The Crosslinguistics of Zero-Marking of Spatial Relations 9783050065304, 9783050062761

Table of contents :
Table of contents
Preface and acknowledgments
Abbreviations
Part A: Introduction and theoretical framework
1 Introduction
1.1 Appetizer
1.2 First examples (of methodology and terminology)
1.3 Organization
2 Foundations
2.1 Generalities
2.2 Previous studies
2.3 (An excursus on) Asymmetries
2.3.1 Syncretism
2.3.2 More or less complex
2.3.3 Markedness – a difficult concept
2.4 On zeroes and related issues
2.4.1 Finding zeroes
2.4.2 Types of zeroes
2.4.2.1 Splits
2.4.2.2 Problematic toponyms
2.4.3 General case-drop – a false friend
2.5 Theoretical and methodological issues for future research
2.5.1 The development of zero-markers
2.5.2 A formal framework: Optimality Theory
2.6 The final touches
Part B: The empirical side of zero-marking of spatial relations
3 Cross-linguistic objets trouvés
3.1 Indo-European
3.1.1 Hittite and sundry old Indo-European languages
3.1.2 (Insular) North Germanic
3.1.2.1 Icelandic
3.1.2.2 Faroese
3.1.2.3 Mainland Scandinavian
3.1.3 Aromunian
3.1.4 Southern Macedonian
3.1.5 Armenian
3.1.5.1 East Armenian
3.1.5.2 West Armenian
3.1.6 Indo-Iranian
3.1.6.1 Kurmancî and Zazakî
3.1.6.2 Gilaki
3.1.6.3 Sivandi
3.1.6.4 Sangesari
3.1.6.5 Persian
3.1.6.6 Kashmiri
3.1.6.7 Punjabi
3.1.6.8 Gujarati
3.1.6.9 Nepali
3.1.6.10 Hindi
3.2 Afro-Asiatic
3.2.1 Semitic
3.2.1.1 East Aramaic
3.2.1.2 Neo-Arabic
3.2.1.2.1 Cairene Arabic
3.2.1.2.2 Sudanese Arabic
3.2.1.2.3 Moroccan Arabic
3.2.1.2.4 Nigerian Arabic
3.2.1.2.5 Cypriot Arabic
3.2.2 Cushitic
3.2.2.1 Somali
3.2.2.2 Dahalo
3.3 Chadic
3.3.1 Miya
3.3.2 Kera
3.3.3 Pero
3.3.4 Mupun
3.3.5 Hausa
3.4 Nilo-Saharan
3.4.1 Lango
3.4.2 Maa
3.5 Fula
3.6 Vai
3.7 Igbo
3.8 Bantu
3.8.1 Yeyi
3.8.2 Kinyamwezi
3.9 Non-Indo-European languages in Asia
3.9.1 Bezhta
3.9.2 Burushaski
3.9.3 Malayalam
3.9.4 Kharia
3.9.5 Khmer
3.9.6 Hmong
3.10 Austronesian
3.10.1 Urak Lawoi’
3.10.2 Malay
3.10.3 Kilivila
3.10.4 Longgu
3.10.5 Mono-Alu
3.10.6 Drehu
3.10.7 Nêlêmwa
3.10.8 Xârâcùu
3.11 Papua-Newguinea
3.11.1 Koiari
3.11.2 Kobon
3.11.3 Bukiyip
3.11.4 Yelî Dnye
3.12 Australia
3.12.1 Guugu Yimidhirr
3.12.2 Ngalakan
3.12.3 Jaminjung
3.13 The Americas
3.13.1 Inside (and on the borders of) the Mesoamerican Sprachbund
3.13.2 Zero-marking of Place and/or Goal in individual Mesoamerican languages
3.13.2.1 Zapoteco of the Isthmus (Oaxaca)
3.13.2.2 Tlapaneco of Malinaltepec (Guerrero)
3.13.2.3 Trique of San Juan Copala (Oaxaca)
3.13.2.4 Huichol of San Andrés Cohamiata (Jalisco)
3.13.2.5 Chinanteco of San Juan de Lealao (Oaxaca)
3.13.2.6 Zoque of Chimalapa (Oaxaca)
3.13.2.7 Huave of San Mateo del Mar (Oaxaca)
3.13.2.8 Náhuatl of Acaxochitlán (Hidalgo)
3.13.2.9 Yucateco of X-Hazil Sur (Quintana Roo)
3.13.2.10 Totonaco of Misantla (Veracruz)
3.13.2.11 Matlatzinca of San Francisco Oxtotilpan (Estado de México)
3.13.2.12 Mexicanero of la Sierra Madre Oriental
3.13.2.13 Guarijío of Arechuyvo (Chihuahua)
3.13.2.14 Seri of Sonora
3.13.2.15 Interlude
3.13.3 Further data from Mayan languages
3.13.3.1 Jacaltec
3.13.3.2 Kanjobal
3.13.3.3 Tojolabal
3.13.3.4 Mam
3.13.3.5 Itzá
3.14 Pidgin and Creole languages
3.14.1 Portuguese-based Creoles
3.14.1.1 Angolar
3.14.1.2 São Tomense
3.14.1.3 Principense
3.14.1.4 Fa d’Ambû
3.14.1.5 Creole of Batavia and Tugu
3.14.1.6 Papia Kristang
3.14.2 Spanish-based Creoles
3.14.2.1 Papiamentu
3.14.2.2 Palenquero
3.14.3 French-based Creoles
3.14.3.1 Haitian
3.14.3.2 Guyanais
3.14.3.3 Seselwa
3.14.4 Dutch-based Creoles
3.14.5 English-based Pidgins and Creoles
3.14.5.1 Pidgins
3.14.5.1.1 Nigerian Pidgin English
3.14.5.1.2 Japanese Pidgin English in Hawaii
3.14.5.1.3 Pidgin English in Nauru
3.14.5.2 Central American Creole English
3.14.6 On zero-marking in Pidgins and Creoles – in general
3.15 Esperanto
3.16 Bridge
4 Case studies
4.1 Zero-marking in contemporary literary French
4.1.1 Background
4.1.2 Something is absent
4.1.3 Privileged street names
4.1.4 Supposed counter-examples
4.1.5 Predicates
4.1.5.1 Verbs combining with Goals
4.1.5.2 Verbs combining with Places
4.1.6 Beyond the VP
4.1.6.1 Semantics and pragmatics
4.1.6.2 Appositions and related phenomena
4.1.6.3 Attributes and secondary predication
4.1.7 Verbs of communication
4.1.8 Where do you live?
4.1.9 Are there preferences for being zero-marked?
4.1.9.1 More data
4.1.9.2 Quantitative aspects
4.2 Zero-marking in contemporary literary Maltese
4.2.1 In medias res
4.2.2 On Maltese spatial prepositions
4.2.3 Where spatial prepositions fail to show up
4.2.4 Statistics
4.2.4.1 The first corpus text (RMT I)
4.2.4.1.1 On location-NPs
4.2.4.1.2 The verbal component of Maltese spatial constructions
4.2.4.2 The second corpus text (EWB)
4.2.4.2.1 Some properties of location-NPs
4.2.4.2.2 Zero-marking of spatial relations and verbs
4.2.4.3 Variation
4.2.4.3.1 Constraints on pluralization
4.2.4.3.2 Specificity and related issues
4.2.4.3.2.1 Definiteness and indefiniteness
4.2.4.3.2.2 Genericity
4.2.4.3.2.3 Demonstratives and location-NPs
4.2.4.3.2.4 Possessed location-NPs
4.2.4.3.2.5 Location-NPs with (and without) attributes
4.2.4.3.3 Individual place names with and without fi ‘in, at, on’
4.2.4.3.3.1 Street names
4.2.4.3.3.2 Toponyms
4.2.4.3.4 Variation with verbs
4.2.5 A final word on the Maltese case
4.3 French and Maltese – similarities and dissimilarities
Part C: Insights and outlook
5 Evaluation
5.1 Goal vs. Place
5.2 Toponyms vs. common nouns
5.3 Obligatory vs. optional
5.4 Interrelations
6 The end is near
6.1 Achievements (which pose questions)
6.2 Zero-marking in the making?
6.3 The OT-framework
6.4 Syntagmatic vs. paradigmatic types of zero
6.5 Complement vs. adverbial
6.6 The Goal bias and mode hierarchies
6.7 Revisiting the object dimension
6.8 A correspondence with noun incorporation
6.9 Conclusions
Appendix
A: Corpus Rajt Malta tinbidel
B: Corpus Tout Maigret IV
C: Corpus Nestor Burma
D: Corpus L-ewwel weraq tal-bajtar
E: Sample languages
F: Maps
Sources
References
Index of authors
Index of languages
Index of subjects

Citation preview

Thomas Stolz, Sander Lestrade, Christel Stolz The Crosslinguistics of Zero-Marking of Spatial Relations

Studia Typologica

Beihefte / Supplements STUF – Sprachtypologie und Universalienforschung Language Typology and Universals Editors Thomas Stolz, François Jacquesson, Pieter C. Muysken Editorial Board Michael Cysouw (München), Ray Fabri (Malta), Steven Roger Fischer (Auckland), Bernhard Hurch (Graz), Bernd Kortmann (Freiburg), Nicole Nau (Poznán), Ignazio Putzu (Cagliari), Stavros Skopeteas (Bielefeld), Johan van der Auwera (Antwerpen), Elisabeth Verhoeven (Berlin), Ljuba Veselinova (Stockholm)

Volume 15

Thomas Stolz, Sander Lestrade, Christel Stolz

The Crosslinguistics of Zero-Marking of Spatial Relations

DE GRUYTER MOUTON

ISBN 978-3-05-006276-1 e-ISBN 978-3-05-006530-4 ISSN 1617-2957 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic Information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. © 2014 Walter de Gruyter GmbH, Berlin/Boston Cover image: Alpha-C/iStock/Thinkstock Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Table of contents

Preface and acknowledgments .......................................................................... xiii Abbreviations ......................................................................................................... xvi Part A: Introduction and theoretical framework 1

Introduction ..................................................................................................

1.1 1.2 1.3

Appetizer ......................................................................................................... 3 First examples (of methodology and terminology) ......................................... 5 Organization .................................................................................................... 10

2

Foundations .................................................................................................. 11

2.1 2.2 2.3

Generalities ..................................................................................................... Previous studies .............................................................................................. (An excursus on) Asymmetries ....................................................................... 2.3.1 Syncretism ......................................................................................... 2.3.2 More or less complex ........................................................................ 2.3.3 Markedness – a difficult concept ....................................................... On zeroes and related issues ........................................................................... 2.4.1 Finding zeroes ................................................................................... 2.4.2 Types of zeroes .................................................................................. 2.4.2.1 Splits ................................................................................ 2.4.2.2 Problematic toponyms ..................................................... 2.4.3 General case-drop – a false friend ..................................................... Theoretical and methodological issues for future research ............................. 2.5.1 The development of zero-markers ..................................................... 2.5.2 A formal framework: Optimality Theory .......................................... The final touches .............................................................................................

2.4

2.5

2.6

3

11 16 18 18 22 30 32 33 38 40 45 50 52 52 56 59

vi

Table of contents

Part B: The empirical side of zero-marking of spatial relations 3

Cross-linguistic objets trouvés ……………………………………….… 65

3.1

Indo-European ………………………………………………………………. 3.1.1 Hittite and sundry old Indo-European languages ………………….. 3.1.2 (Insular) North Germanic ………………………………………..… 3.1.2.1 Icelandic ……………………………………………..… 3.1.2.2 Faroese …………………….............…………………… 3.1.2.3 Mainland Scandinavian ……………………………...... 3.1.3 Aromunian …………………………………………………………. 3.1.4 Southern Macedonian ………………………….............…………… 3.1.5 Armenian ……………………………………………….………...... 3.1.5.1 East Armenian ………………………………………… 3.1.5.2 West Armenian ………………………………………… 3.1.6 Indo-Iranian ……………………………………………………...… 3.1.6.1 Kurmancî and Zazakî ……………………………….… 3.1.6.2 Gilaki ……………………………………………….…. 3.1.6.3 Sivandi ………………………………………………… 3.1.6.4 Sangesari ……………………………………………… 3.1.6.5 Persian ………………………………………………… 3.1.6.6 Kashmiri ……………………………………………..... 3.1.6.7 Punjabi ………………………………………………… 3.1.6.8 Gujarati ……………………………………………..… 3.1.6.9 Nepali …………………………………………………. 3.1.6.10 Hindi ……………………………………………….…. Afro-Asiatic ………………………………………………………………… 3.2.1 Semitic ……………………………………….................................. 3.2.1.1 East Aramaic ……………………..…………………… 3.2.1.2 Neo-Arabic …………………………………….........… 3.2.1.2.1 Cairene Arabic …………………………… 3.2.1.2.2 Sudanese Arabic ………………………..… 3.2.1.2.3 Moroccan Arabic …………………………. 3.2.1.2.4 Nigerian Arabic …………………………… 3.2.1.2.5 Cypriot Arabic …………………………… 3.2.2 Cushitic …………………………………………………….……… 3.2.2.1 Somali ……………………………………………….… 3.2.2.2 Dahalo ………………………………………………....

3.2

68 68 70 70 71 72 73 75 76 76 78 80 80 81 82 82 83 84 85 86 87 87 88 89 89 90 90 91 92 93 94 95 95 96

Table of contents

3.3

vii

Chadic …………………………………………………………..…………... 97 3.3.1 Miya ………………………………………………………………… 97 3.3.2 Kera ………………………………………………………………… 97 3.3.3 Pero …………………………………………………………………. 98 3.3.4 Mupun ………………………………………………………………. 100 3.3.5 Hausa ……………………………………………………………….. 101 3.4 Nilo-Saharan ………………………………………………………………… 102 3.4.1 Lango ……………………………………………………………… 102 3.4.2 Maa ………………………………………………………………… 103 3.5 Fula ……………………………………………………………………….… 105 3.6 Vai …………………………………………………………………………… 107 3.7 Igbo ……………………………………………………………………….… 108 3.8 Bantu ………………………………………………………………………… 109 3.8.1 Yeyi ………………………………………………………………… 110 3.8.2 Kinyamwezi ………………………………………………………… 111 3.9 Non-Indo-European languages in Asia ……………………………………… 112 3.9.1 Bezhta ………………………………………………………………. 112 3.9.2 Burushaski ……………………………………………………….… 113 3.9.3 Malayalam ……………………………………………………….… 114 3.9.4 Kharia ………………………………………………………………. 114 3.9.5 Khmer ……………………………………………………………… 116 3.9.6 Hmong …………………………………………………………....... 116 3.10 Austronesian …………………………………………………………........... 117 3.10.1 Urak Lawoi’ ……………………………………………………..… 117 3.10.2 Malay …………………………………………………………........ 118 3.10.3 Kilivila …………………………………………………………….. 119 3.10.4 Longgu …………………………………………………………….. 119 3.10.5 Mono-Alu ………………………………………………………….. 120 3.10.6 Drehu ………………………………………………………………. 121 3.10.7 Nêlêmwa …………………………………………………………… 122 3.10.8 Xârâcùu …………………………………………………………….. 123 3.11 Papua-Newguinea …………………………………………………………… 124 3.11.1 Koiari ………………………………………………………………. 125 3.11.2 Kobon ………………………………………………………………. 126 3.11.3 Bukiyip …………………………………………………………….. 127 3.11.4 Yelî Dnye ………………………………………………………..… 128

viii

Table of contents

3.12 Australia …………………………………………………………………..... 129 3.12.1 Guugu Yimidhirr …………………………………………………… 129 3.12.2 Ngalakan …………………………………………………………... 130 3.12.3 Jaminjung …………………………………………………………... 131 3.13 The Americas …………………………………………………….................. 132 3.13.1 Inside (and on the borders of) the Mesoamerican Sprachbund .…… 132 3.13.2 Zero-marking of Place and/or Goal in individual Mesoamerican languages …………………………………………… 137 3.13.2.1 Zapoteco of the Isthmus (Oaxaca) …………………..… 137 3.13.2.2 Tlapaneco of Malinaltepec (Guerrero) ………………… 138 3.13.2.3 Trique of San Juan Copala (Oaxaca) ……………….… 139 3.13.2.4 Huichol of San Andrés Cohamiata (Jalisco) …………... 140 3.13.2.5 Chinanteco of San Juan de Lealao (Oaxaca) …………. 141 3.13.2.6 Zoque of Chimalapa (Oaxaca) …………………..…..... 142 3.13.2.7 Huave of San Mateo del Mar (Oaxaca) ……………...... 143 3.13.2.8 Náhuatl of Acaxochitlán (Hidalgo) …………………..... 144 3.13.2.9 Yucateco of X-Hazil Sur (Quintana Roo) …………........ 145 3.13.2.10 Totonaco of Misantla (Veracruz) …………..………….. 147 3.13.2.11 Matlatzinca of San Francisco Oxtotilpan (Estado de México) ………………………………..…… 148 3.13.2.12 Mexicanero of la Sierra Madre Oriental ……………… 149 3.13.2.13 Guarijío of Arechuyvo (Chihuahua) …………………… 150 3.13.2.14 Seri of Sonora ……………………………………..…… 150 3.13.2.15 Interlude ………………………………………………... 151 3.13.3 Further data from Mayan languages ……………………………..… 152 3.13.3.1 Jacaltec ………………………………………………… 152 3.13.3.2 Kanjobal …………………………………………….… 153 3.13.3.3 Tojolabal ……………………………………………… 154 3.13.3.4 Mam …………………………………………………… 155 3.13.3.5 Itzá …………………………………………………..… 156 3.14 Pidgin and Creole languages ………………………………………………… 157 3.14.1 Portuguese-based Creoles ………………………………………….. 158 3.14.1.1 Angolar ………………………………………………… 158 3.14.1.2 São Tomense ………………………………………….... 159 3.14.1.3 Principense …………………………………….……… 160 3.14.1.4 Fa d’Ambû …………………………………………..... 161 3.14.1.5 Creole of Batavia and Tugu …………………………… 162 3.14.1.6 Papia Kristang ………………………………………… 164

Table of contents

ix

3.14.2 Spanish-based Creoles ……………………………………….....…. 166 3.14.2.1 Papiamentu …………………………………….....…… 166 3.14.2.2 Palenquero ……………………………………….…… 168 3.14.3 French-based Creoles ……………………………………………… 170 3.14.3.1 Haitian ………………………………………………… 171 3.14.3.2 Guyanais …………………………………………….… 173 3.14.3.3 Seselwa ………………………………………………… 174 3.14.4 Dutch-based Creoles ……………………………………………..… 176 3.14.5 English-based Pidgins and Creoles …………………………...….... 178 3.14.5.1 Pidgins ………………………………………………… 178 3.14.5.1.1 Nigerian Pidgin English ………………..… 178 3.14.5.1.2 Japanese Pidgin English in Hawaii ……… 179 3.14.5.1.3 Pidgin English in Nauru ……………….… 180 3.14.5.2 Central American Creole English …………………..… 181 3.14.6 On zero-marking in Pidgins and Creoles – in general ……….......… 182 3.15 Esperanto …………………………………………………………………… 183 3.16 Bridge …………………………………………………………………..…… 185

4

Case studies ………………………………………………………….......... 187

4.1

Zero-marking in contemporary literary French …………………………..… 188 4.1.1 Background …………………………..…………………………….. 188 4.1.2 Something is absent …………………………..……………………. 192 4.1.3 Privileged street names …………………………..………………… 193 4.1.4 Supposed counter-examples …………………………..…………… 195 4.1.5 Predicates …………………………..……………………………..... 201 4.1.5.1 Verbs combining with Goals ……………………….….. 201 4.1.5.2 Verbs combining with Places ………………………….. 205 4.1.6 Beyond the VP …………………………..………………………… 208 4.1.6.1 Semantics and pragmatics …………………………..… 208 4.1.6.2 Appositions and related phenomena …………………… 210 4.1.6.3 Attributes and secondary predication ……………….… 213 4.1.7 Verbs of communication …………………………..…………….… 216 4.1.8 Where do you live? …………………………..…………………..… 218 4.1.9 Are there preferences for being zero-marked? ………………………221 4.1.9.1 More data …………………………..……………..…… 221 4.1.9.2 Quantitative aspects …………………………..….......... 223

x

4.2

4.3

Table of contents

Zero-marking in contemporary literary Maltese ………………………….... 225 4.2.1 In medias res …………………………..………………………..…. 225 4.2.2 On Maltese spatial prepositions …………………………..……..… 226 4.2.3 Where spatial prepositions fail to show up ………………………... 227 4.2.4 Statistics …………………………..……………………………..…. 228 4.2.4.1 The first corpus text (RMT I) ………………………….. 228 4.2.4.1.1 On location-NPs ………………………….. 228 4.2.4.1.2 The verbal component of Maltese spatial constructions ……………………..……..… 234 4.2.4.2 The second corpus text (EWB) ………………………… 236 4.2.4.2.1 Some properties of location-NPs ………… 237 4.2.4.2.2 Zero-marking of spatial relations and verbs 240 4.2.4.3 Variation …………………………..…………………… 242 4.2.4.3.1 Constraints on pluralization ……………… 245 4.2.4.3.2 Specificity and related issues …………...… 247 4.2.4.3.2.1 Definiteness and indefiniteness ……………... 247 4.2.4.3.2.2 Genericity ………………… 248 4.2.4.3.2.3 Demonstratives and location-NPs ……………… 251 4.2.4.3.2.4 Possessed location-NPs ...… 254 4.2.4.3.2.5 Location-NPs with (and without) attributes …… 259 4.2.4.3.3 Individual place names with and without fi ‘in, at, on’ ……………………… 262 4.2.4.3.3.1 Street names ……………… 263 4.2.4.3.3.2 Toponyms ………………… 265 4.2.4.3.4 Variation with verbs ……………………… 267 4.2.5 A final word on the Maltese case …………………………..……… 272 French and Maltese – similarities and dissimilarities ……………………… 274

Part C: Insights and outlook 5

Evaluation ………………………..…………………………………..…… 279

5.1 5.2 5.3 5.4

Goal vs. Place ………………………..…………………………………....... Toponyms vs. common nouns ………………………..………………..…… Obligatory vs. optional ………………………..……………………….…… Interrelations ………………………..………………………………….........

280 287 291 296

Table of contents

xi

6

The end is near… ………………………..……………………………..… 301

6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9

Achievements (which pose questions) ………………………..………….… Zero-marking in the making? ………………………..…………………...… The OT-framework ………………………..……………………………...… Syntagmatic vs. paradigmatic types of zero ………………………..…….… Complement vs. adverbial ………………………..………………………… The Goal bias and mode hierarchies ………………………..……………… Revisiting the object dimension ………………………..…………………… A correspondence with noun incorporation ………………………..……..… Conclusions ………………………..…………………………………..........

302 304 307 310 311 312 313 315 316

Appendix A: B: C: D: E: F:

Corpus Rajt Malta tinbidel …………………..……………………………… Corpus Tout Maigret IV …………………..………………………………… Corpus Nestor Burma …………………..…………………………….......… Corpus L-ewwel weraq tal-bajtar …………………..…………………….… Sample languages …………………..……………………………............… Maps …………………..…………………………….................................… Sources …………………..…………………………..............................…… References …………………..…………………………............................… Index of authors …………………..…………………….............……......… Index of languages …………………..……………………….........………. Index of subjects …………………..………………………….............….…

319 340 355 363 367 372 383 385 397 401 405

Preface and acknowledgments

The idea to write this book together developed in the second half of 2011 when the three authors discovered that they had a topic in common, viz. spatial relations. This discovery was made within the framework of the Netzwerk Sprachkontakt und Sprachvergleich/Network Language Contact and Language Comparison at the University of Bremen/Germany with which the three of us happened to be associated with at that time. Since our basic convictions as to theory and methodology can hardly be claimed to coincide to their full extent, our study is a kind of (social) experiment. This book brings together an OT-inspired approach and semantic approach to spatial language, on the one hand, and, on the other hand, a crosslinguistic perspective rooted in functional typology. It has been a very inspiring experience for all three of us because we were given the chance to learn a lot from each other. Clearly, we could not aspire at overcoming all the problems posed by different and sometimes widely divergent terminologies, formalisms, categories, etc. However, it has not been our principal aim to develop a new framework for a common theory on the spot. One minor obstacle that we had to remove concerned the tone of our discussion of the cross-linguistic examples. We had to combine an allergy of the first author to the expression of epistemic modalities with the compulsion of the second author to make explicit the lack of detailed knowledge of the intricacies of each individual grammar. Mentioning the problem here is basically how we solved it. Obviously, we only have a superficial knowledge of the 130-something languages in our sample, but we will not repeat this again and again. The main goal of our project has been to demonstrate that zero-marking of spatial relations is an issue that offers numerous new insights for a wide range of different approaches in linguistics. More importantly, we have set out to show that the phenomenon under scrutiny is not marginal in the world of languages. That is why the empirical component of this study is especially prominent. This prominence of empirical matters explains our choice of sub-title for this book. We conceive of our study as a kind of door-opener for likeminded studies-to-be in which the topic of zero-marking of spatial relations will be elaborated upon. On the one hand, the crosslinguistic distribution of the phenomenon has to be completed and checked against genealogical, typological and areal parameters to allow us to evaluate the synchronic status of zero-marking of spatial relations more thoroughly. More in-depth cor-

xiv

Preface and acknowledgments

pus-based studies are called for as well as diachronic investigations of the genesis and change of zero-marking of spatial relations. It has to be checked in what way frequency or markedness can be evoked as determinants of zero-marking. Preliminary ideas in connection with these and other factors are discussed at various points in this book. We are looking forward to continuing our research program in the future and hope that it finds the approval of other linguists such that they take up the issue and study the phenomenon of zero-marking of spatial relations from as many angles as possible. A word of thanks goes to our audiences at the Linguistische Kolloquium on Language and Space which was held at the University of Bremen in summer-term 2011 and at the Max-Planck-Institute for Evolutionary Anthropology at Leipzig where we presented a draft-version of this study on occasion of the Bremen-Leipzig Meeting in December 2011. We are grateful to Norbert Boretzky, Eleni Buzarovska, Bernard Comrie, Victor Friedman, Paul Haggerty (and his anonymous wife), Martin Haspelmath, Birgit Igla, Marian Markovich, Susanne Michaelis, Stavros Skopeteas, Elisabeth Verhoeven and Søren Wichmann for their helpful comments on the previous versions. Carolin Sophie Ahrens, Deborah Arbes, Sonja Kettler, Nataliya Levkovych, Julia Nintemann, Robin Okrongli, Susanne Schuster and Cornelia Stroh of our Bremen team deserve a very warm handshake for lending us a helping hand by searching the shelves of the local libraries for evidence of zero-marking of spatial relations, by creating the appendices and by repeatedly rearranging the order in which the sections of this book appeared in several pre-final versions. We gratefully acknowledge the support of the University of Bremen whose decision to install the Netzwerk Sprachkontakt und Sprachvergleich/Network Language Contact and Language Comparison has created the most favorable conditions for us to investigate the issue at hand jointly. The authors assume the full responsibility for the ideas exposed in this study and for all shortcomings of form and content which might remain.

Thomas Stolz, Sander Lestrade and Christel Stolz Bremen/Germany and Nijmegen/Netherlands Februar 2014

Abbreviations

A ABL ABS ACC ACT ALL ANAPH ANIM ANT AOR AP ASI AUG AUX B BRIDGE CL COMP COMPL CON COND CONN CONSEC CONT CONV COP D D1 DAT DECL DEF DEM DET DIM

A-set pronominal ablative absolutive accusative actor allative anaphor animate anterior aorist active participle assertive augmentative auxiliary B-set pronominal bridging morpheme class comparative completive contrastive conditional connective consecutive continuative converb copula deixis proximate deixis dative declarative definite demonstrative determiner diminutive

DIR DIST DO DS DUR DUSU EMPH ERG EU EXI EZF F FOC FUT GEN HAB HON HORT ILL IMPER IMPERF INCH INCOMPL IND INDEF INDIC INESS INF INPL INTERR INTR INV IO IRR

directional distal direct object different subject durative dual subject emphatic ergative euphonic clitic existential ezafe feminine focus future genitive habitual honorific hortative illative imperative imperfective inchoative incompletive indicative indefinite indicative inessive infinitive inclusive plural interrogative intransitive invisible indirect object irrealis

Abbreviations

xvi ITV LIG LOC M MIX MOTIV NC NEG NFUT NMLZ NOM NONA NONPAST

NP NSGSU NT OBJ OBL PASS PAST PERF PL PLACE POR POSS POT

PP PPP PPA PRED PRES

intransitive ligature locative masculine mixed gender motive non-contingency negation non-future nominalizer nominative non-agent non past noun phrase non-singular subject neuter object oblique passive past tense perfective plural place possessor possessive potential prepositional phrase past passive participle present active participle predicate present tense

PRET PROG PROX PTCL PTCPL PUNCT R RED REFL REL RELATOR REP

SAE SBV SG SGSU SPEC SUB SUBORD SUBJ

SVO TAM TENSE TOP TOT TR TRL VEN VIS

preterite progressive proximative particle participle punctual realis reduplication reflexive relative relator repetitive Standard Average European subjunctive singular singular subject specific subject subordinator subjunctive subject-verb-object wordorder tense-aspect-mode tense topic totality construction transitive translative ventive visible

Part A Introduction and theoretical framework

1. Introduction

1.1. Appetizer This book explores the cross-linguistic use of zero markers to express spatial relations. To our knowledge no comprehensive stock-taking of this phenomenon has ever been carried out. To fill this gap, we have collected examples of spatial zero-marking from a large variety of languages. We believe our data will show that the use of zero-marking for spatial relations is anything but marginal in languages of the world. Promising as our results may be, however, in the absence of other typologically-minded investigations of the subject matter in hand, our ambitions of necessity must remain modest. Our main goal is to determine whether or not we are dealing with a phenomenon which is (a) wide-spread enough crosslinguistically and (b) displays recurrent structural and/or semantic properties, to warrant a follow-up study which may also connect the observed facts to other areas within the area of spatial language and/or beyond the limits thereof. Thus, we conceive of our study as a proof of concept of larger-scale, more systematic, and more in-depth typological and corpus studies. The strong emphasis on the empirical side does by not imply that theory and methodology are excluded from our line of argumentation. However, some of our ideas in the latter two departments will remain hypothetical until future studies complement ours empirically. Throughout this study we will be concerned with the following three research questions and their interaction: (1) (2) (3)

Is there a difference between the three concepts Place, Goal and Source with respect to the use of zero-marking? Is there a type of reference object that is more susceptible of zero-marking? Is zero-marking obligatory or optional?

Eventually the answers to these questions are meant to allow for generalizations which have universal scope, i.e., they are meant to hold universally unless there are typological, genealogical and/or areal constraints which limit the domain of the generalizations. Therefore, it is necessary that data from as many languages as possible are eventually taken account of to provide robust cross-linguistic foundations for the research project

4

1. Introduction

we are envisaging. We will now briefly introduce the first research question as an appetizer, reserving the discussion of the second and third research question for the following chapter. We first found out about the possibility of spatial zero-marking when (still independently from one another) studying the morphological means languages use to express spatial meaning. As a result of our focus on more morphological rather than lexical means, we were predominantly concerned with a restricted domain of spatial meaning, namely that of mode. In this domain, a distinction is made between locations with a static function (Places, in our terminology), end points (Goals), and starting points (Sources). We noticed that these distinctions were often not explicitly encoded, but left to context. In this connection a quote from Dixon’s (1980: 295) overview of the aboriginal languages of Australia is instructive: Verbs of motion […] can take complements with allative and/or ablative specification; verbs of rest […] may take complements with locative specification. But in fact locative information can be provided for any sentence – we can say that someone laughed/dreamt/died/hit/carried something AT a place. Locative is thus the most unmarked of the three local functions; of the other two functions allative is relatively unmarked, with ablative being most marked of all. These degrees of markedness […] seem to be universal […]. Thus, Dixon proposes a gradient markedness hierarchy locative < allative < ablative which is said to be universal. Replacing his morphosyntactic labels by the conceptual categories used in this book, the assumed hierarchy reads as in Diagram 1.1 Place < Goal < Source Diagram 1: A hypothetical markedness hierarchy One of the reasons we are interested in looking into zero markers in more detail is because of the possibility they offer to study such spatial markedness hierarchies. It is often assumed that conceptual-markedness hierarchies as the above correspond to differences in formal markedness (cf. Horn 1984). The opposition of zero vs. overt marking can be understood as the most extreme and hence clearest case of formal asymmetry on the expression side of linguistic signs. As we know from our previous work that mode distinctions easily allow for zero-marking, it should be possible to determine whether or not Place, Goal and Source indeed differ systematically in this respect. Given Diagram 1, one could expect to find zero-marking predominantly with Place alone. If 1

In the research alluded to above, similar hierarchies are critically evaluated (cf. Stolz 1992a: 77, 83 and 103 and Lestrade 2010: 107–108).

1.2. First examples (of methodology and terminology)

5

one thinks of this diagram as an implicational scale, one could predict next that if Goal is zero-marked the same should apply to Place and that if Source is zero-marked in a given language, the other two spatial relations should likewise attest to zero-marking. Alternatively, one could think of the hierarchy as expressing a difference in frequency with which zero markers are used for the different concepts. Our cross-linguistic survey should help to test the universal validity of Dixon’s hierarchy and determine its bearing on zero markers (i.e., whether it should be seen as an implicational scale or frequency ordering). It is not our intention to spill the beans about the results too early, but what we can disclose already at this point nevertheless is that things are by no means as straightforward as the above quote suggests. The hierarchy has to be revised – most probably by way of introducing parameterized flexibility as to the status of unmarked category.

1.2. First examples (of methodology and terminology) To give the reader an idea of what is meant by zero-marking of spatial relations, we summarize what Wiese (2012: 53–59) reports on the expression of spatial relations by a currently developing variety of German for which the author suggest the label Kiezdeutsch2 – not to be confused with Gastarbeiterdeutsch (‘migrant worker’s German’) of the 1970s (Holm 1989: 618–620). Wiese discusses bare locative NPs (“bloße Ortsangaben”) in this putative variety of German and their equivalents in the spoken and written standard. One of her examples of a bare locative NP is reproduced in (E1). (E1) Spoken German (Wiese 2012: 54) wir we locatum

sind gleich Alexanderplatz be:3PL soon Alexander_Square spatialV ground ‘We will be at the Alexander Square soon.’ (Place)

For the examples we introduce a special format. The first two rows follow standard practice. The third, additional row identifies the important ingredients of our discussion. The ontological system we adopt for our analysis and hence the terms that may appear in this row will be introduced in section 2, for now it suffices to note that this additional level is there. If necessary, the mode distinction of consideration is added between parentheses after the free translation, as is indeed the case in (E1). For obvious reasons, this mostly applies in the case of its zero-marking. Often it turned out to be very difficult to determine to which ontological domain a particular spatial marker belongs on the basis of the data in the grammar. It was not always clear from the examples whether a spatial item marks a constituent as a spatial adjunct in general, or whether it makes a 2

The glossonym is a compound which contains the Berlinian expression Kiez ‘inner-city neighborhood’.

6

1. Introduction

mode or configuration distinction. For this, it is necessary to study the spatial system in which the element figures in much more detail. Unfortunately, our present broad and exploratory set-up does not allow this. Given our research interests, we probably are guilty of overusing the mode label in cases of doubt, although we often simply refrained from categorizing a spatial marker into either category, leaving its analysis for future study. Fortunately, this need not be a problem: Also if the exact meaning contribution of an item is unclear, we can still contrast the contexts in which it is used with those from which it is left out, for example observing a difference between Place and Goal contexts. The grid assigns separate cells to the syntactic words unless bound (spatial) case markers are involved which occupy cells of their own. The morpheme glosses are often presented in the most economic manner, i.e. we avoid hyphenization in those parts of the examples which are irrelevant for the spatial relations as such. With the exception of examples for which our sources use scripts other than Latin-based systems, the examples come in the shape they have in the original. We acknowledge that this practice has the disadvantage that allographies are created if two or more different sources of the same object language opt for different orthographic representations of the data. Moreover we normally respect the morpheme glosses of our sources and thus accept that certain morphemes may be interpreted differently by different sources. In this way, our list of abbreviations is somewhat longer than it could have been had we decided to homogenize the morpheme glosses. Nonetheless, we are confident that our decision is not detrimental to the readability of the examples. Only where morpheme glosses were absent from our sources or problems arose as to their adequacy have we provided morpheme glosses of our own. Similarly, we keep the English translations given in our sources. If English translations were not available, we have provided new ones. For the occasional selfconstrued Dutch and German example we assume full responsibility on the basis of our native-speaker competence in the two languages (Dutch: Sander Lestrade, German: Christel Stolz and Thomas Stolz). For those English examples which are constructed by ourselves we rely on our learner’s knowledge of this language. Unless otherwise stated grey shading is used to identify those cells which are occupied by spatial relators and their morpheme glosses. Boldface marks those elements which are of additional interest in a given explanatory contexts (e.g. determiners). If needs be further typographic strategies are explained for each example individually. Depending on the size of the sentential examples, a grid may be split into several sub-grids. Wiese overheard utterances such as (E1) and dozens of its kind on the train from Potsdam to Berlin. She states that in everyday conversation hardly anybody would consider (E1) ungrammatical or deviant. However, if speakers want to speak or write German by the book, (E1) is certainly not among their options because the NP which fulfills the function of Place is supposed to be governed by a preposition – in this case by an ‘at’ as shown in (E2). In this example an, expressing the config(uration) AT, fuses with the dative form of the definite article dem, expressing the mode Place, to yield am.

7

1.2. First examples (of methodology and terminology)

(E2) Spoken German – standard norm wir we locatum

sind be:3PL spatialV

gleich soon

am Alexanderplatz at:DET:DAT Alexander_Square mode+config ground ‘We will be at the Alexander Square soon.’

Wiese (2012: 54–55) argues that (E1) is common also in the written register – for instance in official announcements of transport services or newspaper articles about public transport issues. In these genres one can find frequently constructions which lack the expected prepositional head as e.g. in (E3). (E3) Written German (Wiese 2012: 55) bei with

allen all:DAT.PL

Fahrten journey:PL

die

3

Bhf Charlottenburg station Charlottenburg REL locatum ground ‘For all trains which terminate at Charlottenburg Station…’ (Place)

enden end:3PL spatialV

Once again the preposition an ‘at’ is missing from the construction. What we expect to read is the PP am Bhf. Charlottenburg ‘at the station Charlottenburg’.4 The preposition pops up in the continuation of (E3) which we present separately as (E4). (E4) Written German (Wiese 2012: 55) besteht EXI

spatialV

an at config

der

Haltestelle Bhf Charlottenburg direkter Anschluss stop station Charlottenburg direct connection ground locatum ‘… there is a direct connection at the stop Charlottenburg Station’ DET:DAT

In contrast to the syntactic function of Bhf Charlottenburg in (E3), this expression is an apposition to the noun Haltestelle ‘stop’. In this combination, the preposition an ‘at’ is less likely to be dropped. There is an alternation of an and Ø which calls for an explanation. The reason for the employment of the preposition in (E4) most probably is to be sought in the word-class membership of the head-noun. Haltestelle ‘stop’ is a common noun whereas Bhf Charlottenburg can be interpreted either as a binary syntagm with a common noun Bahnhof ‘train station’ and a proper name Charlottenburg (i.e. the official designation of one of the major districts of Berlin) or it may be considered a complex proper name (with or without internal structure). As a proper name Bhf Charlottenburg defies the usual morphosyntactic rules that apply to common nouns. Among other things, proper names such as toponyms do not normally take definite articles in German. In the absence of the article, however, the employment of any preposition becomes awkward unless the toponym refers to spatially extended entity (like countries, islands, cities, etc.). Charlottenburg Station however is but a small dot on the city map 3 4

Bhf is the conventionalized abbreviation for Bahnhof ‘train station’. For a similar example from the world of tube stations and railway stations, cf. Lestrade (2010: 144).

8

1. Introduction

of Berlin and thus the place name which refers to this dot is not qualified for combining directly with a spatial preposition. On the basis of a field-study with native speakers of German, Wiese (2012: 55–56) assumes “that it is almost ungrammatical in the spoken register to use articles and prepositions in cases of this kind [i.e. in giving directions]” (our translation).5 On the one hand, this certainly is too bold a conclusion because the author has not conducted any experiment to elicit native-speaker judgments as to the grammatical acceptability of alternative constructions with overt prepositions and articles. To our minds, these alternatives are (almost) always legitimate no matter how stilted they may appear stylistically. On the other hand, Wiese’s results clearly indicate that bare spatial adverbials are a well-established commonality (perhaps not only) in colloquial German that calls for being studied more closely in the near future. For the time being it suffices to assume that, in colloquial (i.e. standard-like spoken and written) German, prepositionless and article-free constructions are restricted to Place NPs and probably to a much lesser extent also Goal NPs which are proper names – more precisely: genuine toponyms. In Kiezdeutsch, this restriction has been loosened considerably such that the NP is no longer required to be a proper name. Prepositions and articles are absent from spatial constructions with common nouns as in (E5)–(E6). Interestingly, examples of bare Source NPs are not discussed in our reference text. (E5) Kiezdeutsch (Wiese 2012: 53) heute today

muss must

isch wieder Solarium I again solarium locatum ground ‘Today I have to go to the solarium again.’ (Goal)

gehen go:INF spatialV

(E6) Kiezdeutsch (Wiese 2012: 53) un and

wenn when

du mal Party you once party locatum ground ‘And when you happen to be at a party…’ (Place)

bist be:2SG spatialV

The standard German versions of (E5) and (E6) would contain the PPs zum Solarium ‘to the solarium’ and auf einer Party ‘at a party’, respectively. Not only are the prepositions zu ‘to’ and auf ‘on’ missing from the Kiezdeutsch examples, but the definite article dem (reduced and fused with zu to yield zum) and the indefinite article einer (= the dative singular feminine of ein ‘a’) does not show up either. Neither Solarium ‘solarium’ nor Party ‘party’ is a proper name. Both are common nouns. Accordingly, Wiese (2012: 58–59) hypothesizes that the spatial constructions of Kiezdeutsch result from a 5

“dass es in der gesprochenen Sprache fast schon ungrammatisch ist, in derartigen Fällen Artikel und Präposition zu benutzen!”

1.2. First examples (of methodology and terminology)

9

generalization of rules which are firmly established in colloquial German. Therefore, she rebukes the commonly held belief that the existence of bare spatial adverbials in their entirety has to be attributed to influence or imperfect learning by non-native speakers of German. What the German case teaches us is that the situation is less straightforward as it seems on superficial inspection. In colloquial German, Place NPs and Goal NPs may come in the shape of bare spatial adverbials. It is important to note that this possibility is barred for Source NPs. It is still largely unclear what the factors are that constrain the absence of prepositions in spatial adverbials of colloquial German. Guessing from our native-speaker intuition we assume that zero-marking depends not only on the toponymical character of the Place NP or the Goal NP but also on the properties of the lexical verb of the clause in which the adverbial occurs. Verbs that have an important spatial dimension, such as existential, posture, and motion verbs (e.g., sein ‘to be’, stehen ‘to stand’, sitzen ‘to sit’, gehen ‘to go’, aussteigen ‘to descend (from a tram, etc.)’, einsteigen ‘to get on (a bus, etc.)’, umsteigen ‘to change trains’) seem to be fine in combination with bare spatial adverbials. This is different with non-spatial verbs some of which are hardly compatible with zero-marked spatial relations. Fully acceptable is an utterance like Ich bin jetzt Hauptbahnhof ‘I am (at the) central station now’ provided it is produced by the speaker while s/he is currently on a means of transportation (train, tram, taxi, etc.). The utterance is not acceptable at all in colloquial German if the speaker is standing inside or outside the building (whereas this is no problem for Kiezdeutsch). If we replace the existential sein ‘to be’ with an action verb like kochen ‘to cook’, an utterance like *Ich koche jetzt Hauptbahnhof ‘I am doing the cooking (at the) central station now’ is not admissible in colloquial German. Ich koche jetzt am/im Hauptbahnhof ‘I am doing the cooking at/in the main station now’, however, is perfectly well-formed and may invite the interpretation that the speaker is working for a fastfood stand inside or outside the building. Whether or not the prepositionless and articlefree version is possible in Kiezdeutsch cannot be decided our empirically insufficient basis of this variety. For the same reasons, we cannot exclude the possibility that Kiezdeutsch allows for zero-marking of Source relations although the data discussed in Wiese (2012) give no evidence of zero-marked Sources. We refrain from pursuing the German case any further in this study because of its cross-linguistic orientation. Nevertheless, the wealth of parameters which are involved in the triggering/blocking of zero-marking of spatial relations in German is symptomatic of the general situation. To determine the entire range of factors which constrain the application of zero-marking and those which enhance it, it is largely insufficient to rely on isolated examples in descriptive grammars. What is needed are large-scale comparative corpus studies because many aspects relevant to zero-marking of spatial relations only come to the fore if certain contextual conditions are met which are systematically neglected in grammatical descriptions. Since electronic corpora of this kind are available for only a very small number of languages of the world, this study of ours must

10

1. Introduction

make do with a check of the extant descriptive grammars in order to initiate the collection of data in the first place. Many grammars do not mention that zero-marking is an option in the language they describe. That is why we also have checked text anthologies and those sentential examples which are meant to illustrate phenomena other than spatial relations. This is the extent of the corpus-linguistic component of what follows. We acknowledge that our results are but provisional and have to be complemented and refined by fully-blown corpus-based follow-up studies.

1.3. Organization On the basis of the sketch of zero-marking of spatial relations in Kiezdeutsch and colloquial German, we can proceed to familiarize ourselves with the theory-borne aspects of our study (including terminological and sundry technical questions). This is one of the topics of section 2, which delimits the scope of our study by way of narrowing down the range of phenomena which are looked at more closely in the empirical sections of this book. In the same section, we address issues of a methodological nature and discuss previous approaches to the problem at hand. To do this properly it is necessary to embed our study of zero-marking in the wider framework of coding asymmetries in the realm of spatial relations (and beyond). Against the backdrop of the asymmetrical organization of the expression systems of spatial relations, we look into the problems posed by the linguistic concept of zero. This discussion concludes Part A of this study. Part B is devoted entirely to the empirical stock-taking of the cross-linguistic evidence of zero-marking of spatial relations. In section 3, we provide evidence of the phenomenon in a wide variety of languages from all over the world which represent several distinct phyla and different language types. Two case-studies are presented in section 4, one of which is dedicated to zero-marking of spatial relations in French (section 4.1) and the other sheds light on related phenomena in Maltese (section 4.2). Part C evaluates the empirical findings by way of putting forward a proposal as to the systematization of the observed facts. We determine in what way our (still preliminary) results may have a bearing on the typological study of spatial relations and on that of coding asymmetries in general. Moreover we identify a set of research questions which should be addressed in follow-up studies which are needed to complement our investigations. Since our own orientation in this study is strictly synchronic, a diachronic perspective on zero-marking of spatial would be highly welcome. Similarly, the point of view of areal linguistics as well as those of language-contact studies must be taken into consideration if one wants to better understand the distribution of zero-marking of spatial relations over the languages of the world.

2. Foundations

Our individual convictions as to the linguistics of space are laid down separately in three monographs of which Goldap (1991) is chronologically the prime mover followed by Stolz (1992a) and much more recently Lestrade (2010). The differences of our linguistic backgrounds notwithstanding, there is general agreement among us that zeromarking of spatial relations is an issue that urgently calls for being paid more attention by linguists. In our previous studies devoted to the grammar of space, we have noticed independently not only that zero-marking of spatial relations is attested in a variety of genetically diverse languages but also that the closer investigation of the phenomenon promises interesting new insights for various branches of linguistics. In this section, we expose those aspects of zero-marking of spatial relations we believe the reader must know before the cross-linguistic data collection is presented in Part B.

2.1. Generalities We use spatial language to talk about the position of things in space, that is, to communicate spatial situations. A simple example is given in (E7). (E7) My coffee cup is on the table. Spatial situations generally involve two participants, namely the locatum, an entity that is to be located, and the ground, a reference object whose position is taken to be known. A configuration function is used to specify a region in relation to the ground, as for example under, above, next to, etc. If such a configuration marker is absent, configuration is most likely zero-marked (see below), the alternative, less-frequent, option being that the locatum coincides with the Eigenspace of the ground, in which case configuration does not apply. The combination of configuration and ground provides the location at which the locatum is located. The relation between locatum and location is often made explicit by means of a verb with an important spatial dimension to it, such as existence (‘be at’), motion (‘go’, ‘put’), or posture (‘sit’), of which the locatum is a core argument. These verbs we will call spatial verbs. Finally, and for our present purposes most importantly, there is the mode function. Intuitively, one could think of the mode

12

2. Foundations

dimension as specifying the movement of the locatum with respect to the location. But theoretically, it is probably more appropriate to say that mode is concerned with the temporal development of the relation between locatum and location. In most cases, this could also be seen as specifying the link between the time interval established by the verb and the subinterval in which the location holds. We seem to have the choice between Goal, Place, and Source mode only (Lestrade 2012). Goal mode applies if the position of the locatum changes such that it ends up in the mentioned location (at the end of the interval expressed by the spatial verb), Source mode expresses that the locatum starts out in this location, Place mode, finally, is used when the locatum stays at the location throughout the spatial situation. The former two modes are often grouped together as they figure in dynamic situations, the latter is typically used in the description of static situations. This latter terminology may be misleading, however, as it is very well possible for a dynamic verb to combine with Place mode, as in the example John is running around in the room. Obviously, one could say much more when communicating spatial situations, for example specifying distance or orientation, but the ingredients just mentioned seem to be most basic. The different components and their relative organization are summarized in Diagram 2.1 As can be seen, configuration and ground together constitute the location that is linked to the spatial verb with the mode function. In the analyses below, we will mostly use this more fine-grained analysis, decomposing location into its constituting subparts without referring to their result.

LOCATUM

SPATIALV

MODE

LOCATION CONFIGURATION GROUND

Diagram 2: Basic elements semantic ingredients of a spatial situation Equipped with the above terminology, let us reconsider example (E7). We can now say that the locatum my coffee cup is located in a location that is defined with respect to the ground the table. This location is specified by means of the preposition on, which thus expresses the configuration. The spatial verb is indicates that no movement of either of the participants is involved in the spatial situation, hence we are dealing with Place mode. As can be seen, there is no separate expression of mode in this example and the mode function could therefore be said to be lexically specified by the verb.

1

The ontological system we use to discuss spatial language has been developed in a recent series of publications (Lestrade 2010, 2012; Reshöft & Lestrade 2013) and is heavily inspired by the references cited therein, among which Stolz (1992a), Svorou (1993), Talmy (2000), Levinson (2006) and especially Bateman et al. (2010) are most noteworthy.

13

2.1. Generalities

(E7’) English My

coffee locatum

cup

is spatialV

on config

the

table ground

In (E8) too, mode could be said to be lexically specified by the verb. The transitive action verb put is indicative of a movement as its object undergoes a change of place and ultimately winds up in a location which it did not occupy previously. Note that the preposition is not indicative of the difference between Place and Goal mode in these examples, although it would if it is replaced with onto in (E8). (E8) English I

put spatialV

the

coffee locatum

cup

on config

the

table ground

This is different with from in the following example, in which the table is marked as the starting point of the coffee cup and it is the configuration that remains underspecified instead. (E9) English the

coffee locatum

cup

dropped spatialV

from mode

the

table ground

The lexically specified downward orientation of the verb, albeit suggestive, is in principle independent from the mode function of the location. As the locatum could also drop on(to) the table, in which case the location would have the function of Goal, this time the preposition specifies mode. Note that, as an alternative way of expression, off can be used in (E9), expressing both mode (Source) and configuration (ON) at the same time. Such portmanteau morphemes are very frequent in languages of the world, as we will see. For ease of reference, we will sometimes use the cover term spatial (relator) to gloss over the exact meaning contribution (mode or configuration or both) of these items. Now consider the pair of examples (E10)–(E11) from Bukharian Uzbek – a Turkic language. (E10) Bukharian Uzbek (Aziz Djuraev, p.c.) siz you.PL locatum

bozor market ground

-ga borasizmi -DAT go:3PL:INTERR mode spatialV ‘Will you go to the marketplace?’

(E11) Bukharian Uzbek (Aziz Djuraev, p.c.) siz you.PL locatum

bozor borasizmi market go:3PL:INTERR ground spatialV ‘Will you go to the marketplace?’

14

2. Foundations

Both versions are grammatically acceptable without any reservations and can also be considered full synonyms of each other. They differ formally however because the dative-case marker of (E10) is absent from (E11).2 The dative case in (E10) expresses the development of the relation between the locatum (the addressee of the question) and the location, the marketplace. If the addressee eventually will be at the marketplace, the answer to the question should be yes. Interestingly, (E11) clearly shows that this spatial relationship need not always be overtly expressed in Bukharian Uzbek. Therefore, Bukharian Uzbek and Kiezdeutsch as well as colloquial German converge as to the optionality of the relators in spatial situations. Their use is by no means excluded and neither is it compulsory. Overt marking and zero-marking of spatial relations are options the speakers of these languages can choose from. This does not mean that the speakers are completely free in their choice, as we will show below. The expression of spatial relations is not the privilege of predicative constructions. The above examples illustrate strategies which involve VP-internal spatial adverbials or prepositional spatial arguments of verbs. It should not go unnoticed however, that spatial relations can also be expressed NP-internally, cf. the English examples in (E12)– (E14). For an elaborate discussion of the issue – without special focus on spatial relations – the reader is referred to Hagège (2010: 232–245). (E12) English – NP-internal Place the

message locatum

in config NP

the

bottle ground

never

reached

left

an

hour VP

was

sunbathing

its VP

destination

(E13) English – NP-internal Goal the

train locatum

to mode

Cardiff ground

late

tonight

NP

(E14) English – NP-internal Source the

girl locatum

from mode

Ipanema ground

on VP

the

beach

NP

With the exception of spatial verbs, there is no problem to identify the usual ingredients of a spatial situation also inside a complex NP.3 However, there are various reasons why we refrain from including NP-internal cases in our study. First of all, the bulk of 2

3

Note that we treat labels such as dative and allative as language-specific labels without any theoretical importance. Instead, we are concerned with the meaning contribution of these markers in terms of Goal, Place, and Source (or a combination thereof). Note that such NP-internal use is an important reason to generalize the definition of the mode function in the way described above, as it otherwise crucially involves a spatial verb.

2.1. Generalities

15

the data discussed in the literature on the grammar of space consists of predicative strategies. As far as we know, there is as yet no substantial discussion of spatial relations and their expression within NPs. Secondly the extant descriptive-linguistic sources from which we draw our data do not systematically address the issue of NP-internal spatial relations so that the large-scale comparison of languages is severely hampered. A third reason for postponing the investigation of NP-internal evidence is the number of problems which arise if we want to determine what counts and what does not count as an instance of zero-marking. In the case of (E12)–(E14), the ground is always the complement of a prepositional phrase which serves as the attribute of the NP representing the locatum. For (E13) and (E14) at least it is possible to replace [NP PP]NP by a compound the initial constituent of which indicates the location followed by the locatum: the Cardiff train and the Ipanema girl. There is no overt marker of the relation which has to be gathered from the context or world knowledge. Are these compounds examples of zeromarking of spatial relations? No matter how the answer to this question might be, it requires a thorough and detailed argumentation which is clearly beyond the scope of this study. We will therefore follow the general practice in the linguistics of space by limiting our discussion to predicative strategies. It is a linguistically very interesting question whether or not there is variation among the languages of the world as to the morphosyntactic representation of spatial situations. From the English renderings of spatial situations so far, bare DPs/NPs seem to be banned, an adpositional relator being obligatory to guarantee the grammatical acceptability of the utterance. The previous examples from colloquial German and Kiezdeutsch suggest, however, that the employment of (adpositional) relators is by no means compulsory even in languages which are relatively closely related to English.4 The overt specification of the character of the relation is subject to variation in cross-linguistic perspective – and, on top of that, also within a given language since not all locata, locations and relations behave necessarily the same. It is worthwhile therefore to determine in what way and to what extent the use of overt relators in spatial situations is subject to restrictions. This study is intended to support this idea by accumulating as much evidence as possible of zero-marking of spatial relations in cross-linguistic perspective.

4

We refer to a kind of Standard English as described in grammars such as Quirck et al. (1978) and Biber et al. (2011) neither of which mention that prepositions can be dropped with spatial relations. We refrain from checking the full array of nonstandard varieties of English for equivalents of the facts reported for colloquial German and Kiezdeutsch. The perusal of the volume edited by Kortmann & Upton (2008) yielded no evidence of bare spatial adverbials in the varieties of English in the British Isles. Since we do not consider English-based Creoles to be varieties of English, data from these languages are discussed separately in section 3 below.

16

2. Foundations

2.2. Previous studies There is as yet no comparable typological survey of this phenomenon. However, this does not mean that zero-marking of spatial relations has gone completely unnoticed in the extant literature. There has been a wealth of studies of spatial language in the past few decades, mostly focusing on spatial adpositions. To give but a glance of the range of aspects which have been investigated (in which the cited references only form a first pointer to the relevant literature): spatial meaning can be decomposed into a dimension of location and one of motion with respect to this location (Jackendoff 1983; Kracht 2002), languages may differ with respect to the frames of reference they use (Levinson 2003), languages differ in their choice of aspects of motion that are encoded by verbs (Talmy 2001), spatial locations can best be formalized using a vector semantics (Zwarts 1997), children exhibit language-particular preferences for spatial-meaning distinctions from a very early age on (Bowerman & Choi 2003) although they may show universal preferences too (Casasola & Cohen 2002; Gentner & Bowerman 2009), spatial adpositions should not be considered as spatial primitives (Herskovits 1986), the way adpositions organize space may differ between languages (Levinson & Meira 2003), and also syntactically, adpositions are anything but simple (Svenonius 2008). The present study will add yet another dimension to this broad field of study, focusing on an aspect of spatial language that has hitherto not systematically been studied, viz. the possibility in many languages to leave the spatial relationship unexpressed without this omission leading to miscommunication. In the literature on the grammar of space, the focus is on constructions which involve phonologically realized spatial markers because, among other things, the grammaticalization paths of these markers can be traced in order to reconstruct semantic maps (Svorou 1993). Zeroes do not lend themselves easily to this kind of study. However, this does not mean that zero-marking of spatial relations is not worth investigating in depth. To the contrary, we start from the idea that the occurrence of zero-marking may tell us something about the internal hierarchy of spatial relations and the organization of the system of which they form part. The above mentioned extensive interest in spatial language notwithstanding, the systematic aspects of zero-marking of spatial relations are still largely unexplored. What is more, the distribution of the phenomenon over the languages of the world has never been determined so that it is impossible to put forward any generalizations and/or typologically meaningful correlations and implications. In the absence of a sufficiently large cross-linguistic data-base, one cannot be certain whether or not there are genetic or areal preferences as to the employment of zero-marking of spatial relations. We consider it a challenge worthwhile facing to try and remedy at least one of these shortcomings in order to prepare the ground of more specialized follow-up studies. Our main purpose is to firmly establish the cross-linguistic occurrence of zero-marking of spatial relations.

2.2. Previous studies

17

A small number of precursors of ours did notice spatial zero-marking as something that goes beyond the incidental and language particular and that deserve to be mentioned here. The complexity of local relators is the topic of a short article by Stolz (1992a) in which zero-marking of spatial relations in Malay, Vietnamese and Hawaiian Japanese Pidgin English is given some thought. In his bulky volume on intransitive predication, Stassen (1997: 57–58) states that there are languages which lack oblique/adverbial marking with locational predicates, but which do feature a support verb with these predicates. Examples of this situation can be found in a fair number of genetically and areally diverse languages. To support this hypothesis, he provides a short list of one or two examples each from Maltese, Mandarin, Vietnamese, Mokilese, Yabêm, Bambara, and Wolof. Also Levinson (2003) discusses strategies of encoding spatial relations in the languages of the world. Apart from morphological case and adpositions, he concedes that there is zeromarking “in isolating languages like Thai” (Levinson 2003: 102) – which is certainly too restrictive a correlation of zero-marking and morphological type. In the Oxford handbook of case, Malchukov & de Swart (2009: 354) mention differential locative marking which they illustrate by an example from Persian where “case marking is exceptionally absent on place names.” In addition, they allude to the optional omission of locative markers in Kwaza and Tariana. In the same handbook, Creissels (2009a: 612) observes more generally that “geographical names” and “[c]ommon nouns characterizable as ‘natural locations’ (such as house, or village)” tend to “have particularities in relation with spatial cases.” One of these particularities consists of having “a ‘lighter’ spatial marking than most other nouns.” The marking may be as light as no marking at all – for which Creissels (2009a: 612–613) refers to the names of towns and countries which lack a locative form in Tswana and the so-called endingless locatives reported for Hittite. The label unmarked allative is employed tentatively for “the use of the absolute form of nouns in allative function without the help of any adposition” (Creissels 2009a: 622). The author claims that this phenomenon can be observed in a number of languages and refers specifically to Maale and Armenian. Lestrade (2013, 2010: 135–176) discusses the optional use of case from a more general perspective, providing additional examples of zero-marking of spatial relations from a selection of languages among which Bukharian Uzbek, Korean, Maricopa, Nepali, and Gujarati are newcomers to the extant list of examples. To our knowledge this is the extent of those studies which give zero-marking of spatial relations some consideration and formulate hypotheses about the factors which favor the employment of this strategy to the detriment of overt marking. To these investigations, one may add dozens of descriptive-linguistic contributions (such as grammars) in which zero-marking of spatial relations is touched upon (often in no more than an aside). Texts of this kind constitute the main resource of our empirical data collection in section 3. At this point, it suffices to mention specifically the work by authors

18

2. Foundations

who scrutinize spatial language in particular. In several of the contributions to Levinson & Wilkins (2006), zero-marking of spatial relations is addressed (though again mostly in passing) for Yelî Dnye (Levinson 2006: 160), Kilivila (Senft 2006: 213) and Yucatec Maya (Bohnemeyer & Stolz 2006: 284). In her monograph on Marquesan, Cablitz (2006: 302) compares Polynesian languages and other sub-groups of Austronesian. She concludes that “not all languages use case-markers or prepositions with place names, in particular Oceanic and Melanesian languages. […] However, in Polynesian languages place names occur with locative prepositions”. As examples of languages with zero-marking she mentions Longgu, Paamese, Kwaiio and Fijian specifically. These and other bits and pieces of information elsewhere in the literature, suggest that zero-marking of spatial relations is a phenomenon in waiting for being investigated linguistically. It is commonly known that it exists. Furthermore, many if not most of the examples given in the extant literature involve toponyms or place nouns as grounds (very often as complements of motion verbs). However, to what extent exactly zero-marking of spatial relations is made use of in the languages of the world and whether or not it is subject to systematic constraints is still largely a mystery. In our opinion, these scattered observations of zero-marking call for a more systematic cross-linguistic study and analysis. As we will show, zero-marking of spatial relations in fact occurs in language after language and seems to be analyzable in terms of the same general principles moreover. The evidence of zero-marking of spatial relations contained in the studies referenced in this section will be presented in section 3 provided that the data meet our criteria for the identification of zero-marking (to be spelled out in section 2.4 below).

2.3. (An excursus on) Asymmetries 2.3.1. Syncretisms Spatial relations form functional paradigms. Accordingly, their morphosyntactic expressions are expected to be in a paradigmatic relationship, too. Adopting the principles of Canonical Typology as outlined by Corbett (2005), we assume that the canonical version of the spatial categories displays a ternary structure with distinct constructions for the relations Place, Goal and Source, cf. Diagram 3. Place ≠ Goal ≠ Source Diagram 3: Canonical paradigm This means that syncretism of two and that of the entire set of spatial relations are instances of divergence from the canonical paradigm. Identical encoding of all spatial relations is not unheard of in the world of languages (Creissels 2009a: 614; Lestrade

19

2.3. (An excursus on) Asymmetries

2010: 96; Pantcheva 2011). The Austronesian language Palauan is a case in point as (E15)–(E17) show clearly. (E15) Palauan (Hagège 1986: 112) ŋmikilíl 3SG:PAST:play spatialV

r

a

sers a garden SPEC ground ‘The child was playing in the garden.’

RELATOR

SPEC

ŋálək child locatum

(E16) Palauan (Hagège 1986: 112) a SPEC

masaháru Masaharu locatum

r

a

RELATOR

SPEC

a

mo

SPEC

FUT

me go spatialV

ʔéi r a fishing RELATOR SPEC ground ‘Masaharu will go fishing tomorrow.’

klúkuk morning

(E17) Palauan (Hagège 1986: 112) mtəbəðí 2SG:extract:3SG spatialV

a SPEC

beáb r rat RELATOR locatum ‘Get the rat out off the hole!’

a SPEC

blsibs hole ground

The domain of the multi-purpose relator r is not confined to spatial relations since it is employed also for cause, instrument, agent and other relations (Hagège 1986: 113). This wide range of contexts in which r is employed can be interpreted as lack of semantic specificity. The relator r thus contributes nothing to telling the three spatial relations apart. Since the relator is too unspecific, the verb alone provides the necessary information. The Palauan r is a semantically empty general relator. One might be tempted to conclude that the three spatial relations under scrutiny are zero-marked because they can be distinguished formally neither among each other nor from other non-spatial relations. However, we do not consider cases of this type to fall into the domain of zero-marking. At least, they are not prototypical. We classify these cases as instances of syncretism or general indistinction. They are thus excluded from our survey. The same applies to systems of spatial relations which are based on a common formal marking strategy for Place, Goal and Source which is distinct from that of other non-spatial relations. Lestrade (2010: 142–143) discusses the case of the Arawakan language Tariana. Examples (E18)–(E20) show that there is a so-called locative suffix which is affixed indiscriminately to Place-NPs, Goal-NPs and Source NPs.

20

2. Foundations

(E18) Tariana (Aikhenvald 2003: 148) napidana uni -se 3PL:go:REP water -LOC locatum spatialV ground mode ‘They went into the water.’

(E19) Tariana (Aikhenvald 2003: 148) nawiki people locatum

pa:putʃita -se nehpanipidana one:CL:CLEARING -LOC 3PL:work:REP ground mode spatialV ‘People were working on a clearing.’

(E20) Tariana (Aikhenvald 2003: 148) hĩ DEM.ANIM

wyaka far

locatum

ground

-se -LOC

kanukaru REL:come:PAST:REL:F

dhumanaka 3SG.F:hear:PRES.VIS

wakunuku 1PL.speech: TOP.NONA

mode spatialV ‘She who came from far away understands our speech.’

The exclusion of syncretistic patterns of the Tariana type does not preclude the possibility that zero-marking applies in the language. As a matter of fact, Tariana provides examples of optional zero-marking of spatial relations since the general locative marker -se can be disposed of under certain conditions which will be discussed in section 3 below. There are languages however which do not require the presence of any kind of relator so that it might appear to be more appropriate to speak of genuine zero-marking in their case. However, the situation is not as straightforward as one might wish. Consider examples (E21)–(E23) from Chinese. (E21) Chinese (Bisang 1992: 86) wŏ I locatum

zài be_at spatialV ‘I am in Beijing.’

Bĕjīng Beijing ground

(E22) Chinese (Bisang 1992: 86) wŏ I locatum

qù go spatialV ‘I am going to Beijing.’

Bĕjīng Beijing ground

(E23) Chinese (Bisang 1992: 86) tā s/he locatum

chū exit spatialV ‘She left the room.’

fángzi room ground

2.3. (An excursus on) Asymmetries

21

In none of the three sentences from Chinese is there a separate relator which indicates mode and/or configuration. The entire functional load is with the verb which, however, does not specify configuration either. The Chinese situation is a challenge to an approach to spatial relations which aims at accomplishing an inventory of all instances of zero-marking. It is a challenge in the sense that it is controversial how the semantic contribution of the verb to the disambiguation of the spatial relations should be characterized. Mode is semantically incorporated in the meaning of the above Chinese verbs. In examples (E21)–(E23) the Goal-NP, Place-NP and Source-NP can be considered direct objects or internal arguments of these verbs (Bisang 1992: 86). If the morphosyntactic relation of these Chinese verbs and their locative objects instantiate zeromarking of spatial relations, then English combinations of verbs and complements like to enter (a room), to exit (a building), to inhabite (a country) may also be taken to represent zero-marking of spatial relations. Since this issue requires a more elaborate discussion, we relegate the solution of the problem to section 2.4.1 below. If we accept for the time being that instances of the Chinese type give evidence of zero-marking, the syncretistic pattern which covers all three spatial relations renders cases of Place = Goal = Source less interesting than those in which at least one of the spatial relations is expressed differently from the other two. This does not mean that complete syncretism is without any importance for a linguistic evaluation of systems of spatial relations. They are less interesting for this study, however, because the egalitarian treatment of all three relations precludes discovering the internal hierarchy along the lines of which systems of spatial relations are organized if the requirements of the above canonical type (cf. Diagram 3) are met. Thus, our attention is drawn to those constellations in which the strategies of encoding the three spatial relations differ on the parameter of formal complexity such that the expression of one or maximally two of the relations involved is achieved without phonologically realized markers. In sum, there are four patterns which arouse our interest throughout the empirical sections 3–4 and their evaluation in section 5, cf. Diagram 4.5

5

Lestrade (2010: 93) argues that Source and Goal “can only be encoded by the same marker if Place is encoded by the same marker too.” Since there seem to exceptions from this rule (e.g. varieties of Laz), Lestrade (2010: 104) assumes that exclusive Source-Goal syncretism must be “a semantically unmotivated diachronic accident”. Interestingly, Classical Nahuatl adds the spatial marker -pa to toponyms and common nouns which already the locative -c(o) or -pan (cf. section 3). In this combination, the NPs can have either a Source or a Goal reading (apart from a perlative reading). Thus, Mexìcopa can mean ‘from Mexico’ if it is combined with quīza ‘to come (out of)’ whereas it means ‘to Mexico’ if combined with yāuh ‘to go’ (Launey 1981: 55; Stolz 1992a: 20–21). This constellation seems to run counter to Lestrade’s above prediction. Future studies will have to reveal whether or not Source-Goal syncretism (with distinct Place) is as rare a bird as it seems on the present state of our knowledge.

22

2. Foundations

Place ≠ Goal ≠ Source Place = Goal ≠ Source Place ≠ Goal = Source Place = Source ≠ Goal Diagram 4: Interesting (syncretistic) patterns

2.3.2. More or less complex Does this self-imposed limitation make sense linguistically? The decision to focus on the patterns exposed in Diagram 4 rests on firm grounds. Even where zero-marking as such is not involved, we can observe that there are more or less pronounced differences as to the complexity of the morphosyntactic encodings of the spatial relations under review. These differences display a surprising consistency in terms of their crosslinguistic recurrence. If there are differences as to the complexity of the marking strategies in a given language, it can be predicted that less complexity associates with Place and/or Goal whereas more complexity overwhelmingly characterizes Source. For expository reasons, we keep our line of argumentation as simple as possible. This means that we refer to differences in complexity exclusively in terms of the number of segments and/or syllables which the marker of a spatial relation comprises. That there are other indicators of markedness and that there are reasons to challenge the concept of markedness is the topic of section 2.3.3 below. To economize on the length of our presentation, we confine the illustration to a small number of examples. For a start, we look at the noun declension of the Ugric language Vogul. Table 1 reproduces part of the paradigm of the noun xāp ‘boat’ which is representative of stems with a single coda consonant (Riese 2001: 24–25). There are three local cases, namely lative -n6 (employed for Goal), locative -t (employed for Place relations) and ablative -nəl (employed for Source relations). Case Lative Locative Ablative

Number Singular xāp-n xāp-t xāp-nəl

Dual xāp-iɣ-n xāp-iɣ-t xāp-iɣ-nəl

Plural xāp-ət-n xāp-ət-t xāpət-nəl

Table 1: Spatial cases in Vogul The suffixes of the lative and locative count exactly one consonantal segment. They are rather short phonologically. In contrast, the exponent of the ablative is relatively com6

On the basis of the data provided in Collinder (1957: 323), Stolz (1992a: 87) assumes a lative suffix -ne which probably reflects an older stage or a different variety since Collinder describes the situation in the Konda branch (= Eastern Vogul) whereas Riese focuses on the Northern branch which is at the basis of the literary variety.

2.3. (An excursus on) Asymmetries

23

plex as the suffix -nəl does not only consist of three segments but it also adds a full syllable to the word-form. This pattern – complex ablative expression vs. simple lative/locative expressions – remains the same across all three number categories. Furthermore, the pattern is the same for all stem classes of Vogul. There is no example of the ablative being encoded in a simpler way than the other local cases. Neither is there evidence of equal complexity/simplicity of their encodings. The Vogul situation can be summarized in the format of a tentative markedness hierarchy, cf. Diagram 5. Place
Goal

Diagram 5: Markedness hierarchy in Vogul If we broaden our empirical horizon, we encounter numerous situations which conform to that of Vogul insofar as Source tends to occupy the rightmost position in the markedness hierarchies established for individual languages. The accumulation of similar patterns is indicative of a cross-linguistic preference for marked Source in contrast to Place and Goal which, however, seem to compete for the rank of unmarked category. To put more flesh on our argument we review briefly a selection of comparable cases from other parts of the world. Turkish provides evidence for the marked status of Source since the ablative case receives the most complex encoding (suffix -dan/-den) whereas the dative-allative is expressed in a morphologically simpler way (suffix -a/-e) with the locative occupying a middle position with the suffix -da/-de, cf. Table 2 which contains the spatial cases of the paradigm of the noun ev ‘house’ (Ersen-Rasch 2012: 50). Case Dative-Allative Locative Ablative

Singular ev-e ev-de ev-den

Number Plural ev-ler-e ev-ler-de ev-ler-den

Table 2: Spatial cases in Turkish The differences in morphological complexity of the expressions or Place, Goal and Source are the same in both numbers and for all stem classes of Turkish. The markedness hierarchy we postulate for Turkish differs from that of Vogul because, for Turkish, the rank order of Place and Goal can be determined, cf. Diagram 6. Goal < Place < Source Diagram 6: Markedness hierarchy in Turkish

24

2. Foundations

In Quechua (Ayacuchano) the situation resembles those reported for Vogul and Turkish since Source is encoded by -manta which is the most complex of all exponents used for spatial relations. It contains five segments and adds two syllables to the stem. The locative suffix -pi and the directional suffix -man are shorter by one syllable each. Since -man comprises three segments and -pi only two, the markedness hierarchy of Quechua inverts the rank order of Goal and Place established for Turkish in Diagram 6. However, there is a second directional suffix -ta (Hartmann 1987: 96).7 The domains of the two directional cases probably overlap. We are not in a position to decide which of the two alternatives should be considered to represent Goal properly. Table 3 presents the spatial cases of the noun wasi ‘house’ (without the second directional case). Case Locative Directional Separative

Singular wasi-pi wasi-man wasi-manta

Number Plural wasi-kuna-pi wasi-kuna-man wasi-kuna-manta

Table 3: Spatial cases in Quechua (Ayacuchano) On this basis, Diagram 7 is a likely candidate for a markedness hierarchy of the spatial relations in Quechua (Ayacuchano). Place < Goal < Source Diagram 7: Markedness hierarchy in Quechua Boro is a Cushitic case-language. Of the three spatial cases directional, locative and ablative, it is the latter which is encoded by a disyllabic suffix the morphemes of the other two cases being monosyllabic (Lamberti 1993: 65–66). It is difficult to determine the hierarchical order of the directional and the locative as the differences are minimal on the phonological level. However, the short vowel in the directional plural as opposed to the long vowel of the locative plural may be understood as evidence of a relatively unmarked Goal. Table 4 surveys the spatial cases of the noun baarì ‘girl’ for the singular and asshúwoótsí ‘people’ for the plural. Case Directional Locative Ablative

Singular baar-uuk baar-uuts baar-uukə

Number Plural asshù-woóts-ok asshù-woóts-iits asshù-woóts-itse

Table 4: Spatial cases in Boro

7

Note that the separative -manta seems be composed of these two directional markers.

25

2.3. (An excursus on) Asymmetries

If we accept the minimal phonological evidence in favor of the unmarked status of the directional, the markedness hierarchy of Boro is identical with that of Turkish, Diagrams 6 and 8. Goal < Place < Source Diagram 8: Markedness hierarchy in Boro For the Daghestanian language Hinukh, it is easy to establish that Place is encoded simpler than Goal and Source. Place is encoded by the essive. The suffixes employed vary according to the noun classes. In Table 5 we present the case markers for four out of seven noun classes. The differences in the complexity of the markers are the same for all seven noun classes – in class I and II the essive marker is even shorter since it consists only of a lateral consonant (Isakov & Xalilov 1994: 184). Noun Class Essive -a: ~ -i: -qo -ho -de

II V VI VII

Case Lative -ar ~ -ir -qor -hor -der

Ablative -as ~ -is -qos -hos -des

Table 5: Spatial cases in Hinukh The markers of the lative case and the ablative case are equal as to their complexity. This allows us to assume the markedness hierarchy in Diagram 9. Goal

Source

Diagram 9: Markedness hierarchy in Hinukh The genetic isolate Basque has a rich inventory of cases. Table 6 selects only those of the spatial cases which are relevant for the issue at hand. It contains definite forms of the noun mendi ‘mountain’ in the inessive, ablative and allative (Villasante 1972: 94–98). Case Inessive Ablative Allative

Singular mendia-n mendi-tik mendi-ra

Number Plural mendie-tan mendie-tatik mendie-tara

Table 6: Spatial cases in Basque

26

2. Foundations

With the number-sensitive suffixes -n and -tan, the inessive is characterized by the simplest case markers in both numbers. In the plural, both Source and Goal are expressed by disyllabic morphemes whereas Place makes do with a monosyllabic marker. In the singular, Source and Goal also boast monosyllabic suffixes. However, the inessive -n is not even syllabic at all. Since the Source marker is longer than that of Goal by one segment in each of the two number categories, we postulate the markedness hierarchy in Diagram 10 which corresponds to that of Quechua (Ayacuchano) in Diagram 7. Place < Goal < Source Diagram 10: Markedness hierarchy in Basque The survey of case marking in Australian languages by Blake (1977) demonstrates that the bulk of the languages display patterns which are in line with the above markedness hierarchies. Table 7 surveys the spatial case morphology of four Australian languages which confirm the marked status of Source.8 Grey shading identifies the case form with the simplest encoding. The locative is always highlighted in this way. Boldface identifies the case forms with the highest complexity. The ablative is marked consistently as especially complex. The allative is as complex as the ablative in three out of four cases. Language Bandjima Munumburu Warrwa Nyulnyul

Locative -la -ra -n -uk

Allative -wali -gu -ngana -jun

Ablative -ŋuṛu -naŋka -nkawu -kun

Table 7: Case suffixes in Australian languages However, Australia hosts several case languages among which we find a number of “violations” of the general tendency towards higher markedness of Source. The language Wardaman (non-Pama-Nyungan) shows that Australian languages may not meet all of our interim expectations. According to Merlan (1994: 63) the spatial case markers of Wardaman are the ones presented in Table 8. Case Locative Allative Ablative

Suffix -ya ~ -ja -garr ~ -warr -ba ~ -wa

Table 8: Spatial cases in Wardaman The allative marker consists of a closed syllable of the CVC type. The markers of the locative and ablative, however, form open syllables of the CV type. Therefore, the abla-

8

No allomorphs are given. The data stem from Stolz (1992a: 80) and McGregor (1994: 26; 1996: 29). Note that McGregor considers the case markers of Warrwa and Nyulnyul to be postpositions.

27

2.3. (An excursus on) Asymmetries

tive marker contains an additional consonantal segment and thus is more complex than the markers of the other spatial cases. In contrast to all other languages discussed above, Wardaman does not assign Source the role of most marked category. In addition, in terms of segmental size, Place and Source are equally complex or simple in comparison to Goal. This situation is captured by the markedness hierarchy in Diagram 11. Place


Source Diagram 11: Markedness hierarchy in Wardaman Table 9 adds examples of supposedly “deviant” patterns of complexity to the list.9 The four Australian languages whose spatial case morphemes are presented in Table 9 do not behave homogeneously. Grey shading identifies the least complex marking which is the Goal marker three times out of four the fourth case being the Source marker. Boldface indicates the most complex morphological marking – and this characterization applies three times to Place and just once to Source but never to Goal. Language Yawarawarga Malag-Malag Minjung Gunin

Locative -nanja -yinŋa -pa -ngɨndalu

Allative -ŋuru -na -e -gu

Ablative -ka -maɲ -kal -yanga

Table 9: Special Australian patterns The Australian evidence cautions us not to jump to conclusions as to the supposedly universal nature of a given markedness hierarchy. What can be stated nevertheless is that on the basis of the data which have come to our notice, Source is most likely to be expressed by the more complex case exponent if there are any differences on the level of morphological complexity at all. Still this observation rests on too small an empirical basis to pass as a robust universal. Many more languages need to be scrutinized in order to formulate a water-tight hypothesis. What seems to be relatively clear for Source is less evident in the case of Place and Goal. With Source being the foremost candidate for the status of marked category, Place and Goal are competitors for the role of unmarked category. Since spatial categories are not only expressed by bound morphological case markers but also by free mor-

9

The data are taken from Stolz (1992a: 80) and McGregor (1993: 36). As for other Australian languages, McGregor classifies the case exponents as postpositions.

28

2. Foundations

phemes like adpositions, Table 10 provides evidence of different phonological complexity of triplets of prepositions or postpositions from five languages.10 Language Chamorro Gə’əz Georgian Kunama Tongan Yoruba

Place gi ba-ši -la i li

Goal para la-ši -tta ki li

Source ginen ‘əmna-dan -nkin mei láti

Table 10: Complexity of spatial adpositions In all cases, the phonologically most complex adposition is that of Source. In several cases, the Source adposition is two syllables long whereas those of Place and Goal are monosyllables. In Chamorro, both Goal and Source are expressed by disyllabic prepositions – with that of Source being longer by a segment. Similarly, the monosyllabic Source postpositions in Georgian and Kunama comprise more segments than their counterparts. Note also that Place and Goal are syncretistic twice whereas Source does not participate in any syncretistic pattern in Table 10. We take this different behavior of Source on the one hand and Place and Goal on the other to corroborate the tendency which we have observed with proper case morphology above.11 As to the hierarchical positions of Place and Goal, the structural solutions of the languages surveyed above are inconclusive insofar as it is sometimes one and sometimes the other spatial relation which can claim to be (relatively) unmarked. In the absence of a reliable cross-linguistic evaluation of the complexity of the morphological and/or morphosyntactic constructions which are employed to express the spatial relations under review, we are forced either to leave the question unanswered for the time being or to provide a convincing qualitative foundation of a language-independent markedness hierarchy. That it is perhaps difficult to establish a qualitative solution to the problem can be gathered from the fact that expressions of Place and Goal are identical in a non negligible number of languages. The languages of the Balkan Sprachbund are a case in point, cf. Table 11.12 The markers are prepositions. Note that, in Rumanian, Source is 10

11

12

The evidence stems from the following sources: Topping (1973: 119–125) for Chamorro, Weninger (1993: 32) for Gə’əz, Fähnrich (1986: 142–144) for Georgian, Bender (1996: 13–15) for Kunama, Broschart (1994: 141–165) for Tongan and Ashiwaju (1968: 77–78) for Yoruba. It goes almost without saying that there are also languages in which the complexity within the triplet of adpositions does not follow the expected pattern as e.g. Welsh Place yn ‘in’, Goal i ‘to’, Source o ‘from’ (Williams 1980: 127–142). The Place expression is more complex segmentally than those of Goal and Source. The difference is minimal (i.e. one segment). In addition, there are also languages like Austro-Asiatic Khasi in which the markers of all three spatial relations are equally complex: Place ha, Goal ša, and Source na (Rabel 1961: 74). The information is drawn from Buchholz & Fiedler (1987: 373–384), Radeva (2003: 309–317), Ruge (2001: 124–131) and Beyrer et al. (1987: 211–212).

29

2.3. (An excursus on) Asymmetries

expressed by a combination of prepositions one of which is identical with the common marker of Place and Goal. Language Albanian Bulgarian Greek Rumanian

Place në v se la

Goal në v se la

Source nga ot apò de la

Table 11: Place-Goal syncretism in the Balkans As Georgian and Yoruba show (cf. Table 10 above), this pattern is not unique to the Balkans. The syncretism of Place and Goal could give rise to the idea that these relations are equally basic in terms of the conceptualization of space. However, it is certainly premature to postulate a conclusion of this kind since no sufficiently extensive survey of the cross-linguistic facts has been undertaken yet. Moreover there are several languages for which the literature reports syncretism of Place and Source whereas Goal receives a distinct expression (Lestrade 2010: 100–103). The sentences (E24)–(E26) in Saami (North) illustrate a case of Place-Source syncretism. (E24) Saami (North) (Nickel 1990: 495) mun I locatum

lean be:1SG spatialV

viesu house ground ‘I am in the house.’

-s -LOC mode

(E25) Saami (North) (Nickel 1990: 495) mun I locatum

boađán come:1SG spatialV

viesu house ground ‘I come from the house.’

-s -LOC mode

(E26) Saami (North) (Nickel 1990: 492) mun I locatum

manan go:1SG spatialV

viesu house ground ‘I go into the house.’

-i -ILL mode

The noun viessu ‘house’ comes in two shapes. In (E24) and (E25), it is inflected for the morphological locative in -s. The spatial relations expressed are those of Place with the copula and Source with a motion verb. In contrast, Goal is expressed by the morphological illative marked by -i in (E26). It remains to be determined whether or not cases of the Saami kind form a minority sizable enough to require a thorough revision of all previously defended markedness hierarchies. It is a serious disadvantage that hard facts

30

2. Foundations

in terms of cross-linguistic statistics are largely missing. Therefore we are still in the impressionistic phase of the investigations.

2.3.3. Markedness – a difficult concept Establishing the markedness hierarchies is by no means a straightforward procedure. Frequently the differences in complexity of the expressions employed for the three categories are controversial especially if everything hinges, for instance, on the monosegmental vs. polysegmental analysis of affricates, geminates, nasal vowels, long vowels, diphthongs, etc. To circumvent the necessity of determining language-independently how the different phonological features can be integrated in a model of morphosyntactic complexity, we focus on the much more transparent case of zero-marking because the complete absence of a marker can be distinguished relatively easily from the presence of a phonologically realized marker (whereas it is difficult at times to decide which of two or three phonologically realized markers is expressed by the more complex means). As soon as we have taken stock of the constellations in which one or maximally two of the three spatial relations Place, Goal and Source require no overt encoding, we will be in a position to reconsider the above type of cases in which all categories display overt markers. On the basis of the evidence from zero-marking, it will be possible to interpret supposedly doubtful or inconclusive sets of data according to a general model. To narrow things down a bit further, our study takes above differences among the spatial relations seriously in the sense that we pay attention predominantly to the two principal contenders for the rank of unmarked category, viz. Place and Goal, whereas Source is discussed only unsystematically. Whether or not this practice is fully justified is a question which should be answered in a separate study. Thus, we start from the alternatives given in Diagram 12. One of the potential orderings is at odds with Dixon’s markedness as schematically represented in Diagram 1 above. This deviant ranking order assumes an unmarked Goal and a relatively marked Place.13 Place < Goal < Source Goal < Place < Source Diagram 12: Alternative ranking orders

13

Among the cases discussed above, there seems to be a slight majority which speaks in favor of the markedness hierarchy Place < Goal < Source. Yet this impression lacks statistical corroboration. The rank order Goal < Place < Source does not appear to be a marginal alternative either. That is why we take both possibilities into consideration until further evidence helps us to determine which of the options is unmarked by default and which is unmarked only under certain conditions.

2.3. (An excursus on) Asymmetries

31

The failure of the markedness hierarchies to be identical is easy to explain. They are based on completely different parameters. The decisive factor for Dixon’s approach is in part conceptually founded, i.e. it has a qualitative explanation. All kinds of actions can be executed at a given Place whereas for Goal and Source to be relevant for a spatial situation motion events must take place. In other words, Goal and Source can be specified only with a relatively small set of verbs (meaning: motion verbs). There is no such restricted to a sub-class of verbs for Place. According to this viewpoint, a lesser degree of markedness correlates with more freedom as to the combinations with predicates. The more marked categories are subject to more restrictions in terms of combinability with predicates. Some authors have argued for a markedness hierarchy among spatial relations on the basis of which the use of spatial zeroes should be predictable. Whether or not there are hidden presuppositions about the differential text frequency of Place, Goal and Source cannot be ascertained on the basis of the above quote from Dixon (1980: 295) alone. According to this author Place can be considered as the most unmarked of the three mode distinctions Place, Source, and Goal because “[Place] information can be provided for any sentence”. That is, whereas Goal and Source require a motion event, Place can be attributed to any event. Although this can be expected to lead to a difference in frequency indeed, it remains to be seen whether this “most unmarked” status also corresponds to the most unmarked expression, i.e. a zero. If our general framework sketched in section 2.5 below is correct, it is predictability rather than frequency that causes the usage of zeroes. If Place can be used anywhere, it could be argued not to be very predictable and hence unlikely to receive zero-marking. Instead, if Goal and Source mode are restricted to motion events, they are more predictable in such contexts and hence more likely to receive a zero expression. It is often argued that Goal is less marked than Source. On the basis of a typological study, Stolz (1992a: 103) observes a markedness hierarchy of mode distinctions in which Source is relatively marked with respect to Goal. Similarly, Lakusta & Landau (2005) show that learners of English often omit Source but not Goal meaning, which is explained by a general cognitive preference for Goal over Source mode. This bias could also be observed in the interpretation of so-called small clauses in English. As argued by Hoekstra (1988), a sentence such as They painted the door green is always interpreted such that the activity of painting leads to the result of the door being green and is never interpreted as an activity starting out with or caused by the door being green. Finally, Wälchli & Zúñiga (2007) explain the preference for the expression of Goals by the fact that Sources are often already given contextually. These interesting observations and explanations notwithstanding, the question is (like for the previous markedness distinction) whether they apply to the use of zero-markers. What is clear nevertheless is that our own language-individual markedness hierarchies in the previous section 2.3.2 are of a different nature. The basis for our decisions is the formal complexity of the morphosyntactic expressions of Place, Goal and Source

32

2. Foundations

independent of their combination restrictions or conceptual properties. One may object that the two approaches cannot come to the same conclusions since their points of departure are much too different. In the light of Haspelmath’s (2006) deconstruction of the linguistic concept of markedness, the incomplete compatibility of the markedness hierarchies discussed above is unsurprising. Given our preoccupation with the expression side of the linguistic signs, we give precedence to only one of several facets of what passes as markedness in the dedicated literature. With a view of establishing an allembracing hierarchy, this is certainly insufficient. If it makes sense at all to keep the concept of linguistic markedness alive, it is paramount to give a comprehensive list of all the factors which must be taken into account and to establish their internal relationship to the extent that a hierarchy of strength or dependencies can be set up systematically. This is an achievement which lies far beyond the scope of this study.14 For purely practical reasons, we continue to use the term markedness in lieu of that of morphosyntactic complexity. Wherever we refer to other kinds of markedness, these are specified by attributes.

2.4. On zeroes and related issues It is by no means an easy task to build up a cross-linguistic data-base of cases of zeromarking of spatial relations. This difficulty should not be taken as evidence of the scarcity of the phenomenon as such. The problem is connected directly to the insufficient documentation of pertinent cases. In descriptive grammars, zero-marking is normally not discussed in a separate section. This also applies to zero-marking of spatial relations. Since zero-marking is often optional and/or restricted to a subset of the nouns which may function as locations, information about supposedly secondary properties of this kind is frequently omitted from grammars. The omissions cannot always be attributed to space restrictions. As the above examples from colloquial German and Kiezdeutsch suggest, zero-marking of spatial relations is often a characteristic of nonstandard varieties or spoken language. Grammars with a normative orientation or an exclusively literary empirical basis are unlikely to mention the phenomenon explicitly. This means that one has to check a descriptive grammar of a given language in its entirety in order to be sure that the phenomenon under scrutiny is or is not attested. Sometimes examples of zero-marking of spatial relations can be found incidentally in chapters which are dedicated to non-spatial topics. 14

We have the experience of several large-scale cross-linguistic projects on subject matters different from that of this study. In Stolz et al. (2006: 166–195; 2007: 69–106), it is shown how a network of mostly formal parameters can be made fertile for determining the markedness relations which hold for semantically/conceptually interconnected categories. However, for the topic at hand, we still lack the empirical foundations to come up with comparable results.

33

2.4. On zeroes and related issues

Another area in which one may strike lucky is the sample texts which may accompany the grammatical descriptions. These examples which are drawn from non-spatial contexts require that we are absolutely certain that we are dealing with zero-marked spatial relations although the grammar itself does not provide the necessary clues.

2.4.1. Finding zeroes A first and crucial question we need to ask ourselves is how to identify an instance of zero-marking. What problems come up can be made clear by a discussion of examples (E27)–(E28) which are the English translations of the examples (E10)–(E11) from Bukharian Uzbek. (E27) English will

you locatum

go spatialV

to mode

the

market ground

(E28) English will

you locatum

go spatialV

home ground

In principle, the two sentences are very similar to one another. Both question the future location of the addressee. That is, also in (E28), the answer should be yes if the addressee is planning to end up somewhere, in this case, in their house. The two sentences differ crucially in their use of the preposition to, however, which is lacking from (E28). Apparently also in English, it is not always necessary to use an overt spatial marker, at least not with all types of ground. Moreover, a spatial relator could also be said to be dependent on the type of verb. Contrast the following two sentences, cf. (E29)–(E30). (E29) English the

man locatum

went spatialV

into mode+config

the

room ground

(E30) English The

man locatum

entered spatialV

the

room ground

Again, both sentences express a very similar meaning. In fact, it is hard to tell the semantic difference between the two. Nevertheless, only in (E29) is the spatial relation between the man and the room overtly expressed by the preposition into. A claim, however, according to which, in the latter example, a zero-marker is used, would ignore the lexical specification of the Goal role of the object of the verb enter in (E30). One probably would like to say that in these cases, it is better not to invoke zeromarking since we are probably dealing with core arguments, the role semantics of which is lexically specified by the verb. We would like to agree, but even in English the distinction

34

2. Foundations

may not always be clear. In addition to the transitive use of enter, one may use a prepositional construction too (cf. Wälchli & Zúñiga 2007, for other examples) as in (E31): (E31) English the

man locatum

entered spatialV

into mode+config

the

room ground

Although the verb enter probably specifies a Goal role for its complement, this apparently does not preclude the use of a specialized marker to express it. The choice now is between analyzing the use of into in (E31) as redundant vs. analyzing (E30) as an instance of zero-marking. This is a problem that is by no means peculiar to English. The Chinese examples (E21)–(E23) above are illustrative of the problem at hand. However, Chinese is far from being the only language which impels us to discuss the issue. Rennison (1997: 169) describes the situation with reference to Koromfe as follows: Quite generally, Koromfe has no noun-related means of distinguishing motion to, from or past something. These semantic functions are carried out by verbs […]. The local semantic functions are borne either by a preposition […] or a postposition […]. Often, the use of one of the verbs of motion alone is sufficient to characterize a following NP as expressing a local semantic function, and so no pre/postposition at all is used. The adpositions are superfluous, in a manner of speaking. If they are used at all they do not convey any information as to the mode of the spatial situation. The Berber language Tamashek displays a number of transitive verbs which express motion/rest, direction and optionally also configuration (Heath 2005: 575), cf. Table 12. Transitive Root -vkkʋ-vflʋ-vwvr-vhʋ-

Meaning ‘to go to’ ‘to leave; to go from’ ‘to be on something’ ‘to be in something’

Table 12: Tamashek transitives with spatial meaning For all four verbs, Heath (2005: 575–576) clarifies that they take direct spatial complements. No prepositions are allowed to intervene. (E32) Tamashek (Heath 2005: 576) nəhá 1PL:be_in spatialV

éhæn SG:house

ground ‘We are in the house.’

35

2.4. On zeroes and related issues

(E33) Tamashek (Heath 2005: 575) tə̀kka 3SG.F:go:PERF spatialV

éhæn SG:house ground ‘She went to the house.’

(E34) Tamashek (Heath 2005: 575) ifæ̀l=əd 3SG.M:leave=CENTRIPETAL spatialV

bæ̀mæko Bamako ground ‘He left Bamako (hither).’

Heath (2005: 272–273) states that [t]here is no specifically ablative or allative preposition. Ablative sense ‘from X’ is expressed indirectly using the transitive verb […] -vfvl- ‘go from, leave (a place)’. Allative ‘to X’ can likewise be expressed by the transitive verb […] -vkkʋ- ‘go to’. With other verbs of motion or conveyance, a non-directional locational adverb such as locative ‘in the desert’ can be glossed contextually as ablative or allative. In contrast, there is the preposition də̀ɣ ~ dæ̀ɣ ‘in, at’ which indicates “position in some space or enclosure, whether or not motion is involved” (Heath 2005: 281). The indiscriminate use of də̀ɣ ~ dæ̀ɣ ‘in, at’ is illustrated by examples (E35)–(E37). (E35) Tamashek (Heath 2005: 282) ə̀zzæʏæn dwell:PERF:3PL.M locatum

dæɣ ǽkall in SG:land config ground ‘They have lived in that country.’

éndæɣ DIST:ANAPH

(E36) Tamashek (Heath 2005: 283) ì DEM

dæ̀ ɣ in locatum spatialV config ‘…one who had sought to put me in(to) a school…’

ittǽræn 3SG.M:seek

ahǐ

æj

DEM:1SG.OBJ

DO.IMPERF

lækkol school ground

(E37) Tamashek (Heath 2005: 282) awa DEM:M:SG

ɣærr indeed

àtæn ǐkkæsæn dæɣ FOC:3M:PL:OBJ 3SG.M:remove:PERF:PL:PTCPL.M.SG in locatum spatialV config ‘…that is precisely what removed them from the land…’

ǽkall SG:land

For the purposes of this study, we opt for considering the optional employment of spatial relators in combination with semantically specified verbs which already incorporate

36

2. Foundations

the basic spatial information an instance of redundancy. If the spatial role could be said to follow from the semantic role assigned by the verb, we will not analyze this as an instance of zero-marking but as a lexically specified (hence overtly expressed) spatial relationship between verb and complement instead. Note, however, that this only holds for specific role assignments, such as INTO for enter and OUT OF for leave, which cannot be overruled with other, overtly marked spatial relations (cf. *leave into the room), although additional path modification may of course be possible (leave the room via the window). In contrast, if some motion verb in principle can combine with both zero-marked TO and FROM, the interpretation depending on the context, we will analyze this as an instance of zeromarking. In practical terms, we rely on the information provided by our sources. If the sources gloss a given static verb as ‘to be in (a place)’ and motion verbs as ‘to go to (a place)’/‘to come from (a place)’ or in similar ways, we discount their combination with spatial arguments as instances of zero-marking. If however the glosses are as unspecific as ‘to be’, ‘to go’ and ‘to come’, we assume that zero-marking applies. Admittedly, this is a rather crude way of handling this crucial problem. However, at this early stage of our investigation, it is technically impossible to implement a more sophisticated procedure unless one is inclined to delay the research considerably. It would be helpful for further studies to draw up an inventory of all those verbs which cross-linguistically are bona fide exemplars of transitive verbs with direct spatial complements. The ungrammaticality of an alternative and overtly expressed spatial relation could be used as a test to distinguish between a complement and zero-marking analysis. If only one spatial relation is allowed by a verb, this relation is lexically specified; if more than one are possible, some of which are covertly expressed, we will analyze the latter as zero-marked. We are very well aware of the difficulty to tell apart these two scenarios, especially using the grammar perusal method in Part B. Indeed, it could be that some of the cases we have identified as zero-marked adjuncts turn out to be verb complements instead (this may well be the case with – among others – our Khmer and Hmong examples in sections 3.9.5–3.9.6 below). More in-depth language-particular studies, as planned in a future project and illustrated already for French and Maltese in section 4, are necessary to solve this issue once and for all. The decision to differentiate between zero-marked adjuncts and verb complements does not say anything about the contrast between going home and going to the marketplace in (E27)–(E28), however. Should we analyze this as a lexical idiosyncrasy of home or as a systematic possibility for a zero? In other words, how big should the class of nouns be that allow zero-marking before we are justified to start talking about “systematic” zero-marking? Probably, we should do so right away. The zero-marked Goal role of home in example (E28) illustrates the phenomenon of present concern as something that is not at all exotic or exceptional, but garden variety instead. Languages may differ with respect to the extent to which they make use of the possibility to use spatial zero-markers. Importantly, this does not have to be generalized to all grounds and all relations to be recognized as such.

2.4. On zeroes and related issues

37

The identification issue is still not completely resolved yet. In examples (E27) and (E29) the preposition into in (E29) conveys more information than the preposition to in (E27). Whereas the inside region of the ground is specified in (E29), the exact location (on or at) with respect to the ground the marketplace is underspecified in (E27). It is up to the hearers to tell the difference from their world knowledge. Should we now say that a zero-marker is used in (E27)? Although we will ignore such cases, our answer should really be yes, as the exact location is left unexpressed. In this work, however, spatial expressions by necessity underspecify some aspects of spatial meaning, as it is impossible to encode all meaning details. For example, spatial expressions generally do not specify the exact distance between locatum and ground, i.e. they do not quantify but have an exclusively qualitative nature. Similarly, English at glosses over the Dutch distinction between aan and op, and both these languages ignore the Japanese FIT TIGHTLY vs. FIT LOOSELY distinction (cf. Bowerman & Choi 2003). Although such partial underspecification most likely follows from the very same principles that motivate the use of zero-markers, we will restrict our analysis and the use of this term for the absence of mode expressions. Thus, we may for example observe that some language uses a zero-marker for Goal but not for Source, glossing over the fact that configuration is zero-marked too. This is by no means a requirement, however. For example, in his discussion of cases such as that of the Dravidian language Koḍava Ebert (1996: 43) argues that [t]he relation of an object (figure) to a second object or place (ground) can be expressed with the help of local relators. The figure noun takes the genitive suffix and the relator is marked by one of the spatial case suffixes -lï for location, -kï for goal, -inja for source. With some relators the local case marker is optional. The examples in Table 13 are drawn from Ebert’s (1996: 43) slightly longer list. They show that Place and Goal mode may be optional also if configuration is still in place. There are no examples of the ablative -inja being optional. Although configuration is expressed explicitly, we analyze these as instances of zero-marking nevertheless, because of our interest in the expression of mode. Obligatory Case Suffix maratï-ra aḍi-lï under-LOC tree-GEN ‘under the tree’ mane-ra oḷ-inja house-GEN inside-ABL ‘from inside the house’ batte-ra bari-kï road-GEN side-ALL ‘to the side of the road’

Optional Case Suffix mangaḷuurï-ra pakka(tï-lï) Mangalore-GEN vicinity(-LOC) ‘near Mangalore’ mane-ra edike(-lï) house-GEN front(-LOC) ‘in front of the house’ mane-ra porame(-kï) house-GEN outside(-ALL) ‘out of the house’

Table 13: Obligatory vs. optional case-marking in Koḍava

38

2. Foundations

Note that in English too, it need not be the case that the configuration dimension is left unexpressed before the mode dimension. As Nikitina (2008) shows, also the Goal specification of the configuration IN can sometimes be omitted. That is, under certain circumstances, in can be used instead of into in the expression of INTO, as for example in Push the box in the room. In these uses, then, it is mode rather than configuration that is zero-marked.

2.4.2. Types of zeroes The example of a zero-marked spatial relation from Bukharian Uzbek we gave in (E10)–(E11) is an instance of what we will call a syntagmatic zero, which contrasts with the paradigmatic version. The latter, we believe, can be shown to derive from the former. A syntagmatic zero contrast with an overt alternative expression with virtually the same meaning. It is the result of a real-time pragmatic choice to leave out the overt expression of some communicated spatial relation, for reasons to be explained in the next section. A paradigmatic zero, on the other hand, is dependent on the paradigmatic contrast with other, overt markers. It is an empty but meaningful slot in a paradigm of spatial markers (comparable to what is called nominative or absolutive case in the structural case paradigms of many languages; cf. de Hoop & Malchukov 2008, Creissels 2009b). As an example of a paradigmatic zero, consider Aribwaungg, a Markham language spoken in Papua-Newguinea (Holzknecht 1989: 145; see also section 3.11 below). In Aribwaungg, the zero locative case contrasts with allative -en. The former marks Place, the latter Goal. By the absence of the Goal marker on a ground argument, the hearer knows that Place is meant. Crucially differently from the former type, Place is not the most likely interpretation because of which a zero-marker can be used, but the zero form simply is the way to encode Place. Thus, a paradigmatic zero derives its meaning from its position in a paradigm of overt markers. To use the wording of Mel’čuk (2002: 242): [a zero marker] must be opposed to non-zero signs, that is, participate in a perceptible semantic – that is, paradigmatic – contrast with overt signs. Informally, a linguistic zero must be an absence of any overt sign in a particular position, this absence being meaningful for the speaker/the hearer. In the literature on linguistic zeroes, there is disagreement about the nature and status of these categories (cf. among other Meier 1961, Lemaréchal 1997 and Rosén 2011: 145). For the purpose of this study, it is largely irrelevant whether one assumes an underlying structural category without phonological realization on the surface or the absence of any category no matter on what structural level. Thus, we are uncommitted as to the two alternative analyses in Diagrams 13–14.

39

2.4. On zeroes and related issues

[object GROUND]marker CONFIGURATION] __ MODE ] Diagram 13: Spatial construction with zero morpheme [object GROUND] marker CONFIGURATION] Diagram 14: Spatial construction without morpheme Diagram 13 represents a construction with two structural slots one of which is phonologically empty but corresponds to an underlying category which is associated to mode (and/or configuration mutatis mutandis). This is the “traditional” zero-morpheme analysis which contrasts with the analysis in Diagram 14. The latter does not presuppose a second slot but takes the surface word-form as sufficient to express the spatial relation under scrutiny. This analysis corresponds to the model of construction morphology (Booij 2010: 16–19). Both solutions are compatible with our investigation – at least for the time being. Zero-marking is of course a more general phenomenon which is not restricted to the realm of spatial relations. In spite of the time-honored and controversial discussion about the signe zéro (Jakobson 1939), there is as yet no cross-linguistic overview of zero-marking and its functional domain. Even approaches like Natural Morphology (Mayerthaler 1981) which are rooted in markedness theory have not attempted to determine with exactitude how wide-spread the phenomenon is and whether it displays traits of systematicity. In this sense, our study fills only part of a huge empirical gap which calls for further studies also of non-spatial subjects. An interesting complementary domain in which zero-markers figure is that of person marking. Cysouw (2003) proposes to acknowledge a zero-marker only if it is paradigmatically equivalent to overt markers with which it is functionally in complementary distribution. Consider his examples from Buriat in Table 14 (Cysouw 2003: 61).15 Overt Person Marker jaba-na-b go-PRES-1SG jaba-na-š go-PRES-2SG

Zero-Marking of Person

jaba-na-Ø go-PRES-3SG

Table 14: Overt marking vs. zero-marking of person in Buriat The absence of an overt person marking on the verb following the present tense marker can be understood as a zero-marker as it fills the slot in the paradigm that corresponds to the third person singular. There are three options for this position, if -b is inserted first person singular is expressed, if -š is inserted second person is expressed, and if the zero-marker is used third person is expressed. For spatial categories, the situation is the 15

The example is originally from Poppe (1960: 57).

40

2. Foundations

same if one of them is expressed obligatorily without a phonologically realized marker whereas at least one of the other relations boasts of a dedicated marker of its own the use of which is compulsory. 2.4.2.1. Splits The situation is less straightforward and therefore also much more interesting than the dichotomy of paradigmatic vs. syntagmatic zeroes might suggest. As an in-between version of syntagmatic and paradigmatic zeroes, a language may generalize its spatial zeroes to a certain class of grounds only. Alternatively, one could think of this as another paradigmatic type of zero that applies orthogonally to the previous one, i.e., differentiated by type of ground object instead of by type of spatial relation. (As we will see in section 3, in fact both dimensions often apply at the same time.) A constellation of facts of this kind constitutes a split because the members of a class – in this case: locations – behave differently as to their morphosyntactic encoding of spatial relations. Splits are linguistically important since they allow the analyst to detect covert distinctions of categories. Splits can be triggered by each of the main ingredients of a spatial situation, i.e. certain properties of the relation, the locatum and/or the location require that a construction is selected which differs from those employed for other relations, locata and/or locations. A cross-linguistically powerful parameter is animacy or empathy (Lehmann et al. 2000: 7–8), meaning: participants that are higher on the animacy scale are treated differently from those which rank lower on the same hierarchy. Accordingly, a location which bears the feature [+animate] may behave different morphosyntactically from a location which is classified as [–animate] (Creissels 2009a: 612–613). Kittilä (2008) studies this phenomenon in connection with differential Goal marking. However, it is not only Goal that is sensitive to animacy. For Sumerian, for instance, Balke (2006: 27) reports that nouns which belong to the “Personenklasse” (which comprises all nouns with the feature [+human]) behave differently from those of the “Sachklasse”, i.e. inanimate nouns. Table 15 is reproduced from Balke (2006: 27). Grey shading highlights those cells which give evidence of zeromarking. Case Dative Comitative Terminative Directional (allative) Locative Ablative

Human Noun -ra -da -še Ø Ø Ø

Inanimate Noun Ø -da -še -e -‘a -ta

Table 15: Case markers in Sumerian Human nouns do not mark any of the spatial relations overtly. Zero-marking is used indiscriminately for Place, Goal and Source in the person class whereas there are dis-

41

2.4. On zeroes and related issues

tinct case markers for each of the relations with inanimate nouns. Zero-marking of spatial relations in Sumerian, thus, does not contribute to identifying a hierarchy of the three relations under review. Another case in point is Krongo – a Kadugli language of Southern Sudan (which however does not involve zero-marking).16 The language employs case-prefixes. The prefix kú- is syncretistic for Place and Goal as opposed to nkı́ - which is employed for Source mode. However, if the Goal-NP is [+animate] the use of the prefix is barred. Animate Goals trigger the use of a serial verb construction (Reh 1985: 285), cf. (E38)–(E39). (E38) Krongo (Reh 1985: 285) ɲáaw

kú-

CONN:M:IMPERF:go

LOC-

spatialV

mode ‘And he goes to the street…’

fúunì street ground

(E39) Krongo (Reh 1985: 285) nyáaw 1/2:IMPERF:go spatialV

àʔàŋ I locatum

àtʊ́ná find mode ‘I go to him.’

ı̀ ʔı̀ ŋ he ground

Reh (1985: 285–286) argues that in constructions of the type illustrated by (E39) àtʊ́ná ‘to find’ is no longer a full blown transitive verb but a “reines Funktionswort” as it occupies the same slot as the locative prefix with inanimate Goals. That is she assumes that the erstwhile serial verb has grammaticalized to a special directional marker for animate Goals.17 Since the animate Goal in (E39) is represented by a pronoun, it is not clear to us whether or not the word-class differences of nouns and pronouns are a factor that determines which of the constructions is to be used. Independent of the correctness of the interpretation, what can be seen by contrasting (E38) and (E39) is the change of marking strategy in correlation to the animacy of the location. Krongo offers additional evidence of a different split. Reh (1985: 286) mentions a second special case, namely the preference of using the dative prefix on the inanimate Goal-NP if the motion verb is càáw ‘to go’. The dative replaces the locative if and only if a certain verb of motion is employed, cf. (E40) as opposed to (E38). (E40) Krongo (Reh 1985: 286) nyáaw 1/2:IMPERF:go spatialV

16 17

àʔàŋ I locatum

òDAT-

mode ‘I go to Tabanya.’

Tábáaɲá Tabanya ground

For a similar scenario in Basque, cf. Stolz (1992a: 43–44). Accordingly, Reh (1985: 285) glosses àtʊ́ná as human directional which suggests that the restriction is to animate locations but to human locations.

42

2. Foundations

The same verb is used in (E38) and (E40). Nevertheless, Goal is encoded differently. On the one hand, this means that the dative is not obligatory with the verb càáw ‘to go’. On the other hand, this verb seems to be the only predicate to allow for a choice. There is another, for our purposes much more important, dimension which is responsible for splits which are caused by NPs, namely the degree to which they are suitable as reference objects. At the extremes, we find NPs that refer to “natural” locations and those that refer to other concepts such as BREAD, PHILOSOPHY, MELANCHOLY, etc. Within the class of natural locations, we distinguish toponyms, proper names for individual referents that are inherently used as locations, from the related classes of place nouns, under which we group together common nouns that refer to landmarks (such as HILLTOP, RIVER-BANK, etc.), prototypical places (OFFICE, SCHOOL, CHURCH, etc.), and configurational nouns (UNDERSIDE, FRONT, SURFACE, etc.). In the main body of our book, toponyms and place nouns are invoked most often because they tend to be readily associated with zero-marking of spatial relations crosslinguistically. It is often the case that zero-marking is restricted to toponyms and place nouns – or to toponyms as in colloquial German. Toponyms have been shown by Lyons (1977: 222–223) to have a special status in the grammatical systems of many human languages. A relatively common situation is that described for the Bantu language Swazi: Names of places or of localities such as those about rivers, mountains or towns, etc. are also locatives without suffix -ini, especially so if the original meaning of the name has been forgotten. (Ziervogel 1952: 149) Swazi toponyms tend to be understood as inherently locative. They do not need to be overtly marked by the otherwise obligatory locative suffix -ini. This means that toponyms are morphosyntactically privileged in spatial situations as they do not have to be marked for the role they play. Unsurprisingly, the special status of toponyms is visible particularly clearly in spatial situations. In the Micronesian language Mokilese, for instance, place names obey morphosyntactic rules which are at variance with those which apply to other locations in spatial situations. Harrison (1976: 203) observes that [a] verb of motion does not require the suffix -oang ‘to’ pointing to the place towards which the motion is directed if that place has been reached or is likely to be reached. […] When the goal of a verb of motion […] is a place name [original boldface] the place name follows the verb directly. When the goal is indicated by some noun other than a place name, it cannot follow the verb of motion directly. In such cases, the goal of a verb of motion must be used in construction with a relational noun of location […].

43

2.4. On zeroes and related issues

Atelic predicates require a directional suffix on the verb as in (E41)–(E42). If the location is represented by an NP which is not a toponym18, an additional relational noun is inserted between the location NP and the verb. This additional relational noun specifies configuration such as pohn ‘on’ in (E38). If the event is telic, the directional marker is inadmissible. If the location is represented by a toponym, verb and location are directly adjacent, cf. (E43). (E41) Mokilese (Harrison 1976: 202) ih s/he locatum

illa go spatialV

-hng pohn -DIR on mode config ‘He is going to the shore.’

oaroahrro shore ground

(E42) Mokilese (Harrison 1976: 206) ngoah I locatum

aluhla -hng go -DIR spatialV mode ‘I am going to(wards) Kolonia.’

Kolonia Kolonia ground

(E43) Mokilese (Harrison 1976: 206) ngoah I locatum

aluhla go spatialV ‘I am going to Kolonia.’

Kolonia Kolonia ground

In atelic events, common nouns which serve as locations may either combine directly with the verb which hosts the directional suffix -oang or a relational noun is intercalated. The intercalation of the general spatial relator in is also possible. In is associated with vagueness (Harrison 1976: 206). (E44) Mokilese (Harrison 1976: 206) ih s/he locatum

indoa come spatialV

-hng -DIR mode ‘He came towards the house.’

umwwo house ground

(E45) Mokilese (Harrison 1976: 206) ih s/he locatum

18

indoa come spatialV

-hng -DIR

in LOC

mode ‘He came towards the house.’

umwwo house ground

There are names of institutions such as sukuhl ‘school’, sidowa ‘store’, pank ‘bank’ (all loanwords from English) which behave like toponyms unless they are made definite (Harrison 1976: 205).

44

2. Foundations

Harrison (1976: 204) argues that the general relator in is unlike the relational nouns which specify configuration. A translation equivalent is offered, namely English at. It remains unclear what role exactly is played by Mokilese in. Locations with the property [+human] usually combine with the relational noun ipen ‘near’ as in (E46). Since ipen ‘near’ marks configuration rather than mode, we may interpret this example as an instance of zero-marking of mode. (E46) Mokilese (Harrison 1976: 204) ngoah I locatum

aluhla ipen walk.PAST near spatialV config ‘I walked to where John was.’

John John ground

For static spatial relations, rules similar to those discussed in connection with examples (E41)–(E43) are valid. The verb mine ‘to exist; to be located’ combines directly with toponyms whereas other NPs need a relational noun to interact with mine, cf. (E47)– (E48). (E47) Mokilese (Harrison 1976: 209) ih s/he locatum

mine be_located spatialV ‘S/he is in Hawaii.’

Hawaii Hawaii ground

(E48) Mokilese (Harrison 1976: 209) John John locatum

mine be_located spatialV

nehn inside config ‘John is in his house.’

imwahu house:POSS.3 ground

The relational noun nehn ‘inside’ in (E48) can be understood as co-encoding mode and configuration although this interpretation is not without problems. If we assume that it is correct, then (E47) is an instance of zero-marking of Place. Mokilese thus proves that the properties of the location and those of the relation are crucial for the morphosyntactic representation of the spatial situation. In what way locata may have a say in the construction of spatial situations is less easy to determine. As far as we can judge on the basis of our data collection, the locatum hardly seems to have an influence on the occurrence of zero-marking. This, however, does not exclude the possibility that locata constrain the range of morphosyntactic solutions for spatial constructions. There is a rich literature on the subject of static verbs of location in connection with different classes of locata. The shape or substance of the reference object of the locatum is decisive for the choice of the suitable verb. A language which proves that the semantic class to which the locatum belongs has

2.4. On zeroes and related issues

45

morphosyntactic repercussions in spatial constructions is Fula (cf. section 3.5). Apart from this West-Atlantic language, however, we have not found further evidence of the same caliber. We assume that the absence of similarly convincing evidence from other parts of the world is largely caused by the descriptive-linguistic practice which seems to focus on the differentiation of the properties of the location-NP. On the basis of the information available in the extant descriptive literature, it is often hard to determine the difference between syntagmatic zero and the paradigmatic version that uses NP type (rather than mode distinction). It can be hypothesized that the paradigmatic type is related to the syntagmatic type via a generalization process. At first, zero-markers are only pragmatically allowed for a small set of grounds for or contexts in which the overt marking of the spatial relation is considered redundant. Over time, this option may spread to other contexts and grounds to become the default expression for some relation. In fact, the in-between zero option could be considered as evidence in favor of such a development, as an intermediate stage of generalization. The generalization process will be discussed in more detail in section 2.5. 2.4.2.2. Problematic toponyms The prominence of toponyms has been alluded to already in the foregoing section 2.4.2.1. We assume that toponyms are privileged frequently insofar as they do not need to be marked specifically for spatial relations. Either they defy being marked at all or they optionally allow for zero-marking. There are languages whose toponyms have an internal structure of the kind that a spatial marker forms part of the place name. With these cases the problem arises how to analyze the internal spatial marker. If it is fully functional syntactically, then no additional external spatial marker is needed. The functionality of the internal spatial marker speaks against zero-marking in the absence of an external spatial marker. Whether or not a spatial marker as component of a toponym is functional is a tricky question. Consider the case of the Pama-Nyungan language Mangarayi. Merlan (1989: 77–78) observes that [i]n locative case function, toponyms and cardinal directions are unmarked […]. Many toponyms are composed of an initial element (sometimes analyzable as a nominal denoting a natural feature or zone, or other meaning, sometimes simply an unanalyzable place name) and fused locative case suffix. […] In toponyms like these, the locative case element is not separable […]. Like other toponyms, these are not further suffixed in locative case function. […] Toponyms are optionally case-marked in allative meaning. The toponym Mañjuṇgan, for instance, contains the locative suffix -gan (preceded by mañjuṇ- ‘sour, rotten’). The locative suffix does not behave like a regular case suffix but is an integral component of the toponym. According to Merlan’s description, the

46

2. Foundations

place name Jembeṛe in (E49)–(E51) can be replaced with the toponym Mañjuṇgan, although only the latter contains a fossilized locative suffix. As to the optional affixation of the allative exponent in (E51), we assume that it is added to Mañjuṇgan which behaves like a stem. If however -gan is replaceable with -ḷama, the fossilization of the locative suffix cannot be complete. In the absence of any evidence to the contrary, we take for granted that the replacement of the two case suffixes is not possible. (E49) Mangarayi (Merlan 1989: 78) Jembeṛe Jembere ground

ganaḷaṇi 3>1INPL:sit spatialV ‘We are living at Jembere.’

(E50) Mangarayi (Merlan 1989: 78) Jembeṛe Jembere ground

wuḷayaj 3PL:go:PP spatialV ‘They went to Jembere.’

(E51) Mangarayi (Merlan 1989: 78) Jembeṛe Jembere ground

-ḷama -ALL mode ‘They went to Jembere.’

wuḷayaj 3PL:go:PP spatialV

We need evidence of the full fossilization of the spatial markers in complex toponyms in order to classify the constructions in which they occur as instances of zero-marking of spatial relations. In Classical Nahuatl, toponyms overwhelmingly display the above mentioned internal structure, i.e. they comprise a spatial relator, in this case a suffix. 19 There is an abundance of such suffixes in the classical language: -c(o), -pan, -cpac, -tech, -tlan, -tlōc, -nāhuac, -tzālan, -nepantlâ, etc. The place name Cuauhnāhuac (= modern Cuernavaca) can be analyzed as cuauh-nāhua-c tree-side-LOC ‘at the side of the trees’ (Launey 1981: 226–227). These toponyms do not have the same syntactic properties as common nouns usually have since they cannot function as arguments of verbs, i.e. they can neither be subjects nor direct objects; they are restricted to the function of adjuncts (Launey 1981: 54–55). This however holds for all nouns which host a spatial affix. The spatial affixes are insensitive to mode, i.e. the same markers are used indiscriminately

19

Morphologically it is doubtful whether we are dealing with bound markers in the strict sense of the term. Several of the so-called suffixes can be detached from their hosts. They are then adpositions which obligatorily bear a possessor prefix. This in turn is reminiscent of erstwhile relational nouns from which the spatial markers of Classical Nahuatl seem to have developed.

47

2.4. On zeroes and related issues

for Place, Goal, and Source (Launey 1981: 55). This is evident from example (E52)– (E54). Example (E55) is meant to show that toponyms behave like common nouns which bear a spatial marker. (E52) Classical Nahuatl (Launey 1981: 54) Mexì Mexico ground

-co -LOC mode ‘S/he lives in Mexico.’

nemi live spatialV

(E53) Classical Nahuatl (Launey 1981: 55) Xōchimīl Xochimilco ground

-co nìcihui -LOC 1SG:hasten mode spatialV ‘I am going to Xochimilco quickly.’

(E54) Classical Nahuatl (Launey 1981: 55) Cuauhnāhua Cuernavaca ground

-c -LOC mode ‘S/he arrives from Cuernavaca.’

àci arrive spatialV

(E55) Classical Nahuatl (Launey 1981: 264) nonīichtequi 1SG:RED:steal spatialV

in

cuezcoma granary ground ‘I frequently steal from (lit. in) the granaries.’ DET

-c -LOC mode

We may assume that the spatial markers indicate configuration in lieu of mode and thus the grey shading in (E52)–(E55) could be dispensed with. In some modern varieties of Nahuatl, the traditional system seems to have undergone changes. This is evident from Huasteca Nahuatl. Goal relations are generally unmarked, cf. (E56)–(E58). (E56) Huasteca Nahuatl (Beller & Cowan de Beller 1981: 5) niyahua 1SG:go locatum

nocha POSS.1SG:house

spatialV

ground ‘I am going to my house.’

(E57) Huasteca Nahuatl (Beller & Cowan de Beller 1981: 5) niyahua 1SG:go locatum

spatialV

tianquis market_square ground ‘I am going to the market-square.’

48

2. Foundations

On common nouns, there is no trace of the erstwhile spatial markers of the classical epoch. They are preserved however on toponyms as in (E58). (E58) Huasteca Nahuatl (Beller & Cowan de Beller 1981: 80) chicueyiya eight:PAST

ninemito Chicontepe -c 1SG:move:ITV.PERF Chicontepec -LOC locatum spatialV ground mode ‘Eight days ago, I went to Chicontepec.’

As with the examples from Classical Nahuatl, it is doubtful that -c(o) can be assumed to be a fully functional spatial marker. In the case of Classical Nahuatl, the doubts are connected to the failure of the spatial markers to encode mode. As to Huasteca Nahuatl, however, the situation is different because -c(o) and other spatial markers of the past are no longer productive and their occurrence is restricted to traditional toponyms. It is attested also on a number of adpositions on which it is fossilized.20 It is safe therefore to argue that the erstwhile spatial markers have completely fused with their host to yield an unanalyzable place name without internal structure. Thus, once more, the grey shading and the morpheme glosses in (E58) are anachronistic. In Huasteca Nahuatl, Goals are zero-marked no matter whether they are represented by place names or other nouns. For Place relations, the situation is more variated. There is zero-marking as well, but this strategy competes with PPs whose head is ipan ‘on’ (< Classical Nahuatl ī-pan ‘its surface = on it’), cf. (E59)–(E60). (E59) Huasteca Nahuatl (Beller & Cowan de Beller 1981: 43) itztoque be:PL

pesojme badger:PL

huan and

miac tosame ipan many dog:PL on locatum config ‘There are many badgers and dogs on my field.’

nomila POSS.1SG:field

ground

(E60) Huasteca Nahuatl (Beller & Cowan de Beller 1981: 15) ama today

ax

titequiti 2SG:work locatum spatialV ‘Aren’t you working in/on your fields today?’

NEG

momila POSS.2SG:field ground

(E61) Huasteca Nahuatl (Beller & Cowan de Beller 1981: 158) amo NEG

20

nitequiti Mexco 1SG:work Mexico locatum spatialV ground ‘I do not work in Mexico-City.’

For instance, the spatial marker -c(o) also survives as integral part of -ijtic ‘on the inside’ (derived from ìtitl ‘belly’) as in calijtic ‘inside the house’.

49

2.4. On zeroes and related issues

There is considerable variation in the use of ipan and Ø. In examples (E62) and (E63), the same location (escuela ‘school’ – a Spanish loan-word) occurs with and without preposition. (E62) Huasteca Nahuatl (Beller & Cowan de Beller 1982: 87) ma OPT

nimomachti escuela 1SG:REFL:learn school locatum spatialV ground ‘Would I only learn at school.’

(E63) Huasteca Nahuatl (Beller & Cowan de Beller 1982: 92) timomachtitiyas 1PL:REFL:learn:LIG:ITV:FUT locatum

ipan on config ‘Let’s go to learn at school.’

escuela school ground

Going by the evidence provided in Beller & Cowan de Beller (1982: 92), it is possible that ipan is used with locations preferably if the verb bears some aspectual or aktionsart specifications such as the itive in -ti-ya in (E63). Note that the employment of the prepositions is independent of the mode implied by the itive, because ipan accompanies the location NP also in static contexts such as the habitual (marked by -nemi) in (E64). (E64) Huasteca Nahuatl (Beller & Cowan de Beller 1982: 92) ne this

telpocatl tequititinemi ipan youngster work:HAB on locatum spatialV config ‘This young man works usually on his field.’

imila POSS.3:field

ground

As far as we can see, the preposition is never used in combination with proper toponyms. The change from the spatial grammar of the classical language to modern Huasteca Nahuatl can be tentatively described as follows: •

• • •

in Classical Nahuatl, spatial adverbials obligatorily bore a spatial marker – either suffixed (by preference) or adpositionally; these markers were insensitive to mode; the so-called locative suffixes of the classical language progressively fell into disuse in the post-classic period; the former spatial suffixes became dysfunctional and unproductive with toponyms whereas they disappeared completely from common nouns; in Goal and Place function, the bare location nouns (either with or without fossilized locative suffix) in combination with verbs whose task it is to specify mode;

50

2. Foundations

•

•

at the same time, the use of prepositions has increased considerably such that in contemporary Huasteca Nahuatl, Place expressions oscillate between zeromarking and the use of prepositions; whether or not the growth of the domain of prepositions is an effect of contact with Spanish is an issue we cannot look into in this study; the latter seems to be quite common, if the verb is overtly modified for aspect and/or aktionsart or if there are lexical NPs which represent the internal arguments of a transitive or ditransitive verb as in (E65).

(E65) Huasteca Nahuatl (Beller & Cowan de Beller 1982: 147) José José

quimama OBJ:load spatialV

etl ipan beans on locatum config ‘José is filling beans into his sack.’

iyoyoncoxtal POSS.3:sack

ground

To our minds, Huasteca Nahuatl is a language which attests to zero-marking of spatial relations. The zero-marking is of the syntagmatic kind, i.e. there is variation since prepositions and Ø are competitors of each other at least in some contexts. Toponyms which were overtly marked for configuration in Classical Nahuatl are zero-marked for both mode and configuration in Huasteca Nahuatl, i.e. zero-marking is an innovation at least in one half of its present domain. We admit that the details need to be looked into more deeply. Nevertheless, we are convinced that Huasteca Nahuatl illustrates the phenomenon which we propose to scrutinize cross-linguistically in Part B.

2.4.3. General case-drop – a false friend In his grammar of Tamil, Beythan (1943: 64–65) terminates the chapter on noun declensions with the remark that “in the colloquial style and in the condensed poetic speech, the case-endings can be suppressed largely” (our translation).21 Similarly, Korean allows for spatial case-markers to be dropped “in casual speech if no emphasis, exclusiveness, or contrast is expressed” (Lestrade 2010: 140–141), which is part of a more general phenomenon since it also affects the markers of nominative, accusative, genitive, dative (Sohn 1994: 231–232). Indeed, the optional use of spatial markers is reminiscent of the much more studied phenomenon of structural case alternation, mostly known as differential case marking (or DCM, for short; cf. amongst many others Aissen 2003, de Swart 2007, de Hoop & Malchukov 2008). Consider the Fore examples (E66)– (E67). In (E66), none of the two NPs is marked for case whereas in (E67) the nonhuman participant bears the ergative suffix.

21

“in der Volkssprache wie in der gedrängten dichterischen Rede, die Kasusendungen weitgehend weggelassen werden können”.

51

2.4. On zeroes and related issues

(E66) Fore (Scott 1978: 115) Yaga: pig

wá man

aegúye 3SG.OBJ.hit.3SG.SUB.IND ‘The man kills the pig.’

(E67) Fore (Scott 1978: 116) Yaga: pig

-wama -ERG

wá aegúye man 3SG.OBJ.hit.3SG.SUB.IND ‘The pig attacks the man.’

There seems to be general agreement in the literature that factors such as prominence, animacy, and volitionality play a role in DCM. For example, without the explicit marking of the opposite, humans are thought to act on lower animates (E66). Morphological case is used if this general principle leads to wrong interpretations, as illustrated in (E67). Although it is very likely that spatial zero-marking has something in common with DCM, two caveats should be noted. • •

First, zero-marking of spatial relations is not restricted to relations which are marked by bound morphological case, but also possible with adpositions. Second, any analysis that aims to cover both types of alternation should be phrased in much more general terms than have hitherto been used, as it is very unlikely that properties such as volitionality play a similar role in spatial language (for an attempt to reconcile the two alternations, cf. Lestrade 2013).

Aristar (1997) proposes an overarching view of the optionality of spatial and structural markers. He argues that pairings of case (or rather: semantic roles) and nominal can be more or less typical. Each case has its own particular typing criteria that different nominals meet to various degrees. Aristar discerns three strategies for dealing with potentially type-incongruent cases and arguments. First, a nominal that does not fit the typing criteria of some case may simply not occur in that case (blocking; cf. also Kiparsky 2004; Karlsson 1986). Second, a nominal that does not fit the typing criteria of some case may receive additional marking (bridging). And finally, an atypical combination of a nominal and a case may be reinterpreted (extension). The second and third strategies often apply together. For example in Yidiny (Aristar 1997: 317), the noun mandi ‘hand’ is more likely to function as Source from which a locatum originates and thus the ablative morpheme is directly attached to the stem as in mandi-m {hand}{ABL} ‘from the hand’. For the noun buña: ‘woman’, however, expectations are different in the sense that a human being is not considered a typical place of origin of something. This is the reason why the ablative suffix is connected to the stem by an interfix -ni- as in buña:-ni-m {woman}-{BRIDGE}-{ABL} ‘because of the woman’. Thus an additional bridging morpheme is used for the type incongruent combination of the ablative case with an animate argument. At the same time, the meaning is extended to the abstract domain.

52

2. Foundations

Aristar (1997) argues that the difference between typing effects of structural and nonstructural cases is that in the former there is a clear alternation of some phonologically realized marker and zero, whereas in the latter there is no such alternation. Instead, he claims, there always is an alternation in the use of a bridging morpheme, as illustrated by the above examples from Yidiny. According to Aristar, the reason that typing has not previously been identified as the coordinating principle in both structural and semantic case is that the alternation in the use of bridging morphemes is harder to detect than the alternation between a marker and zero. He suggests that because of the resulting focus on differential case marking in the structural domain only, linguists were unable to see the overall typing picture. However, typing each combination of case function and filler and marking incongruent pairings is said to explain both structural and semantic case alternations (Aristar 1997: 356). Although we obviously disagree with Aristar’s claim that all spatial DCM is between a spatial case and a more elaborate form, we subscribe to his general point. We too think that semantic roles that inherently follow from the semantics of a nominal may receive less formal marking. In our case, this may eventually even result in a zero expression. However, as said above, we believe that such predictability applies more generally, also exploiting textual and situational context. If in a given language all case markers or the majority thereof are subject to casedrop, the optional absence of morphemes which encode spatial relations is not an independent phenomenon. This means that dropping a Goal, Place, or Source marker in a language that may drop all sorts of markers may not tell us anything as to the internal hierarchy of the spatial relations because their morphological representation is subject to general rules which are dissociated from the realm of spatial language in the first place. For our study, general case-drop is not of primary importance, if at all. Only those structural constellations are of interest for us, which prove that case-drop involves only the spatial relations or a subset thereof. Wherever spatial relations behave like the bulk of the categories of a language’s conceptual system, there is no need to take account of the phenomena because we are dealing with an unspecific kind of zeromarking of relations. In this study, we focus exclusively on those cases for which it can be established that they do not instantiate general case-drop (but cf. Lestrade 2013 for a general account).

2.5. Theoretical and methodological issues of future research 2.5.1. The development of zero-markers Our study is synchronic. Diachronic observations such as made in the foregoing section can be made only unsystematically. Clearly the historical development of zero-marking

2.5. Theoretical and methodological issues for future research

53

– its genesis and its replacement by overt marking strategies – is too important a topic to be passed over tacitly. However, the diachrony of zero-marking deserves a booklength study of its own which is a project for the future. In this section, we will address the developmental questions from a different angle. We will argue that zero-markers are the result of principles of predictability and economy (following insights from, amongst others, Levinson 1983; Grice 1989; Zeevat 2000; Zwarts 2004; Haspelmath 2007). In section 2.5.2, it will be shown how the interaction of these principles can be formalized in an Optimality Theoretic framework. According to Grice (1989) speech participants are expected to observe a rough general principle according to which one should make a conversational contribution such as is required at the stage at which it occurs, by the accepted purpose or direction of the communication event. This principle is called the cooperative principle and tells the speaker to make true and relevant contributions that are intelligible but not too extensive. In spite of his interest in conversation, Grice only spells out the cooperative principle for speakers. But arguably, similar maxims should hold for hearers. If not, a sentence such as He turned on the switch and the motor started cannot be taken care of, as Levinson (1983: 146) argues. The interpretation one generally gets is that the starting of the motor is caused by turning on the switch, which does not follow from any of the Gricean principles. Therefore, Levinson (1983: 146) proposes to add an independent principle of informativeness that says to read as much into an utterance as is consistent with what you know about the world. In later work, Levinson (2000b) proposes simple heuristics, so-called generalized conversational implicatures, which are more explicit on the principle of informativeness. These heuristics can be seen as the hearer’s equivalents to some of Grice’s speaker maxims. The first heuristic implies that what is not said, is not so. The second heuristic requires that what is simply described is stereotypically exemplified, and according to the third heuristic what is said in an abnormal way is not normal. Since it is impossible to capture in language all the semantic details of some meaning intention, the speaker will generally only mention those aspects that the hearer does not know already or cannot tell from the linguistic and situational context themselves. For a simple example, the color of a mug only needs to be mentioned if it differentiates between two differently colored mugs, only one of which is the target. Otherwise, it suffices to ask Could you please give me the mug? We believe such economical language use may hold similarly for functional items such as the spatial markers under scrutiny here. That is, if the spatial relationship between two objects in the sentences is sufficiently clear already, the speaker need not express it explicitly. By the possibility to save the pronunciation effort necessary for this spatial expression the speaker can satisfy an economy constraint that seems pervasive in language and possibly human behavior in general (cf. Zipf 1965). Humans tend to use only as little energy as possible in achieving their (communicative) goals. If they can choose between elaborate and succinct utterances, they will generally opt for the latter. Addition-

54

2. Foundations

ally, and less misanthropically, the use of economical utterances can be motivated by the production bottleneck (see Levinson 2000a, 2000b). Pre-articulation processes have been shown to run faster than the articulation process itself (Anderson 1982; Wheeldon & Levelt 1995: 327) and comprehension can handle increased speech rates without any problems (Mehler et al. 1993). Because of this bottleneck, linguistic coding is costly: It slows down the communication process. By saving some costly and dilatory morphemes, the speaker can speed up the communication process. At first sight, the advantages of economical utterances for the speaker may be obvious. However, this use should of course not go at the cost of communicative success. Generally, repeating or rephrasing an utterance that was too short to be intelligible will cause more speaker effort than directly using a more elaborate successful expression. Thus, a speaker will only leave out those parts that are likely to be recoverable by the hearer, making sure that what needs to be said is said indeed (according to Horn’s 1984 R principle: Say no more than you must). That is, for being economical, the speaker has to exploit the principle of predictability. The importance of predictability in communication has been shown in a number of studies (cf. a.o. Grondelaers et al. 2002; Haspelmath 2008; Bybee 2010; Jaeger & Tily, 2010; Ernestus et al. 2002; Brouwer 2010). For example, Grondelaers et al. (2002) have shown how Dutch speakers use er ‘there’ to prevent hearers from making predictions about what will be said when this is considered futile. Contrast examples (E68)–(E69) below. (E68) Dutch morgen tomorrow

is be.3SG spatialV

er een there a ground ‘There will be a party tomorrow.’

feestje party:DIM locatum

(E69) Dutch in in config

het DET.NT

keukenkast staat kitchen_cupboard stand:3SG ground spatialV ‘There’s a cup in the kitchen cupboard.’

een a locatum

kopje cup:DIM

Whereas kitchen cupboards and posture verbs such as staan ‘to be positioned vertically’ in (E69) could be considered to allow the hearer to effectively narrow down the class of objects that qualifies for the locatum function, temporal adjuncts such as tomorrow and general existential verbs such as to be hardly give any such restriction in (E68): Virtually anything could be going on tomorrow. Thus, Grondelaers et al. (2002) argue, the speaker uses er in this sentence to warn the hearer to give up predicting and to pay careful attention to what they will say next, since whatever is coming is not predictable from the context.

2.5. Theoretical and methodological issues for future research

55

The ultimate reduction, and therefore the maximal economical expression, of a word is of course omitting it altogether. This is only possible if the meaning contribution it would make is completely redundant (or superfluous for present purposes). Therefore, the information by means of which the contribution of the spatial marker can be predicted and because of which it can be omitted should come from sources other than the (possibly degraded) phonological clues of the marker itself. If the only way to tell the relation between a locatum and the ground was via the spatial marker, it could never be left out, at least not without the loss of communicative success. This means that spatial information is not always concentrated in single markers, but at least sometimes is distributed over the clause. Such distributed spatial semantics is exactly what is argued for by Sinha & Kuteva (1995), Zlatev (2003), and Levinson (2003). Whereas traditionally theories of spatial language have mostly focused on verbs and adpositions in isolation, these authors argue against the application of the lexical-semantics approach to spatial expressions. Instead, they propose to analyze spatial meaning as being syntagmatically distributed over both closed-class and open-class items. For example, in our previous English example (E28) enter into, the INTO-meaning is distributed over both the verb and the adposition as the verb enter may assign an INTO-role to its complement also by itself. Differently from these proposals, however, we think that the sentence need not be the only source of information. The context of a spatial utterance is not only textual, but also situational. Following the pragmatic reasoning outlined above, a relation that can be observed directly does not need to be encoded in principle. More generally, any information that the hearer has about a spatial relationship from general world knowledge can be used by the speaker to use a more economic utterance. For example, as illustrated for Tamil in (E70), since the hearer knows that the default relation between a boat and water is for the former to be on the latter, the speaker need not bother encoding this explicitly (note that the spatial relation in this case is underspecified rather than zeromarked). They can simply use a generic spatial marker, the locative -il that generalizes over all spatial relations. (E70) Tamil (Pederson 2006: 405) Paaymarakkapil sailboat locatum

kaTTal -il ninRukkoNTirukkiRatu ocean -LOC stand.CONV.PROG.PRES.3SG ground config spatialV ‘The sailboat is on the water.’

Only if the relation between locatum and location is not predictable by world knowledge, for example in case of a sunken boat, or when the relationship needs to be emphasized, the speaker needs to use a more explicit and more elaborate construction in Tamil. In sum, by making use of the predictability of a spatial relation, the speaker can try to save themselves the effort of producing its expression, thus satisfying economy constraints. If the spatial relation between a locatum and a ground is sufficiently clear from

56

2. Foundations

other information sources already, however, a zero expression for the spatial relation can be used instead. What this means for the cross-linguistic data collection is that the evidence of zero-marking presupposes that the spatial relation which is expressed by zero-marking is recoverable from the context or from world knowledge. Since it is by no means always the same spatial relation that allows for being encoded by a zeromarker, the issue becomes interesting for language typology and universals research. It is the main goal of our study to demonstrate that zero-marking of spatial relations (and beyond) is worthwhile investigating on a grand scale. To this end, we outline a framework in section 2.5.2 which we consider useful for a theory-oriented scrutiny of the problem at hand.

2.5.2. A formal framework: Optimality Theory The above sketched optimization procedure between the speaker’s wish to be economical and the requirement for the utterance to be intelligible can be formalized in bidirectional Optimality Theory (bidirectional OT; Blutner et al. 2006). In addition to the standard OT assumption that language rules are violable constraints Prince & Smolensky (1993/2004), bidirectional OT evaluates candidates on their communicative qualities. That is, both from a hearer and speaker perspective. The use of zero-markers can thus be modeled as a bidirectionally optimal solution for the expression of a meaning. More specifically, the semi-bidirectional version of OT proposed by de Swart (de Swart 2007, in prep.; cf. also Zeevat 2000) will be used here. Differently from standard bidirectional OT, in this version, there is only one meaning for the expression of which in principle only one form is considered. The production of a sentence is constrained by its interpretation. The speaker checks if the optimal candidate from their speaker perspective will indeed lead to the right interpretation. If not, they will resort to a suboptimal form that probably will get the meaning across. This semi-bidirectional version mostly describes the use of syntagmatic zeroes; the paradigmatic zero arguably follows from a process of automatization, in which an optimization process becomes standardized (cf. Lestrade et al. 2012). The constraints we use are inspired by the work of Grice (1989) and Levinson (1983, 2000b) and have been discussed in detail in Lestrade (2010; although we substitute PREDICT for his ADD WORLD KNOWLEDGE (AddW), to better capture its intended general domain). Consider (4). (4) (4a) (4b) (4c)

Constraints FAITHL: interpret linguistic signs PREDICT: enrich the utterance proper with any additional information available to predict the precise meaning. ECONOMY: be economical in expressing what you want to say.

57

2.5. Theoretical and methodological issues for future research

Note that similar constraints have been proposed in other work in OT. For example, in a lexical semantic study of (a)round, Zwarts (2004) uses FIT and STRENGTH. The first assumes that the interpretations should not conflict with the (linguistic) context, the second that stronger interpretations are better than weaker interpretations. Strong and weak interpretations are determined on the basis of inclusion. Strong interpretations include weaker ones (cf. also Hogeweg 2009). Although Zeevat (2000) proposes slightly more specific interpretation constraints, in all cases, the general idea is very much the same. The hearer needs to conform to some general principles for successful communication. Also the principle of ECONOMY is assumed in many OT studies, often in more specific formulations to specifically deal with the variation of concern. The intuitive motivation of it is shared by many linguists. We simply cannot express all semantic details, so choices have to be made. By exploiting the other two constraints, the speaker can save some costly morphemes and speed up the communication process. Thus, the workings of the constraints are closely intertwined. ECONOMY tells the speaker to use as few or as general words as possible given the other principles. PREDICT states that the hearer should enrich the semantics proper of the utterance with any relevant piece of knowledge about the world or the discourse situation. However, this should never happen at the expense of what is being said explicitly, which is stated by FAITHL. As the hearer knows, anything the speaker says is in spite of ECONOMY and is therefore relevant. Finally, note that Gricean principles that ask for relevant and true contributions are not represented in these three constraints, nor are they formulated as independent ones. They are simply assumed to be satisfied at this point. This type of principle pertains to the selection of the input and therefore is of no relevance once some meaning is selected by the speaker. To represent the optimization procedure, constraints are ranked on the basis of importance from left to right in so-called tableaux. Optimality is determined on the basis of violation patterns on these constraints, which are marked by asterisks. The candidate that best satisfies the ranked constraints becomes optimal from a production or interpretation perspective. To be chosen, however, a candidate also needs to lead back to the correct interpretation again. Technically, it needs to be bidirectionally optimal. We will illustrate this with an analysis of the following pair of examples from Maricopa, cf. (E71)–(E72): (E71) Maricopa (Gordon 1986: 48) Kwnho basket locatum

lames table ground

-k -LOC config ‘I put the basket on the table.’

’shvawk 1SG:put:R spatialV

(E72) Maricopa (Gordon 1986: 48) Kwnho basket locatum

lames table ground ‘I put the basket on the table.’

’shvawk 1SG:put:R spatialV

58

2. Foundations

The optimization procedure of (E71) is illustrated in Tableau 1a, in which e in the input stands for the neo-Davidsonian event argument (but any other semantic representation could have been chosen). From a production perspective, the zero-marked form is more economical than the case-marked alternative that by necessity violates ECONOMY more seriously than the former because of the difference in pronunciation effort corresponding to the case morpheme. Hence, the zero-marked option is preferred (as indicated by the single arrow). In the interpretation check for this candidate, FAITHL does not apply, at least not with respect to the semantic role interpretation of the table, since there is no overt marking used. Yet if the context is considered to be insufficient to tell the semantic role of the table, PREDICT too cannot decide between interpretation alternatives (of which only Instrument and Ground are considered here). Thus, the form is considered ambiguous: Both interpretation candidates score equally well. Instead, using explicit marking does lead to the intended meaning, as the interpretation check of the second form candidate shows. Any interpretation of a noun table that is marked for locative other than that as a ground role would ignore the overt spatial marker and hence violate FAITHL. In spite of its violation of ECONOMY, if the speaker wants to be understood, they have to use an overt marking (indicated by the double arrow). This is what Tableau 1 tells us. Prod:

Int a:

Int b:

Put(e) & Agent(e, I) & Theme(e, basket), Ground(e, table) a. table-∅ b. table-loc table-∅ table = Instrument table = Ground table-loc table = Instrument table = Ground

FAITHL

PREDICT

ECONOMY

FAITHL

PREDICT

* ECONOMY

FAITHL *

PREDICT

ECONOMY

Tableau 1: Semi-Bidirectional Analysis of spatial marking in Maricopa (E71) Against this backdrop, consider the optimization procedure of (E72). Again, the zeromarked form is preferred from a production perspective. This time, the speaker thinks that the semantic role of the table follows sufficiently from general knowledge (technically, PREDICT applies and any other interpretation than Ground will violate it). Since its Ground interpretation follows for free, it does not have to be marked explicitly. The zero-marked form, in this case, leads to the correct interpretation and is therefore bidirectionally optimal. Note that also marking this role explicitly would have led to the intended meaning (as Int b in Tableau 1 shows), which however, would have caused an unnecessary violation of ECONOMY, cf. Tableau 2.

59

2.6. The final touches Prod:

Int a:

Int b:

Put(e) & Agent(e, I) & Theme(e, basket), Ground(e, table) a. table-∅ b. table-loc table-∅ table = Instrument table = Ground table-loc table = Instrument table = Ground

FAITHL

PREDICT

FAITHL

PREDICT *

FAITHL *

PREDICT *

ECONOMY * ECONOMY

ECONOMY

Tableau 2: Semi-Bidirectional Analysis of spatial marking in Maricopa (E72) This section has shown how the preference of the speaker to use the most economical form can be formalized in OT. It was illustrated for the choice between zero-marking and an overt marking of the ground role. The former is used only if the speaker thinks the hearer will identify the ground function themselves, otherwise more material will be added until the speaker thinks s/he will be understood. It can be hypothesized that after performing many optimization procedures that lead to the same candidate, speakers eventually generalize over inputs and choose the same output without checking its interpretation. In our OT formalization, this could be represented in the form of the development of a new (high-ranked constraint) member of the FAITHL family of constraints, which for example requires zero to mark typical ground objects when performing a Ground function. Depending on the conditions of such a constraint, i.e., whether it is restricted to a certain type of objects or holds for all ground roles alike, it will account for the in-between or paradigmatic type of zero.

2.6. The final touches There are many issues that we have to skip for reasons of space. We exploit this section to touch upon some of those aspects of zero-marking we believe to deserve being mentioned before we set out to look at the empirical side of the phenomenon. Before we terminate Part A, it is necessary to circumscribe the kind of constructions which form the input of Part B. Since this is an exploratory study, a first in the framework of the linguistics of space, we do not want to exclude too many interesting data from our cross-linguistic collection. On the other hand, we cannot afford applying unlimited tolerance in order not to get lost in the wealth of the empirical data. Accordingly, in section 3, we focus on the prototype of spatial construction and some of its nearest relatives whereas less typical and/or controversial cases are largely ignored. The two case-studies in section 4 however, account also for a number of cases of zero-marking of spatial relations which do not fulfill all of the criteria set up for the prototype.

60

2. Foundations

The prototype is identical to the vast majority of the examples which are provided in the extant descriptive linguistic literature – and constitutes the type of construction that is easiest to identify in text anthologies, etc. By concentrating on the prototype we do not assume that other instances are irrelevant for understanding the nature of zeromarking and the hierarchical order of spatial relations. However, it is appealing methodologically to first develop hypotheses on the safe side. What does the prototype look like? We give precedence to examples which contain a lexically expressed location and an overt locatum which may be either a lexical NP or a pronoun or a pronominal affix on a verb. The construction is predicative, meaning it has a verbal nucleus (which may be realized as Ø in languages which allow for zerocopulas). Furthermore, we preferably include those examples in our discussion which show marking of neither configuration nor mode although frequently our data survey also takes account of cases in which only mode is zero-marked. The data we have studied do not allow for us to homogenize the examples. Thus, the examples cover very different sentence types, modes, polarities, etc. Tenses, aspects, aktionsarten may vary considerably. Many of the sample sentences are simple monoclausal constructions whereas others are complex with an array of additional adverbials, adjuncts and arguments. Whether any of these factors has a bearing on the presence/absence of zeromarking is a highly interesting question which needs to be posed in a separate study. Prototypicality assumptions induce us to be somewhat picky about the types of morphological markers we consider competitors of zero-marking. Bound case markers and adpositions are our favorite candidates since they seem to be particularly wide-spread across the language of the world. We refrain from going into the issue of mode markers on verbs as they seem to require special attention and therefore cannot covered in this study. In connection to adpositions, however, serial verbs give us a tough nut to crack. According to Aikhenvald (2006: 47) “[b]asic motion verbs (‘come’, ‘go’ and ‘move’) are most frequently serialized.” Among (many) other things, these serial verbs indicate direction and orientation, meaning they serve the purpose to give expression to mode. At the same time, serial verbs grammaticalize to adpositions frequently (Aikhenvald 2006: 32). However, it must be determined for each language independently how many verbal properties a serial verb retains and how many adpositional features it has acquired in the process of grammaticalization. If the adpositional character of the serial verbs is well established, they are examples of overt marking of spatial relations. If however, the verby nature of the serial verbs is still strong, chances are that they represent cases of zero-marking unless they can be shown to incorporate mode semantically (cf. section 2.3.2 above). The problem is captured by the semantic differences of the serialized motion verbs of Ewe and Dumo exemplified in (E73)–(E74).

61

2.6. The final touches

(E73) Ewe (Ameka 2006: 138) ɖevía child:DEF locatum

métá NEG:crawl

yi xɔa me go room:DEF inside spatialV ground ‘The child did not crawl into the room.’

o NEG

(E74) Dumo (Ingram 2006: 206) peo daylight

ba and

tae-li SGSU.fall-3SGSU.go_to

deghe 3DUSU locatum

jihwa sago_basket

ghvi 3NSGSU.get

lung-ghya foot-3NSGSU.go

lih-lih yih 3NSGU.go_to~RED sago spatialV ground ‘When daybreak falls, we take the sago basket and walk to the sago bush.’

In the Ewe example (E73), the serial verb is glossed as a pure motion verb without any inherent specification as to mode whereas in (E74) the Dumo serial verb lih ‘to go (to a place)’ is treated as having an inherent Goal meaning component. Unfortunately, the semantics of the serial verbs are not always made as clear as in the Dumo case. There remain too many uncertainties as to the correct semantic interpretation of the serial verbs and the determination of its degree of grammaticalization. These uncertainties impel us not to take proper serial verb constructions into consideration in our data survey although they deserve being studied with attention in the future. To show that the prototype as such does not exhaust the domain of zero-marking of spatial relations, we exceptionally seize the opportunity to provide examples of case marking on adverbs and adpositions, since Creissels (2009a: 613) states that [i]n languages in which nouns have a case inflection including a subsystem of spatial cases, spatial adverbs and adpositions commonly have possibilities of variation similar to noun case inflection, but limited to spatial cases. Interesting as spatial cases on members of these word-classes might be, we do not study this phenomenon systematically in this study none the least because spatial cases on adverbs and adpositions more often than not implies that configuration is overtly expressed – namely by exactly these adverbs and adpositions. These observations mark the end of the introductory Part A. Further information as to the language sample and the methodology are given in the initial section of Part B. Several of the issues discussed above are taken up again in the evaluation and the conclusions of Part C.

Part B The empirical side of zero-marking of spatial relations

3. Cross-linguistic objets trouvés

In Part B, we discuss evidence of zero-marking of spatial relations from a crosslinguistic perspective. Our aim is to demonstrate that we are not dealing with a negligible phenomenon. To the contrary, our data-base is sufficiently large to suggest that languages in which at least some kind of zero-marking of spatial relations is practiced systematically can be found on every continent. Zero-marking of spatial relations is employed in languages of different genetic, areal and typological background. The absence of any major restrictions on its distribution over the world’s languages makes zero-marking of spatial relations a topic worth studying in some detail. In this study we will scratch the surface in order to instigate follow-up studies which scrutinize the issue at hand more thoroughly. To get a better feel of the situation, we draw the reader’s attention to Stassen’s (1997) monograph on the typology of intransitive predication. In this empirically very rich study, there are many examples of zero-marked spatial relation scattered over the chapters over the book. Stassen’s aim is not to determine the differential complexity of the morphosyntax of spatial expressions. However, as a by-product of his search for examples of intransitive predicates, he also assembles a sizable number of cases in which static verbs or motion verbs combine with an NP in the function of ground. This NP is frequently marked either by case-morphology and/or by adpositions. Beside these instances of overt marking of spatial relations, there are also numerous cases of zeromarking. In Table 16, we survey Stassen’s evidence of zero-marked spatial relations (for those languages which are not discussed further in our own cross-linguistic survey). Language Asmat Bambara Biloxi Falor Fongbe Gola Gurenne Kenuz Nubian Lao

Affiliation Central Papuan Mande Siouan West-Atlantic Kwa West-Atlantic Gur East Sudanic Kam-Tai

Place C T C

Goal

C T C T C C

Page 314 58 250 471 134 471 477 517 425

66

3. Crosslinguistic objets trouvés

Language Maranungku Nkore-Kiga Peñoles Mixtec Tiwi Usan Vietnamese

Affiliation Daly Bantoid Mixtecan Tiwi Madang Mon-Khmer

Place

Goal C T C

T C C

Page 414 483 531 144 49 134

Table 16: Evidence of zero-marked spatial relations in Stassen (1997) There are no examples of zero-marked Sources. Cells which host the letters T (= toponym) and/or C (= common noun) indicate that a given spatial relation is zero-marked. The variables T and C refer to the word-class to which the location-NP belongs. The rightmost column identifies the page in Stassen (1997) on which an example of the phenomenon can be found. The languages come in alphabetical order. There are 15 languages from the Americas, Africa, Asia and Australia seven of which attest to zeromarked Place whereas eight give evidence of zero-marked Goal. No language displays zero-marking with both Place and Goal. Five of the examples involve toponyms as location-NP. Twice as many cases involve common nouns in the same function. Since Stassen does not study spatial relations systematically, the picture that emerges from Table 16 cannot be taken to represent the last word on the issue at hand. What it is indicative of nevertheless is (a) (b) (c) (d)

the wide geographical distribution of zero-marking of spatial relations, the almost complete lack of examples of Source in connection to zero-marking, the competition of Place and Goal for the position of unmarked category, the relative importance of toponyms as attractors of zero-marking with either of the two competing spatial relations.

These are the points our subsequent survey and the case studies are meant to clarify by way of adducing as many pieces of evidence as possible. Part B is divided into two sections. Section 3 is dedicated to a cross-linguistic survey of zero-marking of spatial relations whereas section 4 comprises two corpus studies. Information about these case-studies is given in the introductory part of section 4. In section 3, we survey instances of zero-marking of spatial relations in a convenience sample of languages from all over the globe. The languages for which we have ascertained or at least pondered the idea of the existence of zero-marking of spatial relations number altogether 116. This group of languages includes those which have been shown to attest to zero-marking of spatial relations in Part A already, all languages for which we adduce evidence throughout this section (excluding those of Table 16 above) and the languages of the two case-studies in section 4. A full list of these languages, their genetic affiliation and properties is given in the Appendix which includes maps showing the geographical distribution of the languages that attest the phenomenon under review. Our sample does not obey any of the usual prerequisites of balanced samples. This

3. Crosslinguistic objets trouvés

67

means that we purposefully allow for genetic, areal and/or typological biases to get access to as many cases of zero-marking of spatial relations as possible. Nevertheless, the recurrence of the phenomenon in several language families in different parts of the world suffices to prove that the convenience sample is a good starting point for a study that is supposed to prepare the grounds for a new research program. To this end we have combed the extant descriptive linguistic sources for evidence of zero-marking of spatial relations. It is exactly this material evidence (and not its absence) which we are going to discuss in the subsequent sections. Therefore, the methodology for the survey is purely qualitative since we do not make any statements about the statistics of zeromarking (e.g. by determining how many languages attest the phenomenon as opposed to the number of those which give no evidence thereof). Our intention is to take account of as many cases of zero-marking of spatial relations as possible. This goal notwithstanding, no attempt is made at exhausting the subject empirically. The data stem overwhelmingly from languages for which descriptive grammars happened to be at hand at the time of our writing this study. In the absence of properly academic or normative grammars of a given language, we occasionally resort to pedagogical material or grammatically unanalyzed text corpora to complement our inventory of zero-marking of spatial relations. We acknowledge that the mixed character of our data-base is far from being the optimal foundation for far-reaching hypotheses. However, it provides a sufficiently large input for further discussion of the issue at hand. For the analysis of the data we mostly rely on our sources although we reserve ourselves the right to replace their interpretation with a home-made alternative of ours wherever we think it necessary. Owing to the character of most of the descriptive grammars and sundry source texts, the data are presented and analyzed in a relatively traditional structuralist way. We are aware of the problems that arise from comparing data drawn from corpus studies of standardized literary languages and cross-linguistic data which are largely taken from the oral register of less rigorously standardized languages. Information on zero-marking of spatial relations is scattered about the descriptive linguistic literature. As yet, nobody has taken stock of these pieces of evidence – and this means that hitherto it has obviously been too difficult to develop an empirically well-founded theory of zero-marking of spatial relations. In the absence of a sufficiently large data-base, it is very likely that the phenomenon appeared too spurious to deserve any special attention. To prove ideas of this kind wrong, we venture out into the jungle of data. For the present purpose, we will often analyze what is called nominative/absolutive case as the absence of case (cf. de Hoop & Malchukov 2008, Creissels 2009b, Lestrade 2010). Wherever NPs in their nominative (or absolutive) form serve as spatial complements or adjuncts, we consider these cases instances of zero-marking. As to the practical aspects of the organization of this section, we start on our IndoEuropean door-step, in a manner of speaking. From there we move on to Africa, Asia, Oceania, the Americas and close this chapter with a discussion of evidence drawn from Pidgin and Creole languages, on the one hand, and constructed languages, on the other.

68

3. Crosslinguistic objets trouvés

The organization of this section reflects a combination of genetic and geographic principles. As to the controversial genetic affiliation of individual languages or a given (sub-)phylum, we do not claim to contribute anything to solving the problems. Our decisions are guided by purely practical considerations and, thus, are open to revision. The vast majority of the languages we look at are still alive in the 21st century. They have been documented recently (i.e. between 1900 and 2013). Additionally, there are also occasional data from extinct languages and/or earlier stages of a given language. However, retrospection is not prominent in our survey. Unless noted otherwise, the Source function is marked overtly in the languages of our sample and hence we will mostly be concerned with the contrast of expressions for Goal and Place function. Every once in a while, examples of Source relations are included in the discussion in order to show that the postulated differences of Place/Goal on the one hand and Source on the other are indeed attested. As stated in the previous section, the cross-linguistic survey takes stock mostly of those cases which correspond closely to the prototypical construction which involves zero-marking of spatial relations. Prototypicality however does not automatically imply that all of the cases are uncontroversial. To make available as many data as possible for future discussions, we allow on board a number of less clear testimonies such as the very first case to be presented in section 3.1.1 and those of Moroccan Arabic and other Afro-Asiatic languages in sections 3.2.1.2.2–3. The situation in Khmer (section 3.9.5) and Hmong (section 3.9.6) is not absolutely in line with the prototype as sketched in section 2.6 above. Similarly, some of the Pidgin and Creole cases discussed in the sub-sections of section 3.14 might be considered problematic (especially those which are based on evidence from proper Pidgin languages). Nevertheless we are convinced that these cases and others too are either true representatives of the phenomenon at hand or need to be taken account of to better understand the nature of zeromarking of spatial relations. To prove that our conviction corresponds to the facts, a much richer empirical basis is required and thus the clarification of these and similar issues has to be postponed until the data-base has become sufficiently large.

3.1. Indo-European 3.1.1. Hittite and sundry old Indo-European languages We start our survey way back in the distant past – and with an atypical case at that. Creissels (2009a: 613) mentions that [‘e]ndingless locatives’ (which in at least some cases can be analysed as bare nouns stems in locative function) are well-attested in some ancient IndoEuropean languages, but the nouns that have this particularity do not seem to

3.1. Indo-European

69

constitute a natural semantic class […], and the very notion of ‘endingless locatives’ in Indo-European is not entirely clear. In the following paragraphs, we show that Creissels’s reservations against taken the zero-marking of spatial relations in ancient Indo-European languages for granted is well-founded. Neu (1980: 11–12) recapitulates the venerated discussion among Indo-Europeanists about the so-called casus indefinitus of Proto-Indo-European which is believed to be the predecessor of the “endungslose Lokativ” which is attested in a variety of IndoEuropean languages of antiquity. The concept of a locative without ending also implies that this case was insensitive to the distinction of Place and Goal. As far as Brugmann’s (1911: 174–179) documentation of the endingless locative goes, it seems that zeromarking of spatial relations was not integrated into the declension of nouns in the individual Indo-European languages of the past. The instances Brugmann identifies are adverbialized forms which have detached themselves from the regular case paradigm of the nouns from which they derive. We can consider these adverbializations fossilized instances of the casus indefinitus which underwent lexicalization. Often, these new adverbs are not used for proper spatial meanings but tend to reflect temporal meanings like Sanskrit parut ‘(in the) last year’ (in lieu of regular parut-i with the locative exponent -i), Vedic ášman ‘in the winter’ (in lieu of regular ášman-i), etc. Accordingly, Neu’s (1980) study of the phenomenon in Hittite comprises 18 types including endingless locatives with pronouns. For several of the candidates, Neu shows that it is unlikely that they represent zero-marked locatives. The regular locative and directional (allative) of Hittite nouns end in -i and -a, respectively. Bona fide examples of zero-marking include tagan ‘down(wards); below’ (related to tekan ‘earth; ground’), siṷat ‘on a given day’ related to siṷatt- ‘day’), lukat ‘(on the) next morning’ (related to lukkatt- ‘morning[-light]’), ser ‘up; above’, karu ‘formerly, erstwhile, already’. Neu (1980: 55) concludes that with the endingless locative, Hittite continues a very archaic morphological pattern of the postulated Indo-European proto-language. There is evidence that (in contrast to the other Indo-European languages which attest these locative forms) Hittite might have attempted to revive this old pattern to a moderate extent. (our translation)1 The moderate form of productivity which the late expert of Old Anatolian languages assumes for the endingless pattern in later stages of Hittite should not be mistaken as proof of the perseverance of zero-marking of spatial relations in this language. In point 1

“[m]it dem endungslosen ‘Lokativ‘ setzt des Hethitische ein sehr altertümliches Bildungsmuster der zu postulierenden [indogermanischen] Grundsprache fort. Es gibt gar Anzeichen, als habe das Hethitische (anders als die übrigen [indogermanischen] Sprachen, die solche Lokativformen kennen) in bescheidenem Umfang dieses alte Muster neu zu beleben versucht.”

70

3. Crosslinguistic objets trouvés

of fact, what is zero-marked is not the spatial relation as such but the derivation relation of the adverb and its nominal or pronominal basis. Put differently, the Hittite data testify to a morphologically interesting process which derives adverbs from stems of (pro)nouns without overt morphological markers. We are dealing with a word-formation strategy and not with case-marking. This negative judgment applies to all instances of the endingless locative reported for the ancient Indo-European languages. The casus indefinitus is interesting for the purpose of this book nevertheless, because one may take it as evidence of zero-marking of spatial relations in the proto-language. Given that this interpretation is correct, Goal-Place syncretism applies whereas there would be a distinct expression for Source relations. This is a pattern which is in accordance with the cross-linguistic tendencies observed in section 2.3 above and in dedicated work like that of Creissels (2009a: 615). There is a caveat however. The morphosyntactic properties of Proto-Indo-European must remain hypothetical since they can only be reconstructed on the basis of the attested traits of the ancient Indo-European languages. Since the evidence with these is inconclusive and at times even doubtful, we dare not postulate zero-marking of spatial relations for Proto-Indo-European without adding the proviso that pending philological corroboration this interpretation is provisional.

3.1.2. (Insular) North Germanic In contrast to the previous historical flashback, the facts are absolutely clear and indisputable in the modern members of the North Germanic phylum. As we have seen in section 1.1 above, within the Germanic phylum colloquial German and Kiezdeutsch are two varieties for which can be shown to use zero-marking for the expression of spatial relations. This makes it plausible to search for evidence of the phenomenon in other Germanic languages as well. The two insular North-Germanic languages Faroese and Icelandic, for instance, attest to zero-marking of spatial relations although in a way different from those covered by the prototype. 3.1.2.1. Icelandic Faroese and Icelandic spatial adverbs come in triplets which distinguish allative (= Goal), locative (= Place) and ablative (= Source) forms. Place and Source are encoded overtly, whereas the function of Goal lacks any specific marker. For Icelandic, typical triplets are heim ‘home(wards)’ – heima ‘at home’ – heiman ‘from home’ and fram ‘forward’ – frammi ‘in front’ – framman ‘from the front’ (Kress 1982: 97–98), i.e., Place and Source are marked by dedicated suffixes (sometimes in combination with the preposition fyrir ‘for’). The allative forms are morphologically simple. Example (E75) from an Icelandic children’s book is instructive insofar as it contains NPs for Goal and

3.1. Indo-European

71

Place alongside the spatial adverbs inn ‘into’ and inni ‘in(side)’. The case government of the preposition í ‘in’ is accusative with Goal-NPs and dative with Place-NPs. (E75) Icelandic – [Oddi 39] dag day

einn one

þegar when

þeir they locatum

komu come.PRET:3PL spatialV

inn into mode/config

í in config

hádegismatinn lunch.ACC:DEF ground+mode

sátu mamma og pabbi inni í stofu sit.PRET:3PL mother and father in(side) in sitting_room:DAT spatialV locatum mode/config config ground+mode ‘One day when they came in for lunch, mother and father were sitting in the sitting-room…’

The morphologically more complex adverb encodes the locative and thus is employed in combination with expressions of Place whereas the basic form of the adverb hosts no marker of the allative. It is used in combination with PPs which have a complement NP that represents the Goal. It is clear that the stem of the adverb inn- serves the purpose of encoding configuration. However, the additional morphological material is responsible for distinguishing the categories of mode. Since the Goal mode is zero-marked, the expression of configuration suffices to represent both configuration and mode. The situation is very much the same in Icelandic’s sister language Faroese. 3.1.2.2. Faroese Table 17 surveys the full array of triplets attested in Faroese (Thráinsson et al. 2004: 186).2 The form for Goal function morphologically resembles the nominative singular of strong masculine nouns whereas the locative is marked like the dative of the same class and the ablative is syncretistic with the accusative. Basic meaning back East home inside down North South up out West

Goal/allative aftur eystur heim inn niður norður suður upp út vestur

Place afturi eysturi heimi inni niðri norðuri suðuri uppi úti vesturi

Source/ablative aftan eystan heiman innan niðan norðan suðan oman uttan vestan

Table 17: Triplets of spatial adverbs in Faroese

2

For an only remotely similar phenomenon in Turkish, cf. Schroeder (2001).

72

3. Crosslinguistic objets trouvés

Examples (E76)–(E78) are meant to illustrate how the members of a triplet behave syntactically. (E76) Faroese – [HP I Faroese, 174] Hann he

stappaði stuff:PRET:3SG spatialV

ta

inn into mode/config

aftur í hillina behind in shelf:DET:ACC mode/config config ground+mode ‘He stuffed the screaming book deep into the shelf…’

DET

skríggjandi screaming locatum

bókina book:DET:ACC

(E77) Faroese – [HP I Faroese, 68] Pápi mín father my locatum

er be.3SG spatialV

beint directly

har here

inni í in in mode/config config ‘My father is next door in the book-shop…’

bókabúðini book_shop:DET:DAT ground

(E78) Faroese – [HP I Faroese, 171] hann he báðum both

gekk go.PRET

sum like

ein a

froskur frog

við with

hondunum innan fyri troyggjuna hand:DAT.PL:DEF from_inside for sweater:DET.ACC locatum mode+config ground ‘He walked like a frog with both hands (from) inside the sweater.’

On the basis of the above data, it is possible to speak of a recurrent paradigmatic pattern in Faroese. In these triplets, the Goal relation remains unmarked systematically whereas the forms of the locative and the ablative always involve additional morphological material.3 3.1.2.3. Mainland Scandinavian The distinctions which are observed systematically in Icelandic and Faroese spatial adverbs are also attested in continental North-Germanic although not necessarily to the same extent. In Swedish, for instance, triplets of the above kind are found in sizable numbers an example being upp ‘up(wards)’ – uppe ‘above’ – uppifrån ‘from above’ (Teleman et al. 1999: 672–677). In these triplets, the allative is normally the morphologically simplest form, i.e. it gives testimony of zero-marking. However, there are 3

For items like vestur ‘to the west’, the allative form serves as stem for the formation of the locative vestur-i whereas the ablative vest-an is formed on the basis of a stem vest-.

3.1. Indo-European

73

alternative adverb-preposition compounds such as uppåt ‘upwards’ which render the formal expression of the allative as complex as that of the two other categories. Mutatis mutandis examples (E79)–(E80) from Swedish can be taken to be representative of the contemporary situations in Norwegian and Danish as well. (E79) Swedish (Teleman et al. 1999: 675) han he locatum

gick go.PRET spatialV

upp up

till tredje to third mode/config ‘He went upstairs to the third floor.’

planet floor ground

(E80) Swedish (Teleman et al. 1999: 673) han he locatum

er be.PRES spatialV

uppe above

på tredje on third mode/config ‘He is upstairs to the third floor.’

planet floor ground

The adverbial stem upp- is again responsible for encoding configuration whereas the additional morphological elements specify mode. Since Goal is zero-marked the expression of configuration co-encodes the allative automatically. Note that for Source, there are morphological innovations among which we find extra-complex formations such as uppifrån ‘from above’. The examples from North-Germanic are instructive also in another sense. They show that, to understand the workings of zero-marking of spatial relations, it is not sufficient to take account of alternations within the NP/PP domain. Spatial relations are also at issue in other word-classes – for instance, adverbs.

3.1.3. Aromunian Besides French (cf. section 4.1 below), Aromunian is the only other member of the Romance phylum to display a strong tendency towards zero-marking of spatial relations. However, French and its distant relative Aromunian differ as to the scope of zeromarking. In the Aromunian variety spoken at Kruševo in Macedonia, zero-marking is triggered by toponyms, more precisely by names of cities, cf. (E81)–(E82). (E81) Kruševo Aromunian (Gołąb 1984: 208) dúku go:1SG

Mi REFL.1SG

locatum

spatialV ‘I am going to Bitola.’

Bítule Bitola ground

74

3. Crosslinguistic objets trouvés

(E82) Kruševo Aromunian (Gołąb 1984: 144–145) tuc all

a and

bənə́m live:1PL

frásl’i brother:PL locatum

š and spatialV

lukrə́m work:1PL

Krúšuva Kruševo ground

sórme͡ a bəneáʒə š lukreaʒə Skópi̯ a sister:POSS.1SG live:3SG and work:3SG Skopje locatum spatialV ground ‘We brothers all live and work in Kruševo and my sister lives and works in Skopje.’

The prepositionless construction is not compulsory but it is the preferred option for both Place and Goal relations, nevertheless. Alternatively, names of cities as grounds can be part of a PP headed by the prepositions ən ‘in’ or tu ‘in’ as in (E83)–(E84). (E83) Kruševo Aromunian (Gołąb 1984: 208) dúku go:1SG spatialV

Mi REFL.1SG

locatum

ən in mode ‘I am going to Belgrad.’

Biligrádu Belgrad ground

(E84) Kruševo Aromunian (Gołąb 1984: 164) kə that

tu in config

un a

múnte máre áre mountain big EXI ground spatialV ‘…that there is water on a big mountain…’

ápə water locatum

Interestingly, in his bulky grammar of Aromunian, Capidan (1932: 505–507 and 554– 555) does not mention the possibility of dropping spatial prepositions in combination with toponyms. On the basis of the description of the Kruševo variety of Aromunian by Gołąb (1984), which lacks a separate section dedicated to prepositions4, it is evident that zero-marking of spatial relations is possible only with toponyms in Place or Goal function. Toponyms in the function of Source are preceded obligatorily by the preposition di(n) ‘of; from’ as in (E85). (E85) Kruševo Aromunian (Gołąb 1984: 144) pəpə́nl’i améi̯ sə́ntu viníc di grandparent:PL POSS.1SG be.3PL come:PTCPL:PL from locatum spatialV mode ‘My grandparents have come from Gramosta.’

ɣrámosta Gramosta ground

The Aromunian case is in line with a cross-linguistically common pattern according to which Place and Goal are more likely than Source to allow for zero-marking. Moreover, 4

Our examples are drawn from the sizable text section of the descriptive grammar of Kruševo Aromunian.

3.1. Indo-European

75

zero-marking applies only with a sub-class of locations, namely toponyms (more specifically with names of cities since names of countries seem to require the use of a preposition). A third cross-linguistically wide-spread trait is that the Aromunian zeroes instantiate the syntagmatic variety because their use is optional. At the same time, they are paradigmatic in the sense that they contrast with overt prepositions used for Source relations.

3.1.4. Southern Macedonian The Aromunian case described in the previous section has a parallel in southern varieties of Macedonian spoken in the immediate vicinity of Kruševo Aromunian.5 Standard Macedonian does not tolerate zero-marking of spatial relations. Locations which are represented by either common nouns or toponyms are obligatorily part of a PP no matter whether the verb is dynamic or static (Foulon-Hristova 1998: 236–241)6, cf. (E86)– (E87). (E86) Standard Macedonian [HP I Macedonian, 60] i and

sevo all:NT

ova this:NT locatum

može can

da

se

SUBORD

REFL

kupi buy spatialV ‘And all this can be bought in London?’

vo in config

London London ground

(E87) Standard Macedonian [HP I Macedonian, 78] zošto why

odite go:2PL spatialV

vo in locatum config ‘Why do you got to London?’

London London ground

In the south-western varieties of Macedonian however, toponyms in Place and Goal are normally connected to the verb without intervening preposition, cf. (E88). The relator of Source is always expressed overtly. (E88) Southern Macedonian (Koneski 1967: 127) bev be.PAST:1SG spatialV locatum

5

6

Bitola odam Bitola go:1SG ground spatialV locatum ‘I was in Bitola, I am going to Prilep.’

Prilep Prilep ground

We are grateful to Norbert Boretzky and Eleni Buzarovska (p.c.) for drawing our attention to the Macedonian case. Foulon-Hristova (1998: 241) claims that “[d]ans certains cas, la jonction entre deux termes se réalise sans intervention d’une préposition, normalement ou facultativement.” The only example of a spatial situation with optional preposition in her short list is otide (vo) tuǵina ‘he went abroad’. It is unclear to what extent the Macedonian standard makes use of this pattern.

76

3. Crosslinguistic objets trouvés

In accordance with Koneski (1967) and Markoviḱ (2007), Victor Friedman (p.c.) assumes that this dialectal Macedonian feature is the effect of contact with Aromunian7 whose properties have been reported on in the previous section. The Southern Macedonian case thus resembles that of Macedonian Aromunian very closely. In both languages, toponyms behave differently from other nouns which serve as locations. Place and Goal behave differently from Source. Zero-marking is a characteristic of toponyms in Place or Goal function. On the basis of the data to which we have access, it cannot be ascertained conclusively whether or not we are dealing with syntagmatic zeroes as in the Aromunian case. Markoviḱ (2007) who focuses on the Macedonian and Aromunian varieties spoken in the Lake Ohrid area provides a series of examples of toponyms in Place and Goal relations all of which are zero-marked. This consistency speaks in favor of zero-marking as the default solution.

3.1.5. Armenian Within the Indo-European language family, Armenian constitutes an internal isolate although nowadays two distinct varieties are recognized, namely West Armenian spoken mainly in the diaspora and East Armenian spoken in independent Armenia. The many structural similarities of both Armenian varieties notwithstanding, there are some (probably only minor) differences which show up in the realm of spatial language. This gives us reason to look at both East Armenian and West Armenian separately. 3.1.5.1. East Armenian For East Armenian, Dum-Tragut (2009: 82) puts forward the category of the local nominative8 the functions of which are described as follows: The local nominative it [sic!] denotes the place to which an action is directed and usually answers the question ‘where (to)?’. The local nominative is generally used with verbs of motion. From this description, we conclude that we are dealing with a means of expression of the Goal function. The word-form in which the ground occurs is called nominative because it lacks any overt morphological marking. For obvious reasons, we do not gloss the nomina7

8

Norbert Boretzky (p.c.) reports that, in the immediate neighborhood of Macedonian and Aromunian, a variety of Bugurdži – a branch or Romani – displays zero-marking at least of toponymic Goals as in jek gelo Skopje {one} {go:PAST.3SG} {Skopje} ‘one [of three brothers] went to Skopje’. This constellation of facts is suggestive of an areal feature which, however, does not count as a fully-blown Balkanism. For the discussion of a case of emergent zero-marking in a regional variety of Greek, cf. Part C. Creissels (2009a: 622) assumes that “the unmarked allative can be explained by the loss of the nominative vs. accusative distinction in a system originally characterized by an allative-accusative syncretism.”

3.1. Indo-European

77

tive specifically, cf. (E89). Single underlining is used to identify markers of definiteness in the object-language and/or the translations if these markers associate with the ground. We choose highlight issues of definiteness because these will be demonstrated to be important inter alia for our case-study of French in section 4.1 below. (E89) East Armenian (Dum-Tragut 2009: 82) Gyułacinerĕ peasant:PL.NOM:DET locatum

gnac‘in go:AOR.3PL spatialV ‘The peasants went to the field.’

dašt field ground

However, zero-marking of spatial relations is not restricted to the Goal function in East Armenian. There is a morphological locative used to encode static spatial relations of inclusion as in (E90). (E90) East Armenian (Dum-Tragut 2009: 101) Aramĕ Aram:DEF locatum

ē lsaran be.3SG auditorium spatialV ground ‘Aram is learning in the auditorium.’

parapum learn:PTCPL

-um -LOC mode+config

The morphological locative is typical of careful speech and the written register, whereas [i]n colloquial Armenian, this strict meaning of the locative is sometimes replaced by the bare nominative, particularly with place names and in cooccurrence with the copular verb (Dum-Tragut 2009: 102). This means that in everyday spoken East Armenian, not only Goals are zero-marked. Zero-marking also applies to Place as shown in (E91). (E91) East Armenian (Dum-Tragut 2009: 102) Aramĕ Aram:DET locatum

Erewan Eriwan ground ‘Aram is in Eriwan.’

ē be.3SG spatialV

In addition, the morphological locative competes with the morphological dative (with or without the postposition meǰ ‘in’) if the location cannot be understood as a proper container, cf. (E92)–(E93). (E92) East Armenian (Dum-Tragut 2009: 87) nrank‘ they locatum

paṙkac lie:PTCPL

ēin geti ap‘ be.PAST:3PL river:DAT bank spatialV ground ‘They were resting on the river’s bank.’

-i -DAT mode

-n -DEF

78

3. Crosslinguistic objets trouvés

(E93) East Armenian (Dum-Tragut 2009: 101) matitĕ pencil:DEF locatum

grk‘ -i meǰ book -DAT in ground mode The pencil is in the book [i.e. in between the pages].’

ē be.3SG spatialV

The morphological locative is excluded from human nouns (Dum-Tragut 2009: 100; Creissels 2009a: 612). The replacement of the morphological locative with the dative is considered a trait of the colloquial style (Dum-Tragut 2009: 101). Source relations are always overtly encoded by the morphological ablative as in example (E94). (E94) East Armenian (Dum-Tragut 2009: 97) Aramĕ Aram:DET locatum

dursekav Moskva get_out:AOR.3SG Moscow spatialV ground ‘Aram got out of Moscow.’

-yic‘ -ABL mode

Place and Goal behave differently in the sense that zero-marking of Place is more tightly connected to place names whereas the Goal may lack an overt marker also with common nouns which serve as location. The zero with Goals is a clear case of a paradigmatic zero. The prominent role played by place names in the realm of zero-marking of spatial relations is a recurrent theme in the remainder of this cross-linguistic overview. 3.1.5.2. West Armenian In a footnote, Dum-Tragut (2009: 101, footnote 100) remarks that [i]n Western Armenian a locative case does not exist, thus all utterances ‘in a place’ are expressed with this postpositional phrase. One might also regard the increasing use of postpositional phrases in [Modern Eastern Armenian] as a gradually increasing influence of Western Armenian. In this quote, the author refers to the postpositional phrase [Ndative + meǰ]PP exemplified in (E93) above. If Dum-Tragut’s observation were correct, non-toponymic Place would be the spatial relation which requires the most complex encoding strategy in West Armenian because – as in East Armenian – Goals are generally unmarked in the Western variety, cf. (E95)–(E96).9 (E95) West Armenian (Sakayan 2000: 209) yes I locatum

9

Kermanya Germany ground

guzem go:1SG spatialV locatum ‘I want to go to Germany…’

yertal want

In both varieties of Armenian, Source NPs are inflected for ablative. Since we have illustrated the construction already for East Armenian, we skip this issue for West Armenian.

3.1. Indo-European

79

(E96) West Armenian (Sakayan 2000: 208) yertank go:HORT.1PL spatialV locatum

nak͜ h first

sinema hedo cinema then ground ‘Let’s go to the movies first, then to a café.’

sərd͜ jaran café ground

Goals are represented by bare NPs independent of their membership in a specific subclass of nouns. Common nouns and place names behave in identical fashion (Sakayan 2000: 75). It is irrelevant for our purpose that Sakayan (2000: 74–75) distinguishes a nominative function from an accusative function of the zero-marked form. As to the privileged position of Goals, the two varieties of Armenian resemble each other closely. Since Dum-Tragut (2009: 101) is correct in stating that there is no morphological locative in West Armenian, the question arises as to how Place relations are generally expressed. Sakayan (2000: 96) shows that the expression of Place relations forms part of the functional domain of the dative as in (E97) (cf. [E92] above). (E97) West Armenian (Sakayan 2000: 96) gədur roof ground

-i -DAT mode(+config)

təṙc͜ hun bird locatum ‘There was a bird on the roof.’ -n -DEF

mə a

gar be.PAST spatialV

It is likely that the morphological dative is employed if the location is not a proper container, i.e. the same restriction as in East Armenian seems to hold and thus configuration – ON vs. IN – might play a role in the choice of construction. Besides the zero-marked Goals, there are, however, also instances of zero-marked Places in West Armenian as in (E98). (E98) West Armenian (Sakayan 2000: 208) kag͜ hakə city:DEF ground

bidi

mənas stay:2SG.SUB spatialV ‘Are you staying in town?’

IRR

In contrast to colloquial East Armenian which allows for zero-marked Place only with toponyms (cf. [E91] above), this possibility does not seem to be as strictly constrained in West Armenian since the evidence we have found comprises common nouns which, however, represent typical Place concepts such as TOWN, VILLAGE, BEACH, etc. The postposition mec͜ ͜h ‘in’ is included in Sakayan’s (2000: 140) list of postpositions but neither is its use explained nor are any sentential examples given in the source. Whether or not the absence of any proof of the use of this spatial postposition has anything to do with stylistic criteria as mentioned in the previous sub-section for its cognate in East Armenian cannot be ascertained on the basis of the empirical data contained in Sakayan (2000).

80

3. Crosslinguistic objets trouvés

We conclude this section with putting forward the hypothesis that in West Armenian, zero-marking of spatial relations has a wider domain than it has in the eastern variety. In contrast to the latter, West Armenian does not seem to limit zero-marking of Place to proper toponyms. In the absence of tangible proof, we can only speculate that the loss of the inflectional locative from the noun declension has contributed to the growth of the domain of zero-marking of Places in West Armenian. It remains to be determined in what way bare Place NPs interact paradigmatically with postpositional phrases with mec͜ ͜h ‘in’. This is a task, however, which cannot be fulfilled by this study.

3.1.6. Indo-Iranian In this section, we touch upon a relatively long sequence of pieces of evidence which prove that zero-marking of spatial relations is a commonality throughout much of the Indo-Iranian phylum. However, owing to the nature of some of the sources we had access to the presentation is sketchy for several of the languages involved. In several cases, the evidence consists of a single sentence drawn from a linguistic treatise dedicated to topic outside the realm of the grammar of space. We present the data in geographical order proceeding from west to east. For a general survey of the synchrony and diachrony of case-marking in Iranian, we refer the reader to Stilo (2009). 3.1.6.1. Kurmancî and Zazakî In the immediate neighborhood of Armenian, several Neo-Iranian languages are spoken. There are structural differences between these varieties. Among these differences, the possibility of employing zero-marking of spatial relations can be observed. In the Kurdish language Kurmancî, for instance, the oblique case in -ê serves the purpose of marking the Goal complement of a motion verb as in (E99). This rule also applies to toponyms, cf. (E100). In spatial contexts, the oblique case is termed directional by our source. (E99) Kurmancî (Wurzel 1997: 44) Ew s/he locatum

dıçe go.3SG spatialV

mekteb school ground ‘S/he goes to (the) school.’

-ê -OBL mode

(E100) Kurmancî (Wurzel 1997: 44) xwendevan student:PL:DEF locatum

dıçın Amed go:3PL Diyarbakır spatialV ground ‘The students are going to Diyarbakır.’

-ê -OBL mode

In contrast, Zazakî uses the bare noun in structurally similar contexts – including toponyms, cf. (E101)–(E102).

3.1. Indo-European

81

(E101) Zazakî (Selcan 1998: 638) Televey pupil:PL locatum

herey late

sonê go:3PL spatialV ‘The pupils are going to (the) school late.’

mektev school ground

(E102) Zazakî (Selcan 1998: 642) Ça why

tu siya you go.PRET locatum spatialV ‘Why have you gone to Germany?’

Almanya Germany ground

In both of the above varieties, static relations of Place are always marked overtly and the same applies to Source relations: •

•

Kurmancî: o Place: o Source: Zazakî: o Place: o Source:

lı Berlin-ê = [Prep Noblique]PP ‘in Berlin’, ji Berlin-ê = [Prep Noblique]PP ‘from Berlin’, dewe-de = [N Post]PP ‘in the village’, dewe-ra = [N Post]PP ‘from the village’.

In Kurmancî, all spatial relations are overtly marked. However, Goal receives the least complex expression. Goal is also special among the spatial relations of Zazakî. In this language however, Goals are not only expressed with the least effort in terms of phonological segments but they are zero-marked whereas Place and Source require phonologically realized postpositions. Zero-marking applies to toponyms as well as to common nouns which function as locations. 3.1.6.2. Gilaki As far as it is possible to generalize over the small number of relevant examples in our source, the situation in Gilaki (spoken in Iran) resembles that of Zazakî described in the previous sub-section. Examples (E103)–(E104) suggest that Goals are generally zeromarked no matter whether the location is represented by a common noun or a toponym. Places, on the other hand, are always overtly marked by a postposition even if their representative is a toponym. (E103) Gilaki (Bossong 1985: 40) mi my

dil heart

xaye want:3SG

zakana child:PL:ACC

usanim take:1SG

dəryå kənåra sea beach ground ‘I’d like to take my children to the beach.’

bišim SUB:go:1SG spatialV locatum

82

3. Crosslinguistic objets trouvés

(E104) Gilaki (Bossong 1985: 41) xayəm want:1SG

a this

ruzana day:PL:ACC

Raštə -miyan šimebiǰa Rasht -in 2PL:with ground mode ‘I want to pass these days with you in Rasht.’

buguzəranəm SUB:pass:1SG

spatialV

locatum

There was no example of a Source relation. 3.1.6.3. Sivandi The data available for Sivandi (spoken in Central Iran) are even scarcer than those of Gilaki. The two sentential examples in (E105)–(E106) can be interpreted as testifying to a syntagmatic zero with Goals. (E105) Sivandi (Bossong 1985: 50) terāzīā scales locatum

neyeš zimīn put.PAST:3SG ground spatialV ground ‘He put the scales on the ground.’

(E106) Sivandi (Bossong 1985: 50) āmeyne come.PAST:3PL

ārīs bride

dāmārā bebarune ba ābādi feyša bride-groom:ACC SUB:carry:3PL to flat:EZF self:3PL locatum spatialV mode ground ‘They came to carry the bride and the bridegroom to their apartment.’ o and

In (E105), no preposition is used to connect the Goal-NP to the verb whereas, in (E106), the preposition ba ‘to’ is employed as head of the PP whose complement is the Goal-NP. It is possible that zimīn ‘ground’ in (E105) is no longer fully referential so that it could be considered the translation equivalent of English down in combination with certain verbs. In this case it would be doubtful to accept (E105) as proof of zeromarking of spatial relations in Sivandi. Examples of Place and Source were not included in our reference text. 3.1.6.4. Sangesari We must be content with exactly a single example from Sangesari (Northwestern Iran). (E107) contains a toponym (the name of a city). (E107) Sangesari (Bossong 1985: 86) jækæ one

merkæini Moo̤ zderun næmâšun araq bəxord man.OBL:ERG Mâzanderân evening_prayer booze PRET:drink locatum ground spatialV ‘A man had drunk booze in Mâzanderân during evening prayer.’

boo̤ be.PRET

3.1. Indo-European

83

The Place-relation is not made explicit by any morphological means. We assume that this is generally possible with toponyms in this language. For his own non-spatial topic, Bossong (1985: 80–84) elaborates on the special character of Sangesari which let it stand out from all other Iranian languages. The atypical traits of this language include the remarkable development of the originally spatial preposition dar ‘in’ to the marker of the direct object (termed accusative). We understand Bossong’s line of argumentation to mean that Sangesari -de (