework presented above, in this chapter the following points will be made: 1. The opposition between standard and non-standard relativization strategies can be accounted for by means of a number of sociolinguistic principles operating during the process of codification and maintenance of the standard. The linguistic variants which fit these principles are included in the standard variety; the others are not. This issue will be covered in section 5.1. 2. In the languages of the sample, typologically similar relativization strategies sometimes have a different sociolinguistic status. On the basis of the language-specific degree of standardness of each relativization strategy, in section 5.2 a cross-linguistic scale of standardness for relativization strategies in European languages will be proposed. 3. Considering relative constructions as linguistic variants allows the researcher to account for a speaker’s choice of a construction over another on sociolinguistic and functional grounds: the two perspectives do not exclude, but integrate each other. This will be discussed in section 5.3. 5.1. Non-standard vs. standard relative constructions: a sociolinguistic account As stated in section 4.5, the opposition between standard and non-standard strategies in European languages cannot be successfully accounted for in purely functional terms. Since the notion of standard and non-standard is of a sociolinguistic nature, it seems reasonable to try and account for this opposition in sociolinguistic terms. Standard is a codified variety: as mentioned in 1.3.1, this means that there exist linguistic authorities who intervene in the process of codification and maintenance of the standard. Hence, codification is not a random

Non-standard vs. standard relative constructions: a sociolinguistic account 257

process: the linguistic authorities involved in the standardization of a language follow a number of guidelines to establish which of the linguistic variants attested in local varieties should be regarded as standard (cf. also the discussion in 6.5). Stein (1997: 38) identifies three major guidelines valid for the codification of European languages: 1. no variation, 2. imitation of a prestigious (exoglossic)21 model language, 3. anti-orality. The first guideline, NO VARIATION, derives from the necessity of creating a language variety “for public communication over a wider area” (Ammon 20042: 275): in this sense, linguistic uniformity is a precondition for mutual intelligibility. The second guideline, IMITATION OF A PRESTIGIOUS MODEL LANGUAGE, is of a cultural nature: the language of culture – which, in many cases, is a classical language – is often turned to as a model, because it is attributed qualities such as ‘logical’, ‘rational’, ‘transparent’. Stein (1997: 38), who is concerned with English, German and West-European languages, actually speaks of “imitation of Latin models”. Still, this principle was changed to ‘imitation of a prestigious exoglossic model’ as Latin (LAT) was not the only prestigious language which was adopted as a model. In Eastern Europe Ancient Greek (AGR) and – within the so-called Slavia Orthodoxa (Picchio 1984) – Old Church Slavonic (oCSL) had a similar function, at least until the 18th century. Throughout the centuries, these languages were sometimes used along with other European languages that enjoyed areal or international prestige: we may mention here the use of German as a prestigious language in the Czech area in the 13th–15th centuries (Haarmann 1993: 176; Šlosar 2002: 526–528) or the use of French as an internationally prestigious language during the Enlightenment and the first half of the 19th century, when it was employed as court language (e.g. in Russia: cf. Haarmann 1993: 276; Tošović 2002: 421). For sure, the influence exerted by model languages on the formation of European standard varieties strongly varies from language to language. One of the factors the incidence of the exoglossic model seems to correlate with is the way the standard variety is codified. Roughly speaking, we may distinguish between two paths: 1. the standard variety is based on a prestigious variety – typically, the variety spoken in an influential town/region or the one adopted by the ruling classes (Ammon 20042: 275): this variety is then elaborated and embellished taking the model language as a reference;

258 Sociolinguistic issues 2.

the standard variety is developed out of a supraregional dialectal koiné: in the selection of linguistic variants, common interdialectal features are preferred to dialect-specific ones. In the former case, codifiers’ main aim is to create a prestigious variety; in the latter, their main concern is to create a common variety to promote mutual comprehension among speakers with different dialectal background. So, it can be argued that an exoglossic model is preferably referred to in the former case, so as to confer prestige to the new standard variety. As reported in Auer (2005b: 14–15), it may be well the case that both paths of standardization are followed, which leads to the coexistence of two standards: a prototypical example is Norway, where two standards, bokmål and nynorsk, are officially acknowledged in present-day usage. Both of them were codified in the 19th century: bokmål is based on written Danish as it was pronounced by Norwegians, with archaising traits (in this case prestige was pursued); nynorsk derives from a dialectal koiné (mutual comprehension was pursued; see Vikør 20062: 1751). A similar situation is attested in Greek: here, katharévusa, the prestigious, archaising standard, was acknowledged as the official standard until 1976, when it was replaced by dhimotikí, the standard based on a dialectal koiné (Trudgill and Schreier 20062: 1882). A further specification is required. It has been mentioned above that Latin, Ancient Greek and Old Church Slavonic acted as model languages, but this is partially misleading. As a matter of fact, the languages that served as models were not Cicero’s Latin, Aristotle’s Ancient Greek and Cyril and Method’s Old Church Slavonic, i.e. the varieties of these languages that had been used to write ‘classical’ or ‘canonical’ texts. All three languages had developed throughout the centuries and had been adapted to their new role as supraregional languages of culture and of religion (Banfi and Grandi 2003: 41–52, 58–59). Latin followed different paths of development in the different areas in which it was used: for instance, a variety of ‘irishized’ Latin, Hisperic Latin, developed in Ireland between the 7th and the 8th century AD (Mac Giolla Chríost 2005: 70–71). For further details see see Stotz (1994). As for Ancient Greek, two varieties competed in South-Eastern Europe, i.e. Byzantine Greek and the descendant of the Attic koiné which had formed in the 3rd century AD. Eventually, Byzantine Greek disappeared after the capitulation of Byzantium in 1453, whereas the Attic koiné preserved its role as a prestigious variety in the following centuries and served as a basis for katharévusa, one of the two Modern Greek standard varieties (Haarmann 1993: 179). Finally, the denomination ‘Old Church Slavonic’ only refers to a limited number of codices – copies of

Non-standard vs. standard relative constructions: a sociolinguistic account

259

sacred texts written in the 10th–11th centuries – which constitute the socalled ‘canon’ (Trunte 20035: 29–33). From the 11th century onward Old Church Slavonic followed different paths of development, so the model that was used as a reference in the Slavia Orthodoxa was not Old Church Slavonic, but one of the local ‘redactions’ of its descendant, Church Slavonic (CSL, cf. Hock 20065).22 Moreover, all three model languages underwent a process of (re-)codification and maintenance typical of standard varieties, which aimed at avoiding phenomena of language contact with vernacular varieties, so as to preserve the prestige of the language. As for the third guideline, ANTI-ORALITY, it has to be interpreted as an attempt to avoid specific oral traits. According to Stein (1997), it derives from the ideology of essayist literacy: the new variety had to be used first of all for writing – a realm in which the prestigious model language had mostly been used so far – and, like the prestigious model, had to be as logic, rigorous and matter-of-fact as possible. Moreover, (formal) written texts cannot rely on pragmatic inference: all textual and contextual information must be linguistically expressed. In this sense, anti-orality can be understood as a tendency of standard varieties towards what Givón (1979: 97–98) labels as ‘syntactic mode’, which is characterized by careful planning, no time pressure, no communication stress, no face-to-face interaction, no immediately obvious context (Givón 1979: 108; cf. also 4.4). These features largely coincide with the properties of Koch and Oesterreicher’s (1990) distance language (Sprache der Distanz) as opposed to proximity language (Sprache der Nähe). This supports what has been discussed in 2.1, i.e. that constructions typical of the standard variety are more likely to be found in communicative practices close to the pole of conceptional scripturality – where distance language is used – whereas non-standard constructions are more frequently found in communicative practices close to the pole of conceptional orality, where language users rely more strongly on proximity language. From these three guidelines Stein (1997: 38–44) derives some corollaries, which can be summarized in four key features: 1. explicitness, 2. compactness, 3. no redundancy, 4. purism. The first three features concern the relationship between form and meaning of linguistic variants; the last one takes their provenience into account. To start with, EXPLICITNESS implies that each meaning (i.e. each piece of semantic, morphosyntactic and/or pragmatic information) needs to be en-

260 Sociolinguistic issues coded through a linguistic form, in order to avoid any kind of direct reference to the actual speech situation or relying on pragmatic inference. All meanings and relations between meanings must be linguistically expressed. This principle is not to be confused with iconicity (cf. 4.5). Iconicity states that ideally there should be a 1:1-correspondence between form and meaning; instead, explicitness only states that each meaning must be linguistically encoded: conveying more meanings through one form is not excluded – on the contrary, it is favoured, as will become clear through the next key feature, ‘compactness’. COMPACTNESS has to do with economy of means: if an analytic and a synthetic form exist, the synthetic form is preferred; if two different forms are attested, one of which condenses more distinct (morphosyntactic) meanings in itself, this form is preferred. Instead, polyvalent forms are dispreferred because they do not encode distinct meanings with a sufficient degree of explicitness, i.e. there is no univocal correspondence between form and meaning. So, although compactness has to do with economy of means, it cannot be assimilated to the principle of economy (cf. 4.5). In particular, if a compact and an explicit form are in competition, preference is usually granted to the explicit form. This may explain why polyvalent elements, which are usually more compact, are dispreferred. On account of NO REDUNDANCY double encoding of the same meaning through separate morphosyntactic forms and pleonastic elements, i.e. elements that are thought not to encode any meaning, are excluded from the standard variety. Finally, PURISM is at work when it comes to excluding from the standard variety variants borrowed from other languages. These four key features share a common trait: their application leads to the exclusion of some linguistic variants and to the inclusion of others in the standard variety. In fact, the process of standardization of a language is first of all a process of REDUCTION: only a selection of the existing linguistic variants is represented in the standard variety (Ammon 20042: 275). Additionally, since standard varieties are also Ausbau varieties (cf. 1.3.1), the reduction of linguistic variants goes together with the ELABORATION of new linguistic variants – both on a structural (=morphosyntactic) and on a semantic level (Ammon 20042: 281). This entails that in the standard variety, linguistic variants can be attested that do not directly derive from constructions occurring in the local varieties from which the standard variety originates. As mentioned above, the new standard variety had to be rational rather than natural. So, codifiers often banned from the standard variety linguistic variants that did not match the constructions attested in the model language

Non-standard vs. standard relative constructions: a sociolinguistic account

261

or, vice versa, borrowed from the model language constructions that were not attested in the local varieties. As a consequence, “written standards often exhibit an artificial character due to the fact that they – at least in part – are the product of often naive ‘language engineering’” (Van Marle 1997: 14). In Weiß’s (2004: 649–652) opinion, prescriptivism often led to grammatical inconsistency. Weiß illustrates this through the example of double negation: although double negation occurs in all German local varieties, it was excluded from standard German because the model language, Latin, did not possess it; moreover, it was regarded as illogical, since it was argued that two negations actually result in affirmation (Weiß 2004: 652– 656). This suggests that the key feature of ‘no redundancy’ operates rather within the process of codification than within a language itself. Non-regulated language development does not exclude redundancy a priori (cf. Deutscher 2006). Inconsistency is also due to the fact that the key features listed above are mutually interdependent, but none of them seems to be more central than the others: different choices can be accounted for from time to time invoking different key features. So, the results of the codification process can in many respects be described as a “norm of realization” (“norma di realizzazione”, Fiorentino 1999: 164): codifiers start from attested constructions and modify them on the basis of the key features, so as to meet the criteria of ‘logic’ and ‘rationality’ which underlie the codification of a language. The following are some examples of this. 1. The paradigm of the relative pronoun die in standard Dutch, which was described in 4.2.4, actually results from the combination of dpronouns (die, dat) and w-pronouns (wie, wat, waar). The former should be used for SU, DO and IO, the latter for IO, OBL and GEN. In non-standard varieties, a tendency towards the sole use of w-pronouns is attested. That is, non-standard varieties show a higher degree of consistency than their standard counterparts (cf. Van der Wal 2002). 2. The paradigm of the relative pronoun que in standard Portuguese admits the use of que for the relativization of SU and DO; lower positions must be relativized by means of the forms quem or que depending on the animacy of the referent: quem is used for [+animate] referents, que for [–animate] ones. Non-standard varieties show a consistent use of the form que alone also for lower positions of the AH (Andrade Peres and Móia 1995: 269–317). 3. The semantic constraints on the use of relative pronouns in Finnish and Hungarian, described in 4.2.4, are reduced to basic semantic oppositions like [+/–animate] or [+/–human] in non-standard varieties. In

262 Sociolinguistic issues Hungarian, this allows for the elimination of a discrepancy attested in the standard, according to which a different (and higher) number of constraints are in place for the relativization of [–human] referents than for the relativization of [+human] ones. 4. Similarly, the morphosyntactic constraints regulating the use of the relative pronoun and the relative particle in Slovenian and Macedonian, discussed in 4.2.4, prescribe a different distribution of these two strategies on the AH. In non-standard varieties these constraints do not exist: both strategies are used for all positions of the AH. Once more, non-standard varieties prove to be more consistent. If we now try to apply the key features described above to the opposition between standard and non-standard relativization strategies, we will see that they prove to be more apt to account for the exclusion of some constructions from standard varieties than the functional principles reviewed in 4.5. In the following subsections, the three parameters of word order, relative element and syntactic position relativized will be once again considered in order to ascertain how the sociolinguistic principles described above apply to them. 5.1.1. Word order The exclusion of relative constructions showing alternative word orders can be due to different reasons. The postnominal construction in Turkish described in 4.1.1 is considered as non-standard because the postnominal position of the RC contradicts the basic word order pattern of this language, according to which NP-constituents precede the head noun. Then, the feature ‘purism’ may also be invoked: the Turkish relative particle ki is regarded as a Persian borrowing (cf. 3.5.2) – a reason to exclude it from a ‘pure’ version of standard Turkish (Schlyter 20062: 1882). Similarly, the Basque construction introduced by the relative element zein, which is modelled on the relative pronoun of Spanish and French, is excluded from the standard because it is not autochthonous (Trask 1998: 320). Also the correlative strategy described in 4.1.2 is mostly banned from the standard descriptions of the languages in which it is attested: no recent grammars of Hungarian, Latvian or Russian list it among standard relative constructions. This probably has to do on the one hand with the fact that these constructions are considered archaic; on the other, they do not pattern with the structures attested in the prestigious language that served as a model for the formation of the standard, which showed a postnominal

Non-standard vs. standard relative constructions: a sociolinguistic account

263

strategy (see 5.1.2 below). As for Turkish, the correlative construction is mentioned in Kornfilt (1997: 60), who states that this construction is quite rare and is consequently hardly ever mentioned in grammars. The correlative construction in Greek is discussed in Holton, Mackridge, and Philippaki-Warburton (1997: 446–447), who do not make any statement on its sociolinguistic status: so, it can be thought to be variety-independent. Finally, the postposed construction in Dutch and German discussed in 4.1.3 is considered non-standard probably because its internal word order does not pattern with the word order of other subordinate clauses: subordinate clauses in Dutch and German are verb-final, whereas these constructions are verb-second. As Sandig (1973: 42) remarks, other verb-second subordinate clauses, like those introduced by weil ‘because’ or obwohl ‘although’, are also banned from standard German for the same reason. 5.1.2. Relative element Simple relative elements attested in non-standard varieties share a common feature: they generally show a lower degree of explicitness compared to their standard counterparts. Uninflected relative pronouns (cf. 4.2.1) are less explicit than inflected ones: the latter display different case forms for different positions of the AH, whereas the former do not. Similarly, relative particles and specialized relative elements used as relative particles (cf. 4.2.3 and 4.2.2 respectively) lack explicitness as they do not univocally encode the syntactic role of the relativized item. Additionally, all these strategies differ from the default postnominal strategy with a fully inflecting relative pronoun, which was attested in the model languages, Ancient Greek, Latin and Church Slavonic. This strategy was probably regarded as both explicit and compact: it encodes the four features of relative elements (cf. 3.2.1) within a single morphosyntactic unit. The relative pronouns are respectively LAT qui, quae, quod, AGR hós, hḗ, hó and (o)CSL iže, jaže, ježe. As Lehmann remarks, Ancient Greek also possessed a postposed strategy introduced by the relative pronoun hó, hḗ, tó (Lehmann 1984: 141–143), whereas old Latin had a correlative strategy introduced by the relative pronoun qui, which, however, disappeared in classical Latin (Lehmann 1979: 9–10). As for Old Church Slavonic, it developed an additional relative particle out of the uninflected nominative masculine and neuter forms of the relative pronoun, resp. iže and ježe (Vondrák 19282: 490–491; Večerka 2002: 185); moreover, it also exhibited a correlative strategy introduced by wh-pronouns (Večerka 2002: 178–181). However, as was

264 Sociolinguistic issues stated above, these languages underwent a process of maintenance which was aimed at preserving their high prestige. In this sense, the postnominal strategy introduced by a relative pronoun was regarded as the sole relativization strategy admitted (for Old Church Slavonic see Večerka 2002: 181). Language-specific cases will now be analyzed in detail. In standard Bulgarian, two morphosyntactic forms of the relative pronoun compete for DO-relativization and three for IO-relativization (cf. section 4.2.1). As regards DO, the accusative form kogoto is preferred over kojto probably because it encodes this syntactic position more explicitly than kojto, which is identical to the nominative form. For IO-relativization, the dative komuto and na kogoto are preferred over na kojto because komuto is more compact and the preposition na requires the accusative case, which is kogoto, not kojto, at least for [+human] [+masculine] referents. The combination kojto mu is excluded from the standard, as often happens with cases of decumulation(Mišeska-Tomić 2006: 270–271; see also 5.1.3 below). Similarly, in Rumanian the form pe care is preferred over care to relativize DO because the former encodes the syntactic role of the relativized item in the RC more explicitly than the latter. As for IO, the form căruia is preferred over la care because it is more compact. In general, the use of care alone for all positions of the AH is excluded on account of its lack of explicitness (Graur 1963: 183; Nilsson 1969: 11). An attempt to encode individual forms of relative pronouns as explicitly as possible is also made in standard Italian and French (see 4.2.1). In Italian, the form che is dedicated to SU and DO and the form cui preceded by a preposition is used to relativize lower positions of the AH. In French there exist distinct forms for SU and DO, whereas for lower positions of the AH a distinction is made between [+animate] and [–animate] referents. The use of the form che/que for all positions is excluded because it does not encode the role of the relativized item and, in French, it wipes out the difference between [+animate] and [–animate] referents (Renzi, Salvi, and Cardinaletti 2001: 496; Riegel, Pellat, and Rioul 1994: 480). Also the use of specialized relative elements to relativize positions other than the one they usually relativize – SU, LOC or GEN – is generally excluded from codified standard varieties: using the specialized relative element for other positions would imply that it does not univocally encode the syntactic role of the relativized item in the RC any more, which implies lack of explicitness. There are only few exceptions to this: in Dutch, the specialized relative element waar to relativize [–animate] OBL is codified in the standard variety (see 4.2.4 and Donaldson 1997: 72); still, waar

Non-standard vs. standard relative constructions: a sociolinguistic account

265

never occurs alone in these cases, but combines with a stranded preposition, as the example below shows. (5.1) DUT

De

grammofoonplaten [waar ik gisteravond longplays SRE I yesterday.evening met jou naar heb geluisterd,] waren mooi. with you to have listened were nice ‘The longplays I’ve listened to with you yesterday evening were nice.’ (Fontein and Pescher-ter-Meer 19963: 151)

DET

In Bulgarian and Greek the use of the specialized relative element as a relative particle is codified in the standard variety: in Bulgarian it can relativize SU and DO, in Greek it extends to lower positions of the AH. This acceptance can be related to the fact that these languages display a supplementary form of the same specialized relative element: in Bulgarian we have the specialized relative element kădeto/gdeto and the relative particle deto (Tilkov 1983: 409); in Greek we have the specialized relative element ópu in addition to the relative particle pu (Holton, Mackridge, and Philippaki-Warburton 1997: 352). The use of relative particles is codified in the standard variety of the vast majority of the languages in the sample. Still, this use entails constraints that vary from language to language. We may distinguish between four groups: 1. languages in which the use of a relative particle is not codified in the standard; 2. languages in which the use of a relative particle is codified for the relativization of SU and DO; 3. languages in which the use of a relative particle is codified for the whole AH, but a combination of the particle with a resumptive element is required for lower positions; 4. languages in which the use of a relative particle is codified for the whole AH. The four group will be now described separately. GROUP 1. The languages in which the relative particle is not codified are Dutch, Finnish, German, Lithuanian, Latvian and Turkish. As mentioned above, TUR ki is rejected for puristic reasons. In languages that display high inflectional character (Finnish, Lithuanian, Latvian), the particle is excluded probably because it is not thought to be sufficiently explicit for encoding the syntactic role of the relativized item in the RC; in Dutch and German this was probably also due to the influence of Latin, which

266 Sociolinguistic issues served as an exoglossic model for these languages. The use of waar with a stranded preposition in Dutch, described above, constitutes a partial exception. A relative particle cannot be used for any other positions of the AH in standard Dutch. GROUP 2. The languages in which the relative particle is codified for SU and DO-relativization are Albanian, Belarusian, Bosnian/Croatian/ Serbian, Bulgarian, Czech, Lower/Upper Sorbian, Polish, Rumanian and Russian. Catalan, French, Italian, Portuguese and Spanish could also be included: still, this implies interpreting the form que/che as a relative particle, which is not unanimously agreed upon, as was pointed out in 4.2.1 In most of these languages the relative particle coincides with the interrogative pronoun ‘what’. Since the nominative and accusative form of this pronoun overlap with the form of the invariable particle, during the codification process it was probably assumed that the particle encoded the syntactic role of the relativized item in the RC for SU and DO. So, from a morphosyntactic point of view the invariable relative particle can be regarded as a valid representative of the relativized item for these positions. This seems to be supported by the fact that languages in which more than one relative particle is attested (cf. 4.2.3 and 4.4) accept into the standard variety only the ‘what’-based particle: this is for instance the case with BCS što vs. te, CZE co vs. jak, LSO kenž, což vs. ak(o), RUM ce vs. de. However, from a semantic point of view this is questionable: the relative particle coincides with a neuter wh-pronoun, which should be used to refer to neuter (or: [–animate]/[–human]) referents; instead, in all the languages in which a ‘what’-based particle is attested, it can be used also with nonneuter (or: [+animate]/[+human]) referents, for which a non-neuter pronoun would be expected. Codifiers probably considered only the morphosyntactic level and did not take into account the semantic mismatch that arose. As a consequence, less tolerance was shown towards the combination of a relative particle with a resumptive element to relativize lower positions of the AH: in this case the particle does not encode at all the syntactic role of the relativized item. Moreover, this strategy is not compact. This may help explain why the acceptance of the combination of a relative particle with a resumptive element in this group of languages is sometimes controversial: it is excluded in Russian and Upper Sorbian, but it is increasingly regarded as standard in Lower Sorbian (Faßke 1996: 169); in the remaining languages, judgements from grammars and linguistic studies are variable, so a definitive answer on the degree of standardness of this construction cannot be given.

Non-standard vs. standard relative constructions: a sociolinguistic account 267

GROUP 3. In these languages the use of a relative particle is codified for the whole AH, but a combination of the particle with a resumptive element is required for lower positions. These languages are Danish, English, Icelandic, Irish, Macedonian, Maltese, Norwegian, Slovak, Slovenian, Swedish and Ukrainian. Some of them – northern and western ones – are situated at the periphery of the European area under investigation. We may presume that their marginal position made them less subject to the influence of Latin, the model language of Western Europe. Still, there are other, more specific reasons explaining the acceptance of these two strategies in these languages. In Icelandic and Maltese the postnominal strategies introduced by a relative particle or a relative particle combined with a resumptive element are the only strategies attested (Scholten 2002: 326; Borg and AzzoppardiAlexander 1997: 35). Codifiers simply included them in the standard variety, without feeling the need to create new strategies on the basis of exoglossic models. In particular, in Maltese an influence of Classical Arabic can be ruled out for historical and sociolinguistic reasons (Brincat 2003); so we are left with the influence of Arabic dialects, which, without exception, rely on a relative particle or on a relative particle combined with a resumptive pronoun (Georg Bossong, p.c.). As for Irish, the strategies of the relative particle and the relative particle with a resumptive pronoun are preferred over other strategies attested in dialects, like special relative endings on the verb (“Although relative endings occur in Modern Irish dialects, they have in fact been excluded from the Official Standard”, Roma 2007: 247). This is probably due to the fact that relative endings do not exist for all verbs and verb forms, so their application is restricted, whereas the particle can be always applied. In Danish, Norwegian and Swedish the influence of a classical language during the formation of the standard was limited to translations of religious texts (Haarmann 1993: 174–175). This may explain why, although in these languages a postnominal strategy introduced by a relative pronoun is attested, its use has always been restricted to formal and bureaucratic registers (Allan, Holmes, and Lundskær-Nielsen 1995: 202; Strandskogen and Strandskogen 19862: 121; Holmes and Hinchliffe 1994: 220). The main strategies remain the postnominal strategy with a relative particle or with a relative particle and a stranded preposition, which are attested in standard and non-standard varieties. In English, Latin (and French) influence was stronger than in Scandinavian languages: still, it must not have been so strong as to rule out the relative particle and the combination of a relative particle with a resumptive element (preposition), which are attested in pre-

268 Sociolinguistic issues sent-day standard (Biber et al. 1999: 608–617; Carter and Mc Carthy 2006: 571). Instead, the combination of a relative particle with a resumptive element (pronoun) was not accepted into the standard variety on account of purism, as it was regarded as a Celtic borrowing (Roma 2007: 284). In Macedonian, Slovak, Slovenian and Ukrainian the standard variety originated from a dialectal koiné rather than from the adaptation and embellishment of a prestigious language variety: cf. for instance Schweier (20065: 107–108) for Ukrainian. Additionally, three of these languages were officially codified relatively late in time: Slovenian and Slovak in the second half of the 19th century, Macedonian in the first half of the 20th century (Grdina and Stabej 2002: 501; Vintr 20065: 223; Neweklowsky 20062: 1825, 1830–31). This probably entailed that codifiers aimed at creating a common variety for mutual comprehension rather than a prestigious one: in this respect, no exoglossic language was turned to during the process of codification. GROUP 4. Only one language can be catalogued in this group, i.e. Greek. In Greek both the use of a relative particle and the use of a relative particle with a resumptive element are codified for all positions of the AH. With reference to what has been discussed above, this can be explained by the fact that the present-day Greek standard, dhimotikí, is based on a dialectal koiné: the linguistic variants attested in dialects entered this standard more easily than they entered katharévusa, the official standard variety until 1976, which had strongly been modelled on Ancient Greek (Haarmann 1993: 179; Trudgill and Schreier 20062: 1882). One last remark regards combined elements. As was mentioned in 4.2.4, only few combined elements are attested in standard varieties. In particular, the combinations of a simple relative element with a relative or nonrelative particle are systematically ruled out. This can be accounted for by the feature of non-redundancy: except for the zero-marker, simple relative elements always encode the link between the MC and the RC; so, the presence of an additional relative particle or complementizer, which also encodes this link, is probably seen as redundant and consequently not included in the standard variety. As for non-relative particles, in 3.5.4 it was stated that they mostly lost their functional value. So, combinations of a simple relative element with a non-relative particle were accepted in the standard variety of a language only if they could be ascribed a distinguishing function: for instance, in Bulgarian we have a codified opposition between koj ‘who?’ as an interrogative and kojto ‘who’ as a relative pronoun; in Lower and Upper Sorbian the particle -ž is attached to all relative pronouns and specialized relative elements in order to keep them distinct

Non-standard vs. standard relative constructions: a sociolinguistic account

269

from the corresponding interrogative pronouns and adverbs (e.g. LSO źo ‘where?’ vs. źož ‘where’ as a specialized relative element); in Czech the relative pronoun kterýžto (=který+že+to) is a marked alternative for který and has a restricted use as connecting relative (Petr 1987: 534; for the definition of ‘connecting relative’ see 5.2.1.). In Slovenian similar oppositions exist, which, however, do not affect the two most frequent relative elements, kateri and ki (cf. 3.5.4 and Herrity 2001: 102–106). Combinations of relative elements with non-relative particles are discussed from a diachronic perspective in 6.3.5. In other languages, like German and Polish, elements like der da ‘RPRO+other particle’ or co to ‘RPAR+other particle’ were not included in the standard because the non-relative particle was not ascribed a distinguishing function. 5.1.3. Syntactic positions relativized The phenomena attested in the realm of the syntactic positions relativized, which were described in 4.3, are mostly considered non-standard in the languages of the sample. This can be accounted for by three of the key features mentioned above, i.e. explicitness, compactness and no-redundancy. The following tendencies can be identified: 1. if a language displays relativization strategies explicitly encoding the syntactic role of the relativized item, these strategies are preferably regarded as standard (→ explicitness); 2. if a language displays relativization strategies in which the functions of the relative element are encoded by means of a single morphosyntactic unit, these strategies are preferably regarded as standard (→ compactness); 3. if a language displays relativization strategies encoding the syntactic role of the relativized item more than once, these strategies are preferably regarded as non-standard (→ no redundancy). Each tendency will be now analysed separately. TENDENCY 1. This tendency regards strategies in which the syntactic role of the relativized item is not encoded, like those described in 4.3.3 and 4.3.4. They are principally excluded because they are not considered sufficiently explicit. A relative particle only conveys the link between the MC and the RC; the zero-marker encodes none of the parameters expressed by relative elements. In this sense, a relative relation is not univocally encoded when these strategies are used. Since the standard variety requires all se-

270 Sociolinguistic issues mantic and morphosyntactic information to be explicitly expressed, these strategies are ruled out because they rely too strongly on pragmatic inference. TENDENCY 2. Following this tendency, cases of decumulation (4.3.1) are dispreferred if the relevant language exhibits alternative strategies. This is particularly the case with Romance and Slavic languages. As illustrated in 5.1.2, in some Slavic languages (Macedonian, Slovak, Slovenian, Ukrainian) a higher tolerance is attested towards the combination of a relative particle with a resumptive element: this may be due to the fact that pressure from the prestigious model was less strong in these languages. Still, in the great majority of the languages in the sample the relative pronoun strategy is preferred because it is more compact. An exception is represented by Scandinavian languages – Danish, Norwegian and Swedish – in which the relative pronoun has a restricted use even within the standard variety: it is limited to (conceptionally) written communicative practices, as mentioned in the previous section. Being introduced as a borrowing from an exoglossic language (Latin), this strategy remained confined to translations and never gained ground in these languages (Haarmann 1993: 174–175; cf. also 6.3.5). In this sense, the relative pronoun strategy constitutes a non-default strategy, whereas the relative particle and the combination of a relative particle with a stranded preposition are default ones. Finally, if in the relevant language there are no alternative strategies, as is the case with Icelandic and Maltese, decumulation is accepted in the standard variety. At the same time, two additional factors should be considered: the time factor and the Ausbau issue. The Slavic languages mentioned above were codified later than a more prestigious, genetically closely related language which had served as an H-variety in the area where local varieties of these languages were spoken (Bulgarian, Czech, Bosnian/Croatian/Serbian, Russian respectively). So, Slovak developed as an own Ausbau language different from Czech, and the same happened with Ukrainian, Macedonian and Slovenian, which were codified as standard varieties in opposition to Russian, Bulgarian and Bosnian/Croatian/Serbian respectively. The intention to create a variety independent of the neighbouring language may have led to the inclusion in the standard of constructions that were considered non-standard in the neighbouring language, but were accepted in the local varieties on which the new Ausbau variety was based. TENDENCY 3. On account of this tendency, cases of double encoding of the syntactic role of the relativized item are dispreferred. In 4.3.2 it was pointed out that in some languages clitic doubling also occurs in MCs: in

Non-standard vs. standard relative constructions: a sociolinguistic account

271

some of the languages in which this phenomenon is attested it is also codified in the standard; in Italian and French it is admitted only in cases of constituent extraposition. So, one can suppose that if clitic doubling is codified for MCs – with the consequence that one or more objects are double-marked – it is also accepted in RCs. On the contrary, in Italian and French, where clitic doubling is limited to cases of constituent extraposition, clitic doubling was not admitted in RCs as they do not display any extraposed components. As already mentioned in 4.3.2, it must be pointed out that the acceptance of double encoding varies not only cross-linguistically, but also on a language-specific level. So, in Macedonian double encoding in RCs is always allowed both for DO and for IO, irrespective of the form of the relative pronoun, whereas this is not accepted in Bulgarian. (5.2) Каде е студентот кому му ги дадовме книгите? MAC Kade e studentot [komu where is student.DET RPRO.DAT.M.SG mu gi dadovme knigite]? CL.DAT.M.3SG CL.ACC.3PL we.gave books.DET ‘Where is the student we gave the books to?’ (Cramer 2003: 152) (5.3) Каде е студентот на кого му ги дадовме книгите? MAC Kade e studentot [na kogo where is student.DET to RPRO.ACC.M.SG mu gi dadovme knigite]? CL.DAT.M.3SG CL.ACC.3PL we.gave books.DET ‘Where is the student we gave the books to?’ (Cramer 2003: 152) (5.4) BUL

čovekăt, [kogoto (*go) man.DET RPRO.ACC.M.SG CL.ACC.M.3SG vidjaxme] we.saw ‘The man we saw.’ (Mišeska-Tomić 2006: 270)

(5.5) BUL

čovekăt, [komuto (*mu) man.DET RPRO.DAT.M.SG CL.DAT.M.3SG dadoxme pismoto] we.gave letter.DET ‘The man we gave the letter to.’ (Maria Manova, p.c.)

272 Sociolinguistic issues (5.6) BUL

čovekăt, [na kogoto (mu) man.DET to RPRO.ACC.M.SG CL.DAT.M.3SG dadoxme pismoto] we.gave letter.DET ‘The man we gave the letter to.’ (Mišeska-Tomić 2006: 271)

In Macedonian clitic doubling in RCs is allowed both if the ending of the relative pronoun is identical to the form of the resumptive clitic, as in (5.2), and if it is different, as in (5.3). In Bulgarian clitic doubling is not allowed if the ending of the relative pronoun is identical to the form of the resumptive clitic, as (5.4) and (5.5) show. Instead, if the form is different, doubling is possible, as (5.6) exemplifies. As for Spanish, although double encoding is attested both for DO and for IO, it is regarded as standard only in the latter case, at least for what concerns the norm of Spanish spoken in Spain (Bosque and Demonte 1999: 403). So, according to the Spain-Spanish norm, (5.7) would be standard, whereas (5.8) would not; according to the norm of South American Spanish (“rioplatense”, Bosque and Demonte 1999: 405) both sentences would be acceptable. This is probably also connected to phenomena like leísmo, laísmo and loísmo, i.e. the alternation of the clitic le and lo/la for marking direct objects (Bosque and Demonte 1999: 1317–1399). (5.7) SPA

el

profesor [a quien no professor to RPRO.SG not le concedieron la venia docente] CL.DAT.M.3SG they.granted DET permission teaching ‘The professor who was not granted teaching permission.’ (Bosque and Demonte 1999: 404)

(5.8) SPA

el

DET

atracador [a quien lo DET mugger to RPRO.SG CL.ACC.M.3SG vieron por la zona] they.saw around.here ‘The mugger who was seen around here.’ (Bosque and Demonte 1999: 405)

To summarize, this brief review of the degree of acceptance of relativization strategies in European languages confirmed that the opposition between standard and non-standard relative constructions can be better accounted for in sociolinguistic terms: in particular, the four key features

A scale of standardness for relative constructions in European languages

273

expressing the principles that were followed by codifiers during the process of codification of the standard variety proved to be most relevant to explain why typologically similar constructions do not have the same sociolinguistic status in different languages. It may be objected that the guidelines and key features described above have been defined ex post: observing how European languages have been codified led to formulating a number of principles with which all or most languages seemed to pattern. However, in the vast majority of the cases it is impossible to establish with absolute certainty the reasons why some constructions were excluded from the standard variety. In this respect, the considerations made in this section must be regarded as provisional and surely need to be more thoroughly investigated in language-specific studies, particularly when reference works do not provide precise indications on the status and the origin of a construction. Nonetheless, it can be maintained that the opposition between standard and non-standard constructions is far better accounted for by the sociolinguistic principles introduced above than by the functional principles discussed in section 4.5. 5.2. A scale of standardness for relative constructions in European languages In the previous section it was shown that relativization strategies that are typologically related can be regarded as standard or non-standard in different languages. Based on this fact, a scale of standardness for relativization strategies in European languages will be now proposed. When discussing the issue of linguistic variants, Ammon (20042) is quite cautious on the possibility of formulating scales of standardness: in his opinion, “it seems possible, in principle, to develop scales of standardness […] though their general validity may remain questionable.” (Ammon 20042: 278). This, however, does not imply that all attempts to formulate scales of standardness would be useless. On the contrary, a scale of standardness may provide an overview of which constructions are preferably considered standard not only on a language-specific level but also on a cross-linguistic one. Although the position of the individual constructions on the scale of standardness may not always be definitively established due to contrasting judgements on the same construction from different sources, there are never any great shifts: disagreement is usually found for strategies situated next to each other on the scale of standardness, so variation can be assumed to be minimal. For sure, as mentioned above, this scale of stan-

274 Sociolinguistic issues dardness constitutes only an initial proposal to be refined or re-defined when new and more comprehensive data can be examined. The scale of standardness for relative constructions in European languages is presented here: RPRO, SRE > RPAR (high positions of the AH) > RPRO+resumptive element > RPAR+resumptive element > Ø, Ø+resumptive element > RPAR (low positions of the AH) The scale is to be read as follows: if a relativization strategy is regarded as standard in a language, all relativization strategies on its left are also regarded as standard, provided that they are attested in the relevant language. So, if in a language the strategy ‘RPRO+resumptive element’ is considered as standard, then also the strategies ‘RPRO’, ‘SRE’ and ‘RPAR for higher positions of the AH’ – if attested – are considered as standard. A few explicative remarks are in order. First, as the scale has to exhibit cross-linguistic validity, only strategies with significant cross-linguistic distribution were taken into account. So, the parameter ‘word order’ was discarded, since non-postnominal word orders are attested only in a small number of languages, as Tables 3.1 and 4.1 show. As a consequence, all relativization strategies represented on the scale of standardness are postnominal. This implies that if a language exhibits relativization strategies with different word orders only postnominal strategies are taken into account on the scale. Second, the parameter ‘syntactic positions relativized’ was taken into account only if there were patent discrepancies in the syntactic positions that a relative element can relativize. On the scale of standardness this is only the case with ‘RPAR’: a relative particle for the relativization of higher positions of the AH (SU and DO) ranks quite high on the scale of standardness, whereas the same strategy applied to lower positions of the AH appears in the rightmost position. All other strategies can reasonably be regarded to apply to the whole AH, with few exceptions. The most noticeable cross-linguistic exceptions regard the applicability of the three relativization strategies situated in the middle of the scale of standardness (‘RPRO+resumptive’, ‘RPAR+resumptive’, ‘Ø’, ‘Ø+resumptive’) to the position SU: even in languages in which these strategies are considered as standard, SU-relativization by means of these strategies is ruled out. Double marking of SU is probably felt to be quite redundant: in fact, it is accepted neither in MCs nor in RCs. As for the other syntactic positions, there appear to be language-specific constraints, which have been examined in section 5.1.3. Additionally, the acceptance of the strategy ‘RPAR+

A scale of standardness for relative constructions in European languages

275

resumptive element’ strongly varies from language to language and, within individual languages, according to the source consulted: generally speaking, it is admitted if the resumptive element is a clitic pronoun, whereas it is excluded if the resumptive element is a pronominal PP. This is why this strategy is often labelled as ‘standard/non-standard’ in Table 5.1 below. To summarize, the main distinctive element used to determine the degree of standardness of a strategy is the relative element that it exhibits. This is why strategies on the scale of standardness are defined on the basis of relative elements. As the scale shows, both simple and combined relative elements are considered. Still, not all relative elements discussed in Chapters 3 and 4 appear on the scale: only the elements occupying positions 1 to 7 in Table 4.10 were considered, that is, elements occurring at least in five languages. This is because all elements occupying positions 8 to 11 are regarded as non-standard in the few languages in which they are attested: if they were inserted into the scale of standardness, they would occupy the rightmost position, on the right of the strategy ‘RPAR for lower positions of the AH’. In order to demonstrate how the scale of standardness for relativization strategies applies to the languages in the sample, in Table 5.1 relativization strategies are placed on different rows according to their sociolinguistic status. If a strategy is considered standard it is labelled as ‘standard’; if a strategy is considered as non-standard it is labelled as ‘non-standard’; if there are diverging judgements on the sociolinguistic status of a strategy, the label ‘standard/non-standard’ is used; finally, if a strategy is not attested in a language, this is signalled through the symbol ‘–’. The double line marks the boundary between strategies considered as standard and those considered as non-standard. The lower a language appears in Table 5.1, the greater the acceptance shown in this language towards strategies situated in the right segment of the scale of standardness. Two remarks must be added. First, the combination ‘RPRO+resumptive element’ in English seems to be quite rare: it is quoted only in Herrmann (2005: 47) and Biber et al. (1999: 662). If after further investigation this strategy was proven to have significant frequency, then English would not fully pattern with the scale of standardness. Second, the strategy ‘RPAR+resumptive element’ in Basque is attested (and considered standard) only in cases of long relativization (Oyharçabal 2003: 782–783).

276 Sociolinguistic issues Table 5.1. The position of relativization strategies in individual languages on the scale of standardness. Strategy: Languages: EST, HUN FIN, LIT, LTV TUR

RPRO, SRE standard standard –

RPAR RPRO + (high pos. res. el. of AH) – – non– standard non– standard non– standard nonnonstandard standard standard –

DUT, GER

standard

FRE

standard

RUS

standard

ITA, POR

standard

standard

SPA

standard

standard

nonstandard standard

ALB, BUL, RUM

standard

standard

standard

BLR, POL, USO

standard

standard



CZE

standard

standard



CAT

standard

standard

standard

LSO

standard

standard



BCS

standard

standard



ENG

standard

standard

nonstandard

RPAR + res. el. – –

Ø RPAR (+res. el.) (low pos. of AH) – – – –

non– – standard non– – standard non– – standard nonnonnonstandard standard standard non– nonstandard standard non– nonstandard standard standard/ – nonnonstandard standard standard / – – nonstandard standard / non– standard nonstandard standard / – nonnonstandard standard (equi type) standard / – nonnonstandard standard (equi type) standard / – nonnonstandard standard standard/ standard / – nonnonstandard standard

A scale of standardness for relative constructions in European languages 277 Table 5.1. (continued) Strategy: Languages: DAN, NOR, SWE SLK, SLN, UKR MAC GRE ICE, MAL, IRI BAS

RPRO, SRE (formal) standard standard

RPAR RPRO + (high pos. res. el. of AH) standard –

RPAR + res. el. standard

Ø RPAR (+res. el.) (low pos. of AH) standard –

standard



standard





standard (formal) standard –

standard standard

standard standard

standard standard

– –

– standard

standard



standard







standard



standard



standard

The analysis of Table 5.1 suggests a number of observations with reference to three spheres: the degree of standardness of individual elements; the relationship between the degree of standardness of different elements; the relationship ‘standard vs. non-standard’ within individual languages. They will be discussed separately in the next subsections. 5.2.1. The degree of standardness of individual strategies Each column in Table 5.1 provides us with a cross-linguistic overview of the degree of standardness of individual elements. If we consider the column ‘RPRO, SRE’, we will notice that relative pronouns and specialized relative elements are regarded as standard in all languages in which they are attested. In some languages, a further distinction needs to be made: in Danish, Greek, Norwegian and Swedish a relative pronoun does exist, but its use is limited to communicative practices close to the pole of conceptional scripturality (Allan, Holmes, and Lundskær-Nielsen 1995: 202; Holton, Mackridge, and Philippaki-Warburton 1997: 319, 440; Strandskogen and Strandskogen 19862: 121; Holmes and Hinchliffe 1994: 220); this pronoun is G-N-inflecting and typically originates from the interrogative pronoun ‘which’: DAN hvilken, GRE o opíos, NOR hvilken, SWE vilken. This is why the label ‘(formal) standard’ was used in the table. This label partially applies to another group of languages in which more than one relative pronoun is attested: Catalan, Dutch, French, German, Italian, Por-

278 Sociolinguistic issues tuguese and Spanish. In these languages, the relative pronoun deriving from the interrogative pronoun ‘which’ – sometimes combined with a determiner – is more formal than the other attested pronoun(s), which can have different origins. Table 5.2. Relative pronouns in some West-European languages. LANGUAGE CAT DUT FRE GER ITA POR SPA

RELATIVE PRONOUN I qui die qui der che que que

RELATIVE PRONOUN II el qual welke lequel welcher il quale o qual el cual

RELATIVE PRONOUN III el que − − − − o que el que, quien

Table 5.2 illustrates all forms. The label ‘relative pronoun II’ refers to the formal ‘which’-based pronouns, the label ‘relative pronoun I’ refers to the most widespread forms. The relative pronouns in the rightmost column have a limited use (Smits 1989: 246, 206–207, 400). Like in Greek and Scandinavian languages, in these languages, too, the ‘which’-based pronoun is often found in communicative practices close to the pole of conceptional scripturality. This fact is connected with the origin of the ‘which’-based relative pronoun. As Giacalone Ramat (2005: 124) points out, in Italian the first occurrences of this relative element are found in translations from Latin, particularly when LAT qui is used as a CONNECTING RELATIVE. This term is used in Giacalone Ramat (2005: 123), who in turn refers to Lehmann’s (1984: 274) label for this phenomenon as “relativischer Anschluss”: as the example below shows, the relative element la qual is used at the beginning of a new sentence in order to refer back to a participant of one SoA in the preceding co-text (la miglior che vi fosse). (5.9) ITA

la

miglior [che best RPRO/RPAR die-li-la. La qual gave-to.her-her RPRO.F.SG meno-sse-la a brought-REFL-her to DET

vi fosse] there.was quella gentil that gentle casa. home

e and donna prese lady took

e and

A scale of standardness for relative constructions in European languages

279

‘(Teofilo let her choose) the best (woman) that was there and gave her to her. And that lady took her and brought her home.’ (13th century, Racconti esemplari 14, quoted in Giacalone Ramat 2005: 122) So, this pronoun can be regarded as an element facilitating the cohesion between subsequent parts of a text. This function corresponds to the function that Dasinger and Toupin (1994: 370) label as ‘continuing the narrative’ – a function that is prototypically realized through non-restrictive relative relations. So, ‘which’-based relative pronouns had the function to anaphorically resume a previously introduced referent at the beginning of a new clause/sentence. They were preferred to other relative pronouns because they were G-N-inflecting: they could encode both the syntactic role of the relativized item and gender and number agreement, something that other relative pronouns, like ITA che, didn’t do. Hence, they were highly suitable for use in conceptionally written texts, which, as stated in 5.1, require all semantic and morphosyntactic information to be encoded explicitly. The origin of these pronouns also account for the fact that they are mostly found in non-restrictive relative relations, whereas in restrictive relations their use is more restricted or not attested. This is particularly the case in Romance languages, cf. Smits (1989: 245, 305, 321, 350–351, 399). Originally attested in Romance languages – their first occurrences date back to the 12th century (Giacalone Ramat 2005: 124, 132–133) – ‘which’based pronouns soon spread to Germanic languages: the Dutch form welke and the German form welcher are attested in chancellery style from the 13th century onwards (Lockwood 1968: 248; Van der Wal 2002: 29, 33; Maček 1986/1987: 113). In Swedish, the form hwīlkin is attested mainly in translations from Latin and in chancellery style from the 14th century onwards (Noreen 1904: 410; Platzack 2002: 85). Its historical development offers an insight into the present-day use of this pronoun: it arose as a specific written device to improve textual cohesion and it was best suitable to translate the Latin relative pronoun since it had the same morphological features. Moreover, it exhibited all key features that linguistic variants had to display in order to be included into standard varieties: it was explicit, compact and not redundant. Hence, in all languages in which it was attested, it was regarded as standard. More than one relative pronoun is attested also in the vast majority of Slavic languages: this pronouns usually derive from the interrogative pronouns ‘which’ (*ky-), ‘which (of the two)’ (*k(o)t(o)r-/kter-) and ‘of which kind’ (*kak-/(j)ak-). They are reported in Table 5.3. The label ‘relative

280 Sociolinguistic issues pronoun I’ indicates that the relevant form is either more widespread or semantically less specific than ‘relative pronoun II’. ‘Relative pronoun III’ includes forms that are formal and archaic: they are mostly used in conceptionally written communicative practices. Table 5.3. Relative pronouns in Slavic languages. LANGUAGE BCS BLR BUL CZE LSO MAC POL RUS SLK SLN UKR USO

RELATIVE PRONOUN I koji jaki kojto který kotaryž koj(što) który kotoryj ktorý kateri jakyj kotryž

RELATIVE PRONOUN II kakav katory kakăvto jaký kakiž kakov (što) jaki kakoj aký kakršen kotryj kajkiž

RELATIVE PRONOUN III − − − jenž − − − koj − − − −

The relationship between ‘relative pronoun I’ and relative pronoun II’ is not related to diaphasic or diamesic variation, but either to diatopic variation or to a number of semantic constraints, which are analyzed in detail in Křížková (1970). For instance, in Polish który predominates, but in some dialects jaki is preferred (Gołąb and Friedman 1972: 36). In Russian, kotoryj is the more widespread form; still, dialects can be divided into two groups, i.e. northern ones, preferring kotoryj and southern ones, preferably relying on kakoj/jakoj (Avanesov and Orlova 19652: 200). As for semantic constraints, apart from Ukrainian and Belarusian, where the form *kak/*jak- is predominant, in all other languages this form is rather used to express that a participant of the main SoA and a participant of the secondary SoA are analogous or similar, not referentially identical (Křížková 1970: 13–17). Večerka (2002: 175) speaks of ‘total’ vs. ‘qualitative’ identity between the relativized items in the MC and in the RC. So, in (5.10) the relativized item in the RC does not refer directly to a specific glass cabinet, but to a class of glass cabinets which are similar to the one described in the MC. This can also be seen in the lack of number agreement: the relativized item in the MC is singular (omarica ‘cabinet’), whereas the relativized item in the RC, expressed through the relative pronoun, is plu-

A scale of standardness for relative constructions in European languages

281

ral, because it refers to a whole class of objects (cf. also the English translation). (5.10) Na koncu pulta je bila steklena SLN at end desk.of is been made.of.glass omarica [v kakršnih običajno hranijo cabinet in RPRO.PRP.PL usually they.keep bonbone.] sweets ‘At the end of the desk there was a glass cabinet, like those in which usually sweets are kept.’ (Křížková 1970: 15) Unlike ‘which’-based relative pronouns of Romance and Germanic languages discussed above, Slavic ‘which’-based relative pronouns did not originate as specific written devices. This is confirmed by their presence in the whole diasystem of Slavic languages – both synchronically and diachronically. The use of these relative pronouns is not restricted to the (formal) standard variety, but extends to dialects. The only pronouns that nowadays show a restricted use are CZE jenž and RUS koj. The latter is considered formal and archaic; the former is a remainder of the old form iže, which has disappeared in all other Slavic languages, and it is limited to higher registers and GEN-relativization (cf. also section 6.3.1). If we take another look at Table 5.1 (third and seventh column), we will notice that the acceptance of relative particles in standard varieties is biased towards the highest positions of the AH. In 27 of the 34 languages in which a relative particle is attested, it is considered standard only for the relativization of SU and DO. Incidentally, one may remark that the languages in which a particle is not standard belong to different language families and can be both highly inflectional (Finnish, Lithuanian, Latvian, Turkish) or less inflectional (Dutch, German). So, one cannot directly or exclusively relate the presence of a particle to genetic relatedness or to the degree of inflectionality. Similarly, Kurzová claims that die Relativpartikel [wurde] nicht durch den Rückgang der Flexion, sondern durch die auf der syntaktischen Ebene liegenden Tendenzen – Formalisierung des Satzes, Analogie zum konjunktionalen Nebensatz – bedingt […]. (Kurzová 1981: 89)

As far as the relativization of lower positions of the AH is concerned, the use of a relative particle alone is codified only in Greek; in the other 11 languages in which it is attested it is regarded as non-standard. This can be

282 Sociolinguistic issues explained once more by invoking the feature of explicitness. As mentioned in 5.1.2, in languages like Slavic ones, in which the form of the relative particle overlaps with the nominative and accusative form of the neuter interrogative pronoun, its use in positions other than SU and DO is not admitted in the standard variety since it would lead to case mismatches. The combination of a relative pronoun with a resumptive element (illustrated in the fourth column in Table 5.1), though contradicting the feature ‘no redundancy’, is regarded as standard in the vast majority of the languages in which it is attested at least for some syntactic positions, as it was explained in the previous section. Instead, languages strongly differ in terms of the acceptance of the combination of a relative particle with a resumptive element. A quick look at the fifth column of Table 5.1 reveals that in 8 languages this strategy is regarded as standard, in 12 other languages it is regarded as non-standard and in the remaining 11 languages its sociolinguistic status is ambiguous. This may be explained if we consider that this strategy actually matches three of the four features that are relevant to the inclusion of linguistic variants in the standard variety: explicitness, no-redundancy and purism – with the noticeable exception of Turkish, where this strategy is regarded as non-autochthonous. So, the determining criteria to include or exclude this strategy from the standard must have been either compactness – this strategy is less compact than simple relative elements – or the influence of a prestigious exoglossic model on the formation of the standard. In the eleven languages in which this strategy shows an ambiguous status there is probably a competition between an older more restrictive norm and a more recent usage-based norm. This is for instance the case with Low Sorbian, as stated in Faßke (1996: 169): it will be interesting to see if in the long run, the usage-based norm will gain ground. Other factors affecting the acceptance of this strategy in the languages of the sample were mentioned in 5.1.3: the path that was followed to codify the standard variety (prestigious variety vs. dialectal koiné), the time factor (standards codified at a later date are less restrictive) and the Ausbau factor (languages that developed a standard variety in opposition to a neighbouring language show greater tolerance towards constructions that in the neighbouring language are considered as nonstandard). The zero-marker strategy (illustrated in the sixth column of Table 5.1) is considered standard only in a restricted number of languages, but never for SU-relativization, since this would go against the non-pro-drop-character of the standard variety of these languages (cf. also the discussion in 5.1.3)

A scale of standardness for relative constructions in European languages

283

Finally, it needs to be pointed out that the acceptance of different strategies in the standard variety as represented in Table 5.1 means that at least one relativization strategy of the relevant type is regarded as standard. This is particularly evident in the case of relative particles. As was discussed in section 4.2.3, European languages display a full range of relative particles – cf. in particular Table 4.5. Still, not all of the relativization strategies in which these particles are used are regarded as standard. Some examples are: − in Czech, the strategy with the particle co is considered standard, at least for the relativization of SU and DO; the originally dialectal particle jak is nowadays spreading its use to supraregional varieties – still, it has not been accepted into the standard variety yet (Sgall et al. 1992: 138); − in English, that is regarded as standard, whereas as and what are pretty dialectal and, as such, do not belong to the standard variety (Biber et al. 1999: 609) − in Rumanian, the strategy with the particle ce is regarded as standard, whereas the strategy with the particle de is not (Graur 1963: 184ff.); − in Upper Sorbian, the particle kiž is attested in the standard, whereas kož and štojž are not; similarly, in Lower Sorbian kenž and, to some extent, ako belong to the standard variety, whereas hač and (h)až do not (Faßke 1996: 169). 5.2.2. The relationship between strategies If we consider once more the scale of standardness introduced above as well as the horizontal distribution of standardness across relativization strategies in Table 5.1, two observations can be made. First, the table shows that simple elements are included into the standard variety more often than combined elements. This can be represented as follows: simple relative elements > combined relative elements That is, simple relative elements are considered more standard than combined ones. This tendency patterns with the key features of compactness and no-redundancy introduced in 5.1: simple elements are more compact and do not encode the parameters more than once, whereas in combined elements double encoding may occur.

284 Sociolinguistic issues Second, if we consider only simple elements relative pronouns and specialized relative elements tend to be more standard than relative particles and the zero-marker: RPRO, SRE > RPAR, Ø This tendency can be explained by means of another key feature, i.e. explicitness: standard varieties prefer relative elements that possibly encode all of the four parameters. We may now recall the hierarchy of the functions encoded by relative elements introduced in 4.4 and compare it with the above hierarchy: link between MC and RC > syntactic role > gender and number agreement Taking a close look, we may notice a divergence between both hierarchies: the latter hierarchy predicts that in European languages relative elements encoding the link between the MC and the RC are more widespread than elements encoding the other parameters. This fact, applied to simple relative elements, would yield the following corollary hierarchy: RPAR > SRE > RPRO This corollary hierarchy suggests that relative particles are preferably turned to as they encode the minimum information required, i.e. that the RC is linked to the MC. Instead, the ‘standardness’ hierarchy for simple element introduced above predicts that elements that encode only the link between the MC and the RC are dispreferred if compared to elements that encode the link and other parameters. So, it appears that the factors underlying the selection of relativization strategies in standard varieties diverge from the functional and discourse factors on which speakers rely in the selection of relative constructions. This fact has a consequence on a theoretical level, too. As was pointed out in the Introduction, typological studies mostly gather data from standard varieties. Hence, in these studies the relative pronoun strategy turns out to be the default strategy in European languages: see for instance Haspelmath (2001: 1494), Comrie and Kuteva (2005), which was commented in 4.4, or the overview table in Cristofaro and Giacalone Ramat (2007: 88–91). However, if we compare Table 4.10 with Table 5.1, we will notice that the relative pronoun strategy can be considered as default only if the focus is limited to standard varieties; if non-standard varieties are

A scale of standardness for relative constructions in European languages

285

also taken into account, relative pronouns turn out to be at least as default as relative particles, specialized relative elements and, considering combined elements, the combination of a particle with a resumptive element. In this sense, extending the database to non-standard varieties makes the typological representation of European languages different than it is traditionally depicted. 5.2.3. Factors influencing the position of languages on the scale Finally, we may also consider each row in Table 5.1 and see how relative constructions attested in individual languages rank on the scale of standardness. The factors conditioning the position of a language with respect to the range of relative constructions regarded as standard have been already mentioned passim in the previous subsections. They can be summarized as follows: − the influence of a prestigious exoglossic model on the formation of the standard; − the influence of a dialectal koiné on the formation of the standard; − the time period when a language was standardized or re-standardized; − the distinction between strategies considered ‘autochthonous’ and those considered ‘borrowed’. As for the first factor, it seems to have been particularly active in languages that received an early codification, like Dutch, French, German, Italian, Portuguese and Spanish. Here, the Latin model probably contributed to the exclusion of all strategies different from the relative pronoun. 23 Also Estonian, Lithuanian and Latvian received a relatively early codification in the 17th century: here, the German influence must have played a grater role (Druviete 20062: 1867–1868). Instead, languages that were codified more recently, like Albanian, Catalan and Macedonian (all of them in the 20th century), show greater tolerance towards the positions situated in the middle of the scale of standardness. Similarly, languages that have been recently re-codified or languages in which a dialectal koiné played a central role in the formation of the standard are more tolerant towards the right segment of the scale of standardness. This is the case with Slovak, Slovenian, Macedonian and Ukrainian (Lubaś 20062: 1841– 1842; Neweklowsky 20062: 1825, 1830–1831; Schweier 20065: 107–108). The third factor played a role also in Belarusian and Greek. Belarusian was first codified between the 14th and the 17th century. Still, in the following centuries Belarusian local varieties were roofed by exoglossic lan-

286 Sociolinguistic issues guages, i.e. Polish and Russian. After an attempt to re-codify the language in the 1920s–1930s, Russian became once again the roofing language. Nowadays there have been further attempts to re-codify the language as an independent Ausbau variety; still, the standardization process cannot be regarded as complete (Wexler 1992). For this reason, the acceptance of the strategy ‘RPAR+resumptive element’ is ambiguous: it is not accepted in Russian – which serves as a model language for Belarusian – but it is attested in dialects. Atrachovič (1966: 609) qualified this construction as “colloquial”. The Greek situation has been already discussed above: the language underwent a process of re-standardization after 1976, when the prestigious standard, katharévusa, was replaced with the standard based on a dialectal koiné, dhimotikí (Trudgill and Schleier 20062: 1882). This implied that the relativization strategies of dhimotikí, which extend to the rightmost position of the scale of standardness, have been ever since regarded as standard: as a result, Greek is the language whose relativization strategies rank the lowest on the standardness scale (Holton, Mackridge, and Philippaki-Warburton 1997: 443–444). Czech represents an interesting case. Its first codification was carried out by Jan Hus in the 15th century; then, until the 19th century the roofing language was German. In the 19th century Josef Dobrovský attempted a recodification of the language: still, his proposal was based on a translation of the Bible from the 16th century and was consequently strongly archaising. So, the present-day Czech standard (spisovná čeština), which is based on Dobrovský’s proposal, shows little acceptance of strategies different from the relative pronoun. Instead, the common variety based on a dialectal koiné, obecná čeština – which is not regarded as standard, though – shows higher acceptance of the strategies situated in the middle of the scale of standardness (Lubaś 20062: 1839; Petr 1987: 528). Bulgarian underwent a similar process: the attempt of codification in the 13th and 14th century was wiped out by the Ottoman invasion in the 15th century. After independence in 1878 a new standard was codified, which, unlike Czech, partly considered the changes that local and regional varieties had undergone in the meantime (Gutschmidt 2002: 230–232). The case of Czech and Bulgarian shows that the varying acceptance of some strategies – typically, the combination of a relative particle with a resumptive element – depends on ongoing sociolinguistic processes: a standard variety that tends to be archaising and restrictive may be influenced or not by less restrictive non-standard varieties, what can lead to partial re-standardization of the language under the pressure of linguistic use and to the inclusion of new linguistic variants into the standard variety.

A scale of standardness for relative constructions in European languages

287

Finally, the fourth factor played a central role in the codification of Turkish. Until World War I the roofing variety for Turkish local variety was Ottoman, a language that had been subject to strong Persian and Arabic influences. After the language reform of 1928 by Kemal Atatürk a new standard was introduced, which was based on the vernacular of cultivated Istanbul Turks. Atatürk’s main purpose was to try and eliminate all external influences – not only in the realm of lexis, but also in the realm of grammar (Schlyter 20062: 1907). The rejection of the strategy with the relative particle ki can be related to this puristic wave. Similarly, in Basque “relative clauses constructed with relative pronouns occur in the written tradition of all dialects since the beginning of Basque literature in the 16th century” (Oyharçabal 2003: 809): however, between the 19th and the 20th century this strategy was rejected as non-autochthonous, so that its use has strongly decreased ever since and is nowadays “far from being as wide as it was until around 1880” (Oyharçabal 2003: 809). In the whole area where Lower and Upper Sorbian are spoken, a strategy ‘RPRO+particle’ is attested – still, the relative pronoun does not display the Slavic root *kter-/ktor-, but overlaps with the demonstrative pronoun (LSO ten, USO tón ), as the examples below show. (5.11) Źo jo ten luź, [tom ak ja LSO where is DET man RPRO.DAT.M.SG RPAR I som cora pjenjeze dała]? am yesterday money given ‘Where is the man I gave the money to yesterday?’ (Faßke 1996: 170) (5.12) Su to te holcy [z tymi kiž USO are this DET boys with RPRO.INS.PL RPAR sym ja do šule khodźił]? am I to school gone ‘Are they the boys who I went to school with?’ (Faßke 1996: 172) This strategy probably originated from language contact with German local varieties spoken in the same area or in neighbouring regions: in fact, in these varieties – like in standard German – the demonstrative der overlaps with the relative pronoun der; additionally, it can be combined with a particle – de in the Upper Saxon dialect of Leipzig (Fleischer 2004a: 219, 2004b: 66). See the example below.

288 Sociolinguistic issues (5.13) Der Mann, [dem de das Haus GER DET man RPRO.DAT.M.SG PAR DET house gehört.] belong ‘The man to whom the house belong.’ (Fleischer 2004b: 66) In neither Lower nor Upper Sorbian is this strategy regarded as standard: on the one hand, the particle is probably felt to be redundant; on the other, this strategy is the result of language contact and is by no means autochthonous (in no Slavic language can demonstrative pronouns also be used as relative pronouns). A similar case is attested in English. Here, the combination of a relative particle with a resumptive pronoun had been attested since the 9 th century (Roma 2007: 265–266). However, during the process of codification it was excluded from the standard both because it was not attested in Latin, the prestigious exoglossic language, and because it was regarded as a Celtic borrowing, i.e. a borrowing from a non-prestigious variety (Roma 2007: 284). As a result, it is nowadays attested only in dialects (Herrmann 2005: 70–72). If we consider now the top-bottom distribution of the languages in the first column of Table 5.1, we may try and order them by language family or language group, so as to obtain the overview reported in Table 5.4 below. As the table shows, the position of the constructions attested in individual languages on the scale of standardness is mostly only loosely or not at all connected with genetic or areal relatedness. For instance, relativization strategies in Slavic languages may rank very high (Russian) or in the middle (Czech, Polish, etc.) or very low (Slovak, Slovenian, Ukrainian) on the scale, irrespective of the branch (western, eastern, southern) that these languages belong to. Relativization strategies in Balkan languages rank both in the middle (Albanian, Bulgarian, Rumanian) and in the low segment of the scale (Macedonian, Greek). On the other hand, relativization strategies in the majority of Romance languages as well as in Baltic languages rank quite high, while Germanic languages are split into continental Germanic, where relativization strategies rank quite high, and Scandinavian languages plus English, where they rank rather low. Even if there are similarities between genetically related languages, they are not connected with typological proximity, but probably with a similar development of the standard variety. Distinctions between standard and non-standard relativization strategies are inherently language-specific: for instance, all Slavic languages exhibit very similar strategies, but their status varies from lan-

A scale of standardness for relative constructions in European languages

289

guage to language. However, it is possible that different languages resort to the same exoglossic model or that areally close languages may influence each other: this could explain why in some genetically related languages relativization strategies rank together on the scale. Table 5.4. Top-bottom distribution of European languages on the scale of standardness (with respect to the range of constructions regarded as standard). LANGUAGE FAMILY/AREA non-Indo-European Baltic non-Indo-European (continental) Germanic Romance Slavic Romance Balkan Slavic Romance Slavic (Scandinavian and insular) Germanic Slavic Balkan peripheral Indo-European non-Indo-European

LANGUAGES EST, HUN, FIN LIT, LTV TUR DUT, GER FRE RUS ITA, POR, SPA ALB, BUL, RUM BLR, CZE, POL, USO CAT BCS, LSO DAN, ENG, NOR, SWE SLK, SLN, UKR GRE, MAC ICE, IRI BAS, MAL

n 3 2 1 2 1 1 3 3 4 1 2 4 3 2 2 2

high ranking

low ranking

Once more, this fact has consequences on a theoretical level. In typological investigations based on reference grammars only standard relative constructions are taken into account. As mentioned above, even closely related languages may behave quite differently for what concerns the acceptance of relativization strategies. So, if only the constructions attested in the standard varieties are included, misleading conclusions may be reached – for instance, it may be argued that some relativization strategies are attested only in some of the languages in a language family. Instead, an allencompassing analysis covering both standard and non-standard varieties often reveals that the same structures are present in all languages taken into consideration, but that some of them differ in terms of their sociolinguistic status.

290 Sociolinguistic issues This is for instance the case with the strategies of the relative particle and of the relative particle with a resumptive element in Slavic languages. Even in recent cross-linguistic contributions (Sussex and Cubberley 2006: 385–386; Cristofaro and Giacalone Ramat 2007; Fiorentino 2007) these strategies are not said to be attested in all Slavic languages. This is probably because non-standard varieties are considered only partially. However, in other studies, which consider the whole diasystem of Slavic languages, the occurrence of these strategies is always mentioned (Urbańczyk 1939: 27–28; Bauer 1967: 301; Křížková 1970: 24–27; Beličová and Sedláček 1990: 169). 5.3. Standard, non-standard and speakers’ choices The typological analysis conducted in Chapter 4 revealed that encoding of the link between the MC and the RC seems to be the key feature characterizing relativization strategies attested in European non-standard varieties. Instead, the sociolinguistic analysis carried out in this chapter uncovered a strict correlation between the Ausbau character of European standard varieties and their need for explicitness – both from a formal (morphosyntactic) and from a functional (semantic, pragmatic) point of view. Now, since both groups of relative constructions can be accounted for by means of the same functional principles (cf. 4.5), one may query how speakers select a construction instead of another, i.e. which factors actually influence speakers’ choices. According to the tenets of integrative functionalism, each speaker chooses a variant on the basis of a number of sociolinguistic factors related to the communicative situation. That is, the selection of a linguistic construction is carried out considering the context in which the speaker operates. Hence, the first group of factors related to the selection of linguistic constructions has to do with the communicative practice in which one or more speakers are involved: these factors include the communicative situation (the time and place in which the communicative practice is set, the relationship between the speakers, etc.), its oral vs. written character – both on the medial and on the conceptional dimension – and the linguistic variety or varieties (commonly) adopted in this communicative practice (cf. Koch and Oesterreicher 1990: 8–12). A further group of factors is related to speakers themselves. Not all speakers display the same linguistic repertoire, defined as the variety or the varieties that individual speakers can rely on for communicating with other

Standard, non-standard and speakers’ choices

291

speakers. Moreover, speakers may have a varying degree of mastery 24 of the varieties attested in their linguistic repertoire. Also, speakers may have a different degree of awareness of the varieties that they master: this awareness relies on sociolinguistic factors, like schooling, social environment, etc. Awareness of the existence of a plurality of varieties may for instance prevent the speakers from mixing up constructions belonging to different varieties, e.g. dialectal with non-dialectal constructions; vice versa, it may also favour a conscious mixing-up of different varieties in order to attain particular linguistic effects. However, this is an extremely delicate claim and would surely deserve further investigation: in fact, although a speaker may be aware of the difference between linguistic varieties, this does not necessarily entail that s/he will not unconsciously mix them up (see for instance Berruto 2005). Other factors related to speakers are their communicative intention and the functions that they want to convey by means of the linguistic construction that they choose. Table 5.5. Factors influencing speaker’s choice of linguistic variants. FACTORS RELATED TO THE COMMUNICATIVE PRACTICE – communicative situation (time, place, relationship between speakers) – (medial and conceptional) orality vs. scripturality – variety/varieties (typically) adopted

FACTORS RELATED TO THE SPEAKERS

– communicative intentions – communicative functions – linguistic repertoire (variety/varieties mastered)

In Table 5.5 the two sets of factors are summarized and can be compared with each other. As the table shows, the two sets share one factor, i.e. linguistic variety/varieties. This factor has different connotations in the two sets: on the one hand, communicative practices generally require the adoption of particular varieties – typically, the varieties that have been conventionalized for the relevant communicative practice: in fact, we need to bear in mind that communicative practices are “gesellschaftlich herausgebildete konventionalisierte Verfahren zur Bearbeitung rekurrenter kommunikativer Zwecke” (Fiehler et al. 2004: 16), that is to say, they are socially established. They arose out of recurrent forms of linguistic behaviour in a community of speakers, but exist at the same time independently of individual speakers: not all speakers need to have experience of or to master all kinds of communicative practices attested in a community of speakers. On the

292 Sociolinguistic issues other hand, a speaker’s repertoire can consist of several varieties: so, their choice of linguistic variants oscillates between these two poles, i.e. the constraints imposed by the communicative practice in terms of the most appropriate language varieties and the subjective limits of individual linguistic repertoires. This is in line with what is argued in Croft (1995): Most speakers have most variants as part of their grammatical competence, and employ them differently depending on their social position in the speech community and the circumstance of the conversational interaction. […] [T]he sistematicity of an individual’s variable grammar is tied to the individual’s interactions in different social settings […]. (Croft 1995: 518)

Between these two poles speakers can move freely: in particular, speakers who possess several linguistic varieties in their repertoire have a grater possibility of selecting the linguistic variety that best matches the requirements of a particular communicative practice. Once a linguistic variety has been selected, speakers can rely on the constructions attested in this variety to achieve their communicative intentions and, concretely, to encode linguistic functions. For sure, it must be pointed out that the selection of a linguistic variety does not necessarily need to be conscious: the more confident a speaker feels about the use of a communicative practice, the more automatically s/he is likely to select the appropriate variety. Now, how does all this apply to relative constructions in general and to the distinction between standard and non-standard relative constructions in particular? As we saw, relative constructions can be looked upon as linguistic variants and can be attested in different linguistic varieties: only in the standard variety, only in non-standard varieties, or in both. Moreover, in Chapter 4 it emerged that relative constructions occurring in non-standard varieties sometimes encode particular functions, which are differently encoded in standard varieties. For instance, the postnominal strategy introduced by a relative particle is best suitable for afterthoughts (as in Turkish, see 4.1.1), for multiple embedding, to build chains of RCs (see 4.4), or again to link the MC and the RC leaving the syntactic role of the qualified participant unexpressed (see 4.2.2, 4.2.3 and 4.3.3); the correlative strategy is used to introduce and qualify a referent in the preposed RC and to comment on it in the following MC (cf. 4.1.2); the postposed strategy in Dutch and German serves to “signal the importance of the post-modifying clause explicitly” (Weinert forthc.), cf. 4.1.3. In this sense, speakers could select a relative construction instead of another on the basis of the function that they want to convey.

Standard, non-standard and speakers’ choices

293

Beside form-function correlation, further criteria for the selection of relative constructions may be related to grammar-internal factors like the degree of accessibility of the relativized item. This issue is covered in Ariel (1999). Ariel looks for a cross-linguistically valid cognitive explanation for the encoding vs. non-encoding of the syntactic role of the relativized item in the RC (‘zero’ vs. ‘resumptive pronoun’) and argues that this explanation is to be found in the degree of accessibility of the relativized item in the RC to the speaker. Ariel proposes an extended definition of ‘accessibility’ as the concomitant interaction of different factors related to the syntactic-semantic environment of the RC. She identifies five factors: the position relativized, the distance between the head and the relativized item in the RC, the length of the head NP, the degree to which the relativized item in the RC is obligatory, i.e. whether it is an argument of the verb, and the restrictive vs. non-restrictive character of the RC (Ariel 1999: 233). On the basis of a corpus-analysis conducted on a Hebrew spoken corpus, Ariel argues that considering each factor on its own is not insightful enough to justify speakers’ choices; instead, considering the co-occurrence of the five factors, solid correlations can be established between the accessibility of the referent and speakers’ preference towards ‘zero’ or ‘resumptive pronoun’ (Ariel 1999: 238). Still, Ariel does not take into account the existence of linguistic variants. As Croft puts it, “[v]ariation occurs in the speech of single individuals, and hence must be a part of a speaker’s knowledge of the language” (Croft 1995: 518). So, it may well happen that ceteris paribus – i.e. in the same syntactic-semantic environment and in order to convey the same function – two speakers select two different relative constructions belonging to two different linguistic varieties. If this is so, one may argue that speakers operate a two-step selection: first, they select the linguistic variety that they consider more appropriate for the communicative practice in which they are involved; then, they choose among the relative constructions attested in the relevant linguistic variety according to the function that they want to encode or to the syntactic-semantic environment. If a construction encoding a particular function is not attested in the selected variety, alternative constructions are relied on. Construction (5.14) below is considered as non-standard, that is, it is attested in sociolinguistically marked varieties. As such, it is very likely to occur in communicative practices with a low degree of formality, like a chat between relatives, friends or colleagues.

294 Sociolinguistic issues (5.14) l’ unica persona [che l’ incarico di ITA DET only person RPAR DET assignment of telefonar-gli mi emoziona] ring.up.INF-CL.DAT.3SG me excites ‘The only person that I’m excited about having the assignment of ringing him/her up.’ (Fiorentino 1999: 170) Now, if the speaker needed to express the same relation between SoAs within a different communicative practice – for instance, in a formal conversation or in a formal written text – s/he may exclude the construction reported in (5.14) as it is not attested in the standard variety, i.e. the variety that s/he would be expected to adopt in a similar communicative practice. In the Italian standard variety only relative pronouns can relativize lower positions of the Accessibility Hierarchy. So, the speaker should rely on the following construction. (5.15) ?l’ unica persona [l’ incarico di ITA DET only person DET assignment of telefonare alla quale mi emoziona] ring.up.INF to.RPRO.F.SG me excites ‘The only person that I’m excited about having the assignment of ringing him/her up.’ (Fiorentino 1999: 170) Still, as Fiorentino (1999: 170) herself admits, this sentence would be hardly uttered or written, since it displays a high degree of complexity: besides multiple embedding, the relative pronoun is not clause initial, because it relativizes the IO-argument of a clause dependent on the RC. Probably, the speaker would rely on an alternative construction, for instance the following one. (5.16) (è) solo con questa persona (che) mi ITA is only with this person COMP to.me capita di emozionar-mi, se/quando ricevo happens to excite.INF-REFL if/when I.receive l’ incarico di telefonar-le DET assignment to ring.up.INF-CL.DAT.F.3SG ‘Only with this person I happen to be excited when I receive the assignment of ringing him/her up.’

Standard, non-standard and speakers’ choices

295

As mentioned above, the possibility of encoding a particular function through a relative construction attested in the variety adopted by a speaker is by far not the only criterion guiding the choice of a relative construction instead of another. Other criteria, like those identified by Ariel (1999), should be considered, too. Still, these criteria need to be first tested on a language-specific level, so as to attain a representative picture of the contexts in which constructions acting as linguistic variants within a language are more frequently attested. This is what will be attempted in the third part of the work, in Chapter 7 and Chapter 8. Chapter 7 will demonstrate that the postposed strategy in German is particularly frequent when the N serving as a referent has a hyperonymous character (like ‘people’, ‘things’) and that the postnominal construction introduced by the relative particle wo is attested more frequently for linking an MC with an RC without specifying the syntactic role of the relativized item. In Chapter 8, the corpusanalysis will show that zero-marked relatives (asyndetic clause clusters, cf. 3.5.3) are more frequent in non-restrictive relations. Still, in the same chapter it will also be argued that RCs introduced by the zero-marker occur most frequently in the speech of those informants that rely on the varieties razgovornaja reč’ (colloquial Russian) and prostorečie, a lower urban sociolect. Speakers relying on literaturnyj jazyk (standard Russian) use the postnominal strategy with the relative pronoun kotoryj to encode both kinds of relations, restrictive and non-restrictive. In this respect, speakers’ choices are consistent with the linguistic variety adopted and with the relative constructions attested in that variety. We are now left with the opposition of standard vs. non-standard: as mentioned in 1.3.2, this dichotomy was introduced in order to preliminarily delimit the constructions under investigation in this work. Now, after reviewing the typological and sociolinguistic peculiarities of non-standard constructions compared to their standard counterparts, it has emerged that constructions can be variety-bound, i.e. they can be endowed with additional information on the linguistic variety or varieties in which they are most likely to occur. Consequently, the distinction between standard and non-standard has been useful for highlighting a number of constructions which are typically less considered in typological studies. Still, the description of these constructions can be regarded as an initial step towards a full-fledged account of relative constructions: after reviewing the constructions attested in all varieties of a language, it seems appropriate to study their overall incidence, so as to bring to light correlations between the constructions and the varieties in which they occur. In this sense, it is better to abandon the strict ‘standard vs. non-standard’ dichotomy in favour

296 Sociolinguistic issues of a comprehensive view of languages as diasystems composed of several interrelated linguistic varieties. In the light of what has been discussed in this section, Ariel’s conclusion that grammar-internal “[d]efault conventions specify what choice speakers will consider most automatically, what they would tend to choose unless strong indications against it exists” (Ariel 1999: 257) must be further specified: speakers operate a choice on the basis of default conventions within the linguistic variants – in this case, relative constructions – that they think are suitable or can be activated inside the communicative practice in which they are involved.

Chapter 6 Diachronic issues

In Chapter 5, reference was repeatedly made to the development of the standard variety in European languages in order to account for the presentday linguistic status of relative constructions, i.e. why some constructions are regarded as standard whereas others are not. This suggests that the sociolinguistic and the diachronic perspective are closely related and that the latter can help explain the former. So, in this chapter emphasis will be placed on two diachronic issues: 1. Non-standard relative constructions are not innovations of sociolinguistically marked varieties; they are historically attested on a par with constructions that would later be regarded as standard. Evidence for this claim will be provided in sections 6.1 to 6.4. 2. The development of the opposition between standard and non-standard relativization strategies can be described by means of a diachronic model, based on the model proposed in Auer (2005b). This issue will be discussed in 6.5. Before beginning the diachronic analysis, two methodological remarks are in order. First, in this chapter earlier stages of a language will be referred to by means of an ‘o’ before the three-letter code of a language. For instance, ‘oGER’ refers to earlier linguistic stages of German. Still, the symbol ‘o’ has nothing to do with language-specific historical periodization: ‘oGER’ includes Old High/Low German, Middle High/Low German and New High/Low German. That is, ‘o’ generally refers to the time span prior to linguistic codification, which for most of the languages in the sample took place between the 18th and the 20th century. So, ‘historical’ examples, i.e. all examples marked with ‘o’, range from the most ancient linguistic documents to the 17th century. Second, evidence in this chapter will be provided from historical data. As was mentioned in 2.3.8, this work is not philologically oriented, so that the validity of historical data provided by philologists is taken for granted. When possible, indications will be given on the kind of source (e.g. ‘lay’ vs. ‘sacred’) and the century in which the text was written; however, no details about the compilers of the text or the variety used will be given. Moreover, historical data was not equally accessible for all the languages in the sample. Hence, evidence could not be provided for all of them. Still,

298 Diachronic issues when an example can be seen as representative also for languages that are not directly documented, this will be pointed out. This procedure may have two drawbacks. 1. For the most part, the examples found in the sources do not allow researchers to carry out quantitative studies. Hence, it is not always possible to determine the frequency of the constructions attested in earlier linguistic stages; 2. It often happens that a construction is attested only in one source: in this respect, its validity is questionable. However, as stated in 2.3.1, even one single occurrence is looked upon as positive linguistic evidence: that is, this construction was regarded as historically attested in the relevant language. To summarize, the conclusions which will be reached in this chapter based on historical evidence should in principle be regarded as provisional. In fact, it is possible that unusual or rare constructions actually turn out to be the result of errors in writing/copying texts by compilers or in interpreting them by philologists who edited them. This can be brought to light only by means of philological procedures like the collation of various manuscripts, something that would go beyond the scope of the present work: it may be the task of further language-specific studies to check and possibly challenge the statements made in this chapter. 6.1. Non-standard relative constructions: diachronic evidence In this section and in the next ones diachronic evidence for present-day non-standard relative constructions will be provided. As mentioned above, this is aimed to show that these constructions do not constitute innovations of sociolinguistically marked varieties that for some reason deviate from the constructions attested in standard language, but that they have co-existed for centuries with relative constructions that are nowadays regarded as standard (in this regard cf. also Fiorentino 2007: 282–284). The necessity and usefulness of investigating constructions found in non-standard varieties is highlighted also in Guiraud (1966), Sandig (1973) and Albrecht (2003). After reviewing a number of morphosyntactic phenomena usually considered non-standard, Albrecht concludes:

Non-standard relative constructions: diachronic evidence

299

Alle diese Beispiele zeigen, dass viele Substandardphänomene, die den Puristen als barbarische Erneuerungen […] erscheinen, schon vor der Herausbildung der Standardvarietät vorhanden waren und später außerhalb der Norm tradiert wurden. (Albrecht 2003: 25)

That is, the investigation of non-standard varieties is likely to bring to light patterns of persistence – or renewal – of linguistic constructions: if some constructions attested in past stages of a language do not occur in presentday standard varieties, it cannot be automatically assumed that these constructions are archaic or extinct. An all-encompassing check of the varieties in the individual languages should be carried out in order to make claims about the development of linguistic constructions and about ongoing syntactic changes. It needs to be pointed out that this chapter is not directly concerned with the rise and development of relative constructions: the grammaticalization paths leading to the formation of RCs have been already described from a typological perspective in Lehmann (1979, 1984: Ch. VI, 1995) and Heine and Kuteva (2006: Ch. 6, 2007: 224–229). As far as European languages are concerned, Lehmann and Heine and Kuteva describe three grammaticalization channels: the interrogative, the demonstrative and the participial one. These are briefly exemplified below. The interrogative channel is where the correlative strategy first originates. The first step consists in interpreting an interrogative pronoun as an indefinite one, so that the whole construction can be regarded as encoding a conditional relation: see the multiple translations of (6.1). (6.1) LAT

[Qui RPRO.NOM.M.SG

ager field

frigidior colder Licinum Licinian

et macrior and leaner seri be.planted.INF

erit,] ibi oleam will.be there olive.tree oportet. is.necessary ‘If there’s a colder and leaner field, the Licinian olive tree must be planted there.’ ‘Any field that is colder and leaner, there the Licinan olive tree must be planted.’ ‘In a/the field that is colder and leaner, there the Licinian olive tree must be planted.’ lit. ‘Which field is colder and leaner, there the Licinian olive tree must be planted.’ (Cat. agr. 6, 2; quoted in Ramat 1984: 411)

300 Diachronic issues Gradually, the pronoun loses its indefinite value and grammaticalizes to a relative pronoun, encoding the four parameters (link, syntactic role, gender and number agreement). At the same time, the RC becomes postnominal. (6.2) LAT

Mulier [quae se suamque aetatem woman RPRO.NOM.F.SG REFL her.and age spernit,] speculo ei usus est. despises mirror to.her use is ‘A woman who doesn’t care about herself and her age needs a mirror.’ (Pl. Mos. 250; quoted in Lehmann 1979: 12)

The demonstrative channel is based on the juxtaposition of two clauses. The second clause contains a demonstrative pronoun referring to a participant of the SoA encoded in the first clause. This pronoun gradually loses its textual-deictic character and grammaticalizes to a relative pronoun. The demonstrative channel gives rise first to postposed and then to postnominal relative constructions. (6.3) Ingegin liofun imo zehen man riobe, oGER towards ran to.him ten men leprous [thie gistuontun ferro.] those/RPRO.NOM.M.PL stayed far ‘There ran towards him ten lepers, and they kept their distance.’ ‘There ran towards him ten lepers, who kept their distance.’ (9th century, Tat. 111, 1; quoted in Lehmann 1984: 380) (6.4) bl[agoslove]nъ g[ospod]ь, [iže ne oCSL blessed Lord he.PAR/RPRO.NOM.M.SG not dastъ nasъ vъ lovitvǫ zǫbomъ ichъ] will.give us in prey teeth.to their ‘Praise be to the Lord, he has not let us be torn by their teeth.’ ‘Praise be to the Lord, who has not let us be torn by their teeth.’ (11th century, Bible, Ps. 123, 6; quoted in Večerka 2002: 197) In (6.4) and in the following, examples from old Slavic languages are quoted as they appear in the source, so we may come across different transliteration systems: for instance, the Cyrillic letters ‘ъ’ and ‘ь’ are often used to transliterate the original ultra-short /u/ and ultra-short /i/ of Old Slavic, so-called jer” and jer’, as it is the case in (6.4). In examples that

Non-standard relative constructions: diachronic evidence

301

have not been already transliterated in the source, ‘ъ’ and ‘ь’ will be rendered through ” and ’ respectively. As the multiple translations of (6.3) and (6.4) demonstrate, the elements in bold can in principle be interpreted both as demonstrative and as relative pronouns. There are also instances of clause juxtaposition with no demonstrative pronoun, which can be regarded as instances of postnominal RCs introduced by the zero-marker. See the example below. (6.5) mit themo brunnen [Ø thu nu quist] oGER with DET fountain you now speak ‘With the fountain about which you’re now talking about.’ (9th century, Otf. 2, 14, 44; quoted in Lehmann 1984: 379) Finally, the participial channel is based on non-finite verbal forms, which can appear either before or after the referent, giving rise to prenominal or postnominal strategies, as shown in (6.6) and (6.7) respectively. (6.6) OTK

[közi körmäz] kiši eye.POSS.3SG see.not.PTCP person ‘A person whose eyes cannot see.’ (Erdal 2004: 439)

(6.7) CSL

ибо лоучши есть единъ творѧй волю гсдню, нежели тмы законопрестоупных ibo lučši est’ edin” [tvoręj because better is one do.PTCP.PRS.NOM.M.SG volju g(o)s(po)dnju,] neželi tmy zakonoprestupnych will divine than lots criminals.of ‘For one man who does God’s will is better than a plethora of criminals.’ (17th century, sacred text, Rabus 2008: 322)

Both Lehmann and Heine and Kuteva start their analysis from remote linguistic stages in order to see how constructions developed across time, giving rise to the constructions that are attested nowadays. Instead, in this chapter we will move in the opposite direction: we will start from the relative constructions attested in present-day non-standard varieties and look for diachronic evidence. This is because the overall purpose is not to examine how these constructions arose, but only to show that they were attested in earlier linguistic stages and that their structure has not or only minimally changed ever since.

302 Diachronic issues In the next sections data will be once more structured according to the three parameters used to classify relative constructions in European languages: word order (section 6.2), relative element (section 6.3) and syntactic positions relativized (section 6.4). 6.2. Word order 6.2.1. Turkish ki and the postnominal strategy in Basque In 3.5.2 the status of the Turkish particle ki was discussed and its functions as an emphatic particle and as a connector were illustrated. As stated in Auer (1990: 281), even Turkish native speakers cannot always distinguish between the two functions. For instance, in the following clause it is not clear whether ki belongs to the first or to the second clause. (6.8) TUR

İnanıyorum ki herşey düzelecek. I.believe CONN everything will.come.right ‘I believe that everything is going to be all right.’ (Göksel and Kerslake 2005: 112)

As stated in 5.1.1 and 5.2.3, the use of ki to encode relative relations is excluded from present-day standard Turkish on account of purism: ki is regarded as a Persian borrowing. Still, diachronic evidence reveals that the use of a subordinating connector dates back to the first documents written in Old Turkic. Old Turkic (OTK) is the earliest attested Turkic language. Although the oldest documents in Old Turkic were compiled in Central Asia, there is nearly uninterrupted continuity between Old Turkic and present-day Turkish: Old Turkic is regarded here as the direct ancestor of present-day Turkish. Long before intensive contact with Persian, kim – which in Old Turkic and modern Turkish is (also) an interrogative pronoun meaning ‘who’ – was used as a connector in Old Turkic, both in relative and in other subordination relations, as the examples below show (Erdal 2004: 483–494). (6.9) OTK

bu

söz [kim sen aydın] word RPAR you said ‘This word that you said.’ (11th century, Slobin 1986: 281) DEM

Word order 303

(6.10) bo kutsuz kovï tïnlïglar üčün OTK DEM unhappy wretched creatures for tärs biligin ayïg kïlïnčïn ketärmäk alïn false knowledge bad action raise.INF method čavïšïn yarlïkazun, täŋrim, kim ukzun means he.may.order god PAR may.understand bilzünlär may.know.they ‘May he, my lord, for the sake of these unhappy and wretched creatures tell us the means to remove their heresies and sins so that they may understand and know.’ (Erdal 2004: 493) (6.9) is an instance of the postnominal strategy, where kim acts as a relative particle; in (6.10) it introduces a clause with a final meaning, like ki in example (3.121). As Erdal (2004: 436–438) remarks, in Old Turkic the postnominal strategy with kim was used to encode relative relations along with the prenominal strategy containing a non-finite verb form as in example (6.6) above. Hence, the influence of PER ke/ki probably simply reinforced the use of ki(m) as a connector. In texts written from the beginning of the 15 th century onwards, clauses introduced by ki(m) increase exponentially. In these texts all of the uses attested in present-day Turkish can be found (Hanser 1973/1974: 176–197). In (6.11) a postnominal strategy introduced by kim is used to encode a relative relation. (6.11) Bu mezarlar-[kim görürsiz] cemī‘si oTUR DEM graves-RPAR you.see in.totality resūlullāhuŋ aṣḥābından-dur. God’s.messenger.of comrades.POSS.3SG.from-are ‘All these graves that you see are of the comrades of the messenger of God.’ (15th century, chronicle, Hanser 1973/1974: 193) The above example displays strong external influence: all lexical elements, apart from görürsiz, are of Arabic origin and are related to the Islamic religion. Instead, all function words (demonstrative pronoun, case ending, copula) are Turkish. The only exception is aṣḥāb, which is the plural form of Arabic ṣāḥib ‘comrade’. Here, the plural was not formed by means of the suffix -lAr, as in mezarlar (mezar being itself an Arabic word), but through the Arabic vowel template (so-called ‘broken plural’).

304 Diachronic issues In some cases, ki(m) acts as an enclitic particle, generally serving as a connector between autonomous MCs. (6.12) ayıtdı-ki yağı šuŋa dirler-ki biz anlara OTK he.said-CONN enemy DEM.to they.say-CONN we DEM.PL.to yitsevüz öldürürüz anlar bize if.we.can.compete we.kill DEM.PL to.us yitse öldürür didi if.they.can.compete they.kill he.said ‘[The father] said: one calls enemies such people: if we can compete with them, we kill them, if they can compete with us, they kill us – so he said.’, ‘[The father] said that one calls enemies those that if we can compete with them we kill them and if they can compete with us they kill us – so he said.’, ‘[The father] said that one calls enemies those whom we kill if we can compete with them and who kill us if they can compete with us – so he said.’ (14th century, epic, Hanser 1973/1974: 179) In (6.12), the first ki introduces reported speech whereas the second ki allows for multiple interpretations: it either introduces reported speech (first translation) or acts as a general relative connector, similar to the cases discussed in 4.2.2, 4.2.3 and 4.3.3 (second and third translations). The uses of ki in this example are quite similar to examples (3.120) and (3.123). Hanser proposes an all-encompassing interpretation of ki in old Turkish similar to the one offered in Auer (1990: 282) for its use in present-day Turkish: Kim, ursprünglich Fragepronomen oder Relativum in nuce, erst Bestandteil des Nebensatzes, schließlich zur Partikel abgeschwächt, [tritt] als Ankündigungswort an das Ende des Hauptsatzes. Dies würde nun die Ansicht bestätigen, dass das Türkische […] sich eine unterordnende Konjunktion im Laufe der Zeit so angeglichen hat, dass sie diese zu einer hinweisenden Partikel im Vordersatz eines zugeordneten Nachsatzes parataktisch auflöst. (Hanser 1973/1974: 185)

What is certain is that the relative element ki(m), which nowadays is banned from standard Turkish, has been attested in this language and in its ancestor, Old Turkic, for more than a thousand years. As was briefly mentioned in 5.2.3, the Basque postnominal strategy introduced by a ‘which’-based relative pronoun (zein/zoin) or a specialized

Word order 305

relative element (non ‘where’) “occur[s] in the written tradition of all dialects since the beginning of Basque literature in the 16th century” (Oyharçabal 2003: 809). As Haase (1992: 150–153) points out, this strategy probably arose through language contact with French and Spanish: in fact, it is first attested in translations (Hurch 1989: 21). Although it had been integrated in the system of Basque relative constructions, between the 19th and the 20th it was eventually rejected as non-autochthonous (Oyharçabal 2003: 809). 6.2.2. The correlative strategy The correlative strategy is nowadays attested in a restricted group of languages, including Hungarian, Latvian and Russian. This strategy, too, has been attested for several centuries in some of the languages in the sample: as mentioned in 6.1, it constitutes one of the grammaticalization channels for RCs. For instance, this strategy was attested in all Slavic languages, as the examples below show. In (6.13) and (6.14) from old Polish the correlative strategy relativizes SU and DO respectively. Here and in the following, the grapheme stands for a nasal vowel and corresponds to presentday Polish and . (6.13) [Ktory clouek czyni volø othcza oPOL RPRO.NOM.M.SG man does will father.of mego,] tenczy przydze do krolefstwa nebeskego. my that will.come to reign of.heaven ‘The person who is doing my Father’s will shall come into the reign of the heavens.’ lit. ‘Which person is doing my Father’s will, that one shall come into the reign of the heavens.’ (14th century, sacred text, Vondrák 19282: 485) (6.14) [O khtore penandze Mroczek na mø oPOL about RPRO.ACC.PL money Mroczek on me zalowal,] thym on mne do swanthek accused for.them he to.me until holiday roku dal. deadline gave ‘For the money that Mroczek sued me about, he gave me a deadline until the holiday.’

306 Diachronic issues lit. ‘About which money Mroczek sued me, for them he gave me a deadline until the holiday.’ (15th century, tribunal document, Urbańczyk 1939: 20) In (6.15) from old Czech LOC is relativized. (6.15) [K kterémuž tělu najviece ptáků oCZE to RPRO.DAT.M.SG body most birds.of přiletí,] ten má býti najsvětější. will.fly that has be.INF saintest ‘The body to which the most birds will fly has to be saintest.’ lit. ‘To which body the most birds will fly, that one has to be saintest.’ (15th century, chronicle, Lamprecht, Šlosar, and Bauer 1986: 386) Finally, in (6.16) and (6.17) from old Russian SU is relativized; in (6.18) from old Ukrainian the position relativized is DO. It must be pointed out that when quoting old Russian examples from Troickij (1959, 1968) the orthography used by the scholar was preserved, although these examples have certainly been adapted to the modern Russian orthography – a widespread tradition in Soviet linguistics. In fact, neither is the letter jer” (ъ) attached to words ending with a hard vowel, nor is the letter jat’ (ѣ) used, two peculiarities of Russian orthography which were abolished in 1918 – long after the texts that Troickij investigates were written. Cf. for instance (6.25) below, where the original orthography is preserved. (6.16) И которые воровские казаки на боех поиманы, и те все oRUS кажнены смертью. I [kotorye vorovskie kazaki na and RPRO.NOM.M.PL thievish Kazakhs in boech poimany,] i te vse kažneny battles caught and those all condemned smert’ju. death.with ‘And the thievish Kazakhs who were captured during the battles were all condemned to death.’ lit. ‘And which thievish Kazakhs were captured during the battles, they were all condemned to death.’ (17th century, chronicle, Troickij 1968: 78)

Word order 307

(6.17) А без них де в их домех которые были лошади, и те де лошади oRUS пропали без вести, а иные померли. [A bez nich de v ich domech and without them PAR in their houses kotorye byli lošadi,] i te de lošadi RPRO.NOM.PL were horses and those PAR horses propali bez vesti, a inye pomerli. disappeared without news and others died ‘And the horses that were in their houses without them disappeared leaving no trace, and other horses died.’ lit. ‘And which horses were in their houses without them, those horses disappeared leaving no trace, and other horses died.’ (17th century, chronicle, Troickij 1968: 102) (6.18) И мовитъ, ижъ «которую хату купилемъ у Гаврила Кислого, oUKR жителя харкюского, дарую оную дядкови моему Сахнови». I movit”, iž” «[kotoruju chatu kupilem” and he.says COMP RPRO.ACC.F.SG house I.bought u Havrila Kisloho, žitelja charkjuskoho,] at Havril Kislyj inhabitant of.Charkiv daruju onuju djadkovi moemu Sachnovi». I.give her uncle.to my Sachon ‘And he said: I give the house that I bought from Havril Kislyj, resident in Charkiv, to my uncle Sachon.’ lit. ‘And he said: which house I bought from Havril Kislyj, resident in Charkiv, I give it to my uncle Sachon.’ (17th century, lay text, Mel’nyčuk 1962: 115) So, as examples (6.13) to (6.18) demonstrate, a correlative strategy introduced by a relative pronoun existed for several centuries in Slavic languages, where it was used in different communicative practices, both lay and sacred. Additionally, this strategy was already attested in the most ancient documents of Slavic languages, i.e. codices written in Old Church Slavonic dating back to the 10th–11th century (Trunte 20035: 29–33), although its use in these texts is quite marginal (Večerka 2002: 178–179). As illustrated in 4.1.2, the correlative construction is nowadays attested with certainty only in Russian and Ukrainian non-standard varieties. Incidentally, we may observe that old Russian displays some of the variants of the correlative strategy attested in present-day language. So, (6.17) could be

308 Diachronic issues subsumed to Variant 2, where neither the relative pronoun nor the relativized item in the RC is clause-initial (cf. 4.1.2). In other Slavic languages, the correlative strategy is attested nowadays only in constructions with a high degree of idiomaticity, like proverbs. As such, it constitutes a historical remnant and is no longer actively used. See the examples below from Czech, Slovak and Bulgarian. (6.19) [Který kůň ovsa dobývá,] nejméně CZE RPRO.NOM.M.SG horse oat obtains the.least ho jí. it eats ‘The horse which obtains the fodder eats the least of it.’ (Beličová and Sedláček 1990: 182) (6.20) [Ktorá skala nemá miesta,] tá SLK RPRO.NOM.F.SG rock not.has place that vraj machom neobrastá. allegedly moss.with not.forms ‘A stone that has no place doesn’t get covered by moss.’ (Beličová and Sedláček 1990: 182) (6.21) Който поп чете по-малко, той кади по-много. BUL [Kojto pop čete po-malko,] toj RPRO.M.SG priest reads less that kadi po-mnogo. burns.incense more ‘A priest who doesn’t read much burns a lot of incense.’ (Tilkov 1983: 318) (6.22) Която мама много пее, тя рядко сее. BUL [Kojato mama mnogo pee,] tja rjadko see. RPRO.F.SG mum a.lot sings that rarely seeds ‘A mother that sings a lot seeds rarely.’ (Tilkov 1983: 318) The only cases in which the correlative strategy is still used in Bulgarian are those in which kojto can be regarded as an indefinite pronoun, like in (6.23), where the use of kojto is similar to ENG whichever.

Word order 309

(6.23) Който (и) човек да питаш, ще ти каже. BUL [Kojto (i) čovek da pitaš,] šte RPRO.M.SG PAR man COMP you.ask will ti kaže. to.you he.tells ‘Whichever person you ask will tell you.’ (Maria Manova, p.c.) As has been already shown in 4.1.2, the correlative strategy in Slavic languages can also be introduced by a relative particle. This structure is nowadays attested only in Russian. In the past it was attested also in old Russian, but also in old Polish and old Ukrainian, as may be seen in the following examples. (6.24) [Co jedno było królestw chrześćijańskich,] oPOL RPAR once was kingdoms.of Christian wszytki się na religiej pobudowały. all REFL on religion built ‘All the Christian kingdoms that existed once were based on religion.’ lit. ‘How many Christian kingdoms there were once, all of them were based on religion.’ (16th century, sermon, Urbańczyk 1939: 5) (6.25) А что человекъ мой Истомка Суворовъ сынъ Граборуковъ oRUS бѣгаетъ, и тот человекъ женѣ моеи Соломанидѣ, а которые мои иные люди, и тѣхъ всѣхъ людей отпустить на слободу, а женѣ моеи и сыну до нихъ дѣла нѣтъ. A [čto čelovek” moj Istomka Suvorov” and RPAR man my Istomka Suvorov syn” Graborukov” běgaet,] i tot čelovek son of.Graborukij runs and that man ženě moei Solomanidě, a [kotorye wife.to my Solomanida and RPRO.NOM.M.PL moi inye ljudi,] i těch” vsěch” ljudej otpustit’ my other people and those all people set.INF na slobodu, a ženě moei i synu do in freedom and wife.to my and son.to to nich” děla nět”. them business is.not

310 Diachronic issues ‘My subject Istomka Suvorov, son of Graborukij, who is on the run, has to be given to my wife Solomanida; the other people that are mine have to be set free, and my wife and my son do not have anything to do with them.’ (16th century, Zaliznjak and Padučeva 1975: 75) (6.26) A що подъ горою дерево стоитъ, то не Ясково, але намъ oUKR належало. A [ščo pod” goroju derevo stoit”,] to ne and RPAR under mountain tree stands that not Jaskovo, ale nam” naležalo. Jasek’s but to.us belonged ‘And the tree that is found at the bottom of the mountain didn’t belong to Jasek, but to us.’ (17th century, chronicle, Mel’nyčuk 1962: 113) In (6.24) the relative particle also has a quantitative meaning, as shown by the relativized item in the RC, królestw chrześćijańskich, which is in the genitive. This justifies the second translation. In (6.25) the relative particle is used with a [+animate] relativized item, in (6.26) with a [–animate] one. The relativized item in the MC is encoded through a quantifier in (6.24) and through a demonstrative in (6.26). Instead, in (6.25) the relativized item is encoded with the same degree of explicitness through a full NP in the MC and in the RC. The correlative strategy was attested also in languages that nowadays do not use it any more or use it rarely, like German and Turkish. (6.27) darumb [welcher bawm nit gutte frucht oGER so RPRO.NOM.M.SG tree not good fruit bringt] wird abgehawenn produces is cut.down ‘So the tree that does not produce good fruit is cut down.’ lit. ‘So, which tree does not produce good fruit, is cut down.’ (16th century, Luther Bible, Matth. 3, 10; quoted in Ebert 1993: 450) (6.28) [kayu kiši OTK RPRO.NOM person ol tïnlïg DEM creature

ög kaŋ köŋlin mother father heart tamuluk belonging.to.hell

bertsär,] if.harms bolur becomes

Word order 311

‘Any person who breaks the heart of his parents becomes a candidate for hell.’ ‘If some person breaks the heart of his parents, that creature becomes a candidate for hell.’ lit. ‘Which(ever) person breaks the heart of his parents, that creature becomes a candidate for hell.’ (Erdal 2004: 501) In (6.27) the relativized element is encoded only in the MC; instead, in (6.28) it is encoded both in the MC and in the RC through a full NP, like in (6.25). Still, whereas in (6.25) it is encoded by means of the same noun, in (6.28) the relativized item in the RC and the one in the MC are synonyms (‘person’, ‘creature’). 6.2.3. The postposed strategy The postposed strategy in German is attested also in earlier stages of the language, along with the postnominal strategy. So, in (6.29) we have an instance of integrated V2-clause; in (6.30) we have an intermediate strategy: the RC is verb-second, but it is embedded in the MC. (6.29) inti oGER and [der

Judeon daz der heilant was, Jews COMP DET saviour was dar tēta inan heilan] RPRO.NOM.M.SG PAR did him heal ‘And he told the Jews that it was the saviour who had healed him.’ (9th century, Tat. 88, 5; quoted in Ramat 1984: 400)

(6.30) da oGER then doctor, scholar dem

sagata den he.told DET

fieng hie an ein fast wolgelerter began here began DET quite well-read [der was ain pfaff,] zů RPRO.NOM.M.SG was DET clergyman to ersten mall sant Pauls eppistel in DET first time saint Paul’s epistle in teutsche zů lesen German to read.INF ‘Then a well-read scholar, who was a clergyman, began to read for the first time saint Paul’s epistle in German.’ (16th century, chronicle, Ebert 1993: 445)

312 Diachronic issues According to Ebert (1993: 445), constructions like (6.30) are found mainly in dialogues, sermons and chronicles, but not in chancellery style. 6.3. Relative element In this section simple relative elements will be first covered: relative pronouns (6.3.1), specialized relative elements (6.3.2), relative particles (6.3.3) and the zero-marker (6.3.4). Then, diachronic evidence for combined relative elements will be briefly reviewed (6.3.5). 6.3.1. Relative pronouns In 4.2.1 and 4.2.4 two main tendencies of relative pronouns in non-standard varieties were discussed: on the one hand, relative pronouns can display a reduced paradigm; on the other, if there are irregularities in the morphological forms of relative pronouns or in their applicability to the AH, these inconsistencies tend to be levelled out in non-standard varieties. As in the case of word order, these characteristics do not constitute recent developments, but were already attested in earlier linguistic stages. As described in 4.2.1, in Rumanian non-standard varieties the form care remains uninflected and can be used for all positions of the AH. According to Vulpe (1980: 128), uninflected care has occurred along with inflected care at least since the 16th century – the first Rumanian written documents date back to that time – and it was only with the (relatively late) standardization of the language (19th century) that inflected care was selected as the only standard relative pronoun. This can be related to the fact that Şcoala Ardeleană (the Transylvanian School), the initiator of the standardization process in Rumanian, turned to Latin as a model, striving for the maintenance or reintroduction of Latin features in Rumanian (Iliescu 2002: 164). In old Italian, the form che could be preceded by a preposition to relativize lower positions of the AH, as in (6.31), whereas the form cui could also relativize higher positions of the AH, like in (6.32), where DO is relativized. Both uses are not admitted in present-day standard Italian. (6.31) O figliuola oITA o daughter diavolo [di che devil about RPRO

mia […] my io t’ I to.you

questo this ho have

è il is DET parlato.] spoken

Relative element 313

‘My daughter, this is the devil I told you about.’ (14th century, Boccaccio, Decameron 3, 10; quoted in Giacalone Ramat 2005: 122) (6.32) (la fanciulla) mandò la fante e fece oITA DET young.lady sent DET maid and she.let parlare a colui [cui amava] speak.INF to that RPRO she.loved ‘The young lady sent the maid and let her speak with the man whom she loved.’ (13th century, Novellino 99; quoted in Giacalone Ramat 2005: 131) The manifold forms and combinations of relative elements attested in Italian and, in general, in Romance standard and non-standard varieties – both synchronically and diachronically – are actually best understood if we bear in mind that the paradigm of the Latin relative pronoun had already undergone a formal reduction long before Romance languages came into being: the paradigm of the pronoun qui, quae, quod had collapsed into only three forms, qui for SU-relativization, que(m) for DO-relativization and cui for OBL-relativization; the form que could be generally used with all [–animate] referents and was on its way to becoming a relative particle (Schafroth 1993: 67–72). Similarly, the relative pronoun attested in Old Church Slavonic and in old stages of Slavic languages, iže, jaže, ježe, had developed two uninflected forms, iže and ježe, which could be used for all kinds of referents and be combined with resumptive elements (Večerka 2002: 185–186). (6.33) a bьrьvno [iže vь oče tvoemь] oCSL and beam RPAR in eye your ‘And the beam that is in your eye.’ (10th–11th century, sacred text, Vondrák 19282: 490) (6.34) ne žьrǫ kumiremъ bezdušъnymъ, nъ oCSL not I.sacrifice idols.to soulless.to but bogu svojemu, [ježe vь nebesechъ] god.to my.to RPAR in heaven ‘I don’t make sacrifices to soulless idols, but to my God, who is in heaven.’ (10th–11th century, sacred text, Vondrák 19282: 490)

314 Diachronic issues In (6.33) the (masculine) form iže is used to refer to a neuter noun (bьrьvno), whereas in (6.34) the (neuter) form ježe applies to a masculine referent (bog). Examples of invariable iže/ježe, sometimes combined with resumptive elements, are found in Urbańczyk (1939: 48) for old Polish, Mel’nyčuk (1962: 93) for old Ukrainian, Troickij (1968: 209–211) for old Russian and Mirčev (1955: 126) for old Bulgarian. In a fashion that recalls the development of the complementizer que/che in Romance languages, in some Slavic languages the invariable forms of the relative pronoun iže further grammaticalized to complementizers: iže and ježe became iž, ež and že (Vondrák 19282: 490–495). Večerka (2002: 181) remarks that this process is attested in nuce in Old Church Slavonic, although the syntactic and semantic environment of the relevant text passages does not allow us establish with absolute certainty whether iže/ježe act as complementizers. The form že acts as a complementizer in present-day Czech and Slovak, the forms iż and że have the same complementizer function in Polish, the form až in Lower Sorbian (although the development in Lower Sorbian is not agreed on by all scholars, cf. Vondrák 19282: 495). Gallis provides evidence of the same phenomenon from texts written in the Serbian redaction of Church Slavonic (Gallis 1958: 140): in the 13th century the form jere ( LOC > DO > OBL > IO, TEMP. That is, SU and LOC are the two most frequently relativized positions. Interestingly, this mirrors Bernini’s (1989: 90) results on the use of relative constructions in spoken Italian. As mentioned above, C11 – the zero-marked postnominal construction – was considered separately. Cases where the relative element is the zeromarker were kept separate from cases where the relative element is the combination of the zero-marker with a resumptive pronoun. The distribution of this construction in RRR98 is reported in Table 8.6. Again, restrictive and non-restrictive constructions were analyzed separately: of 49 occurrences, 11 are restrictive and 38 are non-restrictive.

Quantitative analysis 401 Table 8.6. The occurrence of C11 (asyndetic clause clusters) in RRR98. POSITION SU DO OBL LOC Total

Ø 6 1 0 1 8

RESTRICTIVE Ø+res. Total 2 8 0 1 1 1 0 1 3 11

% 72.73 9.09 9.09 9.09 100.00

Ø 4 1 1 3 9

NON-RESTRICTIVE Ø+res. Total % 15 19 50.00 0 1 2.63 1 2 5.27 13 16 42.10 29 38 100.00

The following figures are significant: − On the whole, zero-marked RCs turned out to be less frequent than RCs introduced by a relative element (119 vs. 49). Still, considering individual constructions, asyndetic clause clusters (C11) are the second most frequent construction after C1: C1 displays 75 occurrences and C11 displays 49. − Looking at the proportion of restrictive and non-restrictive relations in C1–C10 vs. C11, we may notice that it is reversed. Restrictive relations are prevailingly encoded through C1–C10: 63% of occurrences are restrictive, 37% are non-restrictive. Non-restrictive relations are mostly encoded through C11: 23% of occurrences are restrictive, 77% are non-restrictive. So, the use of asyndetic clause clusters (C11) seems to be connected to the encoding of non-restrictive relations. This is not surprising: as mentioned in 1.1.1, non-restrictive relative relations are akin to asyndetic coordinate clauses in that they are both instances of coordination. The ambiguous status of asyndetic clause clusters between relative and coordination relations suggests that functionally contiguous relations may be encoded by means of similar constructions. − The frequency of the syntactic positions relativized through C11 is slightly different from the frequency determined for C1–C10: SU > LOC > OBL > DO. SU and LOC are once more the most frequently relativized positions. The main results of the quantitative analysis can be summarized as follows: C1, the postnominal strategy with the relative pronoun kotoryj, is the most frequently attested construction in RRR98; SU and LOC are the most frequently relativized positions; constructions in which the RC is introduced by a relative element are more frequently used to encode restrictive relations, whereas asyndetic clause clusters (C11) are more frequently used to encode non-restrictive relations.

402 Relative constructions in a spoken Russian corpus 8.4. Qualitative analysis The relationship between quantitative and qualitative analysis may be considered as a ‘type vs. token’ one. Figures presented in 8.3 were useful to establish which constructions are attested in RRR98, how often they occur and whether their occurrence is to some extent connected with the position relativized or to the nature of the link between SoAs. Instead, in the qualitative analysis each informant’s linguistic behaviour will be examined and described using a number of linguistic and sociolinguistic parameters. This may help clarify why a specific construction has been chosen. The qualitative analysis also reveals that attributing a token to a type is not a straightforward deal: when each construction is considered individually, phenomena like performance errors and contaminations emerge which make the interpretation unclear or ambiguous. Still, these phenomena cannot be disregarded; on the contrary, they provide insightful evidence of speakers’ variable grammars (Croft 1995: 518–520). For the qualitative analysis, constructions C1 to C11 were first divided into three groups: standard and variety-independent ones (C1, C2, C4, C5, C8, C9 and C10), non-standard ones (C3, C6 and C7) and asyndetic clause clusters (C11). This was done on the basis of the indications provided in Švedova (1980: 517–527). Then, the 16 informants were divided into three groups according to the linguistic variety they adopted in the interview.30 Finally, it was ascertained whether any correlations emerged between the variety adopted and the relative constructions used by each group of informants. The first group is composed of informants who, independently of their job, were born into wealthy families, are politically and socially committed or carry out propagandistic activities. Usually, these speakers have a higher level of education – cf. the table in section 8.6. On account of their social status they feel obliged to speak as correctly as possible, i.e. to adopt literaturnyj jazyk or to learn it properly. As they perceive the situation of the interview as formal, they tend to produce articulate sentences. This is the case with speakers 3, 5, 8, 9, 10, 15 and 16. They avoid non-standard constructions and mostly use C1. Taking a closer look, there are slight differences among the informants: speakers 5, 9 and 16 sometimes exhibit a deviating use of relative constructions. Examples are quoted below.

Qualitative analysis 403

(8.9)

Но так как тяжело ему было вспоминать о прожитых годах в семье В.И. Ленина / которые он прожил восемнадцать лет / с двадцать первого по тридцать девятый год / […] No tak kak tjaželo emu bylo vspominat’ but as difficult to.him was remember.INF o prožitych godach v sem’e V.I. Lenina / about lived years in family V.I. Lenin.of [kotorye on prožil vosemnadcat’ let / s RPRO.ACC.PL he lived eighteen years from dvadcat’ pervogo po tridcat’ devjatyj twenty first until thirty ninth god] / year ‘But as he found it difficult to remember the years when he lived in V.I. Lenin’s family, where he lived 18 years, from 1921 to 1939 […]’ (Informant 5, RRR98: 54)

(8.10) каждый день появляется очень много / богатая почта / разбор писем / которые надо дать своевременный ответ // každyj den’ pojavljaetsja očen’ mnogo / bogataja every day appears very much rich počta / razbor pisem / [kotorye nado post review letters.of RPRO.ACC.PL is.necessary dat’ svoevremennyj otvet] // give.INF punctual answer ‘Every day we receive a lot of, a big amount of post, a review of letters, which need to be promptly answered.’ (Informant 5, RRR98: 54) (8.11) Встретился с ихней семьей / всё мне очень много подарили для музея / которые я охотно привез и подарил в нaш музей // Vstretilsja s ichnej sem’ej / vsë mne očen’ I.met with their family all to.me very mnogo podarili dlja muzeja / [kotorye much gave for museum RPRO.ACC.PL ja ochotno privez i podaril v naš I with.pleasure brought and gave in our muzej] // museum

404 Relative constructions in a spoken Russian corpus ‘I met their family. They gave me a lot of things for the museum, which I brought home with pleasure and donated to our museum.’ (Informant 5, RRR98: 58) (8.12) Был на месте, откуда начинаются реки / которая течет в Азию и в Европу // Byl na meste, [otkuda načinajutsja reki / I.was in place SRE begin rivers [kotoraja tečet v Aziju i v RPRO.NOM.F.SG flows in Asia and in Evropu]] // Europe ‘I was at the place where the rivers begin which flow in Asia and Europe.’ (Informant 9, RRR98: 99) (8.13) Много было банков всяких / кроме того, который отец был / других управляющих расстреляли // Mnogo bylo bankov vsjakich / krome togo, a.lot was banks.of any except that [kotoryj otec byl] / drugich RPRO.NOM.M.SG father was others upravljajuščich rasstreljali // directors they.shot ‘There were a lot of banks; except the one where my father was, all other directors were shot.’ (Informant 16, RRR98: 151) (8.12) and (8.13) could be considered as performance errors, perhaps due to imperfect language acquisition: in (8.12) one would expect the relative pronoun to be plural, since the referent is plural; in (8.13) one would expect the form v kotorom to relativize LOC instead of kotoryj, which relativizes SU. As for (8.9) to (8.11), we are probably faced with a fossilized deviating use of the nominative-accusative plural form kotorye: informant 5 uses it in contexts where one would expect a dative (kotorym), as in (8.10), or a prepositive (v kotoroj), as in (8.9). Interestingly, informant 5 also seems to rely on partially deviating forms of connectors, as (8.14) and (8.15) show. The deviating forms are highlighted through ‘(!)’.

Qualitative analysis 405

(8.14) вот в десять часов я находился в Москве / посколько у меня было несколько адресов, с которыми мне надо было встретиться // Vot v desjat’ časov ja nachodilsja v PAR at ten hours I found.myself in Moskve / poskol’ko u menja bylo neskol’ko Moscow since (!) at me was several adresov, [s kotorymi mne nado addresses with RPRO.INS.PL to.me necessary bylo vstretit’sja] // was meet.INF ‘So at 10 o’clock I was in Moscow, since I had a number of addresses where I had to be.’ (Informant 5, RRR98: 54) (8.15) Вот представьте себе / какой наш вождь революции был хороший спротсмен / на то что был и сильно болен // Vot predstav’te sebe / kakoj naš vožd’ PAR imagine REFL which our guide revoljucii byl chorošij sportsmen / na to čto revolution.of was good sportsman despite (!) byl i sil’no bolen // he.was also severely ill ‘Just imagine what a good sportsman our guide of the revolution was, despite being very ill.’ (Informant 5, RRR98: 56) In (8.14) the form poskol’ko is used instead of poskol’ku ‘since’; in (8.15) the concessive connector nesmotrja na to, čto ‘despite’ is reduced to na to, čto, which is not a connector. These speech peculiarities may be explained if we consider that speaker 5 is socially committed and is consequently expected to master the standard variety, i.e. literaturny jazyk. He probably didn’t acquire this variety as his native variety, but learnt it later – without attaining full mastery, though. One may object that speaker 5 is Ukrainian and that Russian is not his native language. This, however, cannot be taken for granted: during the Soviet regime, Russian and Ukrainian were both spoken in Ukraine and it is difficult to tell which of the two languages could be considered as native for Ukrainians (Gutschmidt 20062: 1860–1861). Additionally, relative constructions in Russian and Ukrainian have similar structures, so one would expect speaker 5 not to have any problems mastering relative pronouns. Instead, these problems may be accounted for if assuming that in the Rus-

406 Relative constructions in a spoken Russian corpus sian or Ukrainian variety that served as a native variety for speaker 5, relative constructions are formed through non-standard strategies, like C3. This is only a hypothesis, though: more data on this informant would be needed to confirm it. A further distinction within the first group can be made between speakers using kanceljarit and speakers showing a more or less restricted influence of prostorečie. These two groups are not clear-cut, though: informants 3, 5 and 9 sometimes use kanceljarit; the same informants 5 and 9, however, do exhibit prostorečie features in some passages. Speaker 10 also displays prostorečie features: he omits or repeats prepositions and uses ichnij instead of ich ‘their’. Only speakers 8 and 16 show no tendency to use varieties other than literaturnyj jazyk. The presence of features belonging to different linguistic varieties may be surprising at first. Still, we may suppose that these different varieties entered speakers’ variable grammars (=were acquired) at different times and under different circumstances. This is in line with Croft’s statement that “[a] variable internal grammar can […] evolve within a speaker’s lifetime, as the speaker’s social networks and sense of social identification change over time” (Croft 1995: 520). The second group of informants is composed of speakers 6, 12, 13, 14 and 15, who all adopt razgovornaja reč’ as their main communication variety during the interviews. This emerges when we analyze their use of relative constructions: non-standard constructions also occur in their speech. For instance, the correlative construction (C6) is used by speakers 6 and 12, as reported in (8.16) and (8.17). (8.16) те, которые люди работают на севере, живут здесь […] и нельзя по врaчебным этим соображениям выезжать и на юг отдыхать. te, [kotorye ljudi rabotajut na severe,] those RPRO.NOM.PL people work in north živut zdes’ […] i nel’zja they.live here and is.impossible po vračebnym ėtim soobraženijam according.to medical these ideas vyezžat’ i na jug otdychat’. go.away.INF and in south rest.INF ‘Those people who work here in the north, they live here and one cannot travel and rest in the south, as doctors recommend.’ (Informant 6, RRR98: 74)

Qualitative analysis 407

(8.17) В общем-то все, которые коми находятся или ненцы находятся в тундре, они все и члены колхозов. V obščemto vse, [kotorye komi in general PAR all RPRO.NOM.PL Komi nachodjatsja ili nency nachodjatsja v tundre,] are.found or Nenets are.found in tundra oni vse i členy kolchozov. they all PAR members kolkhozes.of ‘Generally speaking, all Komi and Nenets who live in the tundra are members of kolkhozes.’ (Informant 13, RRR98: 129) In the two examples above, C6 is used in a way that partially deviates from the structure presented in Table 8.3. The RC precedes the MC, but the RC is preceded itself by the demonstrative te ‘those’ in (8.16) and the quantifier vse ‘all’ in (8.17). As discussed in 4.1.2, these instances are subsumed to the same construction type. C3, the postnominal strategy with a relative particle, is used by speaker 6 and speaker 13. In (8.18) the relative particle čto is used with an inanimate referent to relativize SU, which is also accepted in standard; instead, in (8.19) it is used with an animate referent, which is typical of non-standard varieties. (8.18) Высокую, что двенадцать метров над ТЭЦ тоже построили мы […] Vysokuju, [čto dvenadcat’ metrov nad TĖC] high.F.SG RPAR twelve metres over CHP tože postroili my […] also built we ‘We also built the high (chimney), which towers 12 metres over the CHP-station.’ (Informant 6, RRR98: 65) (8.19) И ни один уже что приехал поработал года три-четыре, куплю машину, и только меня здесь и видели! I ni odin uže [čto priechal porabotal and not one already RPAR came worked goda tričetyre,] kupliu mašinu, i tol’ko menja years three four I.buy car and only me zdes’ i videli! here PAR they.saw

408 Relative constructions in a spoken Russian corpus ‘And there have already been many who came and worked three or four years, then bought a car and simply cleared off!’ (Informant 13, RRR98: 131) Speaker 6 presents other peculiarities: in (8.20) he uses a combined relative element, ‘specialized relative element+resumptive element’ (gde tam ‘where there’); in (8.21) he seems undecided between two constructions, which leads to a mix between an adjective phrase, takich vetchich, agreeing in the genitive case with the noun barakov, and an RC in which the relative pronoun is shifted towards the end of the clause. This feature occurs in razgovornaja reč’, as also attested in Lapteva (1976: 301). (8.20) это экспериментальный дом был, где там испытывался немецкий экскаватор. ėto ėksperimental’nyj dom byl, [gde tam this experimental house was SRE there ispytyvalsja nemeckij ėkskavator.] was.tried German excavator ‘This was an experimental house, where we tried to use a German excavator.’ (Informant 6, RRR98: 68) (8.21) Очень много снесли этих бараков, таких ветхих которые, а построили панельные дома, переселили людей. Očen’ mnogo snesli ėtich very a.lot they.demolished this.GEN.PL barakov, takich vetchich barrack.GEN.PL such.GEN.PL very.old.GEN.PL kotorye, a postroili panel’nye doma, REL.NOM.PL and they.built of.panels houses pereselili ljudej. they.let.move people ‘A lot of these barracks that were very old were demolished, prefabs were built, people had to move.’ (RRR98: 72, informant 6) The third group of informants, which includes speakers 1, 2, 4, 7 and 11, display pervasive prostorečie features. Apparently, these informants make no effort to adapt to the variety used by the interviewers. They share the following features, as shown in Table 8.10 in section 8.6: 1. they have a low educational level, if any; 2. they do not occupy a socially relevant position;

Qualitative analysis 409

3. 4.

they usually have humble jobs (cleaners, caretakers, etc.); their nationality is not necessarily Russian: speaker 1 is Komi, speaker 2 is Uzbek, speaker 7 is Belarusian; only speakers 4 and 11 are Russian. However, as mentioned above for speaker 5, a speaker’s nationality cannot be a priori identified with his/her L1. The relative constructions that these informants use are consistent with the language variety they adopt: in fact, most C3-constructions are found among these informants. They are exemplified in (8.22) to (8.24). In particular, (8.24) is an interesting instance of two coordinated relative RCs, C1 and C3: C1 is introduced by the relative pronoun kotoryj, whereas C3 displays a combined relative element, ‘relative particle+resumptive element’ (čto … on), which is used to relativize SU, just like kotoryj in C1. (8.22) Может знаете плиты, что с кружочками были. Možet znaete plity, [čto s kružočkami maybe you.know stoves RPAR with little.rings byli.] were ‘Maybe you remember those cookers that had these little rings.’ (Informant 7, RRR98: 80) (8.23) Ну-у / эти песни и танцы / что они там танцуют / ездят за границу везде / везде золотые медали привозят / Nu-u / ėti pesni i tancy / [čto oni tam PAR these songs and dances RPAR they there tancujut] / ezdjat za granicu vezde / dance they.go behind border everywhere vezde zolotye medali privozjat / everywhere golden medals bring ‘Now, these songs and dances that they dance, they go everywhere abroad, and from everywhere they bring golden medals.’ (Informant 11, RRR98: 108) (8.24) А буровые и бурят, и тут сваи до грунта, который там и что уже вечный он, пробивают и ставят. A burovye i burjat, i tut svai do and drills PAR drill and there piles to grunta, [kotoryj tam i [čto uže ground RPRO.NOM.M.SG there and RPAR already

410 Relative constructions in a spoken Russian corpus večnyj on,] probivajut i stavjat. firm he excavate and set ‘And the drills drill and excavate and set the piles to the ground that is there and is already firm.’ (Informant 7, RRR98: 79) C6 is also attested among these informants. We also find prototypical occurrences here, such as (8.25), along with other cases that are more difficult to interpret. (8.26) could be regarded both as a postnominal RC which underwent extraposition – stariki would be the antecedent and ‘old people went away’ would be the MC – and as a correlative RC: in this case, ‘many went away’ would be the MC and the relativized item stariki would not appear as a full NP in the RC, since it already appears in the previous clause. On account of this ambiguity, the sentence can be interpreted in two different ways, as the multiple translations of (8.26) illustrate. (8.25) И даже вот которые люди здесь были в общем как... вот были, работали […] I daže vot [kotorye ljudi zdes’ byli] and even PAR RPRO.NOM.PL people here were v obščem kak… vot byli, rabotali […] in general how PAR they.were they.worked ‘And even those people who were here, how should I put it, they were, they worked…’ (Informant 7, RRR98: 77) (8.26) Сейчас в основном, все молодежь / старики поуезжали / уже которые на пенсию пошли / многие повыезжали // Sejčas v osnovnom, vse molodež’ / stariki now in basic all youth old.people pouezžali / uže [kotorye na pensiju pošli] / went.away already RPRO.NOM.PL in pension went mnogie povyezžali // many went.away ‘In principle, there are only young people now, the old people who were already pensioners went away, many of them went away.’ ‘In principle, there are only young people now, old people went away, many of those who were pensioners went away.’ (RRR98: 106, informant 11) Finally, there are further instances of contaminations between constructions, like (8.27).

Qualitative analysis 411

(8.27) Ну, у нас, видимо, жильё, наверное, не хватает, чтобы всех, кто приезжающих, заселить. Nu, u nas, vidimo, žil’ë, navernoe, PAR at us evidently housing probably ne chvataet, čtoby vsech not is.enough in.order.to all.ACC.PL [kto priezžajuščich,] zaselit’. REL.NOM.SG come.PTCP.PRS.ACC.PL accommodate.INF ‘We probably don’t have enough buildings to accommodate all those who move to our town.’ (RRR98: 27, informant 2) In (8.27) the informant mixes up a participial RC (C10) – which would not require any further relative element and which correctly agrees in case and number with the referent vsech (‘all’) – with a postnominal RC (C4), requiring a finite verb in the 3rd person singular, as shown in (8.5) above. A final remark regards asyndetic clause clusters (C11). As mentioned in 8.3, they are prevalently used to encode non-restrictive relations. Still, individual informants use C11 with varying frequency. This is displayed in Table 8.7, where each informant’s use of C11 is compared to his/her use of all other constructions. Table 8.7. Occurrences of C1–C10 vs. C11. № 1 2 3 4 5

C1– C10 5 6 12 3 33

C11

Tot.



9 3 2 6 2

14 9 14 9 35

6 7 8 9 10

C1– C10 13 5 4 3 11

C11

Tot.



6 2 0 3 0

19 7 4 6 11

11 12 13 14 15 16

C1– C10 5 2 3 0 6 6

C11

Tot.

1 5 1 3 4 2

6 7 4 3 10 8

If data in Table 8.7 is correlated with the variety adopted by informants, it appears that informants using – or trying to use – literaturnyj jazyk, i.e. speakers 3, 5, 8, 9, 10 and 16, mostly adopt pronoun-marked constructions, sometimes displaying a very high frequency, like informant 5’s thirty C1constructions. The only exception is speaker 9, a woman, with three pronoun-marked and three zero-marked constructions. This may be due to the more emphatic and confidential style that women adopt in Russian face-toface conversation: among other means, confidentiality is expressed through a lower variety, razgovornaja reč’ or prostorečie. The question must be

412 Relative constructions in a spoken Russian corpus left open whether the women interviewed consciously chose this variety or used it as the only one that they fully mastered. As far as informants relying on razgovornaja reč’ and prostorečie are concerned, there are no clear tendencies. In both cases three out of five informants prefer pronoun or particle-marked constructions. As an overall trend, it can be argued that restrictive relations are encoded by means of pronoun or particle-marked constructions, whereas encoding of non-restrictive relations seems to be the main task of the zero-marked construction, at least in razgovornaja reč’ and prostorečie. So, the distinction between restrictive and non-restrictive relatives plays a central role in the selection of constructions. For instance, speaker 6 has ten (out of thirteen) restrictive pronoun or particle-marked constructions and five (out of six) non-restrictive zero-marked constructions. The same applies to speakers 1, 2 and 11: they use pronoun and particle-marked constructions only to encode restrictive relations and zero-marked constructions only to encode non-restrictive relations. To summarize, the qualitative analysis of RRR98 showed that the use of relative constructions is not random, but is in line with the linguistic variety that the informants adopt, as reported in Table 8.8. Table 8.8. Correlating varieties and constructions. VARIETY ADOPTED literaturnyj jazyk

RESTRICTIVE RELATIONS

NON-RESTRICTIVE RELATIONS

pronoun-marked constructions

razgovornaja reč’

pronoun or particle-marked constructions

prostorečie

asyndetic clause clusters

Table 8.8 only reproduces general tendencies. As emerged in the analysis of individual examples, speakers do not necessarily succeed in fully mastering the variety that they adopt. In some cases, we were faced with nonprototypical instances of constructions, resulting from performance errors or contaminations between constructions.

Discussion 413

8.5. Discussion In 8.3 and 8.4 corpus RRR98 was submitted to a quantitative and qualitative analysis. Both proved useful to provide evidence of relative constructions attested in Russian. To start with, the quantitative analysis helped establish how frequent constructions are and whether they are variety-dependent. It emerged that C1 and C5 are variety-independent and – probably as a consequence of this – are among the most frequently used constructions. Other constructions, like C3 and C6, occur exclusively in lower varieties. The distinction between restrictive and non-restrictive clauses also proved to be central to explain the use of asyndetic clause clusters (C11). Then, the qualitative analysis illustrated that construction schemes – as listed in Table 8.3 – are but prototypical models: in spoken language, constructions are realized in many different ways, which may deviate from the model and may not always be easily classified as one or the other construction type. A further result of the qualitative analysis is the possibility to correlate the use of different groups of relative constructions with informants’ linguistic and sociolinguistic features. These correlations are reported in Table 8.9. The table is to be read as follows. The three groups of constructions identified in 8.4 are reported in the first rows of Table 8.9. The frequency of their use, or non-use, allows us to posit a continuum between two poles: pole 1 is characterized by the extensive use of standard constructions and the nearly complete avoidance of all other kinds of constructions; pole 2 is characterized by the use of standard and non-standard constructions – although standard constructions may appear in a non-prototypical form – as well as the use of asyndetic clause clusters, mostly for encoding non-restrictive relations. With respect to their use of relative constructions, the 16 informants can be placed onto this continuum. The position they take on it correlates with a number of social and linguistic parameters. The latter regard the variety that the speakers adopt: speakers close to pole 1 may display the use of kanceljarit, whereas speakers close to pole 2 show prostorečie or dialectal features (see Table 8.2 above). Additionally, these speakers also display a significantly higher frequency of self-corrections and constructions reformulations, which, as was shown in 8.4, may give rise to instances of contaminations between constructions. As for social parameters, the prototypical pole-1-informant is Russian, male, has a high educational level and is socially committed; at the opposite end, the prototypical pole-2-informant is non-Russian, female, has a lower level of schooling and is not socially committed. Of all informants, speaker 3 is the closest to pole 1, whereas

414 Relative constructions in a spoken Russian corpus speakers 7 and 11 are the closest to pole 2. All other informants can be attributed a position on the continuum of each social and linguistic parameter, which accounts for their ‘uneven’ behaviour. Between a linguistic behaviour close to pole 1 and a linguistic behaviour close to pole 2 there is a range of intermediate stages, where relative constructions pertaining to different varieties occur. Table 8.9. Parameters accounting for the use of relative constructions in RRR98. CONSTRUCTIONS standard/variety-independent constructions non-standard constructions asyndetic clause clusters SOCIAL PARAMETERS gender degree social commitment nationality LINGUISTIC PARAMETERS construction reformulations, self-corrections use of kanceljarit prostorečie or dialectal features

Pole 1

Pole 2

use

(partially deviating) use use polarized use (non-restrictive)

avoidance avoidance

man high committed Russian

woman low not committed non-Russian



+

+ –

– +

All this allows us to draw the following conclusions on the use of relative constructions in RRR98: − Relative constructions can be regarded as “linguistic variants” (Ammon 20042), i.e. they can be seen as endowed with information about the linguistic variety in which they are more likely to occur. − There exist variety-bound constructions, like C6, and varietyindependent constructions, like C1 and C2. Variety-independent constructions tend to be more frequent. − The same relation can be encoded by means of different constructions: for instance, in RRR98 non-restrictive relations are encoded through both C1 and C11. The choice of a construction over another depends on the variety adopted by the informant. − When an informant uses a variety or a construction s/he is not very confident with, idiosyncratic instances of ‘deviating’ or non-proto-

Discussion 415

typical constructions may occur. Other instances of non-prototypical constructions can be explained by the nature of spoken discourse, in which self-corrections and reformulations are very frequent (Miller and Weinert 1998: 21): this may give rise to contaminations between constructions. These conclusions confirm that corpus-based analyses prove useful to identify constructions within a language and examine their contexts of use. A number of general construction schemes can be formulated on the basis of reference works, as was done in Table 8.3; then, the corpus analysis helps define the frequency and the distribution of these constructions. Most interestingly, when actual speech production is analyzed, unclear or borderline cases as well as contaminations between constructions often emerge. This hints at the fact that constructions can be reformulated and modified, which may be regarded as one of the sources of linguistic change (to this extent see also Croft 1995: 524). Moreover, the analysis of individual speakers’ language behaviour reveals that constructions are not randomly used: in many cases, their occurrence may be related to languageexternal criteria like speakers’ sociolinguistic features and the communicative situation. In Goldberg’s words, constructions are “form and meaning pairings” (Goldberg 2006: 3): it may be further specified that constructions can also be variety-bound form and meaning pairings. A final point must be made on the usefulness of these kinds of corpusbased studies for typological research. As mentioned in 1.2, traditional typology has a major drawback, that is, it happens not to consider all constructions attested in a language because it relies on data sources describing structures that are found in that language only partially. In this sense, corpora and corpus-based studies help identify constructions that are traditionally neglected and establish how frequent they are. Moreover, if different linguistic varieties are documented in a corpus, it is possible to ascertain whether any of the constructions identified are variety-bound. Finally, the results of corpus-based studies can be compared with descriptions of constructions reported in reference grammars. This can be useful to clarify why some of these constructions are classified as standard and other as non-standard and how this bipartition arose.

416 Relative constructions in a spoken Russian corpus 8.6. The informants The list below reports personal data available about the 16 informants constituting corpus RRR98. As repeatedly pointed out in previous sections, nationality (indicated in the fourth column) cannot be identified with an informant’s L1: that is, non-Russians may have Russian and/or another language as an L1. So, this factor on its own cannot account for the deviating use of relative constructions that some speakers exhibit. Table 8.10. Informants in RRR98. №

Age

Gender Nationality In Vorkuta since M Komi 1947

1

53

2 3

49 27

M M

Uzbek Russian

1956 1956

4

45

M

Russian

1970

5

51

M

Ukrainian

1958

6

51

M

Russian

1954

7 8

52 44

F M

Belarusian Russian

1959 1955

9

60

F

Russian

1958

10

47

M

Russian

1965

11

55

F

Russian

1949

12 13 14 15

46 36 20 29

M M F M

Russian Russian Komi Russian

1957 1947 – –

16

84

F

Russian



Job department foreman driver assistant foreman miner team leader electricianmechanic technician

School or degree –

Speech peculiarities prostorečie

low attending college technical school –

prostorečie kanceljarit

technical school 7 classes –

caretaker mechanic team leader inspector of technical the personnel school department senior garage industrial mechanic school pensioner6 classes cleaner mechanic – mechanic – student – senior scene – shifter pensioner –

prostorečie prostorečie kanceljarit prostorečie prostorečie – kanceljarit prostorečie prostorečie prostorečie – – – prostorečie –

Summary and outlook

The present investigation arose out of the need to provide a comprehensive account of relative constructions in European languages: in particular, special attention was to be devoted to non-standard varieties, which are usually insufficiently documented in areal and typological studies. It is sometimes claimed that the opposition between standard and non-standard is not directly relevant to linguistic investigations in general and to functional typology in particular, because linguistic constructions exist independently of the variety or varieties in which they occur (cf. Miller and Weinert 1998; Henn-Memmesheimer 1997; Cheshire and Stein 1997); however, functional and typological research is strongly biased towards standard constructions. So, in this work it was attempted, on the one hand, to describe and classify relative constructions that are usually underrepresented in linguistic studies on European languages and in typological studies – paying special attention to Slavic languages – and, on the other, to account for their non-standard status. In PART 1 the theoretical and methodological premises of the analysis were discussed. To start with, in CHAPTER 1 a definition of ‘relative relation’ and ‘relative construction’ were introduced that helped to delimit the constructions under investigation. In particular, both restrictive and nonrestrictive relative relations were taken into account as in both cases one SoA involved in the relation qualifies a participant of the other SoA. Then integrative functionalism and the typological-dialectological approach were briefly discussed, which served as a theoretical reference frame for this work: both approaches stress the importance of taking into account non-standard constructions and, more generally, they call for a comprehensive account of linguistic constructions in their context of use. The remainder of Chapter 1 was devoted to the discussion of the term ‘non-standard’, which is regarded as a synonym for ‘sociolinguistically marked’, so as to comprise all constructions labelled as ‘colloquial’, ‘popular’, ‘vernacular’, etc. Existing studies on relative constructions in European languages were also reviewed: it emerged here that in the great majority of these studies non-standard relative constructions are either underrepresented or not paid sufficient attention. So, the ‘standard vs. non-standard’ distinction was adopted a priori as an operative criterion to select the constructions that should be focussed on.

418 Summary and outlook In CHAPTER 2 the issue of data collection was covered. To start with, conceptionally oral communicative practices were identified as the linguistic domain in which non-standard relative constructions are more likely to be found. In conceptional orality (a-la Koch and Oesterreicher 1990) speech elaboration is minimal as speakers’ main concern is immediate speech production rather than speech editing. In this sense, speakers may use the variety which they consider most appropriate for informal peer-topeer communication and this variety may also be a sociolinguistically marked one. The analysis of conceptionally oral communicative practices should ideally be based only on actual speech, i.e. on language corpora. Still, linguistic corpora do not exist for all of the 36 languages constituting the sample of this investigation, nor would it have been possible to scrutinize and evaluate the results of 36 corpus analyses. So, different linguistic sources had to be integrated in order to gather enough evidence for nonstandard relative constructions: grammars, linguistic studies, questionnaires, corpora and the WWW. Each source has its pros and cons, so the reliability of each of them was weighed; additionally, constructions were checked against at least two different sources whenever possible. In CHAPTER 3 the criteria adopted to classify relativization strategies in European languages were described. A relativization strategy was defined as the use of a particular word order and of a particular relative element to relativize one or more positions of the Accessibility Hierarchy. That is to say, relative constructions were classified according to three parameters: word order, relative element and syntactic positions relativized. Existing literature on the three parameters was critically reviewed; then, a slightly different classification was proposed in order to cover all constructions attested in European non-standard varieties. As for word order, the traditional distinction between embedded (prenominal, circumnominal and postnominal) and adjoined (correlative, postposed) strategies was adopted. Similarly, for the syntactic positions relativized, Keenan and Comrie’s (1977) Accessibility Hierarchy was relied on. Instead, greater attention was devoted to relative element, since this proved to be the parameter requiring the most fine-grained classification on account of the extreme variability that European languages show in this respect. Four features encoded by relative elements were first identified: the link between the MC and the RC, the syntactic role of the relativized item in the RC, gender agreement and number agreement with the relativized item in the MC. Then a distinction was made between simple and combined elements, i.e. between elements in which the four features are encoded by means of a single morphosyntactic unit and elements in which they are encoded by means of a

Summary and outlook

419

number of morphosyntactic units. Simple relative elements are relative pronouns, specialized relative elements, relative particles and the zeromarker. Each of them codifies a different number of features. Combined relative elements result out of the combination of simple relative elements with other simple relative elements, with non-relative particles or with resumptive elements. Of the 20 possible combinations between these elements, 12 are attested at least in one of the languages in the sample. Since classifications patterns are only attempts to grasp linguistic reality by means of parameters and categories, reserarchers are likely to be faced with instances of ambiguous or problematic classification. So, in the last section of Chapter 3 some of these instances were reviewed, discussing how they may be best grasped using the parameters introduced in the previous sections. After outlining the theoretical and methodological foundations of the work, PART 2 was devoted to the areal study of non-standard relative constructions in European languages. The study was carried out from a threefold perspective: a typological, a sociolinguistic and a diachronic one. In accordance with the tenets of integrative functionalism, the integration of the three perspectives had the aim to achieve an all-encompassing overview of relativization strategies in European languages: the typological perspective helped establish which strategies are nowadays attested in European non-standard varieties; the sociolinguistic perspective helped clarify which position different relativization strategies occupy in the diasystem of European languages; finally, the diachronic perspective served to unveil the historical continuity of relativization strategies attested in present-day non-standard varieties on the one hand and to shed light on the factors that led to the present-day opposition between standard and nonstandard relative constructions on the other. In CHAPTER 4 relativization strategies were analyzed from a synchronic typological point of view. The peculiarities of non-standard relative constructions were described considering the three parameters ‘word order’, ‘relative element’ and ‘syntactic positions relativized’ separately. As regards word order, languages that in the standard variety exhibit a prenominal construction (Basque, Turkish) also display a postnominal construction in non-standard varieties; some languages that in the standard variety rely on a postnominal construction also possess a correlative construction (Greek, Hungarian, Latvian, Russian, Ukrainian) or a postposed construction (Dutch, German, Turkish) in non-standard varieties. As far as relative elements are concerned, the following phenomena are attested: elements that are inflected in standard varieties may remain uninflected in non-stan-

420 Summary and outlook dard ones; elements that encode the syntactic role of the relativized item in the RC in standard varieties may not encode it in non-standard ones. Additionally, non-standard varieties display a full range of relative particles as well as some instances of paradigm regularization. Finally, most of the combined elements attested in the languages of the sample occur in nonstandard varieties. As regards the syntactic positions relativized, it was attempted to ascertain whether Keenan and Comrie’s (1977) statement that higher positions of the AH are preferably relativized by means of [–case] strategies and lower positions are mostly relativized by means of [+case] ones also holds true for European non-standard relative constructions. The analysis revealed that in European non-standard varieties, strategies that encode the syntactic role of the relativized item in the RC can be applied to all syntactic positions of the AH along with strategies that do not encode the syntactic role and strategies that encode it twice. So, no a priori constraints on the applicability of a strategy to a particular syntactic position seem to exist. Still, further language-specific quantitative analyses may reveal if individual languages show any preference towards a particular strategy to relativize high or low positions of the AH. After describing the peculiarities attested in non-standard varieties, the results were compared with the cross-linguistic patterns of variation identified by Comrie and Kuteva (2005). It emerged that European non-standard relative constructions show an ambiguous behaviour: on the one hand, codifying the syntactic role of the relativized item in the RC does not seem to be one of speakers’ primary concerns, which also happens in the world’s languages. In fact, the frequency with which relativization strategies in European languages convey the four features encoded by relative elements follows the hierarchy link between MC and RC > syntactic role > gender and number agreement That is to say, conveying the link between the MC and the RC is more central than encoding the syntactic role of the relativized item. On the other hand, in European languages one finds that the ‘gap’ strategy, the ‘relative pronoun’ strategy and the ‘pronoun retention’ strategy are attested with nearly the same frequency, which does not match cross-linguistic tendencies. Moreover, the investigation of non-standard varieties did not confirm the results of Cristofaro and Giacalone Ramat’s (2007) survey in two ways. First, it does not appear that “invariable relative elements are exclusively found in Western Europe” (Cristofaro and Giacalone Ramat 2007: 84): relative particles are actually attested in 34 out of 36 languages

Summary and outlook

421

of the sample. Then, the claim that the alternation between strategies making the syntactic role of the relativized item in the RC explicit and ones that do not make it explicit takes place prevalently in Western European languages (Cristofaro and Giacalone Ramat 2007: 84) may also be refuted. In fact, alternation between strategies is attested in both Western and Eastern European languages and can be regarded as one hallmark of relative constructions in Europe. Differences cannot be seen between languages, but between language varieties: adopting Givón’s (1979: 97–98) terminology, we may say that standard varieties preferably rely on a syntactic mode, whereas non-standard ones rely both on a syntactic and on a pragmatic mode. In the final section of Chapter 4 the functional principles underlying non-standard relative constructions (economy, analogy, iconicity) were examined in order to explore the possibility of explaining the opposition between standard and non-standard varieties on functional grounds. Still, it emerged that the same functional principles can be invoked to account for both standard and non-standard relative constructions. The only functional principles that seem to generally grasp the difference between the two groups of varieties can be formulated as follows: − NON-STANDARD: “Make at least the link between the MC and the RC explicit.” − STANDARD: “Make the relation between MC and RC as unambiguous as possible and as explicit as needed.” However, even these very general functional principles cannot account in detail for all strategies found in standard and non-standard varieties. So, in CHAPTER 5 the dichotomy ‘standard vs. non-standard’ was considered again, checking whether it could be based on different terms. In order to do this, the same constructions examined in Chapter 4 were regarded from a sociolinguistic perspective. Each construction was regarded as a linguistic variant, i.e. it was attributed a sociolinguistic status (‘standard’, ‘non-standard’) according to the judgements contained in grammars and linguistic studies. Then, an attempt was made to clarify why some constructions are attested in standard varieties whereas others are attested in non-standard varieties. This was done on the basis of a number of sociolinguistic principles which were active during the codification process of standard varieties: explicitness, compactness, no redundancy and purism. Invoking these principles, the ‘standard vs. non-standard’ opposition can be explained more extensively: linguistic variants that fit these principles are included in the standard variety, whereas those that do not fit them are excluded from it. Additionally, it often happens that the same relativization strategies have

422 Summary and outlook a different sociolinguistic status in different European languages: so, on the basis of the language-specific degree of standardness of each strategy, it was possible to formulate a scale of standardness for relativization strategies in European languages, which reads as follows: RPRO, SRE > RPAR (high positions of the AH) > RPRO+resumptive element > RPAR+resumptive element > Ø, Ø+resumptive element > RPAR (low positions of the AH) Relative constructions in individual languages rank differently on the scale. In some languages only the postnominal strategy introduced by a relative pronoun is regarded as standard; other languages show an increasing acceptance of constructions situated in the middle and in the lower segment of the scale. The different degree of acceptance is related to the codification process: typically, the stronger the influence that a prestigious exoglossic language exerted on the formation of the standard variety, the lower the tolerance towards positions situated on the right of the scale. This can be explained if we consider that the prestigious exoglossic language (Latin, Ancient Greek, Church Slavonic and, later on, also French or German) always displayed a G-N-inflecting relative pronoun, which was granted an outstanding status because it was explicit (it encoded gender and number agreement as well as the syntactic position relativized), compact (it consisted of a single morphosyntactic unit) and non-redundant. As for the other strategies, the reasons underlying their inclusion or exclusion from the standard variety are not always easy to reconstruct: this has also to do with “the fact that they [i.e. standard varieties] – at least in part – are the product of often naive ‘language engineering’” (Van Marle 1997: 14), so they are likely to display inconsistencies, as was repeatedly pointed out also in the typological analysis carried out in Chapter 4 (cf. for instance section 4.2.4). In Chapter 5 it emerged that the present-day sociolinguistic situation is closely connected with the evolution of European languages. In this respect, adopting a diachronic perspective, which was done in CHAPTER 6, proved useful in order to achieve a panchronic view, in which both the development and the present situation of non-standard relative constructions are taken into account. In particular, in Chapter 6 two main points were made. First, evidence was provided for the occurrence of constructions that are nowadays regarded as non-standard in earlier linguistic stages. It emerged that these constructions are historically attested and have developed for centuries along with constructions that are nowadays re-

Summary and outlook

423

garded as standard. Second, the development of the ‘standard vs. non-standard’ dichotomy was examined by means of a diachronic model based on Auer’s (2005b) diachronic-sociolinguistic model. The model allowed us to identify the process of language codification as the turning point for the crystallization of the ‘standard vs. non-standard’ opposition. Before a standard variety existed, relativization strategies did sometimes display an uneven distribution in local varieties: this distribution, though, was either genre-bound or area-bound. Instead, after the codification of a standard variety and the gradual development of a sociolinguistic continuum some strategies entered the standard variety and were regarded as standard, whereas others, attested in sociolinguistically marked varieties, were looked upon as non-standard. In this sense, the distinction between standard and non-standard constructions was at first superimposed on the local varieties roofed by the new standard variety. However, the new standard variety began sooner or later to interrelate with local varieties, thus leading to a new distribution of relative constructions on the sociolinguistic continuum of each language. In this respect, it was pointed out in 5.3 that the ‘standard vs. non-standard’ opposition had been set a priori in order to analyze and discuss a number of constructions that are often underrepresented in linguistic studies. Taking a closer look, it would be better to abandon a strict distinction between standard and non-standard in favour of a comprehensive view of languages as diasystems composed of several interrelated linguistic varieties. In accordance with the tenets of integrative functionalism, this can well represent the starting point for new investigations focussing on the incidence of all relative constructions attested in a language, so as to highlight correlations between constructions, their contexts of use and the varieties in which they occur. This is what was exemplified in PART 3, where two case studies based on a German and a Russian corpus were presented. In CHAPTER 7 the occurrences of (standard and non-standard) RCs in the German presentational construction [es gibt N]MC [REL …]RC (‘There is/are N that…’) were examined. The corpus was constituted by a selection of Web forums. It emerged that the postposed strategy occurs more often when substantives with a hyperonymous character (like ‘people’, ‘things’) are used as referents and that the postnominal construction introduced by the relative particle wo is attested more frequently for linking MC and RC without specifying the syntactic role of the relativized item. In CHAPTER 8 the use of relative constructions was studied in a Russian corpus recorded in the 1980s in Vorkuta, a town in North-Eastern European Russia. This corpus was chosen because the 16 informants offered a

424 Summary and outlook cross-section of the Russian sociolinguistic continuum. The analysis brought to light a number of correlations between relative constructions and their (semantic-syntactic) contexts of use on the one hand and between relative constructions and the variety adopted by the speakers on the other. For instance, it could be established that zero-marked relatives are more frequent in non-restrictive relations than in restrictive ones. Still, this use is subordinate to the variety adopted: RCs introduced by the zero-marker occur most frequently in the speech of those informants who rely on the varieties razgovornaja reč’ (colloquial Russian) and prostorečie (a lower urban sociolect). Speakers adopting literaturnyj jazyk (standard Russian) usually resort to the postnominal strategy with the relative pronoun kotoryj to encode both kinds of relations, restrictive and non-restrictive. In this respect, speakers’ choices are consistent with the linguistic variety that they adopt. To summarize, the case studies seem to confirm the hypothesis presented in Chapter 5 that the use of relative constructions is not random, but varies according to a number of factors related both to speakers’ sociolinguistic features and to the semantic-pragmatic context of use of a construction: that is, the occurrence of relative constructions is best accounted for if one considers both language-external and language-internal factors, weighing their incidence from case to case. All in all, the investigation of relative constructions in European nonstandard varieties carried out in this work uncovered great typological variability, which, unfortunately, is not sufficiently taken into account in traditional typological studies. In this sense, relying on an extended functional framework, i.e. integrative functionalism, proved to be an appropriate choice: not only were more constructions described than those traditionally included in typological studies, but it was possible to establish a link between constructions and their contexts of use. Moreover, the investigation of non-standard relative constructions shed new light onto the European typological panorama: on the one hand, non-standard varieties exhibit strategies that pattern with the tendencies found in the world’s languages; on the other, they also exhibit strategies that can be looked upon as areal features. In any case, the additional strategies described in this work should be included in future typological studies, so as to make the typological panorama more comprehensive and possibly closer to linguistic reality. A final remark concerns possible future research issues. Further research could move in two directions: either within individual European languages or outside Europe as a linguistic area. The first direction would integrate what has been achieved in the present investigation. As a matter

Summary and outlook

425

of fact, this work was cross-linguistically oriented and primarily aimed to compare syntactic structures found in the languages of the European sample. Such a broad-scope investigation inevitably only scratched the surface of language-specific linguistic phenomena. In this respect, one could concentrate on individual languages and conduct a more systematic typological, sociolinguistic and diachronic analysis of the relative constructions attested therein – similarly to the attempt made in Part 3 with German and Russian. These investigations would provide more specific data, which, in turn, would allow researchers to draw more accurate and better-founded cross-linguistic correlations between constructions. The second research direction could aim to situate European languages within the broader panorama of the world’s languages as far as the opposition between standard and non-standard is concerned. This study was focussed on European languages because most of these languages exhibit a full-fledged sociolinguistic continuum, in which it was possible to distinguish between varieties (and constructions) regarded as standard and varieties (and constructions) looked upon as non-standard. Still, this distinction applies to many other languages in the world. Following the claims of the typological-dialectological approach that dialect syntax matters, one could extend the investigation of relative constructions to other world’s languages in which non-standard varieties can be identified. As was done for European languages, constructions found in standard and non-standard varieties could be compared so as to establish whether and in which respect they differ from each other. Moreover, it could be checked whether the principles that guided the formation of standard varieties in European languages also played a role in other world’s languages. Both research directions share a common aim, i.e. to attain an in-depth documentation of the world’s languages: this may help refine cross-linguistic investigations more and more, since the typological-functional framework can be enriched by the findings of sociolinguistic and diachronic research.

Appendix A – The language sample

In the table below languages are classified according to phylum (given in upper case), family (give in small upper case) and branch (given in italic). In the rightmost column the three-letter code is reported which was used throughout the work to refer to each language. INDO-EUROPEAN CELTIC Goidelic ROMANCE West

East GERMANIC

North

West

SLAVIC

West

East

South

BALTIC HELLENIC ALBANIAN

Irish Catalan French Portuguese Spanish Italian Rumanian Danish Icelandic Swedish Norwegian Dutch English German Czech Lower Sorbian Polish Slovak Upper Sorbian Belarusian Russian Ukrainian Bosnian/Croatian/Serbian Bulgarian Macedonian Slovenian Latvian Lithuanian Modern Greek Albanian

IRI CAT FRE POR SPA ITA RUM DAN ICE SWE NOR DUT ENG GER CZE LSO POL SLK USO BLR RUS UKR BCS BUL MAC SLN LTV LIT GRE ALB

Appendix A – The language sample

427

(continued) URALIC FINNO-UGRIC

Ugric

Estonian Finnish Hungarian

EST FIN HUN

South

Turkish

TUR

SEMITIC ARABIC

Maltese

MAL

ISOLATE BASQUE

Basque

BAS

ALTAIC TURKIC

Finnic

Appendix B – The questionnaires

The two questionnaires compiled by the informants (cf. 2.3.4) are reported here. The informants were required to translate only the parts in italic. QUESTIONNAIRE N. 1 1. (Mary is talking with her friend Annika, who has two kids and has left her job) − So you decided to leave jour job… − Yes… You know, I’ve got two little children... although… I’ve known women that I’m always wondering how they get to do everything, working, cooking, cleaning and so on… 2. (at the railway station, two passengers) − What? All trains are delayed?! − There was an accident two hours ago, two trains that one was going to London and the other to Cambridge crashed together. 3. (Sabrina is telling Monica about yesterday’s “girls only”-meeting) − And who was there? − Everybody: Stella, Masha, Rebecca and the other girl, the new one, that I don’t remember the name. 4. (a journalist interviews a famous director) − Mr. Spielberg, what was your purpose in turning this movie? − Well, I mean, first of all I wanted to turn a movie that the characters would be credible, like in everyday life, a movie that you think: “Oh, that happened to me, too!”. 5. (a woman is talking to her husband) − I met John in the bus! − John who? − John, the chap that his car was stolen last week. 6. (at a preparatory meeting, a professor) − Dear colleagues, the conference next week is surely an event that we must try to devote all our efforts to it.

Appendix B – The questionnaires 429 7. (two colleagues working in a team) − Hey, we’ve made it! − What?! − Our project, the project that we’ve worked a lot on it has been accepted!! 8. (Rosemarie is telling her friend about her trip to Germany) − … and we passed in front of an old building and Luke said: “Come on, let’s go and see what’s that”, and it was the old Rathaus. And the guide at the entrance let us in and showed us all the rooms and also the big mayor’s room that now there’s an art gallery in it. 9. (two teachers telling each other about what they’re doing with their pupils) − … and I decided to deal with issues like, you know, the European Union and the integration of different people and cultures, and the pupils were interested and… − Oh, really? Well, I think that these are issues that we must speak more often about them during classes. Pupils must get in touch with reality, they must learn what’s going on out there. 10. (at the swimming pool entrance, the lady at the reception is speaking to a new customer) − And here’s Mark… he’ll show you the swimming pool and the changing rooms where you can change your clothes and the safes that if you want you can put your watch and wallet and mobile there. 11. (two friends are walking on the street) − Look! − What? − That book, in the shop-window! It’s the book that I finished to read it last week, it’s very very funny. 12. (two students talking about their Master thesis) − So I think I’ll focus on Slavic pidgins and creoles. − Fine! It’s a topic that if I’m not wrong Prof. Malek is very fond of it. 13. (Jane has been criticized by her schoolmate because she bought a horrible pen) − You’re right, but now I’ve got a pen that I can write for hours and my hand doesn’t get tired. 14. (at school, the teacher says to the pupils) − And now underline the words that you don’t know the meaning. 15. (a mother to her daughter, on the street) − You shouldn’t have done it. The child that you gave the money is a gipsy.

430 Appendix B – The questionnaires − And what’s wrong about gipsies!? − They’re not good people, they’re people you cannot trust, and they’re always stealing money. − These are only silly prejudices you should get rid of! 16. (Laura tells her neighbour about a colleague who retired) − Of course, of course I miss him, you know... He was… he was a man that we got along well, that you could laugh together, that was always good-tempered... 17. (two girlfriends talking confidentially to each other) − So your mum won’t let you sleep at a friend’s even if you’re 18!? − That’s it! I haven’t been to a party yet that I haven’t got home the same night. − I can’t believe it – if I were you I would rebel… 18. (a couple goes back to the place where they first met and fell in love) − 25 years ago, Bert, do you remember? The moon in the sky, this little lake... we went for a walk around it… − Moon? Which moon? The evening we walked around this lake it was raining! − Oh Bert, how can you be so insensitive! 19. (a woman is talking to her son – or to her nice little cat) − Is it possible, Mike? Every time you’re at home alone you make such a mess in this room!! You scoundrel! 20. (a retired football player is telling a journalist details about his being doped) − And what can you say about that scandal? You were suspected of being doped... − Being doped? Of course we were doped, all of us… they came and said: “Now you’re getting something injected, something good for you”, and they used syringes that they never changed the needle, I must say I was worried but the coach said “Trust me” and we trusted him…

Appendix B – The questionnaires 431

QUESTIONNAIRE N. 2 1. (Mary is talking with her friend Annika, who has two kids and has left her job) − So you decided to leave jour job… − Yes… You know, I’ve got two little children... although… I’ve known women that I’m always wondering how they get to do everything, working, cooking, cleaning and so on… 2. (Sabrina is telling Monica about yesterday’s “girls only”-meeting) − And who was there? − Everybody: Stella, Masha, Rebecca and a girl that I don’t remember her name. 3. (a journalist interviews a famous director) − Mr. Spielberg, what was your purpose in turning this movie? − Well, I mean, first of all I wanted to turn a movie that the audience would be impressed by its resemblance to everyday life, I mean, a movie that you think: “Oh, that happened to me, too!”. 4. (a woman is talking to her husband) − I met John in the bus! − John who? − John, the chap that his car was stolen last week. 5. (Mary tells her fiancé about a car accident she happened to see) − You know, it was just like movies – spectacular somehow, but terrible at the same time… − Jeez! − Yeah, and the people that saw the accident couldn’t yield a single word, they were… I don’t know, kind of shocked… 6. (Rosemarie is telling her neighbour about her trip to Germany) − … and we passed in front of an old building and Luke said: “Come on, let’s go and see what’s that”, and it was the old Rathaus. And the guide at the entrance let us in and showed us the rooms that the Local Council is renovating at the moment… and also the big hall that there’ll be an art gallery in it soon. 7. (two teachers talking to each other) − … and I decided to deal with topics like, you know, the European Union and the integration of different peoples and cultures, and the pupils were interested and… − Oh, really? Amazing, somehow… It’s always so hard to find topics that pupils get interested in, you know? − Yes…

432 Appendix B – The questionnaires − I mean, topics that you speak about them during classes and… − Yes, and pupils don’t get bored. 8. (Jane has been criticized by her schoolmate because she bought a horrible pen) − You’re right, but now I’ve got a pen that I can write for hours and my hand doesn’t get tired. 9. (a mother to her daughter, on the street) − You shouldn’t have done it. The child that you gave the money is a gipsy. − And what’s wrong about gipsies!? − They’re not good people, they’re people you cannot trust, and they’re always stealing money. − These are only silly prejudices you should get rid of!!! 10. (two girlfriends talking confidentially to each other) − So your mum won’t let you sleep at a friend’s even if you’re 19!? − That’s it! I haven’t been to a party yet that I haven’t got home the same night. − I can’t believe it – if I were you I would rebel…

Notes

1.

2.

3.

4.

5.

See, among others, Keenan and Comrie (1977), Comrie (1981: 131–132), Lehmann (1984: 279), Givón (1990: 680), De Vries (2002: Ch. 6) and Cristofaro (2003: 195–196). The topic of restrictive vs. non-restrictive relatives is also discussed in non-typological contributions, like Kleiber (1987), Gapany (2004: Ch. I) and Weinert (2004): in particular, Weinert investigated relative clauses in English and German conversation. The results of a conversational analysis led her to reject the difference between restrictives and non-restrictives and to replace it with the functions that relative clauses have in speech: describing (e.g. “he’s a cat that’s normally friendly”), intensifying/confirming (e.g. “it’s the only one we have”) and identifying (e.g. “the man who drives the bus he’s a good laugh”, Weinert 2004: 46–47). Still, when looking at the examples quoted in Weinert (2004), one may relate, with few exceptions, the first function to non-restrictiveness and the two latter functions to restrictiveness. In this sense, restrictive and non-restrictive relations may also be seen as two superordinated categories, to which a number of more specific subfunctions can be subsumed. This requires making explicit the verb olmak ‘to be’ in the main clause and downgrading it to a non-finite form through the suffix -dIk. The (nominalized) resulting form, olduk, is inserted in a possessive construction: the subject of the clause takes the genitive case (kardeşi-nin), the non-finite verbal form takes the possessive suffix for the third person (olduk+u, yielding olduğ-u). As illustrated above, applying the assertiveness tests to non-restrictive relatives sometimes yields ambiguous results. In (1.14) and (1.20) only one SoA proves to be at issue, which entails that these sentences are instances of coordination. However, the semantic nature of the shared participant in the main SoA (a proper noun), suggests that we are dealing with non-restrictive relations: the participant is sufficiently identified through the proper noun, so the qualifying SoA cannot but provide additional information on the participant (recall Table 1.1). The terms DERANKED and BALANCED apply to the verb of a non-MC compared with the verb of an MC. If the verb of the non-MC shows the same number of time, aspect, mood and person distinctions as the verb in the MC, the verb form is balanced. If it does not, or if it exhibits special markers that cannot occur in MCs, the verb form is deranked (Cristofaro 2003: 54). The term ‘standard’ referred to a language began to be used in the 19 th century and was widely recognized only in the 20th century (Ammon 20042: 273): in Europe, the development of standard was strictly related with historical and political events leading to the rise of nationalism and to the constitution of

434 Notes

6.

7.

national states (cf. Haarmann 1993). ‘Standard’ (or: ‘common’, ‘literary’) language was the variety that represented the nation and its people in opposition to other nations and other peoples. In this sense, standard varieties can be regarded as Ausbau varieties (Kloss 1967: 30): codifiers mark differences from other (standard) varieties and enlarge possible contexts of use, so as to develop a written form for all communicative purposes, including scientific prose, education and technical language. For an overview on the ongoing (terminological) discussion on Ausbau (and Abstand) languages see Haarmann (20042). Some examples: 1. Reference works on Slovak: up-to-date grammars are quite scarce. This is because in the former Czechoslovakia (1919–1993) Czech was the officially acknowledged standard language, whereas Slovak was regarded as a local variety. Accordingly, linguistic studies are minimally concerned with Slovak. Today, Slovak language authorities are making a great effort to revitalize this language, which is one of the official languages in the European Union (Berger 2000; Lubaś 20062: 1842). 2. Autonomist movements in northern Spain: these movements consider Galician as a language – and not simply as a Spanish dialect – which supports the claim that Galicia deserves political autonomy. So, reference works are being quickly compiled in order to increase the visibility of Galician in the European linguistic panorama (Villena Ponsoda 20062: 1807). 3. The political situation in the Balkans: political events may lead in the future to the paradoxical situation that Croatian, Serbian, Bosnian and Montenegrin develop each an own (Ausbau) standard variety, which would be then officially acknowledged as the standard variety of Croatia, Serbia, Bosnia-Herzegovina and Montenegro respectively (Neweklowsky 20062: 1829–1830; Christian Voß, p.c.). This may happen although these languages are very close to each other and annot be regarded as Abstand languages in the terms of Kloss (1967: 30). As for now, linguists prefer to refer to the four languages as a single entity named ‘Bosnian/Croatian/Serbian’ (BCS). This solution is adopted in the present work, too. As Weiß (2004: 663) and Auer (2005b: 10) point out, artificiality is primarily due to rational instances, often connected to the existence of a model language that served for centuries as an exoglossic standard. This was Latin in Western and Central Europe and in the so-called Slavia Romana, i.e. those Slavic countries which were at least until the Reformation under the influence of the Roman Church. To a lesser extent, Latin influenced also Northern Europe. Greek and (Old) Church Slavonic served as an exoglossic standard for Eastern Europe, in particular for Romania and for the so-called Slavia Orthodoxa, i.e. Russia, Belarus, part of Ukraine and most of the Balkan Peninsula. The distinction between Slavia Romana and Slavia Orthodoxa was first introduced in Picchio (1984). Although it has been repeatedly criticized for being too

Notes 435 rough, it will be adopted in this work since it serves to draw the borderline of Latin influence in European languages (cf. also Chapter 6). 8. The term ‘communicative practice’ is taken from Fiehler (2000a, 2000b) and Fiehler et al. (2004), who propose it as an alternative to ‘text’ or ‘text sort’ because it is more comprehensive and highlights the basic communicative and target-oriented function of texts as well as their becoming historically established in each society: “Kommunikative Praktiken sind soziale Praktiken, Formen sozialer Praxis. Es handelt sich um gesellschaftlich herausgebildete konventionalisierte Verfahren zur Bearbeitung rekurrenter kommunikativer Zwecke. Jede Gesellschaft verfügt als Repertoire für die Verständigung über einen spezifischen Satz solcher kommunikativen Praktiken, der sich historisch herausgebildet hat.” (Fiehler et al. 2004: 16). 9. As a matter of fact, in some generativist studies evidence is also gained from elicited data, like questionnaires. Here is one example. For the sake of data comparability, the ASIt (Syntactic Atlas of Italy, URL: http://asis-cnr.unipd.it/, last access: June 30th, 2011) was compiled on the basis of questionnaires. The informants had to translate sentences from Italian into their own dialect. Leaving aside any possible interference between the source and the target language, the structure of the questionnaires clearly reveals that it was first decided which syntactic phenomena should be studied and then the questionnaires were compiled. This procedure has nothing condemnable per se; still, it led to elaborating sentences which are hardly likely to be ever uttered by dialectal speakers. Syntactic phenomena like multiple embedding or different kinds of subordinate clauses are rather typical of elaborated language varieties – like the standard variety – as they are a hallmark of Ausbau. Now, one cannot ignore that generally, dialects are not Ausbau varieties. How can a dialectal informant come to terms with the translation of such sentences? Either s/he will strongly rely on the Italian source sentence, not escaping interference, or s/he would rather provide a paraphrase of the sentence, simplifying it and thus missing the point, because no direct comparison is possible. 10. ‘Default’ strategies are those strategies that are seen as prototypical in the relevant language, although the possibility is left open that other strategies may exist. The term ‘default’ is used here to refer both to the frequency and to the syntactic applicability of strategies: default strategies can either be more frequent or be used to relativize more syntactic positions than non-default ones. For instance, a prenominal strategy (based on non-finite verb forms) exists in German, Finnish and Russian: still, its applicability is more restricted than in Basque and Turkish. In German and Russian it can relativize only subjects; in Finnish it can relativize only subjects and direct objects (Matsumura 1985: 65). In Basque and Turkish this strategy can relativize all syntactic positions (for further details on the Basque strategy see also 3.4). So,

436 Notes

11.

12.

13.

14.

15.

16.

the prenominal strategy can be seen as default in Basque and Turkish and as non-default in German, Finnish and Russian. In order to understand point 3, we need to bear in mind that RCs are seen in generative perspective as MCs which underwent a NP-extraction. This NP leaves a trace, which is the instance of the head NP in the RC – Keenan and Comrie (1977) label it NP rel. If this instance is morphosyntactically expressed through the relative element, the element is gap-constructing, if not, it isn’t. An anonymous reviewer suggested that in this example the element da should be regarded as a locative adverb rather than a particle combining with the relative pronoun. Still, in the context in which the sentence was uttered the locative reading can be excluded. However, it has to be admitted that the status of da in combination with a relative pronoun is ambiguous: although it is diachronically attested, as examples in 6.3.2, 6.3.5 and 6.4.1 demonstrate, it occurs nowadays only in non-standard varieties (cf. Fleischer 2004b: 66, quoted also in 5.2.3) and, as an archaizing feature, in superstandard varieties, especially in Bible translations. This can be explained as follows: while translating the Bible, Luther turned for (linguistic) inspiration to the actually spoken language (“dem Volk aufs Maul geschaut”), thus integrating in his text colloquial constructions. However, later on the language of his Bible translation was felt as an archaic, solemn style (thanks to Georg Bossong for drawing my attention to this fact). Cases in which a construction is nowadays attested in non-standard and superstandard varieties are not untypical: see 6.5.3 for further examples and a brief discussion. Interestingly, TUR çünkü was then further borrowed in colloquial BUL/MAC čunki(m) and ALB tshumë që. In fact, Turkish acted as a superordinate vehicular language in the Balkan peninsula from the 15th to the 19th centuries. The terms attractio relativi and attractio inversa were originally applied to Latin and Ancient Greek. They refer to cases in which the relative pronoun in the RC was case-attracted by the relativized item in the MC (attractio relativi) or, vice versa, the relativized item in the MC was case-attracted by the relative pronoun (attractio inversa). In both cases a syntactic mismatch arises in the RC or in the MC, as one constituent does not display the case required by the predicate (cf. Lehmann 1984: 185, 307). ‘Equi type’ is used in Comrie (1981: 147), ‘equi-case’ in Givón (1990: 666), ‘case matching’, among others, in Fleischer (2006: 226). All termns basically refer to the same phenomenon; the last one will be adopted in the following. ‘Decumulation’ stems from the corresponding French décumul (du relatif), used in Frei (1929: 187) and Guiraud (1966: 41) among others. Although this term is commonly adopted to indicate this phenomenon, it cannot be regarded as felicitous. In fact, scholars never speak of ‘accumulation’ when they refer to the encoding of all four parameters by means of a single morphosyntactic unit. This is probably because the latter case is assumed to be the default one, at least in European languages. Still, looking at Table 3.6 and Table 3.7, it

Notes 437

17.

18.

19.

20.

21.

appears that relative pronouns and specialized relative elements (‘accumulating’ strategies) are as frequent as the combination of a relative particle or a zero-marker with a resumptive element (‘decumulating’ strategies): both are attested in about 30 languages. Tesnière (1959) calls decumulation “disjonction de l’élément translatif et de l’élément anaphorique” (Tesnière 1959: 570). This may be seen as a more neutral definition which better suits the theoretical approach adopted here; however, as the term ‘disjunction’ is quite uncommon in literature on relative clauses, the traditional ‘decumulation’ will be used in the following. The term ‘case recoverability’ is introduced in Givón (1990: 650). Givón bases his typology of RC-formation strategies on the different ways that languages use to encode the “missing argument” in the RC, i.e. the relativized item in the RC. So, ‘case recoverability’ refers to the degree of explicitness with which the speaker encodes the relativized item in the RC as well as to the accessibility of the relativized item in the RC for the hearer. This is the result of a number of experiments conducted by Haig. Turkish native speakers were asked to translate into English a number of Turkish sentences in which the syntactic role of the relativized item in the RC was not encoded. Since in standard English the syntactic role of the relativized item in the RC usually needs to be expressed, the informants were forced to make a decision on the syntactic position relativized. So, Haig could establish which position is preferably relativized when more than one argument is not filled (Haig 1998: 132–140). Antinucci, Duranti, and Gebert conducted an experiment based on a comprehension task: a group of subjects was presented with three relative constructions containing the same lexical material and differing only in the relative element: the zero-marker in the first one, the relative particle that in the second one and the non-standard relative particle what in the third one. The subjects showed greater ease of comprehension in the constructions containing “an explicit ‘signal of interruption’ at the beginning of the RC” (Antinucci, Duranti and Gebert 1979: 165). This is true for the majority of the standard varieties in the sample. A noticeable exception is represented by the prenominal strategy of Turkish and Basque, in which the syntactic role of the relativized item can remain unexpressed. As mentioned above, in the world’s languages non-encoding of the syntactic role of the relativized item in the RC is quite widespread. So, the claim about standard varieties has an areal validity, restricted to the languages of the sample. The terms EXOGLOSSIC and ENDOGLOSSIC are used with reference to a roofing variety (cf. 1.4): if the roofing variety and the local varieties which it is supposed to roof are genetically closely related, the roofing variety is endoglossic; if not, it is exoglossic. For instance, Latin acted as an exoglossic standard for Polish local varieties from the 10th to the 18th centuries; later on,

438 Notes

22.

23.

24.

25. 26.

standard Polish replaced Latin and became the endoglossic standard for Polish local varieties (Lubaś 20062). In fact, standard Polish derives from a group of Polish local varieties, the Masurian ones (Urbańczyk 1939), which are genetically more closely related to other Polish local varieties than to Latin. There existed five so-called ‘redactions’ of Church Slavonic: a Bulgarian, a Russian, a Serbian, a Czech and a Croatian one. The last two disappeared when the western part of the Slavic territory fell under the influence of the Roman church. The Bulgarian and Serbian redactions lost their importance when the Balkan Peninsula was occupied by the Ottomans. Only the Russian redaction preserved its prestige and can be regarded as the actual prestigious language of culture and religion in Slavia Orthodoxa until the 18th century. It is also questionable whether the Russian Church Slavonic redaction should be regarded as an exoglossic or an endoglossic model language: it was surely more closely related to Russian local varieties than Latin was to Polish or Czech local varieties; moreover, it had developed typical East-Slavic traits, which made it closer to Russian local varieties than its South-Slavic-based ancestor Old Church Slavonic. Further information on Old Church Slavonic and Church Slavonic redactions can be found in Trunte (2003 5), Trunte (1998) and Holzer (2002). In Table 5.1 the strategy ‘RPAR (high positions of the AH)’ was marked as standard for Italian, Portuguese and Spanish because, as already stated in 4.2.1, it is impossible to distinguish between the relative pronoun che/que and the relative particle che/que in these languages as far as SU and DOrelativization is concerned. The term ‘degree of mastery’ – discussed in in Coseriu (1988: 148–158) and Lehmann (2007: 25) among others – can be defined as a speaker’s ability to activate, use and/or understand the constructions attested in a linguistic variety and their functions within a communicative practice. That is, the degree of mastery includes both the active and the passive competence of a variety. The forms *kъto and *čьto are used throughout this chapter to refer to all slightly different forms of ‘who’ and ‘what’ in Slavic languages. Schafroth quotes this example from Herman (1963: 68). The Latin text is strongly reduced. The original is quoted and translated here: “Quid est enim praeclarius quam honoribus et rei publicae muneribus perfunctum senem posse suo iure dicere idem, quod apud Ennium dicat ille Pythius Apollo, se esse eum, unde sibi, si non populi et reges, at omnes sui cives consilium expetant, summarum rerum incerti: […]” (free translation: “So, what is more illustrious for an old man who has made his career and has carried out his duty for the state than the chance of rightly saying the same thing that in Ennius’ [Annals] Pythius Apollo says, i.e. that he is the one from whom maybe not peoples and kings, but surely all his fellow citizens insistently try to obtain advice for themselves when they are uncertain about vital questions?”).

Notes 439 27. Cases of double encoding were attested also in Old Church Slavonic (Večerka 2002: 185). They usually appeared in translations of the Bible. Večerka relates this use to a calque of the Ancient Greek source, which, in turn, drew from the Hebrew original text: double encoding would derive from the adaptation of the Hebrew strategy ‘RPAR+resumptive element’ to Ancient Greek, which did not display any relative particle. A similar remark is made in Schafroth (1993: 86 fn. 26) for Latin and Romance languages. However, Schafroth points out that this issue is quite controversial. 28. As was mentioned in 5.1.2, some Slavic languages received a later codification and were less exposed to the influence of a prestigious model language, which explains their greater tolerance towards relativization strategies like ‘RPAR’ and ‘RPAR+resumptive element’. 29. ‘Well, I’m, how shall I say, I’m only an observer and, like people say, in my opinion my language is purely Russian. Most likely, language doesn’t display any kind of this… any kind of variation.’ (Informant 13, RRR98: 128) 30. Since the interviewers use razgovornaja reč’, one would expect the informants to adapt to the communicative context. In fact, some of them do interpret the situation as an official interview and adopt a formal register, quite close to literaturnyj jazyk; others see the interview as an easy-going talk and use razgovornaja reč’. Some informants are continuously looking for the ‘right words’: whether this happens because they are not at ease with the variety that they are trying to adopt or whether it depends on other factors cannot be easily established. To this extent see also Lehmann (2007: 25): “Wenn jemand eine Sprache weniger gut beherrscht, liegt das an minderer Begabung, an minderer bzw. anders orientierter Motivation, am sozialen Umfeld, an schlechter Ausbildung oder vielleicht noch anderen Faktoren? Wer hierüber empirisch forschen will, begibt sich auf vermintes Gelände.”

References

Akimova, Galina N. 1964 Tendencii v razvitii otnositel’nogo podčinenija v sovremennych vostočnoslavjanskich jazykach [Tendencies in the development of relative subordination in contemporary East Slavic languages]. Izvestija Akademii Nauk SSSR. Serija Literatury i Jazyka 23 (2): 138– 144. Albrecht, Jörn 2003 Die deutsche Sprache innerhalb der Architektur europäischer Einzelsprachen. Sociolinguistica 17: 11–31. Alfonzetti, Giovanna 2002 La relativa non-standard: italiano popolare o italiano parlato? Palermo: Centro di studi filologici e linguistici siciliani. Allan, Robin, Philip Holmes, and Tom Lundskær-Nielsen 1995 Danish: a comprehensive grammar. London/New York: Routledge. Ammon, Ulrich 1995 Die deutsche Sprache in Deutschland, Österreich und der Schweiz. Das Problem der nationalen Varietäten. Berlin/New York: Mouton de Gruyter. 20042 Standard variety. In Ammon et al. (eds.), 273–283. Ammon, Ulrich, Norbert Dittmar, Klaus J. Mattheier, and Peter Trudgill (eds.) 20042 Sociolinguistics. An international handbook of the science of language and society. Vol. 1. Berlin/New York: Mouton de Gruyter. 20062 Sociolinguistics. An international handbook of the science of language and society. Vol. 3. Berlin/New York: Mouton de Gruyter. Andrade Peres, João and Telmo Móia 1995 Áreas críticas da língua portuguesa. Lisboa: Caminho. Antinucci, Francesco, Alessandro Duranti, and Lucyna Gebert 1979 Relative clause structure, relative clause perception, and the change from SOV to SVO. Cognition 7: 145–176. Ariel, Mira 1999 Cognitive universals and linguistic conventions: the case of resumptive pronouns. Studies in language 23 (2): 217–269. Arim, Eva, Maria Celeste Ramilo, and Tiago Freitas 2005 Mudança em curso e os média: o caso das relativas [Ongoing changes and mass media: the case of relative clauses]. In Maria Helena Mira Mateus and Fernanda Bacelar do Nascimento (eds), A língua portuguesa em mudança [Changes in Portuguese], 67–80. Lisboa: Caminho.

References 441 ASIt, Atlante Sintattico d’Italia. URL: http://asis-cnr.unipd.it. Last access: June 30th, 2011. Atrachovič, Kandrat K. (ed.) 1966 Hramatyka belaruskaj movy [Belarusian grammar]. Vol. 2. Minsk: Navuka i tėchnika. Auer, Peter 1990 Einige umgangssprachliche Phänomene des Türkischen und ihre Erklärung aus ‘natürlichen’ Prinzipien. In Boretzky, Enninger, and Stolz (eds.), Vol. 2, 271–298. 2000 On-line-Syntax – Oder: was es bedeuten könnte, die Zeitlichkeit der mündlichen Sprache ernst zu nehmen. Sprache und Literatur 85: 43– 56. 2002 Schreiben in der Hypotaxe – Sprechen in der Parataxe? Kritische Bemerkungen zu einem Gemeinplatz. DaF 39: 131–138. 2004 Non-standard evidence in syntactic typology – Methodological remarks on the use of dialect data vs. spoken language data. In Kortmann (ed.), 69–92. 2005a Syntax als Prozess. InList – Interaction and linguistic structures 41. URL: http://www.uni-potsdam.de/u/inlist/issues/41/index.htm. Last access: June 30th, 2011. 2005b Europe’s sociolinguistic unity. In Nicole Delbecque, Johan van der Auwera, and Dirk Geeraerts (eds.), Perspectives on variation: sociolinguistic, historical, comparative, 7–42. Berlin/New York: Mouton de Gruyter. van der Auwera, Johan 1985 Relative that – a centennial dispute. Journal of Linguistics 21: 149– 179. van der Auwera, Johan and Dubravko Kučanda 1985 Pronoun or conjunction – the Serbo-Croatian invariant relativizer što. Linguistics 23: 971–962. Avanesov, Ruben I. and Varvara G. Orlova (eds.) 19652 Russkaja dialektologija [Russian dialectology]. Moskva: Nauka. Bagłajewska-Miglus, Ewa 1991 Der restriktive Relativsatz im Italienischen und Polnischen. Eine vergleichende Untersuchung. Tübingen: Niemeyer. Balliet, Pierre 1997 La relative en Alsace Bossue. Stuttgart: Hans-Dieter Heinz. Banfi, Emanuele and Nicola Grandi 2003 Lingue d’Europa. Elementi di storia e di tipologia linguistica. Roma: Carocci.

442 References Barbiers, Sjef, Hans Bennis, Gunther De Vogelaer, Magda Devos, and Margreet van der Ham 2005 Syntactic atlas of the Dutch dialects. Volume 1: pronouns, agreement, and dependencies. Amsterdam: Amsterdam University Press. Bartnicka, Barbara, Björn Hansen, Wojtek Klemm, Volkmar Lehmann, and Halina Satkiewicz 2004 Grammatik des Polnischen. München: Sagner. Bauer, Jaroslav 1962 K voprosu o vozniknovenii i razvitii tipov složnogo predloženija (na materiale češskogo jazyka) [The issue of the rise and development of different types of complex sentences (evidence from Czech)]. Voprosy jazykoznanija 1962 (6): 89–111. 1967 K vývoji vztažných vět v slovanských jazycích [The development of relative clauses in Slavic languages]. Slavica Slovaca 2: 297–320. Beličová, Helena and Jan Sedláček 1990 Slovanské souvetí [Clauses in Slavic]. Praha: Academia. Bellini, Simona 1998 I pronomi relativi nell’italiano parlato: un’indagine basata sui corpora LIP. Berlin: FU Berlin. Berger, Tilman 2000 Die Rolle des Tschechischen in der heutigen Slowakei. In Panzer (ed.), 179–196. Bernini, Giuliano 1989 Tipologia delle frasi relative italiane e romanze. In Fabio Foresti, Elena Rizzi, and Paola Benedini, L’italiano tra le lingue romanze. Atti del XX congresso internazionale di studi della SLI, Bologna, 25–27 settembre 1986, 85–98. Roma: Bulzoni. Bernini, Giuliano and Paolo Ramat 1992 La frase negativa nelle lingue d’Europa. Bologna: Il Mulino. Bernini, Giuliano and Marcia L. Schwartz (eds.) 2006 Pragmatic organization of discourse in the languages of Europe. Berlin/New York: Mouton de Gruyter. Berruto, Gaetano 1987 Sociolinguistica dell’italiano contemporaneo. Roma: Nuova Italia Scientifica. 1993a Le varietà del repertorio. In Sobrero (ed.), 3–36. 1993b Varietà diamesiche, diastratiche, diafasiche. In Sobrero (ed.), 37–91. 2005 Dialect convergence, mixing, and models of language contact: the case of Italy. In Peter Auer, Frans Hinskens, and Paul Kerswill (eds.). Dialect change, 81–95. Cambridge: Cambridge University Press. 2007 Miserie e grandezze dello standard. Considerazioni sulla nozione di standard in linguistica e sociolinguistica. In Piera Molinelli (ed.), Standard e non standard tra scelta e norma. Atti del XXX Convegno

References 443 annuale della SIG, Bergamo, 20–22 ottobre 2005, 13–41. Roma: Il Calamo. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 Longman grammar of spoken and written English. Harlow: Pearson Education Ltd. Bierich, Alexander 2000 Zur gegenwärtigen Situation der substandardsprachlichen Varietäten im Russischen. In Panzer (ed.), 13–30. Birkner, Karin 2008 Relativ(satz)konstruktionen im gesprochenen Deutsch: syntaktische, prosodische, semantische und pragmatische Aspekte. Berlin/New York: Mouton de Gruyter. Bisang, Walter 2004 Dialectology and typology – An integrative perspective. In Kortmann (ed.), 11–45. Blanche-Benveniste, Claire 1990 Usages normatifs et non normatifs dans les relatives en français, en espagnol et en portugais. In Johannes Bechert, Giuliano Bernini, and Claude Buridant (eds.), Toward a typology of European languages, 317–335. Berlin/New York: Mouton de Gruyter. Boeschoten, Hendrik and Lars Johanson (eds.) 2006 Turkic languages in contact. Wiesbaden: Harrassowitz. Boretzky, Norbert, Werner Enninger, and Thomas Stolz (eds.) 1989–1990 Spielarten der Natürlichkeit – Spielarten der Ökonomie. Beiträge zum 5. Essener Kolloquium. 3 Vols. Bochum: Brockmeyer. Borg, Albert and Maria Azzoppardi-Alexander 1997 Maltese. London/New York: Routledge. Bosque, Ignacio and Violeta Demonte 1999 Gramática descriptiva de la lengua española. Vol. 1. Madrid: Espasa Calpe. Bossong, Georg 1991 Differential object marking in Romance and beyond. In Dieter Wanner and Douglas A. Kibbee (eds.), New Analyses in Romance Linguistics, Selected Papers from the XVIII Linguistic Symposium on Romance Languages 1988, 143–170. Amsterdam: Benjamins Brincat, Giuseppe 2003 Malta. Una storia linguistica. Genova: Le mani. Brunot, Ferdinand 19363 La pensée et la langue. Paris: Masson. Buchholz, Eva 2004 Grammatik der finnischen Sprache. Bremen: Hempen.

444 References Bužarkova, Eleni 2004 Subjunctive relatives in Balkan languages. In Olga Mišeska-Tomić (ed.), Balkan syntax and semantics, 377–404. Amsterdam/Philadelphia: Benjamins. Carter, Ronald and Michael McCarthy 2006 Cambridge grammar of English. Cambridge: Cambridge University Press. Chambers, Jack K. 2004 The search for vernacular universals. In Kortmann (ed.), 180–210. Chambers, Jack K. and Peter Trudgill 19982 Dialectology. Cambridge: Cambridge University Press. Cheshire, Jenny and Dieter Stein 1997 The syntax of spoken language. In Cheshire and Stein (eds.), 1–12. Cheshire, Jenny and Dieter Stein (eds.) 1997 Taming the vernacular. From dialect to written standard language. London/New York: Longman. Cojocaru, Diana 2003 Romanian grammar. URL: http://www.seelrc.org:8080/grammar/ mainframe.jsp?nLanguageID=5. Last access: June 30th, 2011. Comrie, Bernard 1981 Language universals and linguistic typology. Oxford: Basil Blackwell. 1998 Rethinking the typology of relative clauses. Language design 1: 59– 86. 2002 Rethinking relative clause types: the Mediterranean area. In Ramat and Stolz (eds.), 87–98. 2003 Typology and language acquisition: the case of relative clauses. In Anna Giacalone Ramat (ed.), Typology and second language acquisition, 19–38. Berlin/New York: Mouton de Gruyter. Comrie, Bernard and Tania Kuteva 2005 Relativization strategies. In Martin Haspelmath, Matthew S. Dryer, David Gil, and Bernard Comrie (eds.), The world atlas of language structures, 494–501. Oxford: Oxford University Press. Corpus LIP. Banca dati dell’italiano parlato. URL: http://badip.uni-graz.at/. Last access: June 30th, 2011. Coseriu, Eugenio 1988 Sprachkompetenz. Tübingen: Francke. Cramer, Christina E. 2003 Macedonian. A course for beginning and intermediate students. Madison: University of Winsconsin Press.

References 445 Cresti, Emanuela and Massimo Moneglia 2005 C-ORAL-ROM: integrated reference corpora for spoken Romance languages (with DVD-ROM: Encrypted multimedia corpus). Amsterdam/Philadelphia: Benjamins. Cristofaro, Sonia 2003 Subordination. Oxford: Oxford University Press. 2006 The organization of reference grammars: a typologist user’s point of view. In Felix K. Ameka, Alan Dench, and Nicholas Evans (eds.), Catching language. The standing challenge of grammar writing, 137–170. Berlin/New York: Mouton de Gruyter. Cristofaro, Sonia and Anna Giacalone Ramat 2002 Relativization patterns in Mediterranean languages, with particular reference to the relativization of time circumstantials. In Ramat and Stolz (eds.), 99–112. 2007 Relativization strategies in the languages of Europe. In Ramat and Roma (eds.), 63–93. Croft, William 1995 Autonomy and functionalist linguistics. Language 71 (3): 490–531. 2001 Radical construction grammar. Oxford: Oxford University Press. Cuendet, Georges 1939 Sur l’expansion de la particule relative. In Mélanges de linguistique offerts à Charles Bally, 93–100. Genève: Georg et Cie. Damourette, Jacques and Edouard Pichon 1927–1952 Des mots à la pensée. Essai de grammaire de la langue française. 7 Vols. Paris: D’Artrey. Dasinger, Lisa and Cecile Toupin 1994 The development of relative clause functions in narrative. In Ruth A. Berman and Dan I. Slobin (eds.), Relating events in narrative: a cross-linguistic developmental study, 457–514. Hillsdale and Hove (NJ): Erlbaum. Dekeyser, Xavier 1984 Relativicers [sic!] in early Modern English: a dynamic quantitative study. In Fisiak (ed.), 61–88. Desző, László 1982 Typological studies in old Serbo-Croatian syntax. Köln/Wien: Böhlan. Deutscher, Guy 2006 The unfolding of language: the evolution of mankind’s greatest invention. London: Arrows books. Dmitriev, Pëtr A. 1972 Pridatočnye predloženija s dvustoronnimi sintaksičeskimi svjazami [Subordinate clauses with double syntactic links]. Južnoslovenski filolog 29 (1/2): 243–274.

446 References Donaldson, Bruce C. 1997 Dutch: a comprehensive grammar. London/New York: Routledge. Downing, Bruce 1978 Some universals of relative clause structure. In Joseph Greenberg (ed.), Universals of human language. Vol. 4: Syntax, 375–418. Stanford: Stanford University Press. Druviete, Ina 20062 The Baltic States. In Ammon et al. (eds.), 1864–1873. DuBois, John A. 1985 Competing motivations. In John A. Haiman (ed.), Iconicity in syntax, 343–366. Amsterdam/Philadelphia: Benjamins. Dudenredaktion (ed.) 20098 Duden. Die Grammatik. Mannheim/Wien/Zürich: Dudenverlag. Duličenko, Aleksandr D. 20065 Das Russinische. In Rehder (ed.), 126-140. Dürscheid, Christa 1999 Zwischen Mündlichkeit und Schriftlichkeit: die Kommunikation im Internet. Papiere zur Linguistik 60 (1): 17–30. Dürscheid, Christa 2003 Medienkommunikation im Kontinuum von Mündlichkeit und Schriftlichkeit. Theoretische und empirische Probleme. Zeitschrift für Angewandte Linguistik 38: 37–56. Ebert, Robert Peter 1993 Syntax. In Oskar Reichmann and Klaus-Peter Wegera (eds.), Frühneuhochdeutsche Grammatik, 313–485. Tübingen: Niemeyer. Endzelin, Jānis 1923 Lettische Grammatik. Heidelberg: Winter. Engel, Ulrich 2004 Deutsche Grammatik. Neubearbeitung. München: Iudicium. Erdal, Marcel 2004 A grammar of Old Turkic. Leiden/Boston: Brill. Erkman-Akerson, Fatma and Şeyda Ozıl 1998 Türkçede niteleme: sıfat işlevli yan tümceler [Qualification in Turkish: adjectival subordinate clauses]. İstanbul: Simurg. Erelt, Marti 2003 Syntax. In Marti Erelt (ed.), Estonian language, 93–128. Tallinn: Estonian Academy Publisher. Faßke, Helmut 1996 Sorbischer Sprachatlas. Vol. 15: Syntax. Bautzen (Budyšin): Domowina. Ferguson, Charles A. 1959 Diglossia. Word 15: 325–340.

References 447 Fiehler, Reinhard 2000a Gesprochene Sprache – Gibt’s die? Jahrbuch der ungarischen Germanistik 2000: 93–104. 2000b Über zwei Probleme bei der Untersuchung gesprochener Sprache. Sprache und Literatur 85: 23–42. Fiehler, Reinhard, Birgit Barden, Mechthild Elstermann, and Barbara Kraft 2004 Eigenschaften gesprochener Sprache. Theoretische und empirische Untersuchungen zur Spezifik mündlicher Kommunikation. Tübingen: Narr. Fiorentino, Giuliana 1999 Relativa debole. Sintassi uso storia in italiano. Milano: Franco Angeli. 2007 European relative clauses and the uniqueness of the relative pronoun type. Italian journal of linguistics 19 (2): 263–291. Fisiak, Jan (ed.) 1984 Historical syntax. Berlin/New York: Mouton. Fleischer, Jürg 2002 Die Syntax von Pronominaladverbien in den Dialekten des Deutschen. Eine Untersuchung zu Preposition Stranding und verwandten Phänomenen. Stuttgart: Franz Steiner Verlag. 2004a A typology of relative clauses in German dialects. In Kortmann (ed.), 211–244. 2004b Zur Typologie des Relativsatzes in den Dialekten des Deutschen. In Franz Patocka and Peter Wiesinger (eds.), Morphologie und Syntax deutscher Dialekte und historische Dialektologie des Deutschen, 60– 83. Wien: Praesens. 2006 Dative and indirect object in German dialects: evidence from relative clauses. In Daniel Hole and André Meinunger (eds.), Datives and similar cases, 213–238. Amsterdam/Philadelphia: Benjamins. Fontein, A.M. and A. Pescher-ter Meer 19963 Nederlandse grammatica voor anderstaligen. Utrecht: Nederlands centrum buitenlanders. Forgács, Tamás 2001 Ungarische Grammatik. Wien: Praesens. Fox, Barbara A. 1987 The noun phrase accessibility hierarchy reinterpreted: subject primacy or the absolutive hypothesis? Language 63: 856–870. Fox, Barbara A. and Sandra A. Thompson 2007 Relative Clauses in English conversation: Relativizers, frequency, and the notion of construction. Studies in language 31 (2): 293–326. Frei, Henri 1929 La grammaire des fautes. Paris: Geuthner.

448 References Friedman, Victor A. 2002 Macedonian. München: Lincom Europa. 20062 Albania. In: Ammon et al. (eds.), 1874–1880. Gadet, Françoise 1997 Le français ordinaire. Paris: Armand Colin. Gallis, Arne 1958 Flektiertes Relativum und Relativum generale, insbesondere im Serbokroatischen. Scandoslavica 4: 137–148. Gapany, Joël 2004 Formes et fonctions des relatives en français. Bern: Peter Lang. Gärtner, Eberhard 1998 Grammatik der portugiesischen Sprache. Tübingen: Niemeyer. Gärtner, Hans-Martin 2001 Are there V2 relative clauses in German? Journal of comparative German linguistics 3: 97–141. Genadieva-Mutafčieva, Zara, Stan’o Georgiev, Elena Georgieva, Aksela Lazarova, and Ruselina Nicolova 19983 Gramatika na săvremennija bălgarski knižoven ezik [Grammar of modern standard Bulgarian]. Vol. 3 (2): Sintaksis. Čast vtora [Syntax. Part 2]. Sofija: Abagar. Gessner, Emil 1894 Das spanische Relativ- und Interrogativpronomen. Zeitschrift für romanische Philologie 18: 449–497. Giacalone Ramat, Anna 1982 Explorations on syntactic change (relative clause formation strategies). In Anders Ahlqvist (ed.), Papers from the 5th international conference on historical linguistics, 283–292. Amsterdam/Philadelphia: Benjamins. 2005 Persistence and renewal in the relative pronoun paradigm: the case of Italian. Folia linguistica historica 26 (1/2): 115–138. Givón, Talmy 1979 From discourse to syntax: grammar as a processing strategy. In Talmy Givón (ed.), Discourse and syntax, 81–112. New York: Academic Press. 1990 Syntax: a functional-typological introduction. Vol. 2. Amsterdam/ Philadelphia: Benjamins. Göksel, Aslı and Celia Kerslake 2005 Turkish. A comprehensive grammar. London/New York: Routledge. Gołąb, Zbigniew and Victor A. Friedman 1972 The relative clause in Slavic. In Perenteau, Levi, and Phares (eds.), 30–46.

References 449 Goldberg, Adele 2006 Constructions at work: the nature of generalization in language. Oxford: Oxford University Press. Graur, Aleksandru (ed.) 1963 Gramatica limbii române [Romanian grammar]. Vol. 2. Bucuresti: Editura academiei republicii socialiste rumâne. Grdina, Igor and Marko Stabej 2002 Slowenisch. In Okuka (ed.): 495–508. Guiraud, Pierre 1966 Le système du relatif en français populaire. Langages 1 (3): 40–68. Günthner, Susanne 2002 Zum kausalen und konzessiven Gebrauch des Konnektors wo im gesprochenen Umgangsdeutsch. Zeitschrift für Germanistische Linguistik 30: 310–341. Gutschmidt, Karl 2002 Bulgarisch. In Okuka (ed.), 219–234. 20062 Die ostslawische Region. In Ammon et al. (eds.), 1851–1863. Haarmann, Harald 1993 Die Sprachenwelt Europas. Frankfurt/New York: Campus. 20042 Abstandsprache – Ausbausprache. In Ammon et al. (eds.), 238–250. Haase, Martin 1992 Sprachkontakt und Sprachwandel im Baskenland. Hamburg: Buske. Haig, Geoffrey 1998 Relative constructions in Turkish. Wiesbaden: Harrassowitz. Hanser, Oskar 1973/1974 Die Nebensatzgrammatik des Türkischen, untersucht an ausgewählten Beispielen. Wiener Zeitschrift für die Kunde des Morgenlandes 65/66: 155–218. Haspelmath, Martin 1998 How young is Standard Average European? Language sciences 20 (3): 271–287. 2001 The European linguistic area: Standard Average European. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds.), Language typology and language universals. An international handbook. Vol. 2, 1492–1510. Berlin/New York: Mouton de Gruyter. Heine, Bernd and Tania Kuteva 2005 Language contact and grammatical change. Cambridge: Cambridge University Press. 2006 The changing languages of Europe. Oxford: Oxford University Press. 2007 The genesis of grammar: a reconstruction. Oxford: Oxford University Press.

450 References Henn-Memmesheimer, Beate 1997 The patternings of nonstandard syntax in German. In Cheshire and Stein (eds.), 232–249. Hentschel, Elke and Harald Weydt 20033 Handbuch der deutschen Grammatik. Berlin/New York: Mouton de Gruyter. Herman, József 1963 La formation du système roman des conjonctions de subordination. Berlin: Akademie-Verlag. Herrity, Peter 2000 Slovene. A comprehensive grammar. London/New York: Routledge. Herrmann, Tanja 2005 Relative clauses in English dialects of the British Isles. In: Kortmann, Herrmann, Pietsch, and Wagner, 21–124. Hinrichs, Uwe 1994 Balkanische Konvergenz und das Standard/Nonstandard-Problem. In Norbert Reiter, Uwe Hinrichs, and Jiřina van Leeuwen-Turnovcová (eds.), Sprachlicher Standard und Substandard in Südosteuropa und Osteuropa. Beiträge zum Symposium vom 12.–16. Oktober 1992 in Berlin, 96–113. Wiesbaden: Harrassowitz. Hock, Wolfgang 20065 Das Altkirchenslawische. In Rehder (ed.), 35-48. Hoekstra, Jarich 2002 Relativisation in Frisian. In Poussa (ed.), 63–75. Holmes, Philip and Ian Hinchliffe 1994 Swedish. A comprehensive grammar. London/New York: Routledge. Holton, David, Peter Mackridge and Irene Philippaki-Warburton 1997 Greek: a comprehensive grammar of the modern language. London/ New York: Routledge. Holzer, Georg 2002 Altkirchenslawisch. In Okuka (ed.), 187–202. Hualde, José Ignacio 1992 Catalan. London/New York: Routledge. Hurch, Bernhard 1989 Hispanisierung im Baskischen. In: Boretzky, Enninger and Stolz (eds.), Vol. 1, 11–36. Iliescu, Maria 2002 Rumänisch. In Okuka (ed.), 145–166. Janaš, Pětr 1976 Niedersorbische Grammatik. Bautzen (Budyšin): Domowina. Johanson, Lars 2006 Turkic language contact in a typology of code interaction. In Boeschoten and Johanson (eds.), 4–26.

References 451 Joseph, Brian D. and Irene Philippaki-Warburton 1987 Modern Greek. London/New York: Routledge. Karlsson, Fredrik and Kirk P. H. Sullivan 2002 The use of relativization in the regional dialect of Swedish spoken in Burträsk. In Poussa (ed.), 97–107. Katičić, Radoslav 1986 Sintaksa hrvatskoga književnog jezika [Syntax of standard Croatian]. Zagreb: Globus. Keenan, Edward L. 1985 Relative clauses. In Timothy Shopen (ed.), Language typology and syntactic description. Vol. 2: Complex constructions, 141–168. Cambridge: Cambridge University Press. Keenan, Edward L. (ed.) 19872 Universal grammar: 15 essays. London/Sydney: Croom Helm. Keenan, Edward L. and Bernard Comrie 1977 Noun Phrase Accessibility and Universal Grammar. Linguistic inquiry 8 (1): 63–99. 1979a Data on Noun Phrase Accessibility Hierarchy. Language 55 (2): 333–351. 1979b Noun Phrase Accessibility Hierarchy revisited. Language 55 (3): 649–664. 19872 Noun Phrase Accessibility Hierarchy and Universal Grammar. In Keenan (ed.), 3–45. Keenan, Edward L. and Sarah Hawkins 19872 The psychological validity of the Accessibility Hierarchy. In Keenan (ed.), 60–88. Kenesei, István, Robert M. Vago, and Anna Fenyvesi 1998 Hungarian. London/New York: Routledge. Kleiber, Georges 1987 Relatives restrictives et relatives appositives: une opposition “introuvable”? Tübingen: Niemeyer. Kloss, Heinz 1967 Abstand languages and ausbau languages. Anthropological linguistics 9 (7): 29–41. Koch, Peter and Wulf Oesterreicher 1985 Sprache der Nähe – Sprache der Distanz: Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte. Romanistisches Jahrbuch 36: 15–43. 1990 Gesprochene Sprache in der Romania: Französisch, Italienisch, Spanisch. Tübingen: Niemeyer.

452 References König, Ekkehard and Peter Siemund 2007 Speech act distinctions in grammar. In Timothy Shopen (ed.). Language typology and syntactic description, 276–324. Cambridge: Cambridge University Press. Kordić, Snježana 1999 Der Relativsatz im Serbokroatischen. München: Lincom Europa. Kornfilt, Jaklin 1997 Turkish. London/New York: Routledge. Korš, Fëdor J. 1877 Sposoby otnositel’nogo podčinenija. Glava iz sravnitel’nogo sintaksisa [Types of relative subordination. A study in comparative syntax]. Moskva: Univ. Tipografija. Kortmann, Bernd 1997 Adverbial subordinators. A typology and history of adverbial subordinators based on European languages. Berlin/New York: Mouton de Gruyter. 2002 New prospects for the study of dialect syntax: impetus from syntactic theory and language typology. In: Sjef Barbiers, Leonie Cornips, and Susanne van der Kleij (eds.), Syntactic microvariation, 185–213. Amsterdam: SAND. Kortmann, Bernd (ed.) 2004 Dialectology meets typology. Berlin/New York: Mouton de Gruyter. Kortmann, Bernd, Tanja Herrmann, Lukas Pietsch, and Susanne Wagner 2005 A comparative grammar of British English dialects: agreement, gender, relative clauses. Berlin/New York: Mouton de Gruyter. Křížková, Helena 1970 Relativní věty v současných slovanských jazycích [Relative clauses in contemporary Slavic langauges]. Slavia 39: 10–40. Kručinina, Irina N. 1968 Konstrukcii s mestoimeniem kotoryj v sovremennom russkom jazyke [The constructions with the pronoun kotoryj in contemporary Russian]. Voprosy jazykoznanija 1968 (2): 82–88. Kuno, Susumo 1976 Subject, theme and the speaker’s empathy – A reexamination of relativization phenomena. In Charles N. Li (ed.), Subject and topic, 417– 444. New York: Academic Press. Kunstmann, Pierre 1990 Le relatif-interrogatif en ancien français. Genf: Librairie Droz. Kurzová, Helena 1981 Der Relativsatz in den indoeuropäischen Sprachen. Hamburg: Buske.

References 453 Kuz’mina, Irina B. 1993 Sintaksis russkich govorov v lingvogeografičeskom aspekte [The syntax of Russian dialects from a geolinguistic perspective]. Moskva: Nauka. Lambertz, Maximilian 1959 Lehrbuch des Albanischen. Vol. 3: Grammatik der albanischen Sprache. Berlin: Deutscher Verlag der Wissenschaften. Lambrecht, Knud 1994 Information structure and sentence form. Cambridge: Cambridge University Press. Lambton, Ann K. S. [1953] 1974 Persian grammar. Cambridge: Cambridge University Press. Lamprecht, Arnošt, Dušan Šlosar, and Jaroslav Bauer 1986 Historická mluvnice češtiny [Czech historical grammar]. Praha: Statní pedagogické nakladatelství. Lapteva, Ol’ga A. 1976 Russkij razgovornyj sintaksis [The syntax of spoken Russian]. Moskva: Nauka. Laskowski, Roman 1991 Status gramatyczny wskaźnika syntaktycznego zespolenia w zdaniach względnych z co [The grammatical status of the indicator of syntactic junction in relative clauses with co]. In Maciej Grochowski and Daniel Weiss (eds.), “Words are physicians for an ailing mind”, 271–277. München: Sagner. Lehmann, Christian 1979 “Der Relativsatz vom Indogermanischen bis zum Italienischen”. Die Sprache 25: 1–25. 1984 Der Relativsatz. Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: Narr. 1988 Towards a typology of clause linkage. In John Haiman and Sandra A. Thompson (eds.), Clause combining in grammar and discourse, 181–225. Amsterdam/Philadelphia: Benjamins. 1995 Relativsätze. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, and Theo Vennemann (eds.), Syntax. Ein internationales Handbuch. Vol. 2, 1199–1216. Berlin/New York: Mouton de Gruyter. 2007 Daten – Korpora – Dokumentation. In Werner Kallmeyer and Gisela Zifonun (eds.), Sprachkorpora – Datenmengen und Erkenntnisfortschritt, 9–27. Berlin/New York: Mouton de Gruyter. Lešnerová, Šarka and Karel Oliva 2003 Česká vztažná souvětí s nestandardní strukturou” [Czech relative clauses with a non-standard structure]. Slovo a slovesnost 44: 241– 252.

454 References Lewis, Geoffrey 20002 Turkish Grammar. Oxford: Oxford University Press. Lockwood, William B. 1968 Historical German syntax. Oxford: Clarendon Press. Löffler, Heinrich 2005 Wieviel Variation verträgt die deutsche Standardsprache? Begriffsklärung: Standard und Gegenbegriffe. In Ludwig M. Eichinger and Werner Kallmeyer (eds.), Standardvariation. Wieviel Variation verträgt die deutsche Sprache? IDS-Jahrbuch 2004, 7–27. Berlin/New York: Mouton de Gruyter. López del Castillo, Lluís 1999 Gramàtica del català actual. Sintaxi i morfologia. Barcelona: Edicions 62. Lötscher, Andreas 2004 Dialektsyntax oder Syntax der gesprochenen Sprache? Eine Fallstudie anhand von Nebensatzproblemen im Schweizerdeutschen. Zeitschrift für Dialektologie und Linguistik 71: 156–179. Lubaś, Wladyslaw 20062 Die westslawischen Sprachen. In Ammon et al. (eds.), 1837–1850. Maček, Dora 1986/1987 Ambivalence of relative clause structure – A diachronic view. Studia romanica et anglica Zagrabiensia 30/31: 103–115. Mac Giolla Chríost, Diarmait 2005 The Irish language in Ireland. From Goídel to globalisation. London/New York: Routledge. van Marle, Jaap 1997 Dialect versus standard language: nature versus culture. In: Cheshire and Stein (eds.), 13–34. Maslova, Elena and Giuliano Bernini 2006 Sentence topics in the languages of Europe and beyond. In Bernini and Schwartz (eds.), 67–120. Matsumura, Kazuto 1985 Two types of relative clauses in Finnish. Gengo kenkyu [Language studies] 81: 60–82. Mathiassen, Terje 1997 A short grammar of Latvian. Columbus (OH): Slavica Publishers. Matzinger, Joachim 2006 Der altalbanische Text Mbsuame e Krështerë (Dottrina Cristiana) des Lekë Matrënga von 1592. Eine Einführung in die albanische Sprachwissenschaft. Dettelbach: Röll. Mauri, Caterina 2008 Coordination relations in the languages of Europe and beyond. Berlin/New York: Mouton de Gruyter.

References 455 Maxwell, Daniel 1979 Strategies of relativization and NP accessibility. Language 55 (2): 352–371. Mc Gonagle, Noel 1988 Irish grammar. A basic handbook. New York: Hyppocrene. Mel’nyčuk, Oleksandr S. 1962 Istoryčnyj rozvytok systemy vidnosnych sliv v ukrajinskij movi [The historical development of the system of relative elements in Ukrainian]. Slov-jans’ke movoznavstvo 4: 80–121. Menz, Astrid 2006 On complex sentences in Gagauz. In Boeschoten and Johanson (eds.), 139–151. Metzeltin, Michael 1979 Altspanisches Elementarbuch. Vol. 1: Das Altkastilische. Heidelberg: Winter. Meyer-Lübke, Wilhelm 1899 Grammatik der romanischen Sprachen. Vol. 3: Romanische Syntax. Leipzig: Reisland. Miller, Jim and Regina Weinert 1998 Spontaneous spoken language. Syntax and discourse. Oxford: Clarendon Press. Miller, Jim and M. M. Jocelyne Fernandez-Vest 2006 Spoken and written language. In Bernini and Schwartz (eds.), 9–64. Mirčev, Kiril 1955 Istoričeska gramatika na bălgarskija ezik [Bulgarian historical grammar]. Sofija: Narodna Prosveta. Mišeska-Tomić, Olga 2006 Balkan Sprachbund morpho-syntactic features. Dordrecht: Springer. Murelli, Adriano 2009 ‘Where’-based relative elements in Balkan languages. Balkansko Ezikoznanie / Linguistique Balkanique 49 (3): 183–195. 2010 Subject clitics in Lombard dialects. Evidence from the ASIS (Syntactic Atlas of Northern Italy). URL: http://www.freidok.uni-freiburg.de /volltexte/7549/. Last access: June 30th, 2011. 2011 Das Geheimnis, das oder was du mir verraten hast? — das oder was als Relativpronomen. In Grammatik in Fragen und Antworten. URL: http://hypermedia.ids-mannheim.de/pls/public/fragen.ansicht?v_typ= e&v_id=4531. Last access: June 30th, 2011. Murelli, Adriano and Bernd Kortmann 2011 Non-standard varieties in the areal typology of Europe. In Bernd Kortmann and Johan van der Auwera (eds.), The languages and linguistics of Europe: a comprehensive guide, 525–544. Berlin/Boston, Mass.: De Gruyter Mouton.

456 References Mustanoja, Tauno F. 1960 A Middle English syntax. Part 1: Parts of speech. Helsinki: Société Néophilologique. Neweklowsky, Gerhard 20062 Die südslawische Region. In Ammon et al. (eds.), 1824–1836. Nilsson, Elsa 1969 Les termes relatifs et les propositions relatives en roumain moderne. Etude de syntaxe descriptive. Lund: Berlingska Boktryckeriet. Noreen, Adolf 1904 Altschwedische Grammatik. Halle: Max Niemeyer. Okuka, Miloš (ed.) 2002 Lexikon der Sprachen des europäischen Ostens. Klagenfurt: Wieser. Oyharçabal, Bernard 1989 Pro-drop and the resumptive pronoun strategy in Basque. In László Marácz and Pieter Muysken (eds.), Configurationality. The typology of asymmetries, 63–83. Dordrecht: Foris. 2003 Relatives. In José Ignacio Hualde and Jon Ortiz de Urbina (eds.), A grammar of Basque, 762–823. Berlin/New York: Mouton de Gruyter. Panzer, Baldur 1992 Zur Geschichte der russischen Standardsprache: Identität, Kontinuität, Entwicklung. Sociolinguistica 6: 1–10. Panzer, Baldur (ed.) 2000 Die sprachliche Situation in der Slavia zehn Jahre nach der Wende. Frankfurt am Main: Peter Lang. Paul, Herrmann 199824 Mittelhochdeutsche Grammatik. Tübingen: Niemeyer. Penke, Martina and Annette Rosenbach 2004 What counts as evidence in linguistics? Studies in language 28 (3): 480–526. Peranteau, Paul M., Judith N. Levi, and Gloria C. Phares (eds.) 1972 The Chicago Which Hunt. Papers from the Relative Clause Festival, April 13, 1972. Chicago: Chicago Linguistic Society. Perini, Mário A. 2002 Modern Portuguese. A reference grammar. New Heaven/London: Yale University Press. Petr, Jan (ed.) 1987 Mluvnice češtiny [Czech grammar]. Vol 3: Skladba [Syntax]. Praha: Academia. Picchio, Riccardo 1984 Guidelines for a comparative study of the language question among the Slavs. In Riccardo Picchio and Harvey Goldblatt (eds.), Aspects

References 457 of the Slavic Language Question. Vol. 1, 1–42. New Haven: Yale Concilium on International and Areal Studies. Pittner, Karin 2004 Wo in Relativsätzen – eine korpusbasierte Untersuchung. Zeitschrift für Germanistische Linguistik 32: 357–375. Platzack, Christer 2002 Relativization in the Germanic languages, with particular emphasis on Scandinavian. In Poussa (ed.), 77–96. Poussa, Patricia (ed.) 2002 Relativization on the North Sea littoral. München: Lincom Europa. Pugh, Stefan M. and Ian Press 1999 Ukrainian. A comprehensive grammar. London/New York: Routledge. Rabus, Achim 2008 Die Sprache ostslavischer geistlicher Gesänge im kulturellen Kontext. Freiburg i. Br.: Weiher. 2010 Die Entwicklung der Relativisatoren im Ruthenischen. In Björn Hansen and Jasmina Grković-Major (eds.), Diachronic Slavonic syntax: gradual changes in focus, 155–168. München/Berlin: Kubon & Sagner. Radeva, Vasilka 2003 Bulgarische Grammatik. Hamburg: Buske. Ramat, Paolo 1984 ‘Es war ein König in Thule (...), Dem sterbend seine Buhle...’: on the rise and transformation(s) of morphosyntactic categories. In Fisiak (ed.), 393–416. Ramat, Paolo and Thomas Stolz (eds.) 2002 Mediterranean languages: papers from the MEDTYP workshop, Tirrenia, June 2000. Bochum: Universitätsverlag Dr. N. Brockmeyer. Ramat, Paolo and Elisa Roma (eds.) 2007 Europe and Mediterranean as linguistic areas: convergences from a historical and typological perspective. Amsterdam/Philadelphia: Benjamins. Ramza, Taccjana and Claudia Hurtig 2003 Belarussische Grammatik in Tabellen und Übungen. Hramatyka belarusskaj movy ŭ tablicach i praktykavannjach. München: Sagner. Rehder, Peter (ed.) 20065 Einführung in die slavischen Sprachen. Darmstadt: Wissenschaftliche Buchgesellschaft.

458 References Rendahl, Anne-Charlotte 2001 Swedish around the Baltic sea. In Östen Dahl and Maria Koptjevskaja-Tamm (eds.), The Circum-Baltic languages. Typology and contact, 137–178. Amsterdam/Philadelphia: Benjamins. Renzi, Lorenzo, Giampaolo Salvi and Anna Cardinaletti (eds.) 2001 Grande grammatica italiana di consultazione. Nuova edizione. Vol. 1: La frase. I sintagmi nominale e preposizionale. Bologna: Il Mulino. Riegel, Martin, Jean-Christoph Pellat, and René Rioul 1994 Grammaire méthodique du français. Puf: Paris. Roma, Elisa 2007 Relativization strategies: early Celtic-English contact. In Ramat and Roma (eds.), 245–288. Romaine, Suzanne 1980 The relative clause marker in Scots English: diffusion, complexity, and style as dimensions of syntactic change. Language in society 9: 221–237. 1984 Towards a typology of relative-clause formation strategies in Germanic. In Fisiak (ed.), 437–470. Rösler, Irmtraud 2002 Relative clauses in Low German (15th–16th century). In Poussa (ed.), 51–62. Rounds, Carol 2001 Hungarian. An Essential Grammar. London/New York: Routledge. Rusu, Valeriu (ed.) 1984 Tratat de dialectologie românească [A handbook of Rumanian dialectology]. Craiova: Scrisul românesc. Salzmann, Martin and Guido Seiler 2010 Variation as the exception or the rule? Swiss relatives, revisited. Sprachwissenschaft 35 (1): 79–117. Sandig, Barbara 1973 Zur historischen Kontinuität normativ diskriminierter syntaktischer Muster in spontaner Sprechsprache. Deutsche Sprache 3: 37–57. Sansò, Andrea 2005 The Pavia typological database 1.0. URL: http://www-3.unipv.it/ paviatyp/. Last access: June 30th, 2011. Šapiro, Abram B. 1953 Očerki po sintaksisu russkich narodnych govorov [An outline of Russian dialect syntax]. Moskva: Izdatel’stvo Akademii Nauk SSSR. Scherer, Carmen 2006 Korpuslinguistik. Heidelberg: Winter. Schlyter, Birgit N. 20062 Turkey. In: Ammon et al. (eds.), 1903–1910.

References 459 Scholten, Daniel 2000 Einführung in die isländische Grammatik. München: Philyra. Schroeder, Christoph 1997 ‘Relative’ ki-clauses and the structure of spoken Turkish. In Kâmile İmer and N. Engin Uzun (eds.), Proceedings of the 8th international conference on Turkish linguistics, August 7–9, 1996, 347–363. Ankara: Ankara Üniversitesi Basımevi. Schwartz, Arthur 1971 General aspects of relative clause formation. Working Papers on Language Universals 6: 139–171. Schweier, Ulrich 20065 Das Ukrainische. In Rehder (ed.), 94-103. Sergieva, Natal’ja S. and Aleksandr S. Gerd (eds.) 1998 Russkaja razgovornaja reč’ evropejskogo severo-vostoka Rossii. Sbornik tekstov [Colloquial spoken Russian of North-Western European Russia] Syktyvar: Syktyvarskij universitet. Sgall, Petr, Jirí Hronek, Alexandr Stich and Ján Horecký (eds.) 1992 Variation in language. Code-switching in Czech as a challenge for sociolinguistics. Amsterdam/Philadelphia: Benjamins. Širjaev, Evgenij N. 1986 Bessojuznoe složnoe predloženie v sovremennom russkom jazyke [Asyndetic clause clusters in contemporary Russian]. Moskva: Nauka. Slobin, Dan I. 1986 The acquisition and use of relative clauses in Turkic and Indo-European languages. In Dan I. Slobin and Karl Zimmer (eds.), Studies in Turkish linguistics, 273–294. Amsterdam/Philadelphia: Benjamins. Šlosar, Dušan 2002 Tschechisch. In Okuka (ed.), 513–534. Smits, Reinier J.C. 1989 Eurogrammar. The relative and cleft constructions of the Germanic and Romance languages. Dordrecht: Foris. Sneyders de Vogel, Kornelis 1919 Syntaxe historique du français. Groningen/Den Haag: Librairie Wolters. Sobrero, Alberto A. (ed.) 1993 Introduzione all’italiano contemporaneo. La variazione e gli usi. Roma/Bari: Laterza. Sornicola, Rosanna 1981 Sul parlato. Bologna: Il Mulino. Stark, Elisabeth 2009 Romance restrictive relative clauses between macrovariation and universal structures. PhiN. Philologie im Netz 47: 1–15. URL:

460 References http://web.fu-berlin.de/phin/phin47/p47t1.htm. Last access: June 30 th, 2011. Stein, Dieter 1997 Syntax and varieties. In Cheshire and Stein (eds.), 35–50. Stojkov, Stojko 19933 Bălgarska dialektologija [Bulgarian dialectology]. Sofija: Izdatelstvo na bălgarskata akademija na naukite. Stolz, Thomas 2006 Europe as a Linguistic Area. In Keith Brown (ed.), Encyclopaedia of language and linguistics. Vol. 4, 278–295. Amsterdam: Elsevier. Stotz, Peter 1994 Le sorti del latino nel Medioevo. In Guglielmo Cavallo, Claudio Leonardi, and Enrico Menestò (eds.), Lo spazio letterario del Medioevo 1. Il Medioevo latino. Vol. 2: La circolazione del testo, 153–190. Roma: Salerno. Strandskogen, Rolf and Åre-Berit Strandskogen 19862 Norwegian. An essential grammar. London/New York: Routledge. Sulkala, Helena and Merja Karjalainen 1992 Finnish. London/New York: Routledge. Sussex, Roland and Paul Cubberley 2006 The Slavic languages. Cambridge: Cambridge University Press. Švedova, Natalija Ju. (ed.) 1980 Russkaja Grammatika [Russian grammar]. Vol. 2: Sintaksis [Syntax]. Moskva: Nauka. Szczegielniak, Adam 2004 Two types of relative clauses in Slavic – Evidence from reconstruction and ellipsis. In Michal Temkin Martínez, Asier Alcázar, and Roberto Mayoral Hernández (eds.), Proceedings of the thirty-third western conference on linguistics, 373–384. Department of Linguistics, California State University: Fresno. URL: http://www.csufresno. edu/linguistics/WECOL%20Web/WECOL%20volumes%20online_fi les/WECOL%202004%20final%20electronic%20proceedings.pdf. Last access: June 30th, 2011. Tesnière, Lucien 1959 Éléments de syntaxe structurale. Paris: Klincksieck. Tilkov, Dimităr 1983 Gramatika na săvremennija bălgarski ezik [A grammar of contemporary Bulgarian] Vol. 3: Sintaksis [Syntax]. Sofija: Izdatelstvo na bălgarskata akademija na naukite. Toman, Jindřich 1998 A discussion of resumptives in colloquial Czech. In Željko Boškovič, Frank Stevens, and William Snyder (eds.), Formal approaches to

References 461 Slavic languages: The Connecticut meeting 1997, 303–318. Ann Arbor: Michigan Slavic Publications. Tompa, Jószef 1968 Ungarische Grammatik. The Hague/Paris: Mouton. Topolińska, Zuzanna 1984 Gramatyka współczesnego języka polskiego [A grammar of contemporary Polish]. Vol. 1: Składnia [Syntax]. Warszawa: Państwowe wydawnictwo naukowe. 1998 Polish że – all-powerful introducer of new clauses. In Maciej Grochowski and Gerd Hentschel (eds.), Funktionswörter im Polnischen, 219–237. Oldenburg: BIS. Tošović, Branko 2002 Russisch. In Okuka (ed.), 409–436. Trask, Robert L. 1998 The typological position of Basque: then and now. Language Sciences 20 (3): 313–324. Troickij, Veniamin I. 1959 K istorii sojuza čto v russkom jazyke [The history of the conjunction čto in Russian]. In Vladimir V. Vinogradov (ed.), Slavjanskoe jazykoznanie. Sbornik statej [Slavic linguistics. Collected articles], 160–169. Moskva: Izdatel’stvo Akademii Nauk SSSR. 1968 Otnositel’noe podčinenie v jazyke russkoj pis’mennosti XVI–XVII vekov [Relative subordination in Russian written documents of the 16th–17th centuries]. Kazan’: Izdatel’stvo Kazan’skogo universiteta. Trudgill, Peter and Daniel Schreier 20062 Greece and Cyprus. In Ammon et al. (eds.), 1881–1888. Trunte, Nikolaos H. 1998 Slavenskij jazykъ. Ein praktisches Lehrbuch des Kirchenslavischen in 30 Lektionen. Zugleich eine Einführung in die slavische Philologie. Vol. 2: Mittel- und Neukirchenslavisch. München: Sagner. 20035 Slověnьskyj językъ. Ein praktisches Lehrbuch des Kirchenslavischen in 30 Lektionen. Zugleich eine Einführung in die slavische Philologie. Vol. 1: Altkirchenslavisch. München: Sagner. Urbańczyk, Stanisław 1935 Wyparcie staropolskiego względnego jen, jenże przez pierwotnie pytajne który [The substitution of Old Polish relative jen, jenże through the originally interrogative który]. Kraków: Nakład polskiej akademii umiejętności. 1939 Zdania rozpoczynane wyrazem co w języku polskim [The clauses beginning with the expression co in Polish]. Kraków: Nakład polskiej akademii umiejętności. 19622 Zarys dialektologii polskiej [An outline of Polish dialectology]. Warszawa: Państwowe wydawnictwo naukowe.

462 References Usikova, Rina P. 2003 Grammatika makedonskogo literaturnogo jazyka [A grammar of standard Macedonian]. Moskva: Muravej. Večerka, Radoslav 2002 Altkirchenslavische (altbulgarische) Syntax. Vol. 4: Die Satztypen: der zusammengesetzte Satz. Freiburg i. Br.: Weiher. Vikør, Lars 20062 Scandinavia. In Ammon et al. (eds.), 1747–1753. Vintr, Josef 20065 Das Slowakische. In Rehder (ed.), 214-229. Villena Ponsoda, Juan Andrés 20062 The Iberian Peninsula. In Ammon et al. (eds.), 1802–1810. Vondrák, Wenzel 19282 Vergleichende slavische Grammatik. Vol. 2: Formenlehre und Syntax. Göttingen: Vandenhoeck & Ruprecht. de Vries, Mark 2002 The syntax of relativization. Utrecht: LOT. Vulpe, Magdalena 1980 Subordonarea în frază în dacoromâna vorbită [Subordination in spoken Daco-Rumanian]. Bucureşti: Editura Ştiinţifică şi Enciclopedică. van der Wal, Marijke J. 2002 Relativisation in the history of Dutch: major shift or lexical change? In Poussa (ed.), 27–35. Weinert, Regina 2004 Relative clauses in spoken English and German – Their structure and function. Linguistische Berichte 197: 3–51. forthc. Postmodifying verb-second clauses in German (ms.). Weinrich, Harald 1993 Textgrammatik der deutschen Sprache. Mannheim: Duden. Weiss, Daniel 1993 Die Faszination der Leere. Die moderne russische Umgangssprache und ihre Liebe zur Null. Zeitschrift für slavische Philologie 53: 48– 82. Weiß, Helmut 2004 A question of relevance. Some remarks on standard languages. Studies in language 28 (3): 648–674. Westedt, Adela 1997 Los relativos en el habla culta de Madrid. Berlin: FU Berlin. Wexler, Paul 1992 Diglossia et schizoglossia perpetua – The fate of the Belorussian language. Sociolinguistica 6: 42–51.

References 463 Wheeler, Max W., Alan Yates, and Nicolau Dols 1999 Catalan. A Comprehensive Grammar. London/New York: Routledge. Wingender, Monika 2003 Überlegungen zur Weiterentwicklung der Theorie der Standardsprache. In Wolfgang Gladrow (ed.), Die slawischen Sprachen im aktuellen Funktionieren und historischen Kontakt, 133–152. Frankfurt am Main: Peter Lang. Zagorska Brooks, Maria 1975 Polish Reference Grammar. Paris/The Hague: Mouton. Zaliznjak, Andrej A. and Elena V. Padučeva 1975 K tipologii otnositel’nogo predloženija [The typology of relative clauses]. Semiotika i informatika 6: 51–102. Zemskaja, Elena A. 1973 Russkaja razgovornaja reč’ [Colloquial spoken Russian]. Moskva: Nauka. 1987 Russkaja razgovornaja reč’ [Colloquial spoken Russian]. Moskva: Russkij Jazyk. Zemskaja, Elena A., Margarita V. Kitajgorodskaja, and Evgenij N. Širjaev 1981 Russkaja razgovornaja reč’: obščie voprosy, slovoobrazovanie, sintaksis [Colloquial spoken Russian: general issues, word formation, syntax]. Moskva: Nauka. Zemskaja, Elena A. and Dmitrij N. Šmelëv (eds.) 1984 Gorodskoe prostorečie. Problemy izučenija [Urban prostorečie. Problems of its investigation]. Moskva: Nauka. Zifonun, Gisela 2001a Grammatik des Deutschen im europäischen Vergleich: Der Relativsatz. Mannheim: Institut für Deutsche Sprache. 2001b Grammatik des Deutschen im europäischen Vergleich: Das Pronomen. Teil I: Überblick und Personalpronomen. Mannheim: Institut für Deutsche Sprache. 2003 Sprachtypologische Fragestellungen in der gegenwartsbezogenen und der historischen Grammatik des Deutschen, am Beispiel des Relativsatzes. In Anja Lobenstein-Reichmann and Oskar Reichmann (eds.), Neue historische Grammatiken. Zum Stand der Grammatik-Schreibung historischer Sprachstufen des Deutschen und anderer Sprachen, 59–85. Tübingen: Niemeyer. Zifonun, Gisela, Ludger Hoffmann, and Bruno Strecker 1997 Grammatik der deutschen Sprache. 3 Vols. Berlin/New York: de Gruyter.

Index of languages

Accadian, 82 Albanian, 53, 97, 99, 124, 127, 213, 219–220, 224, 226, 237, 241– 242, 266, 276, 285, 288–289, 329 Ancient Greek, 54, 257–258, 263, 268, 348, 422, 436, 439 Arabic, 54, 81–82, 267, 287, 303 Basque, 24, 30, 44, 54, 74, 76, 78– 79, 82, 95, 97, 99, 103, 115–116, 118–119, 123, 151–152, 154– 155, 158, 234–236, 240, 252, 262, 275, 277, 287, 289, 302, 304–305, 350–352, 419, 435– 437 Belarusian, 53, 74, 76, 97, 99, 123, 147–148, 220, 241–242, 266, 276, 280, 285–286, 289, 320, 356–358 Bosnian/Croatian/Serbian, 53, 64, 79, 82, 91–92, 97, 99, 123–124, 126–129, 147, 181, 184–186, 192–195, 198, 204, 221, 237, 242–243, 256, 266, 270, 276, 280, 289, 314–316, 326–328, 353–354, 434, 438 Bulgarian, 43, 53, 64, 92, 94, 97, 99, 102, 124, 126, 140–141, 167– 168, 172–174, 177, 182, 184– 186, 189, 191, 193–195, 198, 205, 212, 226, 241–242, 251, 264–266, 268, 270–272, 276, 280, 286, 288–289, 308– 309, 314, 317, 436, 438

Catalan, 24, 38, 97, 99, 107, 212, 224–227, 234, 236, 241–242, 266, 276–278, 285, 289 Chinese, 82 Church Slavonic, 54, 259, 263, 301, 314, 317–319, 348, 354–356, 368, 422, 434, 438 Czech, 43, 53, 82–84, 88–89, 94, 96–97, 99, 113, 116, 124–125, 136–137, 167, 193–195, 199, 200–202, 219–220, 232, 242– 243, 257, 266, 269–270, 276, 280–281, 283, 285, 288–289, 306, 308, 314, 317, 319–320, 325–327, 333, 336–337, 353– 354, 367, 434, 438 Danish, 53, 90–91, 96–99, 102, 104– 105, 116, 123, 136–137, 206, 208, 214, 219, 241–242, 248, 258, 267, 270, 277, 289, 331 Dutch, 35, 43, 53, 82, 97–99, 106– 107, 151, 168–171, 184, 187, 189, 197, 208–210, 240, 242– 243, 252, 261, 263–266, 276– 279, 281, 285, 289, 292, 321, 325, 328, 330, 350, 368, 419 English, 13, 16–17, 33, 35, 39, 43, 53, 56, 60–61, 63–64, 67, 76–79, 81–84, 89, 94, 97, 99, 106, 114, 123, 130, 136–137, 144, 154, 168, 177–182, 184–185, 193, 202–203, 267, 275–276, 281, 283, 288–289, 308, 320–321, 324–325, 329–330, 338, 340, 343, 358, 372

Index of languages Estonian, 40, 54, 97, 119–120, 184, 276, 285, 289 Finnish, 43, 54, 64, 97, 195, 211– 212, 250, 252, 261, 265, 276, 281, 289, 435–436 French, 29, 35, 39–40, 43, 52–53, 56, 62, 66–67, 83–84, 90–91, 94, 97, 99, 103–104, 107, 109, 123, 141–142, 155, 177–185, 189– 191, 199, 202, 215–216, 225– 227, 241–242, 248, 251, 257, 262, 264, 266–267, 271, 276– 278, 285, 289, 305, 321, 323, 328, 338–343, 350, 359–360, 422, 436 Gagauz, 54, 157, 167–168 Galician, 29, 52, 83, 434 Genoese, 43, 113 German, 33, 35, 39–40, 43, 52–54, 58, 60, 62, 64, 67–69, 72, 75–77, 80–81, 94, 97–100, 103–107, 113, 124, 145, 151, 154–155, 168–171, 183–184, 186–191, 195–199, 214–215, 235, 237, 240, 242–243, 250, 252, 257, 261, 263, 265, 269, 276, 278– 279, 281, 285–289, 292, 295, 297, 300–301, 310–311, 321– 325, 328, 330–334, 337–338, 343, 345, 348, 350–351, 358– 359, 368–369, 372–386, 419, 422–423, 425, 433, 435–436 Germanic languages, 38–39, 41, 53, 123, 141, 197, 250, 279, 281, 288–289, 321, 324–325, 333, 337, 350, 352–353 Gothic, 81 Greek, 4, 39, 43, 52–53, 64, 97, 99, 106, 113, 127, 151, 158–159, 184, 189–190, 198–199, 212– 213, 223, 226, 229, 241–242,

465

250, 258, 263, 265, 268, 277– 278, 281, 285–286, 288–289, 364, 366, 419, 434 Hungarian, 40, 43, 54, 61, 75, 78, 90, 97, 151, 154, 158–159, 168, 211–212, 250, 252, 261–262, 276, 289, 305, 419 Icelandic, 53, 97, 99, 104–105, 241– 242, 248, 267, 270, 277, 289 Irish, 53, 97, 99, 213, 250, 267, 277, 289, 329, 331 Italian, 32–33, 40, 44, 53, 64–66, 69, 72, 79, 81, 83, 94, 97, 99, 105, 107, 115, 123, 142–143, 177– 183, 186, 191–192, 199, 203– 204, 216–217, 220–222, 225, 227–229, 231, 233–234, 241– 243, 251, 264, 266, 271, 276– 279, 285, 289, 294, 312–313, 320, 329, 339–340, 345–346, 359–363, 372, 400, 435, 438 Latin, 43, 54, 162, 166–167, 183, 257–258, 261, 263, 265, 267, 270, 278–279, 285, 288, 299– 300, 312–313, 316–319, 321– 322, 332, 338, 342–344, 348, 350–354, 357, 368, 422, 434– 439 Latvian, 43, 53, 57, 89, 97, 151, 158–159, 184, 191, 241–242, 262, 265, 276, 281, 285, 289, 305, 419 Lituhanian, 53, 97, 184, 191, 241– 242, 265, 276, 281, 285, 289 Lower Sorbian, 53, 97, 99, 103, 117, 120, 122, 140, 155, 193–194, 234, 236, 242–243, 265–266, 268–269, 276, 280, 283, 287– 289, 314, 359, 368

466 Index of languages Macedonian, 28, 30, 52–53, 94, 97, 99, 103–104, 124, 126, 140–141, 155, 174, 193, 195, 198, 210, 212, 215, 226, 241–242, 252, 262, 267–268, 270–272, 277, 280, 285, 288–289, 436 Maltese, 54, 97, 99, 213–214, 235– 236, 270, 277, 289 Norwegian, 28, 53, 97, 99, 104, 108, 123, 136, 214, 241–242, 248, 258, 267, 270, 277, 289, 369 Old Church Slavonic, 257–259, 263– 264, 300, 307, 313–314, 318, 328, 333, 434, 438–439 Old Turkic, 54, 129, 301–304, 310– 311, 350 Persian, 54, 129, 132–133, 136, 262, 287, 302–303, 350 Polish, 39, 53, 56, 58, 64, 91–92, 94, 97, 99, 101, 104–105, 124–125, 139–140, 184, 193, 195, 206, 215, 220, 224–225, 227, 237, 241–242, 266, 269, 276, 280, 286, 288–289, 305, 309, 314, 317, 326, 335–336, 353, 357, 359, 367, 437–438 Portuguese, 53, 64, 66, 91–94, 97, 99, 114, 116, 119, 122–123, 142, 147, 199–200, 220, 226, 228– 229, 241–242, 261, 266, 276– 278, 285, 289, 359–360, 438 Rhaeto-Romance, 29, 52 Romance languages, 28–29, 41, 53, 66, 106, 123, 145, 181, 250, 270, 279, 281, 288–289, 313–319, 328–329, 338, 340–342, 349, 359, 368 Rumanian, 38, 53, 64, 97, 99, 124– 127, 142, 172, 174–177, 182,

184, 191, 193, 195, 212–213, 223–224, 226, 232, 237, 241– 242, 250–251, 264, 266, 276, 283, 288–289, 312, 317, 339– 340 Russian, 3, 15–18, 23–25, 30, 33, 43–44, 52–53, 64–67, 75–76, 79, 83–84, 86, 92, 94, 96,–97, 99, 101–102, 104–105, 109, 117– 118, 120–123, 136–139, 141, 143–144, 147, 151, 160–166, 168, 184, 206, 222, 231, 238– 242, 247, 262, 266, 270, 276, 280–281, 286, 288–289, 295, 305–307, 309, 314–316, 319, 329–337, 344–345, 353–357, 362–367, 372–373, 387–416, 419, 423–425, 435–436, 438– 439 Rusyn, 102 Ruthenian, 320, 356 Sardinian, 83 Schwyzertütsch, 43, 348 Slavic languages, 38–44, 53, 92, 106, 123, 137–138, 141, 145, 147, 167, 193, 248, 250, 270, 279–282, 288–290, 300, 305, 307–309, 313–320, 325–329, 331, 335, 353, 357, 367–368, 417 Slovak, 53, 67, 97, 99, 167, 241– 242, 267–268, 270, 277, 280, 285, 288–289, 308, 314, 364, 434 Slovenian, 43, 53, 82, 97, 99, 113, 141–142, 145–146, 167, 209– 210, 252, 262, 267–270, 277, 280–281, 285, 288–289 Spanish, 29, 39–40, 43, 53, 66–67, 92, 94, 97, 99, 107, 109, 115, 123, 142, 155, 199, 212, 223, 226, 241–242, 262, 266, 272,

Index of languages 276, 278, 285, 289, 305, 323, 328, 339, 340, 343, 350, 359– 360, 372, 434, 438 Swedish, 28, 53, 57, 81–82, 94, 97, 99, 104, 109, 123, 136, 142, 207–208, 214, 241–242, 248, 267, 270, 277, 279, 289, 321, 324, 329, 332, 344, 351 Turkish, 13–16, 43, 52, 54, 73, 77– 79, 97, 99, 109–110, 129–136, 151–158, 166–168, 228–232, 240, 242, 247, 252, 262–263, 265, 276, 281–282, 287, 289, 292, 302–304, 310, 329, 350, 352, 364, 419, 435–437

467

Ukrainian, 30, 53, 75–76, 97, 99, 106, 117, 123, 151, 158–160, 167, 241–242, 267–268, 270, 277, 280, 285, 288–289, 306– 307, 309–310, 314–315, 319– 322, 326, 353, 356–358, 367, 405–406, 419 Upper Sorbian, 30, 53, 67, 97, 99, 103, 140, 155, 193–194, 241– 242, 266, 268, 276, 280, 283, 287–289, 359, 368 Welsh, 43 Yakut, 74 Yiddish, 82

Index of authors

Akimova, Galina N., 101–102 Albrecht, Jörn, 298–299 Alfonzetti, Giovanna, 191–192, 199, 203, 227–228, 362, 372 Allan, Robin, 90, 116, 206, 219, 267, 277 Ammon, Ulrich, 27, 31–32, 255, 257, 260, 273, 346, 349, 414, 433 Andrade Peres, João, 91–92, 199, 261 Antinucci, Francesco, 246–247, 437 Ariel, Mira, 148, 217, 222, 293, 295–296 Arim, Eva, 199, 229, 359 Atrachovič, Kandrat K. Auer, Peter, 3, 13, 29, 30, 35, 45, 48, 52, 70, 129, 133–135, 149–150, 158, 203, 252, 258, 297, 302, 304, 347–349, 366, 389, 423, 434 van der Auwera, Johan, 79, 316 Avanesov, Ruben I., 23, 160, 280, 389 Azzoppardi-Alexander, Maria, 235, 269 Bagłajewska-Miglus, Ewa, 367 Balliet, Pierre, 98 Banfi, Emanuele, 258 Barbiers, Sjef, 28, 98, 209 Bartnicka, Barbara, 58 Bauer, Jaroslav, 141, 290, 306, 317, 319–320, 326–327, 336–337 Beličová, Helena, 121, 167, 201, 290, 308 Bellini, Simona, 65, 372 Berger, Tilman, 364, 434

Bernini, Giuliano, 35, 145, 166, 183, 187–188, 191–193, 216–217, 237–238, 376, 400 Berruto, Gaetano, 29, 32, 79, 221, 291, 363 Biber, Douglas, 60, 268, 275, 283, 358 Bierich, Alexander, 388–389 Birkner, Karin, 372 Bisang, Walter, 27 Blanche-Benveniste, Claire, 181, 191, 359 Borg, Albert, 235, 269 Bosque, Ignacio, 272 Bossong, Georg, 142, 267, 436 Brincat, Giuseppe, 267 Brunot, Ferdinand, 339 Buchholz, Eva, 195 Bužarkova, Eleni, 127 Cardinaletti, Anna, 58, 264 Carter, Ronald, 268 Chambers, Jack K., 29–30, 32, 54 Cheshire, Jenny, 34–36, 417 Cojocaru, Diana, 175 Comrie, Bernard, 25, 38, 40, 42–44, 61, 63, 74, 79, 95, 112–113, 115, 148, 186, 216, 219, 248–250, 284, 399–400, 418, 420, 433, 436 Coseriu, Eugenio, v, 438 Cramer, Christina E., 141, 174, 271 Cresti, Emanuela, 66, 178–179, 360 Cristofaro, Sonia, 7–13, 17–19, 25, 38, 40, 56, 59, 62, 79, 83–84, 86, 93, 114–115, 143, 217, 249–250, 284, 290, 325, 420–421, 433

Index of authors 469 Croft, William, 6, 8–9, 27, 292–293, 402, 406, 415 Cubberley, Paul, 91–92, 290 Cuendet, Georges, 193, 217, 339 Damourette, Jacques, 341 Dasinger, Lisa, 8, 222, 246, 279 Dekeyser, Xavier, 321, 325, 329, 335 Demonte, Violeta, 272 Desző, László, 314–315 Deutscher, Guy, 261 Dmitriev, Pëtr A., 192 Dols, Nicolau, 234 Donaldson, Bruce C., 179, 208, 264 Downing, Bruce, 25, 42, 61, 74 Druviete, Ina, 285 DuBois, John A., 26 Duličenko, Aleksandr D., 102 Duranti, Alessandro, 246–247, 437 Dürscheid, Christa, 50, 69, 377 Ebert, Robert Peter, 310–312, 323– 325, 330–331, 334, 337–338, 345, 350 Endzelin, Jānis, 57, 159, 191 Engel, Ulrich, 375 Erdal, Marcel, 301–303, 311 Erelt, Marti, 119–120, 184 Erkman-Akerson, Fatma, 131, 135, 154, 156–157, 166–167 Faßke, Helmut, 103, 117, 193–194, 243, 266, 282–283, 287, 359 Fenyvesi, Anna, 75, 78, 90, 212 Ferguson, Charles A., 30, 347 Fernandez-Vest, M. M. Jocelyne, 34, 36, 48 Fiehler, Reinhard, 291, 435 Fiorentino, Giuliana, 38, 40–42, 44, 79, 203–204, 228, 233–234, 236, 261, 290, 294, 298, 317, 342, 359, 368, 372, 386

Fleischer, Jürg, 103, 106–107, 145, 197, 235, 243, 287–288, 331, 358, 436 Fontein, A.M., 188, 265 Forgács, Tamás, 212 Fox, Barbara A., 114, 372 Frei, Henri, 436 Freitas, Tiago, 199, 229, 359 Friedman, Victor A., 75, 117, 127, 140, 192, 194–195, 201, 210, 218, 221, 280 Gadet, Françoise, 104, 142, 180– 181, 202 Gallis, Arne, 128, 314–316, 326, 328, 353–354 Gapany, Joël, 62, 90, 94, 107, 109, 122–123, 181, 185, 189–190, 199, 216, 225–226, 241, 433 Gärtner, Eberhard, 147, 199, 220 Gärtner, Hans-Martin, 168–171, 375–376 Gebert, Lucyna, 246–247, 437 Genadieva-Mutafčieva, Zara, 195 Gerd, Aleksandr S., 161, 206, 222, 316, 372, 387 Gessner, Emil, 328, 343 Giacalone Ramat, Anna, 20, 38, 40, 62, 79, 83–84, 86, 93, 114–115, 217, 249–250, 252, 278–279, 284, 290, 313, 316–317, 320, 325, 338, 350, 420–421 Givón, Talmy, 8–9, 84, 231, 251, 259, 421, 433, 436–437 Göksel, Aslı, 110, 129–134, 136, 156, 302 Gołąb, Zbigniew, 75, 117, 127, 192, 194–195, 201, 210, 221, 280 Goldberg, Adele, 6, 415 Grandi, Nicola, 258 Graur, Aleksandru, 59, 264, 283 Grdina, Igor, 268

470 Index of authors Guiraud, Pierre, 189, 298, 317, 338, 341, 350, 436 Günthner, Susanne, 124 Gutschmidt, Karl, 286, 366, 388, 390, 405 Haarmann, Harald, 257–258, 267– 268, 270, 317, 434 Haase, Martin, 305 Haig, Geoffrey, 77, 110, 132–136, 142, 153–154, 156–157, 228– 232, 247, 437 Hanser, Oskar, 303–304, 350, 352 Haspelmath, Martin, 149, 284 Hawkins, Sarah, 113 Heine, Bernd, 155, 299, 301, 348 Henn-Memmesheimer, Beate, 34, 35, 417 Hentschel, Elke, 375 Herman, József, 438 Herrity, Peter, 141, 145–146, 210, 269 Herrmann, Tanja, 28, 114, 144, 178– 180, 193, 202, 207–208, 213, 220, 223, 243, 275, 288, 358, 372 Hinchliffe, Ian, 94, 109, 142, 207, 214, 267, 277, 364 Hinrichs, Uwe, 32–33 Hock, Wolfgang, 259 Hoekstra, Jarich, 197, 209 Hoffmann, Ludger, 58, 60, 358, 375 Holmes, Philip, 90, 94, 109, 116, 142, 206–207, 214, 219, 267, 277, 364 Holton, David, 159, 263, 265, 277, 286, 364–365 Holzer, Georg, 438 Hualde, José Ignacio, 224–225 Hurch, Bernhard, 74, 305 Hurtig, Claudia, 74, 357 Iliescu, Maria, 176, 312

Janaš, Pětr, 122, 140, 234 Johanson, Lars, 74, 129 Joseph, Brian D., 4, 159, 189, 223, 229 Karjalainen, Merja, 195 Karlsson, Fredrik, 207–208, 363– 364 Katičić, Radoslav, 128–129 Keenan, Edward L., 25, 38, 40, 61, 63, 112–115, 148, 216, 219, 250, 399–400, 418, 420, 433, 436 Kenesei, István, 75, 78, 90, 212 Kerslake, Celia, 110, 129–134, 136, 156, 302 Kitajgorodskaja, Margarita V., 389 Kleiber, Georges, 433 Kloss, Heinz, 30, 434 Koch, Peter, 20, 32, 48–51, 67, 259, 290, 418 König, Ekkehard, 10 Kordić, Snježana, 56, 221, 256 Kornfilt, Jaklin, 129, 131, 135–136, 166, 263 Korš, Fëdor J., 44 Kortmann, Bernd, 27–28, 52, 124 Křížková, Helena, 145, 210, 280– 281, 290 Kručinina, Irina N., 121, 160, 162– 163 Kučanda, Dubravko, 29, 316 Kuno, Susumo, 188 Kunstmann, Pierre, 323, 328 Kurzová, Helena, 37–38, 44, 79, 191, 281 Kuteva, Tania, 42–44, 155, 248– 250, 284, 299, 301, 348, 420 Kuz’mina, Irina B., 389 Lambertz, Maximilian, 124, 319 Lambrecht, Knud, 8 Lambton, Ann K. S., 132–133

Index of authors 471 Lamprecht, Arnošt, 306, 317, 327, 336–337 Lapteva, Ol’ga A., 24, 75, 98, 117, 137–138, 160–162, 164, 239, 389, 408 Laskowski, Roman, 195 Lehmann, Christian, 7, 9, 23, 25, 39, 42, 44, 55–56, 58–59, 65, 67, 74, 76, 78–85, 108, 114, 132, 158, 162, 166–167, 253, 263, 278, 299, 300–301, 324, 328, 331, 338–339, 344, 368, 433, 436, 438–439 Lešnerová, Šarka, 200–201, 219 Levi, Judith N., 61 Lewis, Geoffrey, 129 Lockwood, William B., 279, 250– 351, 358, 369 Löffler, Heinrich, 33 López del Castillo, Lluís, 24 Lötscher, Andreas, 188, 190 Lubaś, Wladyslaw, 285–286, 364, 434, 438 Lundskær-Nielsen, Tom, 90, 116, 206, 219, 267, 277 Mac Giolla Chríost, Diarmait, 258 Maček, Dora, 147, 279 Mackridge, Peter, 159, 263, 265, 277, 286, 364–365 van Marle, Jaap, 33, 51, 261, 422 Maslova, Elena, 166, 183, 187–188, 376 Mathiassen, Terje, 89 Matsumura, Kazuto, 435 Matzinger, Joachim, 329 Mauri, Caterina, 7, 10–11, 13 Maxwell, Daniel, 25 Mc Carthy, Michael, 268 Mc Gonagle, Noel, 213 Mel’nyčuk, Oleksandr S., 159–160, 307, 310, 314–315, 322, 326 Menz, Astrid, 157, 168

Metzeltin, Michael, 323, 340 Meyer-Lübke, Wilhelm, 338–340 Miller, Jim, 34, 36, 37, 48, 139, 169–170, 181, 195, 203, 240, 250, 362, 415, 417 Mirčev, Kiril, 314 Mišeska-Tomić, Olga, 126–127, 173–174, 198, 224, 226, 264, 271–272 Móia, Telmo, 91–92, 199, 261 Moneglia, Massimo, 66, 178–179, 360 Murelli, Adriano, 28, 191, 197, 361 Mustanoja, Tauno F., 321 Neweklowsky, Gerhard, 268, 286, 354, 434 Nilsson, Elsa, 124, 176–177, 264 Noreen, Adolf, 57, 59, 279, 332, 351 Oesterreicher, Wulf, 20, 32, 48–51, 67, 259, 290, 418 Oliva, Karel, 200–201, 219 Orlova, Varvara G., 23, 160, 280, 389 Oyharçabal, Bernard, 24, 95, 115– 116, 118–119, 123, 152, 155, 234, 275, 287, 305, 352 Ozıl, Şeyda, 131, 135, 154, 156– 157, 166–167 Padučeva, Elena V., 24, 74, 79, 86, 120, 160, 310, 365 Paul, Herrmann, 322, 325, 338, 343 Pellat, Jean-Christoph, 91, 94, 264 Penke, Martina, 55–57, 63–64 Peranteau, Paul M., 61 Perini, Mário A., 200 Pescher-ter Meer, A., 188, 265 Petr, Jan, 88, 96, 116, 125, 137, 159, 195, 243, 269, 286 Phares, Gloria C., 61

472 Index of authors Philippaki-Warburton, Irene, 4, 159, 189, 223, 229, 263, 265, 277, 286, 364–365 Picchio, Riccardo, 257, 434 Pichon, Edouard, 341 Pietsch, Lukas, 28 Pittner, Karin, 62, 191, 272, 275– 376, 386 Platzack, Christer, 96, 98, 102, 104, 137, 208, 279, 329, 332, 334, 351 Press, Ian, 106, 357 Pugh, Stefan M., 106, 357 Rabus, Achim, 301, 320, 353, 356– 357 Radeva, Vasilka, 141, 241 Ramat, Paolo, 35, 299, 311, 324– 325, 330, 334 Ramilo, Maria Celeste, 199, 229, 359 Ramza, Taccjana, 74, 357 Rendahl, Anne-Charlotte, 208 Renzi, Lorenzo, 58, 264 Riegel, Martin, 91, 94, 264 Rioul, René, 91, 94, 264 Roma, Elisa, 267–268, 288, 329, 331, 343 Romaine, Suzanne, 196, 321, 324– 325 Rosenbach, Annette, 55–57, 63–64 Rösler, Irmtraud, 321, 323, 331, 334, 338 Rounds, Carol, 212 Rusu, Valeriu, 124, 175–177 Salvi, Giampaolo, 58, 264 Salzmann, Martin, 236 Sandig, Barbara, 263, 298 Sansò, Andrea, 63, 142, 146, 210, 214, 236 Šapiro, Abram B., 105, 117 160, 231, 389

Scherer, Carmen, 65, 68, 378 Schlyter, Birgit N., 262, 287, 352 Scholten, Daniel, 267 Schreier, Daniel, 258, 268, 286 Schroeder, Christoph, 134 Schwartz, Arthur, 25, 61, 74 Schweier, Ulrich, 268, 285 Sedláček, Jan, 121, 167, 201, 290, 308 Seiler, Guido, 236 Sergieva, Natal’ja S., 161, 206, 222, 316, 372, 387 Sgall, Petr, 33, 195, 201, 283 Siemund, Peter, 10 Širjaev, Evgenij N., 138–139, 238, 389, 397 Slobin, Dan I., 135–136, 302 Šlosar, Dušan, 257, 306, 317, 327, 336–337 Šmelëv, Dmitrij N., 147, 365, 388– 389 Smits, Reinier J.C., 8, 38–39, 43, 76, 79, 81, 92, 104, 178, 186, 248, 278–279 Sneyders de Vogel, Kornelis, 343 Sornicola, Rosanna, 203, 231 Stabej, Marko, 268 Stark, Elisabeth, 342, 359 Stein, Dieter, 34–36, 51, 257, 259, 352, 417 Stojkov, Stojko, 198 Stolz, Thomas, 149 Stotz, Peter, 258 Strandskogen, Åre-Berit, 108, 214, 267, 277 Strandskogen, Rolf, 108, 214, 267, 277 Strecker, Bruno, 58, 60, 358, 375 Sulkala, Helena, 195 Sullivan, Kirk P. H., 207–208, 363– 364 Sussex, Roland, 91–92, 290

Index of authors 473 Švedova, Natalija Ju., 58–59, 122, 316, 395, 397–398, 402 Szczegielniak, Adam, 101 Tesnière, Lucien, 218, 437 Thompson, Sandra A., 372 Tilkov, Dimităr, 265, 308 Toman, Jindřich, 200–202, 232 Tompa, Jószef, 159 Topolińska, Zuzanna, 58, 125, 220 Tošović, Branko, 257 Toupin, Cecile, 8, 222, 246, 279 Trask, Robert L., 119, 154–155, 262, 350, 352 Troickij, Veniamin I., 17–18, 25, 102, 167, 306–307, 314, 317, 330, 332, 335–337, 344–345, 353–356 Trudgill, Peter, 29–30, 32, 54, 258, 268, 286 Trunte, Nikolaos H., 259, 307, 438 Urbańczyk, Stanisław, 104, 139, 206, 215, 224–225, 248, 290, 306, 309, 314, 317, 326, 335– 336, 353, 367, 438 Usikova, Rina P., 103, 140, 212, 215 Vago, Robert M., 75, 78, 90, 212 Večerka, Radoslav, 263–264, 280, 300, 307, 313–314, 316–317, 333, 439 Vikør, Lars, 258 Villena Ponsoda, Juan Andrés, 434 Vintr, Josef, 268, 364

Vondrák, Wenzel, 83, 167, 263, 305, 313–314, 317, 319, 333 de Vries, Mark, 25, 41–43, 61, 74, 78–79, 81–84, 94, 100 Vulpe, Magdalena, 124, 126, 185, 223–224, 232, 241, 312 Wagner, Susanne, 28 van der Wal, Marijke J., 197, 208– 209, 261, 279, 350 Weinert, Regina, 34, 36, 37, 139, 168–171, 181, 195, 203, 250, 292, 362, 373, 384, 415, 417, 433 Weinrich, Harald, 375 Weiss, Daniel, 164 Weiß, Helmut, 34–36, 261, 434 Westedt, Adela, 107, 199, 223, 372 Wexler, Paul, 286 Weydt, Harald, 375 Wheeler, Max W., 234 Wingender, Monika, 30 Yates, Alan, 234 Zagorska Brooks, Maria, 58 Zaliznjak, Andrej A., 24, 74, 79, 86, 120, 160, 310, 365 Zemskaja, Elena A., 18, 25, 118, 138, 147, 162, 164–166, 239, 247, 365, 388–389 Zifonun, Gisela, 38–40, 44, 58, 60, 70, 79, 106, 168, 324–325, 358, 375

Index of subjects

Accessibility Hierarchy, 112–115, 250, 399–401, 418 see also syntactic positions relativized adposition, 108, 141–145 agreement, gender and number, 80– 90, 108–111, 243–245, 279, 284, 418–422 analogy, principle of, 252, 422 anaphora, scale of, 22–23, 108 anti-orality, principle of, 257, 259 argument filling, 228–229 assertiveness tests, 11–19, 433 asyndetic clause clusters, 138–139, 238–240, 397–402, 411–414 at issue, 11 attractio relativi, 163, 436 attribution, 80–86 Ausbau, 30, 260, 270, 435 case matching, 233–236, 246, 252, 345–346, 436 recoverability, 148, 228–233, 246, 363, 437 chains of clauses, 158, 203–206, 252 clitic doubling, 226–227, 270–272 codification, 30–35, 256–261, 273, 285–288, 346–347, 367–368, 421–422 communicative practice, 48–51, 65– 70, 259, 290–296, 348, 352, 360, 368, 418, 435 compactness, 269–270, 279, 281, 283, 352, 421 complementizer, identitiy of a particle with the, 82, 94–95, 102,

104–105, 110, 123–127, 181– 182, 314 connecting relative, see relative element connector, 127–130, 181–183, 185, 191, 201, 237, 314, 321, 344, 383–384 contextual information, 13, 232–233, 259 coordination, syndetic/asyndetic, 18, 20–21, 96, 170, 237, 240 coreferential, 23 decumulation, 217–222, 252, 264, 270, 334–340, 436–437 diaglossia, 347–348, 366 dialect, 28–33, 48, 51, 54, 347–350, 358–365 loss, 347, 366–367 dialectology, 26–28 diglossia, 347–348 distance language, 49–52, 259 double encoding, 222–228, 253, 260, 270–272, 283, 340–343 economy, principle of, 27, 251–253, 260 equi-case, equi type, see case matching explicitness degree of, 22, 25, 75, 86–87, 108, 148, 158, 166, 254, 263, 421, 437 principle of, 259–260, 263, 269, 284, 290, 352, 421

Index of subjects 475 gap construction, 80–86 grammaticalization, 123–124, 140, 157, 167, 191, 299–300, 305, 314, 328–329, 358, 367–368

203, 218, 243–247, 252, 254, 284, 290, 418, 410–421 long relativization, 29, 186, 237, 275, 344, 383, 385

iconicity, principle of, 27, 252–253, 260, 421 illocutionary force, 10–18 inconsistency, 35, 210, 261, 312, 421 incremental syntax, 203 integrated verb second clauses, 168– 171, 311–312, 374, 376–381, 383–385 integrative functionalism, 27–28, 290, 271, 386, 417, 419, 423– 424 IV2, see integrated verb second clauses

marked, sociolinguistically, 33, 52, 60, 255, 293, 297–298, 349, 359, 362–363, 417 mastery of a language, 291, 392, 405, 438 matrix clause, 22 model language, 257–261, 263, 267, 357, 368, 434, 438–439

juxtaposition, 18, 49, 95, 300–301, 330 koiné, 258, 268, 282, 285–286, 349, 366 language variety definition, 29 high vs. low, 30, 270, 347–348 prestigious, 30, 257–258, 268, 347–350, 352 standard vs. non-standard, 1, 29– 37, 47–48, 251, 254–277, 283–284, 252–296, 346–369, 414, 421 linguistic variant, 27, 255 choice of, 290–296 elaboration of, 260 reduction of, 260, 346–347 link between the matrix clause and the relative clause, 22–25, 85–88, 110–111, 124, 127, 182–188,

no encoding, 228–223, 343–344 no redundancy, 259–261, 269, 282, 352, 421 no variation, principle of, 257 non-relative particle, 98, 104–105, 110–111, 243–247, 253, 268– 269, 331–334, 419 non-standard definition, 33 see also language variety orality conceptional, 69, 259, 268, 372, 385 vs. scripturality, 48–52, 291, 377, 418 other particle, see non-relative particle paradigm regularization, 137, 207– 212, 252, 420 participant coding, 22, 26, 73 preferred interpretation, 230–232 projective structure, 203, 246 proximity language, 49–52, 259 purism, 259–260, 282, 299, 421

476 Index of subjects recoverability, see case referent, 7–8, 22–23 relative clause adjoined, 74 avoiding of, 237–240 circumnominal, 74 correlative, 43, 75–76, 116–122, 151, 158–168, 240, 252–253, 262–263, 292, 299, 305–311, 314–315, 318–320, 351, 367, 396, 398, 406–407, 418–419 definition, 22–23 embedded, 74 extraposed, see postposed free, or headless, 75–76, 141, 196, 326, 396 postnominal, 74, 76, 78–79, 116– 117, 119, 122, 151–158, 240, 262–264, 274, 300–305, 314– 320, 352–354, 374–375, 378– 381, 383–384, 395–396, 398, 401, 418–419, 422–424 postposed, 75–79, 151, 154, 168–171, 240, 263, 293, 300, 311–312, 216, 374, 378–384. 418, 423 prenominal, 74, 76–79, 116, 118–119, 151–158, 228, 235, 247, 301, 350 preposed, see correlative relative construction meaning, see relative relation form, 6–7, 21–26 relative element, 26, 37–40, 73, 79– 111, 115–122, 172–216, 240– 250, 263–269, 275, 283–285, 312–334, 353–363, 418–420 combined, 85, 98–112, 116, 139– 145, 212–216, 218–219, 222, 243–248, 268–269, 275, 282– 283, 331–334, 418–420 connecting, 278–279, 327

simple, 85, 87–97, 139–145, 243–250, 263, 283–284, 312– 331, 419 specialized, see specialized relative element triple combination of, 214–216, 243–244 relative particle, 38–39, 81–82, 88, 93–99, 103–104, 108, 111, 116, 145–148, 176–177, 179, 181– 183, 191, 193–206, 210, 218– 219, 241–242, 244–249, 265– 268, 273–277, 281–285, 314– 315, 318, 324–329, 359, 381– 383, 419–420, 422, 438–439 definition, 93 polyvalency of, 123–129, 241, 248, 253, 259 relative pronoun, 25, 39–44, 79–85, 88–90, 96–97, 99–101, 108, 111, 116–117, 119–120, 140–145, 172–182, 208–213, 222, 243– 244, 248–249, 253, 261–264, 274–281, 284–285, 300–301, 312–321, 333, 368, 381–383, 398, 401, 419–420, 422 definition, 88 G-N-inflecting, 89, 172, 175, 177–178, 209, 243, 277, 279, 314, 319, 350, 422 G-inflecting, 89–90, 178, 243 non-G-N-inflecting, 89–90, 175, 178, 200, 209, 243 ‘which’-based, 41–42, 278–279, 281, 304, 314, 316–319, 321, 329, 340–341, 350–353 relative relation, 85–86, 127, 136, 204, 240, 376, 395, 417 definition, 7–8 restrictive vs. non-restrictive, 7– 21, 169–171, 238–240, 295, 361–362, 399–401, 412–414, 417, 424, 433

Index of subjects 477 relativization strategy, 26–27 default, 78, 284–285, 435 [+case] vs. [−case], 40, 112, 115, 216–217, 250, 420 [++case], 250–253 autochthonous vs. borrowed, 282, 285–288, 350–352 relativization pattern, see relativization strategy relativized item definition, 22–26 syntactic role of, 84–88, 108– 111, 182–192, 216–218, 222, 228, 233, 245–246, 248–253, 263–264, 269, 284, 334–335, 340, 343, 345, 420–421 replication, 155, 348 resumptive element, 38, 56, 82–83, 98–100, 105–111, 113, 145–148, 188–190, 209–210, 212–213, 218–219, 222, 227, 243–248, 250, 265–268, 272, 274–276, 282, 331, 335–340, 422 resumptive pronoun, see resumptive element semantic predictability, 231–232 sociolinguistic continuum, 1, 32, 70, 348–349, 358, 366, 389–390, 423–425 specialized relative element, 81, 88, 90–93, 96–97, 99, 102, 111, 116, 182–193, 222, 244, 249, 263– 264, 274–277, 284, 321–323, 333, 419, 422 definition, 90 standard definition, 30 endoglossic, 347–351, 369, 437– 438

exoglossic, 257–258, 282, 285, 289, 347–351, 369, 437–438 see also language variety standardization, 30, 52, 257–258, 260, 285–286, 325, 351–352 standardness degree of, 36, 198, 256, 266, 275, 277–283, 422 scale of, 2, 256, 273–277, 285– 290, 422 state of affairs, 7–13 syntactic positions relativized, 26, 38–39, 73, 112–122, 145–148, 183–184, 216–240, 250–252, 269–274, 334–346, 382, 399– 401, 418–420 syntactic role of the relativized item, see relativized item recoverability of, see case recoverability subordination, 9–11, 19, 21, 80–81, 85 topic, 187, 202–205, 246–247, 364 typological-dialectological approach, 27–28, 47–48, 417, 425 word order, 26, 73–79, 115–122, 151–171, 240, 247, 252, 262– 263, 301–312, 418–419 zero-marker, 88, 95–97, 99, 108, 111–112, 116, 136–139, 144– 145, 207–208, 218–219, 244, 246–247, 254, 269, 274–277, 295, 301, 329–331, 363–364, 366, 397–401, 411–412, 419, 422, 424 definition, 95