Simplicity and Typological Effects in the Emergence of New Englishes: The Noun Phrase in Singaporean and Kenyan English 9783110521801, 9783110516593

The book is based on a detailed corpus-based investigation of the structure of noun phrases (NPs) in Singaporean English

192 52 2MB

English Pages 360 [362] Year 2017

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Simplicity and Typological Effects in the Emergence of New Englishes: The Noun Phrase in Singaporean and Kenyan English
 9783110521801, 9783110516593

Table of contents :
Acknowledgements
Contents
List of Figures
List of Tables
List of Abbreviations
1. Introduction
Part A: New Englishes and the Structure of NPs
2. New Varieties of English
3. Modelling language change in New Englishes
4. Kenyan and Singaporean English
5. The English NP — structure and variation
Part B: NPs in Kenyan English and Singaporean English
6. Methodology, corpus handling and statistics
7. Studying NP modification in Singaporean English and Kenyan English
8. Conclusion
Bibliography
Appendix A
Appendix B

Citation preview

Thomas Brunner Simplicity and Typological Effects in the Emergence of New Englishes

Topics in English Linguistics

Editors Elizabeth Closs Traugott Bernd Kortmann

Volume 97

Thomas Brunner

Simplicity and Typological Effects in the Emergence of New Englishes The Noun Phrase in Singaporean and Kenyan English

ISBN 978-3-11-051659-3 e-ISBN (PDF) 978-3-11-052180-1 e-ISBN (EPUB) 978-3-11-051954-9 ISSN 1434-3452 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 on the Internet at http://dnb.dnb.de. © 2017 Walter de Gruyter GmbH, Berlin/Boston Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Acknowledgements This book is a revised version of my dissertation submitted at the University of Regensburg in 2015. My heartfelt thanks go to my supervisor Prof. Edgar W. Schneider for inspiring this project, for his constant, invaluable advice and for motivating and encouraging me throughout my work on the project. I am also grateful to my colleagues Dr Alexander Kautzsch, Dr Sarah Buschfeld and Dr Florian Schleburg, for their expertise and support. Thanks are due to Dr Christian Rapold, who checked my glosses of African languages, Dr Lisa Lim, who gave advice on Malay and Prof. Stefan Th. Gries, for help with statistical problems in the final phase of the project, and, of course, to Duncan Steele and Jawad A. Deo, for the wonderful LATEX editor which they have developed. I would also like to express my gratitude to Prof. Bernd Kortmann and Prof. Elizabeth Closs Traugott for including my book in the TiEL book series, and to Julie Miess and Olena Gainulina at De Gruyter for their help in producing the book. Many thanks to Catherine Stone, who meticulously proofread the manuscript and made numerous valuable suggestions. Above all, I would like to thank my friends and my family, for their unwavering support and for keeping my spirits up.

Contents Acknowledgements List of Figures

XI

List of Tables

XIV

List of Abbreviations 1 1.1 1.2 1.2.1 1.2.2 1.2.3 1.3 1.3.1 1.3.2 1.4

V

XIX

Introduction 1 The nature of New Englishes 1 Why NPs in Singaporean English and Kenyan English? 3 Analysing language contact on the basis of NPs: rationale and aims 3 The analysis of simplification on the basis of NPs: rationale and aims 4 The key factors of NP variation addressed 5 Theoretical backdrop of the study 6 The approach taken to New Englishes 6 The approach taken to language 7 The outline of the study 9

Part A: New Englishes and the Structure of NPs 2 2.1 2.1.1 2.1.2 2.1.3 2.1.4 2.1.5 2.2 2.2.1 2.2.2 2.3

New Varieties of English 13 The phenomenon of New Englishes 13 The definition of “New Englishes” 13 Models of New Englishes 14 An overview of linguistic innovations in New Englishes 18 An overview of explanatory devices for innovations in New Englishes 19 The field of SLA and the study of New Englishes 20 Language contact and simplification as major processes of change 21 Language contact in New Englishes 22 Evidence of simplification in New Englishes 26 Summary: New varieties of English 30

VIII

3 3.1 3.1.1 3.1.2 3.1.3 3.2 3.2.1 3.2.2 3.2.3 3.2.4 3.3

Contents

Modelling language change in New Englishes 33 Language contact and simplification in the field of SLA 33 L1 transfer in SLA 33 Simplification in SLA 41 Summary: Language contact and simplification in the field of SLA 45 46 An evolutionary model of language change for New Englishes Evolutionary approaches to language change 47 The key assumptions in Croft’s theory of language change 47 Problematising Croft’s (2000) evolutionary approach 51 Modelling linguistic processes in the emergence of New Englishes in the light of Croft’s (2000) theory of language change 54 Summary: Modelling language contact and simplification in New Englishes 56

4 4.1 4.1.1 4.1.2 4.2 4.2.1 4.2.2 4.3

Kenyan and Singaporean English 57 Singaporean English 57 Singaporean English as a variety 57 The local languages in Singapore 69 Kenyan English 83 Kenyan English as a variety 83 The local languages in Kenya 94 Singaporean English and Kenyan English in perspective

5 5.1 5.1.1 5.1.2 5.1.3 5.1.4 5.2 5.2.1 5.2.2

The English NP — structure and variation 108 The structure of the English NP 108 The model by Bache (2000) 108 Evaluating Bache’s (2000) model 110 An adapted model of the NP 112 Syntactic phenomena and problems in the NP 116 Variation in the English NP 119 Factors influencing the form and complexity of modification 119 Factors influencing the choice between pre- and postmodification 124 NP complexity 127 Problematising linguistic complexity 128 The major metrics of NP complexity 130 Discussion of the main metrics of syntactic complexity 135 How NP complexity will be operationalised 136 Summary: Delineating the variable context in NPs 144

5.3 5.3.1 5.3.2 5.3.3 5.3.4 5.4

106

Contents

5.4.1 5.4.2

IX

Methodological conclusions from the discussion of NP variation 144 The variable contexts in the NP 145

Part B: NPs in Kenyan English and Singaporean English 6 6.1 6.1.1 6.1.2 6.1.3 6.2 6.2.1 6.2.2

Methodology, corpus handling and statistics 151 General methodology: Corpus, software and statistics 151 The International Corpus of English 151 Software 154 Statistical methods 155 Corpus handling in the study of NP modification patterns 157 Choice of corpus sections 157 Corpus processing 158

7

Studying NP modification in Singaporean English and Kenyan English 165 Hypotheses on NP structures in Singaporean English and Kenyan English 165 Hypotheses related to language contact 165 Hypotheses related to simplification 168 Register-specific areas of interest 169 Testing hypotheses about language contact phenomena 170 Language contact in private dialogue () 170 Language contact in student essays () 205 Language contact in S1A and W1A — a comparison 245 Discussion and interpretation: Language contact in NPs in Singaporean English and Kenyan English 246 Testing hypotheses about simplification 248 The complexity of NPs in private dialogue () 249 The complexity of NPs in student essays () 266 Simplification in S1A and W1A — a comparison 287 Discussion and interpretation: Simplification in NPs in Singaporean English and Kenyan English 288 Summary and discussion: Language contact and simplification across varieties and registers 291

7.1 7.1.1 7.1.2 7.1.3 7.2 7.2.1 7.2.2 7.2.3 7.2.4 7.3 7.3.1 7.3.2 7.3.3 7.3.4 7.4

8 8.1

Conclusion 295 The outline of the study

295

X

Contents

8.2 8.2.1 8.2.2

The implications of the results for the study of New Englishes 297 Methodological implications 297 Implications with regard to the phenomenon of New Englishes 299

Bibliography

301

Appendix A

HCFA results

Appendix B B.1 B.1.1 B.1.2 B.2 B.2.1 B.2.2 B.3 B.3.1 B.3.2

321

Some overviews and additional analyses 329 An overview of the two datasets from S1A and W1A 329 The dataset from S1A 329 The dataset from W1A 330 Syntactic position and the choice between pre- and postmodification 331 Syntactic position and the choice between pre- and postmodification in S1A 331 Syntactic position and the choice between pre- and postmodification in W1A 334 Semantic class of the head and complexity levels of NPs 336 Semantic class of the head and complexity levels of NPs in S1A 337 Semantic class of the head and complexity levels of NPs in W1A 339

List of Figures Fig. 3.1 Fig. 3.2

The psychotypology hypothesis 40 An altered version of Croft’s (2000) model of language change

Fig. 4.1

L. Lim’s (2007: 4587) overview of the historical relevance of indigenous languages in Singapore 71

Fig. 5.1 Fig. 5.2 Fig. 5.3 Fig. 5.4 Fig. 5.5 Fig. 5.6

Berlage’s (2014: 36) analysis of the NP the books 130 Berlage’s (2014: 5) analysis of the NP a man of great honour Complexity hierarchy of premodifiers 137 Complexity hierarchy of postmodifiers 139 NP structures as schematic linguemes 145 NP modifiers as schematic linguemes 146

Fig. 6.1

Bar chart showing the interaction of modification type and variety

Fig. 7.1

Premodification vs. postmodification in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 171 The frequencies of modifier subtypes in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 173 The co-occurrence patterns of all pre- and all postmodifier types in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 177 The lengths of premodifiers in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 179 The lengths of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 183 The formal realisations of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 186 The formal realisations of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 187 The formal realisations of Postmod. 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 190 Complexity levels of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 192 Complexity levels of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 194 The complexity level of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 195 The complexity levels of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 196 The number of premodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 199 The number of postmodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 200 The positions of NPs embedded at level 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 201

Fig. 7.2 Fig. 7.3 Fig. 7.4 Fig. 7.5 Fig. 7.6 Fig. 7.7 Fig. 7.8 Fig. 7.9 Fig. 7.10 Fig. 7.11 Fig. 7.12 Fig. 7.13 Fig. 7.14 Fig. 7.15

54

130

156

XII

Fig. 7.16 Fig. 7.17 Fig. 7.18 Fig. 7.19 Fig. 7.20 Fig. 7.21 Fig. 7.22 Fig. 7.23 Fig. 7.24 Fig. 7.25 Fig. 7.26 Fig. 7.27 Fig. 7.28 Fig. 7.29 Fig. 7.30 Fig. 7.31 Fig. 7.32 Fig. 7.33 Fig. 7.34 Fig. 7.35 Fig. 7.36 Fig. 7.37 Fig. 7.38 Fig. 7.39 Fig. 7.40 Fig. 7.41

List of Figures

Premodification vs. postmodification in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 206 The frequencies of modifier subtypes in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 207 The co-occurrence patterns of all pre- and all postmodifier types in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 212 The lengths of the whole premodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 214 The lengths of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 217 The lengths of Mod. III in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 219 The lengths of the whole postmodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 222 The formal realisations of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 224 The formal realisations of Mod. III in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 226 The formal realisations of Postmod. 1 in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 229 The complexity levels of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 231 The complexity levels of Mod. III in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 234 The complexity levels of the whole premodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 235 The complexity levels of the whole postmodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 237 The number of premodifying slots occupied per NP in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 238 The number of postmodifying slots occupied per NP in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 240 The positions of NPs embedded at level 1 in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 241 The lengths of NPs in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 250 The lengths of whole NPs by syntactic position and variety in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 251 The complexity level of whole NPs in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 254 The correlation of NP complexity and syntactic position in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 256 The numbers of pre- and postmodifiers in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 259 Number of NPs embedded into other NPs per variety 261 The lengths of NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 267 The correlation of syntactic position and NP length in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 269 The complexity level of whole NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 273

List of Figures

Fig. 7.42 Fig. 7.43 Fig. 7.44 Fig. B.1 Fig. B.2 Fig. B.3 Fig. B.4

The correlation of NP complexity and syntactic position in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 276 The numbers of pre- and postmodifiers in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 279 Number of NPs embedded into other NPs per variety 281 Pre- vs. postmodification by variety and syntactic position in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 333 Pre- vs. postmodification by variety and syntactic position in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 334 NP complexity levels by semantic class of head noun in S1A 337 NP complexity levels by semantic class of head noun in W1A 339

XIII

List of Tables Table 1.1

Factors impacting on NP structure

Table 2.1

Varieties of English along Schneider’s (2007) evolutionary cycle.

Table 3.1 Table 3.2 Table 3.3

Error rates in relative clauses among learners of English 35 Logit values (head-final vs. head-initial compounds) 38 Mean values of syntactic complexity measures 43

Table 4.1 Table 4.2 Table 4.3

Population by ethnic group 64 “Languages literate in” according to the 2010 census in Singapore 65 The languages spoken most frequently at home according to the 2010 census of population in Singapore 66 Languages of Singapore according to the Ethnologue 70 NP structures in Bazaar Malay, Standard Malay and Hokkien 76 NP structures in Baba Malay, Old Baba Malay, Standard Malay and Hokkien 79 Language first learned during childhood and still used according to the 2006 Kenya National Adults Literacy Survey 87 Language spoken at home most of the time according to the 2006 Kenya National Adults Literacy Survey 87 Heine and Möhlig’s (1980: 61) overview of patterns of second-language knowledge in Kenya 88 Language use outside schools in Kenya 91 Kenyan languages by language phyla 96 Speaker numbers of Kenyan languages 96 Speaker numbers of Kenyan languages 97 Speaker numbers of Kenyan languages 98 NP structures of the Niger-Congo languages of Kenya 101 NP structures of the Nilo-Saharan languages of Kenya 103 NP structures in the Afro-Asiatic languages of Kenya 104 Key properties of Singaporean English and Kenyan English 106

Table 4.4 Table 4.5 Table 4.6 Table 4.7 Table 4.8 Table 4.9 Table 4.10 Table 4.11 Table 4.12 Table 4.12 Table 4.12 Table 4.13 Table 4.14 Table 4.15 Table 4.16 Table 5.1 Table 5.2 Table 5.3 Table 5.4 Table 5.5 Table 5.6 Table 5.7

5 17

The structure of the NP according to Bache (2000: 160; 171; 238–9) 109 A model of the NP geared towards the present study 113 The token frequencies of complex premodifiers across newspapers (cf. Mazaud 2004: 242) — frequencies normalised per 100,000 words 126 Berlage’s (2014: 44) NP complexity hierarchy for NPs involving up to one postmodifier 132 A complexity hierarchy of NP modification patterns involving up to one postmodifier, from Berlage (2014: 255) 133 A complexity hierarchy of NP modification patterns involving more than one postmodifier, from Berlage (2014: 255) 134 The complexity levels of whole NPs 141

List of Tables

XV

Table 6.1 Table 6.2 Table 6.3 Table 6.4 Table 6.5

ICE texts and text categories 152 ICE corpus sections 157 The output of the R routine for NP extraction NPs extracted from 162 NPs extracted from 162

Table 7.1

Premodification vs. postmodification in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 171 The frequencies of modifier subtypes in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 173 The co-occurrence patterns of all pre- and all postmodifier types in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 175 The co-occurrence patterns of all pre- and all postmodifier types in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 176 The lengths of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 179 A TukeyHSD post hoc test of the mean lengths of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 180 The lengths of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 181 The lengths of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 181 Pairwise Wilcoxon rank sum test of the lengths of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 182 The lengths of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 184 Pairwise Wilcoxon rank sum test of the mean lengths of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 184 The formal realisations of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 186 The formal realisations of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 187 The formal realisations of Postmod. 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 190 The complexity levels of Mod. II in S1A 192 Complexity levels of Mod. III in S1A 193 The complexity levels of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 195 Complexity levels of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 196 The number of premodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 198 The number of postmodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 199 The position of NPs embedded at level 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 201 The functions of NPs embedded at level 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 202

Table 7.2 Table 7.3 Table 7.3 Table 7.4 Table 7.5 Table 7.6 Table 7.7 Table 7.8 Table 7.9 Table 7.10 Table 7.11 Table 7.12 Table 7.13 Table 7.14 Table 7.15 Table 7.16 Table 7.17 Table 7.18 Table 7.19 Table 7.20 Table 7.21

161

XVI

Table 7.22 Table 7.23 Table 7.24 Table 7.25 Table 7.26 Table 7.27 Table 7.28 Table 7.29 Table 7.30 Table 7.31 Table 7.32 Table 7.33 Table 7.34 Table 7.35 Table 7.36 Table 7.37 Table 7.38 Table 7.39 Table 7.40 Table 7.41 Table 7.42 Table 7.43 Table 7.44 Table 7.45 Table 7.46

List of Tables

Effects of language contact in S1A 203 Premodification vs. postmodification in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 206 The frequencies of modifier subtypes in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 208 The co-occurrence patterns of all pre- and all postmodifier types in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 209 The lengths of the whole premodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 214 Pairwise Wilcoxon rank sum test of the mean lengths of the whole premodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 215 The lengths of the whole premodifying string by variety in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 215 The lengths of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 217 A TukeyHSD post hoc test of the mean lengths of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 218 The lengths of Mod. III in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 218 Pairwise Wilcoxon rank sum test of the mean lengths of Mod. III in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 219 The lengths of Mod. III by variety in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 220 The lengths of the whole postmodifying strings in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 221 Pairwise Wilcoxon rank sum test of the mean lengths of the whole postmodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 223 The formal realisations of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 224 The formal realisations of Mod. III in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 226 The formal realisations of Postmod. 1 in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 228 The complexity levels of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 231 The exact types of formal realisation at different complexity levels of Mod. II in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 232 The complexity levels of Mod. III in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 234 The complexity levels of the whole premodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 235 The complexity levels of the whole postmodifying string in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 236 The number of premodifying slots occupied per NP in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 238 The number of postmodifying slots occupied per NP in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 239 The position of NPs embedded at level 1 in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 241

List of Tables

Table 7.47 Table 7.48 Table 7.49 Table 7.50 Table 7.51 Table 7.52 Table 7.53 Table 7.54 Table 7.55 Table 7.56 Table 7.57 Table 7.58 Table 7.59 Table 7.60 Table 7.61 Table 7.62 Table 7.63 Table 7.64 Table 7.65 Table 7.66 Table 7.67 Table 7.68 Table 7.69 Table 7.70 Table 7.71 Table 7.72 Table 7.73

XVII

The functions of NPs embedded at level 1 in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 242 Effects of language contact in W1A 243 Effects of language contact in S1A and W1A 246 The lengths of NPs in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 249 Pairwise Wilcoxon rank sum test of the mean NP lengths in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 250 The lengths of whole NPs by syntactic position and variety in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 252 The numbers of subjects and non-subjects in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 252 The impact of Register and Variety on the mean lengths of the whole premodifying string 253 The complexity levels of whole NPs in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A (cf. 5.3.4.2 for the definition of complexity levels) 254 The correlation of NP complexity and syntactic position in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 257 The numbers of pre- and postmodifiers in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 259 Number of NPs embedded into other NPs per variety (cf. 5.3.4.2.2) in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 261 The complexity rankings of NPs in S1A 264 The lengths of NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 267 Pairwise Wilcoxon rank sum test of the mean lengths of NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 268 The lengths of whole NPs by syntactic position and variety in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 268 The numbers of subjects and non-subjects in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 269 The impact of Register and Variety on the mean lengths of the whole premodifying string 270 Post-hoc test — length of the premodifying string by Function and Variety 271 The complexity level of whole NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 273 The correlation of NP complexity and syntactic position in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 277 The numbers of pre- and postmodifiers in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 279 Number of NPs embedded into other NPs per variety (cf. 5.3.4.2.2) in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 281 The complexity rankings of NPs in W1A. 284 The varieties with the most obvious effects of simplification by register 287 Language contact and simplification by register/modality and variety 292 The susceptibility of varieties to language contact and simplification in S1A and W1A 293

XVIII

List of Tables

Table A.1

HCFA results shown in figure 7.1 (p. 171) — The overall frequencies of pre- and postmodified NPs in S1A 322 HCFA results shown in figure 7.2 (p. 173) — Modifier subtype and the choice between pre- and postmodification in S1A 322 HCFA results shown in figure 7.3 (p. 177) — The co-occurrence patterns of subtypes of pre- and postmodifiers in S1A 322 HCFA results shown in figure 7.7 (p. 187) — Formal realisations of Mod. III in S1A 323 HCFA results shown in figure 7.10 (p. 194) — Complexity levels of Mod. III in S1A 323 HCFA results shown in figure 7.15 (p. 201) — Number of NPs embedded as preand postmodifiers in S1A 323 HCFA results shown in figure 7.32 (p. 241) — Number of NPs embedded as preand postmodifiers in W1A 324 HCFA results shown in figure 7.35 (p. 254) — Hierarchical complexity levels in S1A 324 HCFA results shown in figure 7.36 (p. 256) — The correlation of syntactic position and complexity level in S1A 325 HCFA results shown in figure 7.37 (p. 259) — Numbers of pre- and postmodifiers in S1A 325 HCFA results shown in figure 7.38 (p. 261) — Levels of NP embedding in S1A 325 HCFA results shown in figure 7.41 (p. 273) — Hierarchical complexity levels in W1A 326 HCFA results shown in figure 7.42 (p. 276) — The correlation of syntactic position and complexity level in W1A 326 HCFA results shown in figure 7.44 (p. 281) — Levels of NP embedding in W1A 327 HCFA results shown in figure B.1 (p. 333) — Syntactic position and the choice between pre- and postmodification in S1A 327 HCFA results shown in figure B.3 (p. 337) — Semantic class of the head noun and the complexity levels of NPs in S1A 328 HCFA results shown in figure B.4 (p. 339) — Semantic class of the head noun and the complexity levels of NPs in W1A 328

Table A.2 Table A.3 Table A.4 Table A.5 Table A.6 Table A.7 Table A.8 Table A.9 Table A.10 Table A.11 Table A.12 Table A.13 Table A.14 Table A.15 Table A.16 Table A.17

Table B.1 Table B.2 Table B.3 Table B.4 Table B.5 Table B.6 Table B.7 Table B.8

Number of NPs embedded into other NPs per variety 329 Modification types in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 330 Proportions of NPs embedded into other NPs per variety 330 Modification types in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 331 Pre- vs. postmodification by variety and syntactic position in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 332 Pre- vs. postmodification by variety and syntactic position in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A 335 NP complexity levels by semantic class of head noun in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A 338 NP complexity levels by semantic class of head noun in W1A 340

List of Abbreviations AP ass BZM CLAWS cls comp.dir def EIC gen HCFA ICE ICE-EA ICE-GB ICE-SING IndSAfE int intj L2 lp Mod. I Mod. II Mod. III mpm nmp nom pass perf pm prox punya rel rl SD SLA Tukey HSD VPs WALS

adjective phrase associative Bazaar Malay Constituent Likelihood Automatic Word-tagging System noun class directional complement definite Principle of Early Immediate Constituents genitive Hierarchical Configural Frequency Analysis International Corpus of English ICE-East Africa ICE-Great Britain ICE-Singapore Indian South African English intensifier interjection second language linking particle in Cantonese (cf. Matthews and Yip 2011: xxiii) Premodifier, class I Premodifier, class II (descriptive) Premodifier, class III (classification) modifying phrase marker name particle nominaliser passive perfective aspect marker possessive marker proximal Malay-derived particle punya in Bazaar Malay relative pronoun relative Standard Deviation Second Language Acquisition Tukey’s Honestly Significant Difference Verb Phrases World Atlas of Language Structures

1 Introduction This study is a linguistic investigation of the structure of noun phrases (NPs) in Singaporean English, Kenyan English, and, for comparative purposes, British English. What is the reason for the focus on this specific area of syntactic variation, and the interest in these particular language varieties, which belong to the group of so-called “New Englishes”? This introduction will outline the linguistic motivation for this research question and prepare the ground for the study by sketching its theoretical and methodological basis. 1.1 starts by elaborating on the research question against the backdrop of the phenomenon of New Englishes. 1.2 justifies the choice of NPs as an area of variation and, at the same time, warrants the specific interest in Singaporean English and Kenyan English. 1.3 deals with the theoretical grounding of the study. 1.4 provides an overview of the study.

1.1 The nature of New Englishes Kenyan and Singaporean English are two members of the set of varieties of English known as “New Englishes” (cf. 2). They have both emerged in the aftermath of British colonialism. Having been transplanted to Singapore and Kenya in the course of the 19th century, English slowly took root in both regions, and, today, despite the countries’ independence since 1965 (Singapore) and 1963 (Kenya) respectively, has made considerable inroads in both societies. In Singapore and Kenya, English is nowadays assigned the status of an “official language” by the constitution, plays an important role for international communication and, in many domains of life, is even used within the two countries, with its entrenchment, however, stronger in Singapore than in Kenya (cf. 4.3). In the following, two main reasons for why linguists have been intrigued by such varieties will be sketched. (a) First, they are marked by a particular kind of sociolinguistic ambiguity: On the one hand, New Englishes are not usually spoken as L1s, but serve primarily as L2s in certain domains of life, while other contexts are dominated by local indigenous languages¹. On the other hand, however, both Singaporean English and Kenyan English fulfil important functions in key domains of the societies, including politics, the media, the economy, and, to an extent, people’s private lives (cf. above; cf. also 2.1.1). The present-day status of such post-

1 As explained below (cf. 4.3), the view as an L2 is unanimously true for Kenyan English only; in Singapore, in contrast, English has made some inroads as an L1.

2

1 Introduction

colonial varieties, hence, is intermediate between that of “a second language, on the one hand, and more native-like patterns of indigenous transmission and use, on the other” (Sharma 2005: 194). This property raises numerous questions of a broadly sociolinguistic nature, including e.g. issues of “language policy and pedagogy, […] cultural evaluation and sociopsychological integration and […] pragmatic evolution” (Schneider 2007: 2). (b) The second key motivation for the systematic study of New Englishes is more linguistic than the first one. New Englishes are marked by the effects of “structural nativisation”, which, during the evolution of such varieties, has brought about a wide variety of “locally characteristic linguistic patterns” (Schneider 2007: 5–6) at all levels of linguistic organisation. They comprise, for instance, innovations in phonology, such as the loss of diphthongs, morphology, including e.g. the loss of the nominal plural morpheme, or syntax, such as the lack of subject-verb inversion in interrogative clauses (cf. 2.1.3). Although both of these perspectives are highly relevant, the present study will concentrate on the innovative structural properties of New Englishes mentioned in (b). To be exact, the study will focus on two major subclasses of structural changes which have turned out to be particularly formative (cf. 2.2). As will be seen, both of them can be related directly to the L2-status of the varieties at hand mentioned above. (a) Since they are usually spoken alongside local indigenous languages, New Englishes have been frequently found to show evidence of cross-linguistic influence (cf. Schneider 2007: 107–9). (b) Linguistic constructions in New Englishes are often viewed as simpler in structure than corresponding British English expressions. This property has been attributed to restrictions of language production typical of the use of an L2 (cf. Schneider 2007: 102–3). It is these two processes of change which will take centre stage in the analysis of NP syntax in the present study. The main question pursued is the extent to which the structure of NPs, a well-defined area of English syntax, shows evidence of language contact in New Englishes, and to what extent they can be viewed as simpler than in the mother variety British English. This motivates the inclusion of British English in the analysis. According to Schneider’s (2007: 33–55) theory of New Englishes, New Englishes partake of a large-scale process of language change, which, despite many differences, can be considered as fundamentally similar across varieties of New Englishes and thus allows for a ranking of the varieties spoken around the world in terms of their evolutionary stages. From this point of view, Singaporean En-

1.2 Why NPs in Singaporean English and Kenyan English?

3

glish is deemed more advanced than Kenyan English in the literature (cf. 4.3 for a detailed justification of this stance). It will be interesting to see to what extent the effects of language contact and simplification to be found in the syntax of NPs in Singaporean English and Kenyan English can be related to these differences in evolutionary stages.

1.2 Why NPs in Singaporean English and Kenyan English? In what follows, it will be argued that both the choice of NP syntax and the focus on Singaporean English and Kenyan English are direct consequences of the interest in the processes of language contact and simplification.

1.2.1 Analysing language contact on the basis of NPs: rationale and aims The syntax of NPs has been singled out as a field of variation subject to language contact because of its pattern of cross-linguistic variation. On the basis of the structure of their NPs, the languages of the world can be classified into two major groups (cf. 5.1). One finds, for one thing, languages where NP dependents such as determiners or modifiers, for the most part, precede the NP head, with e.g. the Sinitic language family a prominent example of this type. At the same time, a group of languages can be identified where the majority of dependents follow the head by default (cf. 4.1.2.2; 4.2.2.2); Bantu languages spoken in Africa are wellknown to represent this subtype. Singaporean English has always been in contact with Sinitic languages, while Kenyan English is spoken in an area where (among other language phyla), Bantu languages have played a dominant role (cf. 4.1.2; 4.2.2). Consequently, due to their typological backgrounds, Singaporean English and Kenyan English allow for a particularly reliable analysis of language contact. If one of the two varieties shows one effect, while the other variety consistently exhibits a contrary effect, both of which are in line with the respective local languages, the pattern of variation can be ascribed to language contact with a high degree of certainty (cf. Gut 2011: 117). This approach greatly minimises the danger of mistaking an effect as being due to language contact, which is, in fact, due to other factors such as language-internal developments (cf. 2.2.1.2). The guiding hypothesis in the analyses of language contact in the present study, thus, will be that premodified (or head-final) NPs shown in (1.1)–(1.2) will be preferred in Singaporean English.

4

(1.1) (1.2)

1 Introduction

a cultural concept the performance forms

Kenyan English, in contrast, will be assumed to exhibit a preference for postmodified (or head-initial) NPs as in (1.3)–(1.4). (1.3) (1.4)

the cashbox full of money the most common name given to a lingua franca spoken throughout west Africa from Sierra Leone to the Gabon

1.2.2 The analysis of simplification on the basis of NPs: rationale and aims With regard to simplification, too, there is a clear motivation for the choice of the NP as a field of variation and the choice of the two varieties. For one thing, NPs come in different degrees of linguistic complexity. The NP in (1.5), for instance, would qualify as an excessively short, hierarchically simple case. (1.5)

I wasn’t up today I don’t drink uh so in far as socialising my socialising was very minimal

(1.6), in contrast, which contains several nested PPs and a coordination, needs to be ranked as considerably more complex, both in terms of length and with regard to syntactic hierarchy (cf. 5.3.2). (1.6)

The comparative analysis of the syllabuses of the series and the subject

The syntactic complexity setting apart (1.5) from (1.6) may serve as a proxy of cognitive complexity. Complex structures such as (1.6) can be plausibly viewed as more difficult to produce and comprehend than simple ones as in (1.5) (cf. 5.3.2.2). This is the reason why the understanding of simplification tendencies in New Englishes can benefit from the analysis of NP structures. Speakers of a New English variety are likely to make use of NPs which are simpler than those in a native variety (cf. 1.2) as they are easier to process and to produce. The plan to investigate complexity differences provides a second motivation for the specific interest in Singaporean English and Kenyan English, which is related to their distinct evolutionary stages (cf. 1.1). As New English varieties evolve along the stages presupposed by Schneider’s (2007) theory (cf. 2.1.2), the degree of

1.2 Why NPs in Singaporean English and Kenyan English?

5

entrenchment of English in a society can be assumed to grow, and, along with it, the average linguistic competence of speakers will rise. For this reason, the need for simplification is likely to subside in the course of a variety’s development. With Singaporean English being in phase 4 of Schneider’s (2007) developmental model and Kenyan English commonly assigned to phase 3 (cf. 4.3), the choice of these two varieties will allow for an empirical assessment of the actual degree of correlation between the tendency for simplification and the developmental stage of the varieties at hand.

1.2.3 The key factors of NP variation addressed Summing up, the present study will revolve around the influence of the factors identified in table 1.1 on the structure of NPs in Singaporean English and Kenyan English. The main aim of the corpus analysis is the identification of the role of these factors in the use of NPs in Singaporean English and Kenyan English. Table 1.1. Factors impacting on NP structure

typology of substrates evolutionary stages

Singaporean English

Kenyan English

head-final NPs phase 3

head-initial NPs phase 4

Beyond these influences, which are of central interest, there is, however, a range of further factors which have a bearing on the structure of NPs in language use, including e.g. the syntactic position of the NP, the semantics of the head noun or register and modality (spoken vs. written) (cf. 5.2.1.2). While all of these factors will be taken into account in the analyses below, the influence of register and modality on NP structures, which is known to be extraordinarily strong (cf. 5.2.1.2), will be addressed in a systematic way by taking into account two distinct registers in the analyses. The same type of inquiry will be conducted both in informal conversations, which represent one end of the register continuum, where NP structures tend to be largely simple, and academic essays, which are known to rely on very complex NP structures. It will be interesting to see which patterns of language contact and/or simplification affect the informal register only and to what degree they also appear in formal language.

6

1 Introduction

1.3 Theoretical backdrop of the study The investigation of NPs in the present study presupposes a specific view of New Englishes as well as a particular linguistic take on language as a whole, which will now be explicated briefly.

1.3.1 The approach taken to New Englishes The approach taken to contact and simplification has to be seen in the bigger picture of the recent variationist research tradition in New Englishes along the lines of Schneider’s (2007) Dynamic Model of the evolution of New Englishes, which presupposes that the structure of today’s new varieties can be described as the result of a shared, largely uniform process of linguistic evolution (cf. 2.1.2). This view allows for a meaningful comparison of varieties as far apart geographically and as dissimilar in terms of substrates as Singaporean English and Kenyan English, and forms a prerequisite for all linguistic conclusions drawn from such an inquiry. At a more specific level, the study can be viewed as a reaction to a range of recent desiderata identified in the literature. (a) For one thing, it reacts to calls for more comparative studies of New Englishes taking into account several varieties “that have an L1 with […] different structural properties” (Gut 2011: 117); many previous studies have merely compared the structures of one New English variety with the same structure in British Standard English (cf. Olavarrìa de Ersson and Shaw 2003; Bao and Hong 2006; Sharma 2005, to cite but a few) or have relied on the comparison of a feature in one single New English variety with the corresponding feature in the relevant substrate language (cf. e.g. Alsagoff and C. L. Ho 1998). In either case, however, it is difficult to prove with certainty that a given linguistic phenomenon is, in fact, due to language contact and that other factors, such as universal processes of SLA, or variety-internal developments, do not play a role. (b) Earlier phases of research have frequently focussed on overt effects of structural nativisation, such as the presence or absence of phonemes or morphemes (cf. 2.1.3). It has been only since the advent and spread of comparable computer corpora (cf. 6.1.1) that researchers have started to take into account nativisation in the area of subtle preferences of usage “below the level of anybody’s awareness” (Schneider 2007: 87). The present study seeks to provide further evidence of structural nativisation with regard to this type of variation.

1.3 Theoretical backdrop of the study

7

1.3.2 The approach taken to language 1.3.2.1 A corpus-based approach to language The present-study, for one thing, subscribes to the corpus-based paradigm, according to which the analysis of a computerised, restricted but representative sample of language is accepted as a valid way to study the usage norms of a whole speech community (cf. McEnery, Xiao, and Tono 2006: 4–5). In particular, a corpus-linguistic approach is viewed as an important methodological device in the study of New Englishes, which can help to reveal preferred usage patterns on a statistical basis.

1.3.2.2 A cognitive, usage-based approach to language This study subscribes to a both cognitive and a usage-based understanding with regard to language in general and the structure of NPs. Croft and Cruse (2007: 1) identify the following views as the smallest common denominator of the “cognitive linguistic enterprise” (Langacker 1999: 13). (a) Language, rather than being an autonomous cognitive system, is based on general cognitive abilities of human beings, which are also used for other purposes (cf. Croft and Cruse 2007: 1). (b) Language structure is conceptual in nature; in particular, syntactic or morphological structures cannot be explained away as epiphenomena, as in formal approaches, but stand in an iconic relationship to cognitive concepts (cf. Croft and Cruse 2007: 1). (c) The usage-based hypothesis implies that language structure, rather than being an abstract phenomenon in its own right, “emerge[s] from language use” (Croft and Cruse 2007: 1). These tenets have a number of ramifications for the interpretation of corpus data below. For one thing, viewing language as being based on general cognitive abilities leads to an objection against categorical judgements in grammar. Just as we are able as human beings to deal with the complexity of reality, which eschews sharp categorisation, linguistic phenomena, too, are a matter of more or less (cf. Croft and Cruse 2007: 77–9). In the present study, this cognitive view will be capitalised on e.g. in the analysis of the syntactic structure of the NP, where distinctions viewed as categorical in traditional grammar will be treated as gradient phenomena (cf. 5.1.4.1). Furthermore, if language structure is an effect of language use, language must be treated as a “structured inventory of conventional linguistic units” (Langacker 1987: 57) rather than a phenomenon derived from a combination of lexical items

8

1 Introduction

through abstract rules. In construction grammar, such conventional linguistic units of meaning and form go under the label of constructions (cf. Hilpert 2014: 12–3), which are defined as follows by Goldberg (2006: 5). Any linguistic pattern is recognized as a construction as long as some aspect of its form is not strictly predictable from its component parts or from other constructions recognized to exist. In addition, patterns are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency.

The syntactic structures of NPs such as (1.5) or (1.6) (p. 4) do not have cognitive reality by themselves but only gain linguistic relevance by virtue of their status as constructions entrenched through their use. In addition, constructions can be posited at varying levels of generality or abstractness (cf. Langacker 1987: 46; Croft and Cruse 2007: 262–3). First, concrete NPs such as (1.7)–(1.8) can be viewed as constructions. (1.7) (1.8)

soft drink the problem of definition

At the same time, however, such concrete NPs instantiate higher-ranking, purely schematic constructions, with (1.7), for instance, representing a subtype of premodified NP described exemplarily in (1.9), while (1.8) is cognitively related to a general postmodified schema as shown in (1.10). (1.9) [[AP][head]] (1.10) [[head][PP]] As a consequence, the use of NPs by speakers of Singaporean English and Kenyan English can be described in terms of the choice from a complex taxonomic network of NP constructions with varying degrees of concreteness or abstraction, with certain subtypes of NPs (e.g. head-final or head-initial ones) more or less entrenched in the minds of speakers, depending on the respective substrate languages. A second corollary of this view of language is the eminent role of the usage frequency of a given NP structure (or linguistic construction in general). The more frequently a construction is used, the stronger its cognitive representation in the network of constructions. This view is rendered plausible by a wide range of psycholinguistic studies which have demonstrated that “token frequencies correlate with degree of entrenchment” (Gries 2012: 49; cf. also Bybee 2006), e.g. by virtue of shorter reaction times of test subjects to highly frequent linguistic items (cf. Caldwell-Harris, Berant, and Edelman 2012).

1.4 The outline of the study

9

1.4 The outline of the study The study falls into two major parts. Part A, to start with, lays the theoretical groundwork by addressing the theoretical issues necessary for the analysis of language contact and simplification in Singaporean English and Kenyan English. – Chapter 2 discusses the notion of New Englishes, previous models for the analysis of their emergence and present status and singles out language contact and simplification as two of the main players in their evolution. – Chapter 3 is dedicated to the development of a comprehensive model of language change which allows for the explanation of language contact and simplification in New Englishes. The chapter has two main components: (a) 1. Both language contact and simplification will first be discussed from the point of view of SLA, which accounts for linguistic innovations at the level of individual language users. Results from SLA are considered useful for the study of New Englishes due to the character of New Englishes as second languages, which renders them similar (albeit not equal) to the type of variety studied in the field of SLA (cf. 2.1.1). Furthermore, it is assumed that any process of language change starts out as an idiolectal change in individuals before it it is transferred to the communal language norm (cf. 3.2.2.2). Furthermore, results from SLA are helpful as numerous studies in this field have dealt with the use of NPs. 2. Thereafter, Croft’s (2000) evolutionary theory of language change will be adapted for the purposes of the present study. It represents a fully-fledged theory of language change which explains in a detailed way how innovations translate from the level of individual speakers to the level of the speech community. Furthermore, it is in keeping with the cognitive approach to language (cf. 1.3.2.2). – Chapter 4 provides linguistic accounts of Singaporean English and Kenyan English with regard to their emergence and present-day status, including an analysis of the NP structures of the relevant substrate languages. – Chapter 5 deals with the syntactic structure of the NP and the variation of NP structure in language use. In addition, a set of complexity metrics is devised which can be used to rank NPs and their modifiers according to their complexity. Part B is dedicated to a corpus-based study of NP structures in Singaporean English and Kenyan English. – Chapter 6 provides the methodological background for the corpus study, including details on the corpus, the technicalities of corpus handling and the statistical methods used to describe results.

10

1 Introduction

– Chapter 7 is dedicated to the analysis of language contact and simplification in NP modification patterns in Singaporean English, Kenyan English and the mother variety to both, British English. In this chapter, the basic hypotheses of language contact and simplification are broken down into five sub-hypotheses to allow for easier operationalisation. The analysis deals with two distinct corpus sections containing spontaneous conversations (ICE-S1A) and formal academic language (ICE-W1A) taken from the relevant components of the International Corpus of English (cf. 6.2.1). – Chapter 8 provides a conclusion of the study and summarises the results.

Part A: New Englishes and the Structure of NPs

2 New Varieties of English The present study is an inquiry into grammatical variation in two varieties of English known as “New Englishes”. In this chapter, I seek to lay the foundations for the corpus-study in 7 by characterising these varieties in linguistic terms. I will start out by defining the notion of New Englishes, by discussing models which capture their emergence and present state and by giving a concise overview of their most frequent linguistic features as well as presenting basic explanatory models for linguistic innovation in New Englishes (cf. 2.1). 2.2 will establish language contact and simplification as two major processes in the linguistic evolution of New Englishes, and 2.3 will provide a summary of the key points of the chapter.

2.1 The phenomenon of New Englishes 2.1.1 The definition of “New Englishes” Kenyan English and Singaporean English are typical examples of the varieties covered by the term “New Englishes”, which was introduced into the field by Pride (1982) and John Talbot Platt, Weber, and M. L. Ho (1984). The notion of New Englishes is frequently defined by the following four criteria. (a) New English varieties have developed in the wake of (mostly British) colonialism in countries where English was not originally spoken. Their spread is due to English language teaching and their use as a medium of instruction in the educational system of those countries. (b) They are spoken as L2s in multilingual settings alongside other local languages. (c) They fulfil important functions in the regions in which they are spoken, particularly in the official domain, such as the media, politics and education, but also, to a greater or lesser degree, in people’s private lives. (d) New varieties of English have gone through a process called nativisation, i.e. they have developed innovations at various levels of linguistic organisation, running the gamut from novel lexical items via grammatical preferences to high-ranking pragmatic principles (cf. John Talbot Platt, Weber, and M. L. Ho 1984: 2–3). This notion of “New Englishes” has been criticised for a number of reasons, including a lack of objectivity, since the attribute ‘new’ merely “reflects […] a shift of attention in western, Anglocentric scholarship” (Schneider 2007: 3). Similarly,

14

2 New Varieties of English

Indian English, which is usually viewed as a “New English”, for instance, is in fact older than Australian English, which is considered a traditional L1 variety (cf. Mesthrie and Bhatt 2008: 3; originally in Kachru 1983). Alternative terms suggested in the literature, which highlight other features, include… – “World Englishes”, which is associated with Kachru’s quasi-political agenda of raising the awareness for new varieties in a postcolonial world (cf. Schneider 2007: 3). For the present purpose, however, this term is overly general, as it covers all varieties of English spoken “in the world”, ranging from the traditional cases of L1s spoken in the UK and the US, postcolonial varieties like Kenyan English and Singaporean English to pidgins and creoles (cf. Schneider 2013: 133). – “Postcolonial Englishes” (cf. Schneider 2007), which stresses the sociohistorical background of this type of variety, is likewise excessively wide in reference as it encompasses varieties such as Australian English, New Zealand English and American English, which are nowadays viewed as classic L1s because sociohistorical circumstances have led to a linguistic dominance of the colonisers and a lower degree of language contact than in the varieties at hand (cf. Schneider 2007: 118). Despite this, the notion of “New Englishes” will be used in the present study: It refers to the very type of variety under discussion and, what is more, stresses their nature as L2s as well as their multilingual background, both of which will take centre stage in the present study.

2.1.2 Models of New Englishes Among the many attempts to capture the variability of present-day English in theoretical models, the approaches which have gained greatest currency are, beyond doubt, those of Kachru (1985) and Schneider (2007) (cf. Buschfeld 2011: 72; cf. also Buschfeld 2013)¹. According to Kachru (1985), the varieties of English fall into an “Inner Circle”, where English is spoken as a native language, such as in the US, the United Kingdom or Canada, a “[n]orm-developing” (Kachru 1985: 17) “Outer Circle”, where English is spoken as a second language in the aftermath of colonisation as in Kenya or Singapore, and an “Expanding Circle”, which includes

1 Further models have been suggested by Strang (1970), McArthur (1987), Görlach (1990), Gupta (1997), Moag (1992), Melchers and Shaw (2003), Schneider (2003, 2007) and Mesthrie and Bhatt (2008). Space precludes an evaluation of all them; cf. Buschfeld (2011: 44-69) and Buschfeld (2013) for a thorough discussion.

2.1 The phenomenon of New Englishes

15

China, Russia, or Japan, where English is used as a foreign language (EFL; cf. Kachru 1985: 13). This model, although widely cited and much-used in the past (cf. Bruthiaux 2003: 159), has a range of drawbacks: It fails to allow for varieties which are ambiguous between circles; Buschfeld’s (2011: 70) results on English in Cyprus, for instance, cast doubt on Kachru’s (1985) clear-cut distinction between the Outer (=ESL) and the Expanding Circle (=EFL), as Cyprus cannot be assigned unanimously to any of these variety types (cf. also Bongartz and Buschfeld 2011). Most significantly, though, it draws primarily on political rather than concrete linguistic criteria (cf. Schneider 2007: 14; Mesthrie and Bhatt 2008: 29) and lacks a diachronic, evolutionary perspective on the varieties. These features strongly restrict its usability in the present context. It is especially the latter two respects which recommend Schneider’s “Dynamic Model” of the evolution of New Englishes for the present study. Unlike the Three Circles model, Schneider’s (2003, 2007) approach to New Englishes is inherently diachronic. What lies at the heart of the Dynamic Model is the insight that the evolution of all new varieties of English can be accounted for by fundamentally similar principles, across the highly diverse linguistic and cultural backgrounds in which they have emerged (cf. Schneider 2003: 234). A second basic assumption made in the model is that there is an intimate relationship between social and linguistic identity. New varieties arise through a series of linguistic adaptations which interact closely with the type of identity alignment on the part of the colonisers (referred to as the “stl strand” by Schneider) and the indigenous population (designated as “idg strand” in the Dynamic Model; cf. Schneider 2007: 95). Schneider assumes that in the development of new varieties of English around the world, a series of five evolutionary stages can be distinguished, each of which is marked by a specific interaction of identity factors with linguistic developments. In the following, the five phases will be sketched. Phase 1: Foundation — The evolution of a new variety of English starts with English being transplanted to a territory where it has not been spoken before. Initially, language change happens predominantly on the part of settlers, whose dialects typically undergo koinéisation. Communication between the stl and idg strands is limited to encounters e.g. in trading contexts. Cross-linguistic influence remains restricted to individual lexical items such as place names (cf. Schneider 2003: 244–5). Phase 2: Exonormative Stabilisation — In the second phase, the stl community gradually settles in and adapts to the new environment, even though the identity of its members is still primarily rooted in “a common territory of origin and a feeling of culturally belonging there” (Schneider 2003: 245). Consequently, speakers

16

2 New Varieties of English

orient themselves towards the acrolectal norm of the language used in the distant home country. Despite that, on the part of the settlers, the new environment brings about early linguistic innovations such as borrowings from indigenous languages for objects of local significance. Among the indigenous population, bilingualism starts to spread slowly and gradually. Such instances of linguistic contact between the two strands lead to early forms of contact-induced language change and can be viewed as the first tentative effects of what Schneider (2003: 246) refers to as “structural nativisation”, which, however, starts in earnest only in phase 3 (cf. Schneider 2003: 245–7). Towards the end of this phase, the settlers’ identity could be described as “English cum-local” (Schneider 2003: 246). Phase 3: Nativisation — In Phase 3, the process of identity rewriting proceeds at an even greater pace. The settlers accept the colony as their new home country, which may find expression in political independence. At the same time, the “stl and the idg strands become closely and directly intertwined” (Schneider 2003: 247), which prompts all inhabitants of the colony, in particular the members of the stl strand, to adapt to each other, in both cultural and linguistic terms. This large-scale identity shift brings about the considerable linguistic effects of structural nativisation which was initiated in phase 2. Innovations in the idg strand which have developed through the mechanisms of language contact are gradually taken over by the stl strand, in a complex process of negotiation and accommodation. Divergent views among conservatively minded and more progressive speakers within the stl strand may manifest themselves in a so-called “complaint tradition” (Schneider 2003: 248), in which concerns over falling linguistic norms are voiced in public. As regards linguistic consequences, Schneider (2003: 248) mentions an even stronger tendency for lexical borrowing and the development of a local accent. This phase is also rife with grammatical innovations, which tend to happen at the interface between grammar and lexis (cf. 2.1.3; Schneider 2003: 247–9). Phase 4: Endonormative Stabilisation — This phase is marked by a thorough sense of independence from the distant home country, which is frequently engendered by “Event X” (Schneider 2003: 250), some kind of cataclysmic political event which brings an acute awareness of isolation from the home country. In the emergence of New Zealand English, Great Britain’s accession to the European Union in 1973 “without softening provisions for New Zealand” (Schneider 2007: 131) may have had such effects on the development of a sense of nationhood in New Zealand. Within the emergent nation, the importance of ethnic boundaries is diminished, and the emphasis shifts to the collective experience of a shared nationhood. Independence as a nation is closely intertwined with a sense

2.1 The phenomenon of New Englishes

17

of linguistic independence: The novel linguistic norms which have developed during the earlier evolutionary phases are “accepted as adequate also in formal usage” (Schneider 2003: 250) and slowly gain prestige. Schneider emphasises the homogeneity exhibited by new English varieties at this stage, which can be associated with a sense of national coherence typical of such a phase of transition. Varieties at this stage are frequently codified in dictionaries and become a medium for literary creativity (cf. Schneider 2003: 249–53). Phase 5: Differentiation — The final stage in the evolution of a new English variety brings about differentiation into regional and social dialects. According to Schneider (2003: 253), this linguistic effect emanates directly from aspects of social and political identity — “once […] one’s global, external position is safe and stable […] this allows for more internal diversification” (Schneider 2003: 253). The new variety need not, however, function as an L1 but may co-exist with other local languages (cf. Schneider 2003: 253–4). The evolutionary stages of the varieties covered by Schneider (2007) are shown in table 2.1. Table 2.1. Varieties of English along Schneider’s (2007) evolutionary cycle. Phase 1

Phase 2

Phase 3

Phase 4

Phase 5

Fiji English

Philippine English Malaysian English Indian English Kenyan English Tanzanian English Nigerian English Cameroon English

Singaporean English South Afr. English Bajan Jamaican English

Australian English New Zealand English Canadian English

The Dynamic Model has been criticised on several counts. Mesthrie and Bhatt (2008: 35–6), for instance, object that the all-important concept of identity in the Dynamic Model is primarily construed in terms of nationhood and does not place due emphasis on other aspects of identity relevant to linguistic choices, such as class or social status. Furthermore, they argue that the assumption of an “Event X” actually amounts to overestimating a single political incident, which may, in fact, bear little relevance for lower socioeconomic classes. Finally, varieties of English in West Africa may have progressed directly from phase 3 (nativisation) to phase 5 (differentiation), which goes against the model’s assumption of subsequent phases (cf. Mesthrie and Bhatt 2008: 35). In a similar vein, Buschfeld

18

2 New Varieties of English

(2011: 76) states that the Dynamic Model fails to encompass intermediate cases between ESL and EFL like English on Cyprus. These problems should certainly be taken seriously, even though I would argue that they do not detract from the model’s value as a theoretical backdrop for the analysis in the present study. A certain degree of generality and abstraction is inevitable in a theoretical model accounting for as complex a phenomenon. A model which is applicable as widely as the Dynamic Model will always capture some aspects of reality better than others. Schneider (2007: 57) states explicitly that the model describes a prototypical process and “is not intended to account for all observable details, nor does it apply equally well to all individual instances of the process it describes” (Schneider 2007: 29). On the positive side, the Dynamic Model recommends itself by offering a detailed, sociolinguistically informed breakdown of the diachronic evolution of New Englishes, which is capable of classifying borderline cases. In addition, it provides an interface to concrete linguistic correlates of the evolution of New Englishes, which is indispensable for a corpus-based study. What also recommends the model is the degree of acceptance in the literature. In studies of new varieties of English, the Dynamic Model is currently among the approaches used most widely. It has been adopted in numerous studies on New Englishes (cf. e.g. Sand 2004; Mukherjee and S. Hoffmann 2006; L. Lim 2007; Mukherjee and Gries 2009; Mukherjee and Gries 2010; van Rooy 2010; T. Hoffmann 2011; Bongartz and Buschfeld 2011; Yao and P. Collins 2012, Zipp and Bernaisch 2012 to cite but a few; Schneider 2014: 12–6). In addition, recent studies have documented repeatedly that the degree of structural nativisation in concrete areas of grammar is, in many cases, in line with the predictions made by the Dynamic Model about evolutionary stages of varieties (cf. Mukherjee and Gries 2009: 46).

2.1.3 An overview of linguistic innovations in New Englishes Linguistic innovations in New Englishes run the gamut from phonology to highranking pragmatic and stylistic principles. To mention but a few, Mesthrie and Bhatt (2008: 118–26), for instance, note a widespread tendency for the obliteration of the distinction between long and short vowel phonemes, with the result, for instance, of a coalescence of the kit and fleece vowels. Speakers of New Englishes frequently borrow lexical items from indigenous languages such as makan ‘food’ in Singaporean English, which originates in Malay. Kortmann and Szmrecsanyi (2004: 1188–9) collate a list of the most frequent non-standard grammatical features found in their corpus of L2 varieties of English, which includes e.g. a “lack of inversion in main clause yes/no questions” (You get the

2.1 The phenomenon of New Englishes

19

point? You liked India?, from Indian South-African English), the irregular use of articles (Then he thought, what about getting girl [to marry] from India?) or a lack of difference between Present Perfect and Simple Past (Some of us have been to New York years ago.). As far as pragmatics is concerned, Mesthrie and Bhatt (2008: 136–9) mention the use of a range of novel discourse particles in Singaporean English (cf. 4.1.1.3). As the present study will deal with both an informal spoken and a highly formal written register, it is of particular interest that there is also variation at the level of registers. Relevant studies have noted, first, a tendency for intervarietal convergence in formal, especially written registers (cf. Schneider 2007: 82; cf. also Mair’s (2007: 84); Bao and Hong 2006: 110–1; Xiao 2009: 443–4; van Rooy, Terblanche, et al. 2010: 346) and a tendency for greater formality in informal registers, which van Rooy, Terblanche, et al. (2010: 346) attribute to the sociolinguistic fact that in New English varieties, “the formal registers are more frequently used than the informal registers, because other languages customarily serve the private functions of the speakers”. This overview, of course, only scratches the surface of the degree of variability to be found; cf. e.g. Kortmann, Burridge, et al. (2004) for an in-depth treatment.

2.1.4 An overview of explanatory devices for innovations in New Englishes Beyond merely listing innovative features, it is crucial to identify the principles of language change at work in New Englishes. According to a general overview by Schneider (2007: 99–112), the linguistic development of New Englishes involves three principal processes: (a) continuity, i.e. speakers of New Englishes, first and foremost, take over a substantial “core” of English grammar, both in terms of standard and vernacular usage; (b) innovation, i.e. the “emergence of new elements for internal reasons” (Schneider 2007: 99–100), and (c) contact, i.e. crosslinguistic influence from local languages spoken in the region. Innovation may involve both the simplification of linguistic structure and restructuring, i.e. “the systematic rearrangement and reinterpretation of constituents and constituent sequences in language evolution” (Schneider 2007: 105). Language contact leads to borrowing or calquing; likewise, it can serve as a catalyst to such processes as simplification or restructuring (cf. above). In addition to these structural reasons for borrowing, Schneider (2007: 88–90) cites the following three forces which likewise drive evolution in New Englishes: There are, for one thing, cultural factors which may play a role in language change, for instance with regard to the borrowing of lexical items. Olavarrìa de Ersson and Shaw (2003: 159) have even suggested that the Indian English preference for complementation patterns foregrounding the receiver in verbs of fulfilling

20

2 New Varieties of English

(e.g. provide) in Indian English may be due to South Asians’ “view[ing] the individual as a part or a small object in a larger whole” (Olavarrìa de Ersson and Shaw 2003: 159), which sets them apart from Europeans; this, however, is highly speculative. A second important force is second-language acquisition, which, for instance, frequently leads to simplification (cf. Schneider 2007: 89). Finally, there may be varieties in which speakers “consistently select […] forms and patterns that conform to an overarching language type, with many innovations being typologically similar in nature and strengthening one specific parameter” (Schneider 2007: 89), which, over time, transforms the typological character of a variety as a whole; Mesthrie (2006, 2012), for instance, suggests that due to their Bantu substrates, which are syntactically highly explicit, African Englishes have an overarching tendency to “spell out the syntax in full” (Mesthrie 2012: 99), i.e. to use redundant syntactic elements such as cataphoric them in (2.1); he labels this tendency “anti-deletion”. (2.1)

Yes, most of them, I call them confused scholars. (Mesthrie 2012: 92)

Asian Englishes, in contrast, tend to delete such elements (cf. Mesthrie 2012: 97– 9), which fits in with the structure of the (Sinitic) substrate languages in this region, which are replete with syntactic “zero” elements.

2.1.5 The field of SLA and the study of New Englishes With Schneider (2007) ascribing an important role to Second Language Acquisition (cf. 2.1.4), a short detour is in order in order to explore further the relationship between the field of SLA and New English scholarship. The question of the relevance of SLA for the study of new varieties of English has been under discussion since the early phases of research. Williams (1987: 163), for instance, states that linguistic features of New Englishes “strongly resemble forms found in learner languages” and explicitly characterises such varieties as “an acquisitional phenomenon” (Williams 1987: 163). At the same time, however, there has always been an awareness of the differences between language users in a New Englishes context and the prototypical test subjects known from SLA studies. Learners in SLA aim at “native-like competence”, while speakers of New Englishes primarily interact with other speakers using their variety, without necessarily sticking to a standard norm. The input available in prototypical SLA situations is a native variety, while it is a non-native variety in New Englishes contexts. In addition, indigenised varieties of English are used alongside other languages, whereas the prototypical language learner in

2.2 Language contact and simplification as major processes of change

21

SLA acquires English in a monolingual English environment. In the same vein, learners in SLA contexts will combine the aim of learning a language with a desire to become a full member of the relevant culture; in postcolonial situations, speakers are likely to acquire English with an instrumental motivation, while retaining their ethnic background. Finally, what is known as “interlanguage” in classical SLA contexts is a largely individual phenomenon, whereas New English varieties have a societal dimension (cf. K. K. Sridhar and S. N. Sridhar 1986: 5– 7; Mesthrie and Bhatt 2008: 157). Consequently, despite the strong similarities of linguistic features between SLA and New Englishes, there are undeniable differences. The study of SLA fails to pay attention to the specific sociolinguistic dimensions in which New Englishes are used, including the negotiation of linguistic norms, standardisation and aspects of social identity. For this reason, results from SLA cannot be mapped wholesale onto New Englishes. On the other hand, though, there are two arguments in favour of “bridging the paradigm gap” (K. K. Sridhar and S. N. Sridhar 1986: 3) between the two fields. For one thing, it would certainly be wrong to assume that the two types of language acquisition differ in a degree which makes them wholly incomparable. In the early stages of a New English variety, it is marked by cognitive processes which are fundamentally similar to those typical of SLA interlanguage (cf. Winford 2003: 16). If complemented by considerations about the social dimension of language change, results from SLA can be highly beneficial to the study of New English varieties. Furthermore, from a methodological point of view, there is no doubt that speakers of New Englishes represent “one of the most significant segments of second-language acquirers in the world today” (K. K. Sridhar and S. N. Sridhar 1986: 3). Either discipline would benefit greatly from a greater mutual awareness of the other discipline’s results. These conclusions will be taken into consideration in the discussion of language contact and simplification from an SLA perspective provided in chapter 3.1.

2.2 Language contact and simplification as major processes of change In the present study, two key processes of language change in New Englishes, which can be viewed as highly basic for all kinds of contact-induced language change (cf. Croft 2000: 202; 204; 2003), will be singled out, as they are particularly relevant in the study of NP structures: Language contact and simplification. As will be shown in more detail, either process has been documented amply in New Englishes.

22

2 New Varieties of English

2.2.1 Language contact in New Englishes It does not come as a surprise that language contact has been described in New Englishes in virtually all areas of linguistic organisation. Space precludes a full account; suffice it to mention some pertinent studies for the major fields of linguistic organisation. Then, the notion of language contact will be problematised briefly.

2.2.1.1 Evidence of language contact in New Englishes As for phonology, Wiltshire (2005) attributes the retention of the distinction between /v/ and /w/ in Indian English, which has otherwise been lost in this variety of English, to Tibeto-Burman substrate languages (cf. Wiltshire 2005: 284). Instances of lexical influence from substrate languages abound. In the field of morphosyntax, Bao (2005: 237) finds systematic influence from Chinese in the aspectual system of Singaporean English, which makes use of the analytic aspect marker already, as in I see the movie already ‘I saw the movie’. Sharma (2009) analyses the extension of the meaning of progressive forms in Indian and Singaporean English, finding that speakers of Indian English have a tendency to overuse the -ing form. In Standard English, progressives can be used for progressive and delimited verbal actions with a habitual character, as in (2.2). In Indian English, in contrast, the progressive form is also used for habituals without temporal delimitation and stative verbs, which are illustrated in (2.3)–(2.4) (cf. for details Sharma 2009: 181); this can be put down to the rich set of imperfective markers available in Hindi (cf. Sharma 2009: 177). (2.2) (2.3) (2.4)

I’m eating meat these days. Generally only dry-cleaning clothes are coming. Some people are thinking it’s a bad job. (Sharma 2009: 181)

It is of particular relevance for the present study that cross-linguistic influence has also been found in syntax. The striking one-relative construction in Colloquial Singaporean English shown in (2.5), for instance, has been attributed to the influence from Chinese relative clauses (cf. (2.6)), where the relative pronoun de (rp in (2.6)) follows the relative clause rather than preceding it (cf. Alsagoff and C. L. Ho 1998: 145). (2.5)

That boy pinch my sister one very naughty. ‘The boy who pinched my sister is very naughty.’

2.2 Language contact and simplification as major processes of change

(2.6)

23

Nie jie-jie de neige haizi hen huaidan. pinch sister rp that child very naughty ‘That child who pinched my sister is very naughty.’

Although far from complete, this concise survey shows that language contact has been attested widely in New Englishes.

2.2.1.2 Problematising cross-linguistic influence in New Englishes A general controversy in the New English literature revolves around the roles of language contact and universal factors of change (such as simplification). A number of researchers (particularly those aiming at a large-scale bird’s-eye view of New Englishes) have foregrounded universal processes in the evolution of such varieties, drawing primarily on the impressive degree of correspondence between varieties. In Kortmann, Burridge, et al.’s (2004) Handbook, for instance, out of the twelve L2 varieties of English taken into account, an amazing eight share a set of 19 morphosyntactic features. With regard to phonology, too, a large number of features has been found to recur across New Englishes (cf. 2.1.3; Mesthrie and Bhatt 2008: 118–30). The high degree of intervarietal correspondence leads to Mesthrie’s (2004: 1141) argument that… [i]t is prima facie implausible, areal linguistics notwithstanding, that over a thousand languages should induce the very same (or very similar) influences. This would be tantamount to claiming the languages of Africa-Asia are the same in structure, united in their differences from English.

Along similar lines, Mair (2003) has suggested a universalist notion of “Angloversals”, i.e. “joint tendencies observable in the course of the standardization of postcolonial varieties of English which cannot be explained historically or genetically” (Kortmann and Szmrecsanyi 2004: 1192), including e.g.the lack of inversion in yes-no questions, the use of me instead of I in coordinate subjects, the nondifferentiation of adjective and adverb forms, or never as a past tense negator (cf. Kortmann, Burridge, et al. 2004: 1192–3). At the same time, however, others have cautioned against sweeping conclusions with regard to such universals, stressing that “substrate-superstrate interactions can go a long way towards explaining emergent systems in postcolonial varieties of English” (Sharma 2009: 192) and warning that universal tendencies should only be posited once cross-linguistic influence has been ruled out with due certainty (cf. Sharma 2009: 192). Sharma (2009: 170) demonstrates that a closer look at the grammatical conditioning of three putative universals across New Englishes, past tense omission, the over-extension of the meaning of the progressive

24

2 New Varieties of English

and copula omission in Indian and Singaporean English reveals that a substrate explanation is a great deal more plausible than an explanation in terms of universal markedness (cf. Sharma 2009: 192). Similarly, Mesthrie’s (2012: 99) largescale tendencies of “deletion” and “anti-deletion” are likely to be substrate effects (cf. 2.1.4). In sum, these positions should be viewed as complementary: Substrate influence does play a natural role in varieties spoken primarily alongside other languages, but it is hardly surprising that universal processes should have a part to play, too, as many aspects of L2 acquisition are known to be universal across learners (cf. 3.1). Universalist explanations should not be rejected blankly as long as universality can be explained plausibly on the basis of general principles of language change and without taking recourse to innatist principles (cf. e.g. Chambers 2004; cf. also Winford 2009: 227; Thomason 2008). This discussion has two ramifications for the present study: First, it is necessary to devise a method of telling apart the two types of factors in a reliable way. Second, the study can, to an extent, contribute to this discussion by gauging the relative influence of language contact and universal simplification tendencies.

2.2.1.3 A problem in the analysis of New Englishes: Finding a benchmark In the analysis of innovations in New Englishes, a major methodological problem arises. In order to tease out effects of language change, the varieties in question are usually analysed by means of a comparison to their mother variety, which, in the case of Singaporean English and Kenyan English, is British English. This approach has become especially widespread with the availability of comparable corpora such as the ICE family (cf. 6.1.1). However, there are two major problems with regard to this practice. – The variety transplanted to the relevant colonies in the formative colonial years was 19th century English, which differs from the variety spoken nowadays on the British Isles. Indeed, putative innovations found in New English varieties may actually be retentions from earlier forms of British English. Mukherjee and S. Hoffmann (2006: 161), for instance, identify a range of verbs used in ditransitive double-object complementation patterns in a large corpus derived from Indian newspapers, which, from a modern standard English perspective, are used with a to-phrase only; examples include to present, to provide and to supply. Colleman (2011: 399), however, demonstrates convincingly that in 18th-century British English, numerous verbs which, from today’s perspective, allow the to-dative only, were used in the double-object construction

2.2 Language contact and simplification as major processes of change

25

(e.g. deliver in little Betty […] delivered him a card; similarly: design, return, repeat, recommend etc.; cf. Colleman 2011: 400). Whether Mukherjee and S. Hoffmann’s (2006) findings are actually Indian innovations, or just reflect older usage patterns, remains open to discussion². As regards the typological focus of the present study, i.e. preferences for pre- or postnominal structures, it should be kept in mind, for instance, that the usage frequency of nominal premodifiers such as (2.7), must be assumed to have been very low in the 18th century and has soared from the 19th to the 20th century, especially in academic writing and newspaper language (cf. Biber and Gray 2011: 231). (2.7)

mortality rate (Biber and Gray 2011: 231)

The frequency of this construction in present day English, which will play an important role in the analyses in 7, cannot be assumed to be equal to its frequency in the early 1800s, when English was brought to Singapore (cf. Mesthrie and Bhatt 2008: 45). – Furthermore, there is no doubt that settlers in New English contexts used a vernacular rather than a standard variety of English. Stock features of New Englishes, such as the pluralisation of non-count nouns, or the use of double negation (cf. 2.1.3), for instance, may well have been a feature of British nonstandard varieties brought over to colonies rather than actual innovations in New Englishes (cf. Mesthrie and Bhatt 2008: 45). To an extent, these problems cast doubt on the widespread practice in the study of New Englishes to use findings from British English as a benchmark for the assessment of the development of new varieties. Instead, it would seems advisable to take into account 18th and 19th century usage in the study of grammatical innovations in New Englishes. However, consulting a corpus of 19th century English in each and every study of New Englishes is a tall order, considering e.g. the lack of modern text types and registers, or, indeed, the lack of spoken data from periods earlier than the 20th century. For this reason, this study will follow the wellestablished practice of relying on present-day British English, despite the fact that such an approach, to some degree, masks the historical dimensions of language use.

2 I would like to thank Elizabeth Closs Traugott (p.c. June 2017) for pointing out this problem in Mukherjee and S. Hoffmann’s (2006) study to me.

26

2 New Varieties of English

2.2.2 Evidence of simplification in New Englishes The claims for linguistic structures of New Englishes to be simpler than the corresponding structures in standard varieties go back to the early days of research into New Englishes (cf. e.g. Williams 1987). In this section, key results on simplification in New Englishes will be collated, starting with studies that focus on the local complexity of individual linguistic features, and moving on to analyses which aim at a global assessment of the complexity of varieties. Importantly, the discussion will be based on an informal, intuitive notion of complexity; a full-blown definition of this concept will be provided in 5.3.

2.2.2.1 Studies with a local focus Kadenge (2009) and Simo Bobda (2007) set out to detect simplification in the phonology of New Englishes. Many African Englishes, for instance, reduce the sixmembered system of short monophthongs in English to a system of five phonemes (cf. Kadenge 2009: 169). Simo Bobda (2007: 415) finds a widespread tendency for the monophthongisation of diphthongs in syllable onsets in African Englishes, e.g. through the use of [eː] for the face lexical set (cf. Simo Bobda 2007: 412), the disyllabification of triphthongs, as in [eːə] for /eɪə/ in player, or consonant cluster reduction in syllable onsets and codas, as in [en ɔf jiɛ] for end of year (cf. Simo Bobda 2007: 415; 418). Williams (1987), De Klerk (2003) and Wong (1983) look into the simplification of morphosyntax in different varieties of New Englishes. The features analysed by Williams (1987) include invariant tag questions like isn’t it which lack the subject concord found in Standard English, the omission of redundant grammatical markers, as in all the suicide case, where the plural -s is omitted due to the plural quantifier all (cf. Williams 1987: 170; 176), or the preference for a transparent mapping between content and form. She stresses that these features of New Englishes are the cumulative result of SLA-induced simplification, which took place due to adult SLA in the early days of these varieties (cf. Williams 1987: 163). De Klerk (2003) queries a corpus of spoken Xhosa English, likewise detecting various kinds of evidence for simplification, including the coalescence of mass and count nouns in forms like equipments (cf. De Klerk 2003: 230), which she puts down to economy, or effects of hyperclarity, like the explicit marking of habitual aspect by used to, as in The high schools they also used to get books (cf. De Klerk 2003: 237).

2.2 Language contact and simplification as major processes of change

27

Wong (1983), finally, in much the same vein, finds evidence of morphosyntactic overgeneralisation in colloquial Malaysian English, such as the omission of expletive there and it, as in No need to trouble him now (cf. Wong 1983: 132) or an invariant question tag isn’t it replacing the “‘disproportionately burdensome’ system [of question tags, T.B.] of standard formal English” (Wong 1983: 135). Huber (2012) looks for evidence of simplification in the factors determining relativiser choice in written Ghanaian English. He assumes that the system of relativisers is a prime example of redundant marking which is likely to undergo simplification, as there is a bewildering set of variants with low communicative import, including who/m and which, that and a zero relativiser. In addition, most languages of the world have a smaller set of relativisers (cf. Huber 2012: 220). His most noteworthy finding concerns the influence of the animacy of the referent on relativiser choice: Ghanaian English shows a markedly stronger preference for that over who/m for inanimate referents than British English as one descends the animacy hierarchy (from human via animate to inanimate); inanimate referents trigger that in 68.9% of cases in ICE-GH; in ICE-GB, they do in a mere 59.8% of cases (cf. Huber 2012: 229–30). He concludes that this result is due to the simplicity of invariant that and the complexity of the alternation between which/who/m. Steger and Schneider (2012) take as their starting point the much-cited observation that New Englishes prefer simple one-to-one, i.e. iconic, mappings of meaning and form over complex, non-iconic structures. They focus on complement clauses used after verbs like allow, cause, enable, expect, love, prefer, require and want, which can be either finite or non-finite, as can be seen in the following example sentences (cf. Steger and Schneider 2012: 165). (2.8)

(a) I want [ Ø him Ø to do that]. (b) I want that he should Ø do that.

(2.8) (b) can be viewed as more iconic, and hence simpler, than (2.8) (a), as in the less iconic version of the clause in (2.8) ((a)), both tense and agreement, modality, and a complementiser are missing, all of which are coded explicitly in the more iconic version (2.8) ((b)) (cf. Steger and Schneider 2012: 165). They hypothesise that speakers of New Englishes should exhibit a preference for such iconic (i.e. simple) patterns. One of the most striking results from their study of corpus data from Great Britain, Singapore, India, Hong Kong and East Africa is that speakers of New English varieties frequently show insecurity and rely on selfcorrections with regard to the choice between the finite and non-finite patterns, as in Defendant promised that to took the uh uh sorry promised to take the boy out

28

2 New Varieties of English

(cf. Steger and Schneider 2012: 176). Furthermore, the proportions of complement clauses exhibiting an explicit subordinating conjunction that are higher than average in certain New English varieties, particularly in Indian and East African English (cf. Steger and Schneider 2012: 180–1). Finally, the verb expect stands out by showing a higher frequency of explicit modality markers in the complement clause (cf. Steger and Schneider 2012: 181–2). Such and similar observations bring them to the conclusion that New Englishes are, on the whole, simpler in that they prefer iconic over non-iconic structures (cf. Steger and Schneider 2012: 187). These studies are important in having paved the way to an understanding of morphosyntactic simplification in New Englishes, and in having established the cognitive effects of SLA as a plausible explanation for simplification. It is somewhat less convincing, though, that some of them rely on anecdotal observation as their prime method, that they hardly go beyond the compilation and classification of examples, and that they lack a clear definition of complexity. In addition, linguistic features considered as examples of simplification tend to be highly heterogeneous.

2.2.2.2 Studies with a global focus One recent line of research has given a new twist to this kind of inquiry. Kortmann and Szmrecsanyi (2009b) gauge complexity on the basis of the following four measures, which are computed for four variety types, on the basis of data taken from Kortmann, Burridge, et al.’s (2004) Handbook of Varieties of English: traditional L1 varieties (such as British English), high-contact L1 varieties (such as Irish or Welsh English), L2 varieties (including New Englishes) and Pidgins/Creoles. “Ornamental rule complexity” refers to the number of redundant grammatical features and rules which offer no “added communicative bonus” (Kortmann and Szmrecsanyi 2009b: 269). Examples include the use of the pronouns she or her for inanimate referents, or the Northern Subject Rule (cf. Kortmann and Szmrecsanyi 2009b: 273). L2 varieties (which include New Englishes) have the lowest score for ornamental complexity, whereas traditional L1 varieties have the highest values, which accords well with the idea of grammatical simplicity in New Englishes. “Complexity deriving from irregularities and low transparency” refers to the “text frequency of irregular, lexically conditioned grammatical morphemes […] in naturalistic discourse” (Kortmann and Szmrecsanyi 2009b: 269). Features which count as simplifying in this respect include the coalescence of inflectional forms in the

2.2 Language contact and simplification as major processes of change

29

verbal paradigm, as in He had went instead of He had gone or the regularisation of the reflexive pronoun paradigm, as in hisself — theirselves / theirself (cf. Kortmann and Szmrecsanyi 2009b: 273). It comes as a surprise, however, that L2 varieties have a markedly low number of simplifying features, with L1 varieties, in fact, scoring considerably higher. In this respect, New Englishes cannot be viewed as simpler than L1 varieties. “L2-acquisition complexity”, in turn, is defined as the difficulty posed by grammatical features for learners of English as a second language (cf. Kortmann and Szmrecsanyi 2009b: 271). This type of complexity is operationalised by a set of features which are typical characteristics of learner language: the omission of inflectional endings, the overgeneralisation of morphological endings, such as he goed or the use of zero copula verbs (cf. Kortmann and Szmrecsanyi 2009b: 274). In this respect, too, L2 varieties do not stand out as particularly simple: They have L2 simplicity indices which are as low as those of L1 varieties; instead, it is Pidgins and Creoles that exhibit a high proportion of such features. “Grammaticity” is defined as the number of free grammatical markers (e.g. function words like the, prepositions or conjunctions etc.), which is measured in a socalled “analyticity index”, and bound grammatical markers (such as -s in sing-s), which is captured in a “syntheticity index”. In this respect, L2 varieties pattern as expected. They invariably exhibit the lowest values for both of these indices in comparison to traditional and high-contact L1 varieties, which means that a low proportion of words in these varieties carries synthetic grammatical morphemes, and comparatively few of them are function words. This leads Kortmann and Szmrecsanyi (2009b: 278) to conclude that they have a very general tendency to leave grammatical contrasts unmarked, no matter whether it is about analytic or synthetic marking. In a nutshell, it can be said that Kortmann and Szmrecsanyi (2009b) have greatly furthered the understanding of simplification in New Englishes by using a differentiated and well-founded notion of complexity and by showing that simplification only affects selected complexity metrics. New Englishes do reduce redundant marking (i.e. ornamental rule complexity) and have a lower incidence of both synthetic and analytic grammatical markers. They do not, however, have many simplifying features which reduce irregularity. Despite that, however, the focus of the complexity measures used in the Handbook of Varieties of English is clearly morphosyntax (cf. Kortmann and Szmrecsanyi 2009b: 270). Kortmann and Szmrecsanyi (2009b), hence do not address the notion of hierarchical complexity used in the present study to rank NP structures (cf. 5.3).

30

2 New Varieties of English

2.2.2.3 Simplification and evolutionary stage The question of the correlation of simplificatory tendency with evolutionary stages, which plays an important role in the present study due to the differences in developmental phase between Singaporean English and Kenyan English, has received little attention. In the literature reviewed, the only comprehensive study to have addressed this question explicitly is Mesthrie’s (1992) sociolinguistic study of Indian South African English (IndSAfE), which is based on interviews with 150 speakers from the Natal province in South Africa. In his sample, Mesthrie (1992: 65) identifies a “polylectal continuum” within IndSAfE. He finds “pre-basilectal” speakers, who have a very basic, “makeshift” (Mesthrie 1992: 65) command of English and struggle “expressing themselves even about domestic topics”. Basilectal speakers know “core or skeletal features of language structure” (Mesthrie and Bhatt 2008: 187) including e.g. relative clauses, syntactic devices for topicalisation, prepositions, and a tense system (cf. Mesthrie 1992: 191). Mesolectal speakers, in turn, have a more strongly explicit syntactic framework at their disposal, which involves e.g. subordinators or question words. “Acrolectal speakers”, finally, can approximate Standard English in formal situations (cf. Mesthrie 1992: 46). Crucially, Mesthrie does not treat these stages of competence as synchronic facts about IndSAfE but, under the apparent time hypothesis (cf. Milroy and Gordon 2003: 35), argues that the four lects can be viewed as developmental stages of IndSAfE, akin to the developmental stages assumed in the Dynamic Model. His study of IndSAfE allows a plausible conclusion. New varieties of English which progress from an early to a later developmental stage are marked by a rise in linguistic complexity. In a similar vein, Steger and Schneider (2012: 183) (cf. 2.2.2.1) find a lower frequency of (non-iconic, and thus complex) raising constructions after verbs like allow and want in Kenyan, Indian and Hong Kong English, all of which are in phase 3, while the rate found for Singaporean English, which is in phase 4, is on a par with British English. They attribute this effect to more “recent acquisition” (Steger and Schneider 2012: 188) in the less advanced variety. In sum, although rarely addressed, there are clear indications that a gain in evolutionary stage in New Englishes leads to a rise in complexity.

2.3 Summary: New varieties of English In the following, the main implications of this chapter’s discussion for the present study will be collated. The discussion of language contact in New Englishes (cf. 2.2.1) allows the following conclusions. First, with language contact having

2.3 Summary: New varieties of English

31

been shown to be detectable at a variety of levels of linguistic structure, it is assumed that it will also affect preferences for head-initial or head-final NP structures in Singaporean English and Kenyan English (cf. 4). The exact envelope of variation will be explicated below (cf. 5.2). Second, there are clear indications that language contact should not be invoked as a blanket explanation. The high degree of similarity across New Englishes suggests that a sizeable number of innovations must be due to universal processes of change. The present study has to make sure that cross-linguistic influence and universal factors can be teased apart. For this reason, the following chapters will develop a set of variables which allow a reliable distinction between effects of language contact and universal simplificatory factors (cf. 5.4) and, in addition to that, adopt a comparative methodology by analysing the same structure in several new varieties of English with different substrates (cf. Gut 2011: 117), in line with Kortmann and Szmrecsanyi’s (2009: 31) call for studies “combin[ing] careful, intralingual-philological, variationist analysis with the broad, abstractive bird’s eye perspective that is the hallmark of language typology”. Finally, the study will describe in detail the distinct language-typological background of Singaporean English and Kenyan English (cf. 4). Fourth, in studies of cross-linguistic influence, the exact workings of the process of language contact are hardly ever spelt out in detail. The results by Williams (1987), however, have shown that many innovations in New Englishes are akin to effects from SLA (cf. 2.1.5). A plausible explanation of this phenomenon must explain language contact in the idiolects of individual speakers and account for its translation to the level of the speech community (cf. Siegel 2008: 105). This has two ramifications with respect to the present study. Assumptions about language contact phenomena in New Englishes should be complemented by results from the study of SLA, which can help to assess the potential for cross-linguistic influence on NPs in individual learners (cf. 3). Moreover, an indispensable part of this study will be a theory of language change which models the relationship between language contact among individual language learners and its effect on the communal linguistic norm. Such a framework will be suggested in 3. With regard to simplification in New Englishes (cf. 2.2.2), the following conclusions are in order. First, New Englishes have been shown to be simpler than native varieties in a variety of respects, including phonology, morphosyntax at large and specific syntactic constructions. Given this strong degree of evidence, it is more than likely that structures like NPs, too, are susceptible to simplification in New Englishes. Second, the studies cited above have failed to provide a coherent theory of complexity. What is more, the majority of studies on complexity in New Englishes have dealt with broadly morphosyntactic phenomena, which are of limited use with regard to the complexity of NPs. What is needed, therefore is

32

2 New Varieties of English

a plausible, well-founded hierarchical theory of complexity, which allows for the analysis of structures like NPs (cf. 5.3). Third, there are indications that a variety gains in complexity as it moves along the evolutionary cycle of the Dynamic Model. The corpus analysis in 7 will deal with this question, which will be substantiated on a theoretical basis in the chapter on language contact (cf. 3). Fourth, like language contact, simplification needs to be explained as a process starting out in individuals, which subsequently translates to the level of the speech community. This, again, is a reason for the use of a fully-fledged theory of language change, along with evidence from the study of SLA, which provides the crucial individual perspective on this process of language change (cf. 3).

3 Modelling language change in New Englishes It is important to note that calling a linguistic phenomenon found in a New English variety e.g. “simplification” or “borrowing” (cf. 2.1.4) is shorthand for a highly complex statement. Any linguistic innovation in such varieties needs to be viewed as the effect of an intricate diachronic process involving both the idiolects of individuals and the dynamics of the whole speech community. What needs to be taken into account for the sake of a comprehensive account of a language variety is both the “psycholinguistic process […] in individuals” (Siegel 2008: 105) bringing about the innovation in question and the principles governing its spread to the communal linguistic norm (cf. 2.2.2.1). In order to account for these two processes, I will first collate core findings on L1 transfer and simplification gained in the study of Second Language Acquisition (SLA). Subsequently, the language use of individuals will be related to the linguistic norm by invoking a model of language change for New Englishes.

3.1 Language contact and simplification in the field of SLA The aim of this section is to describe the processes of language contact and simplification which have to be expected to play a role at the level of individuals using English as a second language (cf. 2.1.5). This area has been dealt with extensively in SLA research. In order to narrow down the sizeable body of research available on these topics, I will lay particular emphasis on the effects of L1 influence and simplification in the field of syntax, and, above all, NP structures, while disregarding the many results in other levels of linguistic structure. In the following, L1 influence and simplification will be dealt with in turn. 3.1.1 L1 transfer in SLA In the field of SLA, the notion of transfer refers to “the influence resulting from similarities and differences between the target language and any other language that has been previously (and perhaps imperfectly) acquired” (Odlin 1989: 27). The concept of transfer has been called “the most discussed theoretical construct in L2 acquisition research”(Schwartz 1999: 211), as it has always been intimately related to the linguistic theory in vogue at the time. It would go beyond the domain of this study to cover the relevant debates in detail. Suffice it to say that, broadly speaking, the early days of SLA research were under the influence of the Contrastive Analysis Hypothesis (cf. Lado 1957: 2), i.e. the assumption that the best

34

3 Modelling language change in New Englishes

predictor of transfer us the similarity or dissimilarity of the two languages in question. This led to a tendency to overrate the role of the L1 in transfer. As of (roughly) the 1970s, the pendulum swung the other way, with many researchers marginalising the influence of the first language. L2 learning was likened to L1 learning, and L1 influence tended to be explained away as a temporary “production strategy” rather than a learning strategy leading to “real” progress (Krashen 1983: 142). In today’s SLA, there is a more balanced middle-of-the road view in between the two extremes (cf. Ellis 2008: 364), based on the assumption that neither the L1 nor universal tendencies alone affect SLA, but that “transfer — in some mentalist instantiation — works together with universal processes” (Schwartz 1999: 213; cf. Ellis 2008: 397–400 for a comprehensive overview). 3.1.1.1 Effects of L1 transfer in SLA According to an overview compiled by Gut (2011: 105), L1 transfer can have the following effects. (a) direct lexical borrowings (b) nonstructural, conceptual transfer at levels such as information packaging (c) hypercorrection (d) a general facilitating effect in language learning on the basis of the knowledge of typologically similar languages (e) avoidance patterns (f) preference patterns for particular structures known from the L1 (g) the blending of L1 and L2 structures All of these effects have been documented amply in the literature (cf. e.g. Færch and Kasper 1986: 59 for direct lexical borrowing; Carroll et al. 2000: 459–60 on conceptual transfer in the description of spatial scenes by speakers of English, German and French; Ibrahim 1978: 209 for hypercorrection in the spelling of Arabic learners; Ringbom 1978 on Finnish and Swedish learners of English). However, none of these effects can be mapped onto NP usage in the present study. It is factors (e)–(g) which are of relevance for the corpus analysis. Below, relevant NP-related results with regard to these three effects will be summarised briefly. Avoidance patterns — Schachter (1974) looks into relative clause formation in English essays written by Iranian, Arab, Chinese, Japanese and American students (cf. Schachter 1974: 207). Table 3.1 shows the numbers of correct and incorrect relative clauses produced by her test subjects. Taken at face value, these data suggest that Persian and Arab students are more error-prone than Chinese and Japanese ones, as their error rates are considerably higher. Rather than dwelling on these

3.1 Language contact and simplification in the field of SLA

35

Table 3.1. Error rates in relative clauses among learners of English with different L1 backgrounds (Schachter 1974: 209)

Persian Arab Chinese Japanese American

Correct

Error

Total

% errors

131 123 67 58 173

43 31 9 5 0

174 154 76 63 173

25 20 12 8 0

error rates, however, Schachter (1974) directs the reader’s attention to a different phenomenon. If one disregards the error rates in table 3.1 and focuses on the overall raw frequencies of relative clauses by speakers of the two languages, it emerges that the overall frequencies of relative clauses in the production of speakers of Persian and Arab is a great deal higher than among speakers of Japanese and Chinese. The reason for this avoidance pattern of relative clauses, Schachter suggests, is most likely to be L1 influence: In both Chinese and Japanese, relative clauses occur in prenominal position, which leads to an aversion to postnominal ones in English (cf. 4.1.2.2.2). Numerous follow-up studies to Schachter (1974) have buttressed the existence of avoidance phenomena in SLA (cf. Kleinmann 1977: 96 for the English passive, the present progressive or non-finite verb complementation among Arabic, Spanish and Portuguese learners of English; Seliger 1989: 31 for the avoidance of passives by speakers of Hebrew; Dagut and Laufer 1985: 77 and Laufer and Eliasson 1993 for phrasal verbs). Preference patterns — L1 influence may likewise lead to the over-production of syntactic structures in the L2. Schachter and W. Rutherford (1979), for instance, look into transfer phenomena in discourse organisation, finding an overuse of such it-extrapositions as in (3.1) in the L2 English of Japanese learners and a preference for there-existential constructions as in (3.2) in the English production of Chinese learners. (3.1) (3.2)

It is believed that sweet flag leaves contain the power to expel sickness and evil. There is a small restaurant near my house in my country. (Schachter and W. Rutherford 1979: 4)

This preference can be traced directly to Chinese and Japanese clause-level topiccomment structures, which, in English, can be glossed as Airplanes, the 747 is big

36

3 Modelling language change in New Englishes

(cf. Schachter and W. Rutherford 1979: 6–7). Likewise, L. Collins (2002: 53) reports L1-induced preference patterns in the use of the English present perfect among French learners¹. Blending of L1 and L2 structures — Another phenomenon hailing from transfer which can be found in L2 syntax is the structural blending of L1 and L2 structures. Matthews and Yip (2003), for instance, find striking prenominal relative clause constructions in the English of bilingual children in Hong Kong, which are modelled on prenominal relative clauses in Cantonese (cf. 4.1.2.2.2); examples are shown in (3.3)–(3.4). (3.3) (3.4)

You buy the motorbike ‘the motorbike that you bought’ (uttered by a child at the age of 2;07) Timmy take that one, I want. ‘I want the one taken by Timmy.’ (produced by a child aged 3;03)

Although such mixed structures, which clearly violate English syntactic norms, are likely to disappear soon in these children’s linguistic development, it is highly interesting to see that bilingual children acquiring two languages as dissimilar as English and Cantonese produce such mixed structures in the early phase of language acquisition. It is, in principle, conceivable that children in Singapore, who grow up in a very similar multilingual situation, will show such L1 effects (cf. 4.1.2.2). Klein and Perdue’s (1992, 1993) study of language acquisition among European immigrants likewise provides clear evidence for the use of NP structures modelled directly on the speaker’s L1. Perdue (1993: 49–52) looks into the use of N+N compounds in the L2 Dutch of Arabic and Turkish immigrants to the Netherlands. Two Moroccan speakers of L1 Arabic regularly use the head-initial prepositional constructions shown in (3.5)–(3.6)(a) rather than the usual Dutch N+N constructions shown in (3.5)–(3.6)(b). (3.5)

(a) kleren

van baby clothes of baby

‘baby clothes’ (b) babykleren ‘baby clothes’

1 The overproduction of a given structure may also result from the avoidance of another structure. The lack of relative clauses in Schachter’s (1974) study, for instance, entails an overuse of main clauses (cf. Odlin 1989: 37).

3.1 Language contact and simplification in the field of SLA

(3.6)

37

(a) auto van police

car

of

police

‘police car’ (b) politie-auto ‘police car’

As this preference is not detectable among the Turkish learners at hand, it is ascribed to a “clear source language effect” (Perdue 1993: 51). In addition, as shown in table 3.2 (p. 38), L1-induced preference for either head-initial or head-final NP structures can be replicated in all the immigrant groups under scrutiny, covering three head-initial L1s (Arabic, Italian and Spanish) and three head-final L1s (Turkish, Punjabi and Spanish; cf. Perdue 1993: 55; the logit values in table 3.2 are zero whenever the frequencies of head-final and head-initial constructions are equal; they are positive if head-final constructions are more frequent, while they turn out to be negative if head-initial constructions prevail). Both avoidance patterns, preference patterns and structural blending clearly play a role in SLA and, for this reason, can also be expected to play a role in the usage of speakers of New Englishes. In addition, the studies cited document that NP structures are affected strongly by all three processes.

3.1.1.2 Constraints on transfer in SLA In the SLA literature, a number of constraints have been made out which restrict the degree of L1 transfer. There is, for one thing, a set of non-structural, subjective factors: Both personality, the learners’ individual aptitude for language acquisition, their motivation, the specific type of task or linguistic challenge a learner has to face and the quality and type of language instruction are known to have an impact on transfer (cf. Ellis 2008: 379 Odlin 1989: 129–50). In a corpus-based study, however, these factors can hardly be controlled for. Another factor which has been found to play a major role is the social context of interactions. ‘Focused’ situations, i.e. situations in which speakers are acutely aware of “linguistic and other norms that distinguish one group from another” (Odlin 1989: 144), encourage the use of standard forms, and hence prevent negative transfer (cf. Odlin 1989: 148; Ellis 2008: 380–1). Transfer is more likely to occur in unfocused, informal situations (cf. Odlin 1990: 147–8). This factor is relevant to the present study as the corpus analysis below relies on the analysis of two registers, both a very informal and a very formal one (cf. 6.2.1). What is even more relevant from the point of view of the present study, though, are three further factors.

cycle 1 cycle 2 cycle 3 mean value

0.571 0.302 0.000 0.291

FA

Learner 1.609 -0.167 1.099 0.847

MO

Arabic (head-initial) ER 2.132 2.615 2.615 2.454

MA

Turkish (head-final)

2.197 3.219 1.609 2.341

Dutch

L1 NPs in L1

L2

2.197 0.336 0.511 1.015

AN 1.946 1.466 2.944 2.119

LA

Italian (head-initial)

3.434 2.944 3.661 3.346

MA 0.000 2.833 2.565 2.699

RA

Punjabi (head-final)

English

0.847 3.045 2.197 2.030

MA

1.946 2.565 1.609 2.040

LE

Finnish (head-final)

FE 0.435 0.000 1.609 0.681

NO -0.847 1.526 1.526 0.735

Spanish (head-initial)

Swedish

Table 3.2. Logit values of the numbers of head-final vs. the numbers of head-initial compounds, from Perdue (1993: 55)

38 3 Modelling language change in New Englishes

3.1 Language contact and simplification in the field of SLA

39

The level of linguistic organisation — There have been claims that basic clause patterns are non-transferable, including the positions of subject, verb and object or, indeed, the order of modifier and noun. W. E. Rutherford (1983), for instance, fails to detect L1 transfer of basic syntactic features in essays produced by Chinese, Spanish, Japanese, Korean and Arabic learners of English, while detecting numerous cases of L1-induced topicalisations or dummy subjects (cf. W. E. Rutherford 1983: 367). This leads him to conclude that basic word order is all but “untransferable” (W. E. Rutherford 1983: 368; cf. Zobl (1986) for a similar argumentation), while “it is […] discourse and not syntax that gives gross overall shape to interlanguage”. The evidence against such claims of non-transferability of basic word order, however, is overwhelming. Apart from the fact that from a usage-based point of view, the formalist claims made by both authors are questionable in their very premises (cf. 1.3.2), there is a wide variety of language contact scenarios in which basic word order transfer must have taken place (cf. Odlin 1990: 98–9). Bamboo English, which has originated from language contact between Korean (which is of the SOV type) and English (which is SVO), has retained the SOV order from Korean; Hawaiian Pidgin English is SOV, which must be viewed as a retention of the SOV pattern in the substrates Japanese and Korean; Dyirbal, having OAV/AOV word order, is known to have developed AVO word order in its phase of attrition during the 20th century through contact with English SVO. The claim that basicword order transfer is principally impossible, thus, must be viewed as an overstatement. The reasons for the dearth of evidence for transfer of such features in the literature may be the lack of data taken from the earliest stages of language acquisition, where L1 transfer is strongest (cf. Odlin 1990: 109). In addition, Siegel (2008: 114) suggests that due to the salience of word order, learners may have a particularly high metalinguistic awareness for it, which prevents L1 transfer (cf. Odlin 1990: 109–10) on a large scale, without, however, excluding it categorically. Consequently, in this study, it will be assumed that language contact in basic NP syntax is possible. The the similarity of the languages in question has likewise been claimed to affect the degree of transfer. This issue carries particular relevance in the context of New Englishes, where the languages in contact tend to be extremely distant from a typological point of view. On the one hand, it stands to reason that the structural and typological similarity of languages raises the chances for language transfer. The studies by Ringbom (1978, 1992), for instance, give ample evidence that learners with Swedish as L1 have the edge in learning English over learners with a Finnish background due to the possibility of positive transfer from Swedish to English (cf. Ringbom 1992: 91). From this point of view, the chances for language

40

3 Modelling language change in New Englishes

contact in New Englishes would not seem to be overly high, as the indigenous languages of the respective countries tend to be strongly different from English². However, this assumption cannot be upheld without reference to actual linguistic structures and, above all, to learners themselves. Kellerman (1979, 1983) suggests a model of transfer which is based on “psychotypology”, i.e. the “learner’s perception of language distance”(Kellerman 1983: 114) rather than actual linguistic distance. A speaker’s psychotypology of a given language depends both on the speaker’s estimation of the degree of language-specificity of a feature in question, and on the perceived (rather than the actual) typological closeness of the two languages involved. The relationship between these two dimensions is illustrated in figure 3.1. Close • • • • Specific • Neutral

• • • •

• • •

• •

Distant •

Fig. 3.1. Gass and Selinker’s (2008: 149) visualisation of Kellerman’s (1979, 1983) psychotypology hypothesis (slightly adapted); the amount of dots symbolises the probability of transfer

According to Kellerman (1979, 1983), learners will rank linguistic items on a scale from language-neutral to language-specific. Core grammatical structures, basic writing conventions or “internationalisms” like suffixations in -ion will be perceived as language-neutral, while anything idiosyncratic, such as idiomatic expressions, proverbs, as well as e.g. inflectional morphology or phonology will be deemed high in specificity (cf. Kellerman 1977: 107, Kellerman 1979: 40–1). As can be seen in figure 3.1, the probability of transfer is inversely proportional to both variables. An idiom like kick the bucket, which will be viewed as highly languagespecific, will only be transferred in languages considered reasonably close to each other, such as, for instance, English and Dutch. A strongly neutral item such as the syntax of a default sentence like the ball rolled down the hill, in contrast, has

2 It would be wrong to conclude that there is always a positive correlation between structural similarity and cross-linguistic transfer. On the contrary, the formal and semantic similarity of structures between languages may “manifest itself in so-called idiomatic disbelief” (Laufer and Eliasson 1993: 39), which, for instance, leads Dutch learners of English to avoid phrasal verbs English phrasal verbs which have precise matches in Dutch (cf. Hulstijn and Marchena 1989: 248).

3.1 Language contact and simplification in the field of SLA

41

a higher probability of being transferred, even in languages gauged as distant (cf. Gass and Selinker 2008: 147). Kellerman’s (1979, 1983) psychotypology hypothesis, thus, suggests that the basic syntax of NPs, which speakers will arguably deem rather unspecific to their language, as it is part of basic syntax, may be subject to contact influence even if the languages involved are typologically unrelated and strongly dissimilar, as is the case with New Englishes (cf. Gass and Selinker 2008: 144–51 for this line of argumentation). The relationship between L2 skills and L1 transfer has been subject to a great deal of controversy in SLA research, which is relevant in the present context due to the fact that during the emergence of New Englishes, competence levels can be expected to rise (cf. 2.2.2.3). The discussion in the literature boils down to two main positions. On the one hand, there is an agreement that early learners are particularly prone to transfer (cf. Gut 2011: 111–2; Taylor 1975: 83 for relevant results). Other authors, in contrast, have argued that “the possibilities of transfer increase as knowledge of the second language increases” (Klein 1986: 27), since a certain proficiency level is indispensable for a speaker to recognise similarities between languages (cf. e.g. Chan 2004). Consequently, it can be assumed that while early learners are particularly likely to carry over L1 structure into their L2 usage (cf. also Gut 2011: 111–2), the degree of L1 influence neither decreases linearly, nor subsides altogether among highly proficient learners (cf. Ellis 2008: 294)³. With regard to New Englishes, this means that even high skill levels of English, which can be expected among speakers of advanced new varieties such as Singaporean English, do not preclude language contact.

3.1.2 Simplification in SLA In addition to transfer, SLA is also well-known to bring about the simplification of the structure of the target language, with the typical “first victim of reduction” (Winford 2003: 218) being inflectional morphology. Studies on the inflectional simplicity of interlanguage abound (cf. e.g. Fathman 1977: 34; cf. Ellis 2008: 80–2 for a synopsis). Due to the focus of the present study, however, the emphasis will be on the simplification of syntactic structures. In the following, major empirical results on simplification in SLA will be complemented by a brief discussion of theoretical debates. 3 I will not deal with the wide range of studies on the relationship between age/exposure and phonological acquisition; a useful overview is given by Odlin (1989: 137–8); cf. also Jiemin (2012).

42

3 Modelling language change in New Englishes

3.1.2.1 Structural/syntactic simplification — exemplary studies Out of the wide range of inquiries into simplification in SLA, the following three stand out as particularly relevant to the present study, as they address either NP structures or aspects of hierarchical complexity (cf. 5.3.2.1). Cooper (1976) analyses syntactic complexity in student essays, tapping into the production of 40 college students enrolled in German classes at Florida State University at five skill levels and a control group of professional native German writers (cf. table 3.3; Cooper 1976: 178). Having segmented his corpus into T-units, i.e. “one main clause plus whatever subordinate clauses happen to be attached to or embedded within it” (Hunt 1970: 196), Cooper computes a set of indices shown in table 3.3, including the proportion of subordinate clauses per T-unit (“subordinate clause ratio”), the average length of T-units, the number of T-units per orthographic sentence (“coordination ratio”) as well as the average length of clauses and sentences. The most relevant of his metrics with regard to the present study is the frequency of “complex nominals”; complex nominals are defined as NPs involving adjective phrases, possessives, NPs modified by prepositional phrases, relative clauses, participles and appositions, nominal clauses and gerunds (cf. (3.7)–(3.10))⁴. (3.7) (3.8) (3.9) (3.10)

der grosse [sic] Hund der Mann, der ein Bier trank der Wagen unter dem Baum Er denkt, er hat den Dieb gesehen. (Cooper 1976: 180)

It does not come as a surprise that all of the indices are consistently (and, in fact, significantly, cf. Cooper 1976: 181) lowest for Sophomores, and the majority of them increase steadily across the skill levels (with the exception of the coordination ratio). A gain in language skills goes hand in hand with an increase of linguistic complexity. More advanced users of German use longer syntactic units, more subordination and more coordination, and, crucially, more phrasal modification as they gain in proficiency.

4 Cooper’s (1976) notion of “complex nominal” (which is rooted in a formalist understanding of language) is wider than the idea of a complex NP used in the present study because it encompasses clauses in nominal function as well as ing-forms.

3.1 Language contact and simplification in the field of SLA

43

Table 3.3. Mean values of syntactic complexity measures (cf. Cooper 1976: 178)

Sophomores Juniors Seniors Graduate Students Native Germans

Average clause length

Subordi nation ratio

Average T-Unit length

Coordi nation ratio

Average sentence length

Average freq. of compl. nominals

7.3 7.4 8.5 9.9 10.7

1.2 1.4 1.5 1.4 1.7

8.7 10.3 12.5 14.0 18.4

1.2 1.2 1.2 1.2 1.3

10.3 12.6 15.2 16.9 23.0

8.1 15.5 18.9 26.2 32.6

Larsen-Freeman (1983) looks into the oral production of 25 L2 university students studying English as a second language, who hail from a variety of L1 backgrounds (cf. Larsen-Freeman 1983: 289). Having assigned them to one of five competence levels, she investigates the complexity of their utterances on the basis of their retelling of a picture-based narrative. Her complexity measures include the total number of error-free T-units, the average number of words per T-unit, the average number of words per error-free T-unit as well as the percentage of error-free T-units. As it turns out, there are strong statistical correlations between the speakers’ competence levels and all of these metrics (cf. Larsen-Freeman 1983: 291). Although not dealing with NPs, thus, she confirms that an overall gain in language skills is accompanied by a rise in hierarchical syntactic complexity. Klein and Perdue (1992, 1997) — One of the most comprehensive studies on complexity reduction in SLA is Klein and Perdue’s (1992) large-scale study of language acquisition among immigrant workers in Europe. Their 40 test subjects, who reside in Great Britain, Germany, the Netherlands, France and Sweden, come from a variety of linguistic backgrounds and acquire the respective national languages at a very basic level, the so-called “Basic Variety” (cf. 3.1.1.1). This stage of acquisition, a largely homogeneous, highly effective means of communication, is marked by utter simplicity. There is a complete lack of “free or bound morphemes with purely grammatical function” as well as “complex hierarchical structures” (Klein and Perdue 1997: 332), with the NPs found merely involving “some adjectives”, “a few prepositions” and “no complementizers” (Klein and Perdue 1997: 312) as can be seen in the examples taken from Klein and Perdue’s (1997) corpus (cf. (3.11)–(3.13)).

44

3 Modelling language change in New Englishes

(3.11) stealing bread girl (Klein and Perdue 1992: 310) (3.12) Charlie give present for young children (Klein and Perdue 1992: 315) (3.13) after comeback the brigade fire (Klein and Perdue 1992: 328) The study, thus, likewise confirms the intimate relationship between skill level and grammatical/hierarchical complexity. It can be assumed that in the earliest stages of new English varieties, language skills akin to the “Basic Variety” must have played a role.

3.1.2.2 Debating the notion of complexity in SLA In the literature on SLA, a debate has recently been sparked about an asymmetry in the use of different types of linguistic variables for the purpose of operationalising syntactic complexity, which is highly relevant to the present study. On the basis of a meta-analysis involving 27 studies of syntactic complexity in interlanguage, Ortega (2003: 497) has argued that all too many of these studies have relied on length measures of syntactic units, such as the length of T-units, clauses and sentences used by Cooper (1976: 179), or Larsen-Freeman’s (1983: 291) number of words per T-unit/error-free T-unit mentioned in 3.1.2.1, and measures of the degree of subordination, such as the proportion of subordinate clauses per T-unit (cf. Cooper 1976: 179). To Ortega’s (2012: 139) mind, however, there is a dearth of studies which operationalise syntactic complexity in terms of the frequency of “linguistically sophisticated” syntactic forms, which are represented e.g. by Cooper’s (1976: 176) coordination ratio, the complex NPs mentioned in 3.1.2.1 or Larsen-Freeman’s (1983: 291) number of error-free T-units and percentage of errorfree T-units mentioned in 3.1.2.1). Interestingly, recent studies have made the point that it is, in particular, NPs which should be taken into account in this context. Lu (2010: 491), for instance, emphasises the importance of the frequency of complex NPs (i.e. NPs modified by adjectives, possessives, PPs or any clausal structure) for the assessment of L2 competence in a meta-analysis of studies of syntactic complexity in SLA writing. In his sample, their frequency rises proportionally to language skill levels from 1.415 via 1.453 to 1.651 (𝑝 = 0.000) (cf. Lu 2010: 490), which leads him to the conclusion that… [s]tudents at higher proficiency levels tend to produce longer clauses and T-units, not as a result of increased use of dependent clauses or complex T-units, but as a result of increased use of complex phrases and complex nominals. For ESL writing instruction, this observation points to the importance of helping students engage with complexity more at the phrasal level and less at the clausal level as they advance to higher levels of proficiency. (Lu 2010: 491)

3.1 Language contact and simplification in the field of SLA

45

Likewise, Ortega (2012: 139) compares two texts composed by an L2 learner of Spanish at a US university at an interval of six weeks, finding that one of the most striking differences between the two texts, which can be correlated to a gain in linguistic competence, is a rise in the complexity of NP modification patterns. The type of complex NP shown in (3.14)–(3.15), which have been taken from the text written at the later point in time, are strikingly absent from the texts written in the earlier phase. (3.14) imaginando un mundo japonés

imagining lo mío the mine

a

world

completamente diferente de Japanese totally different from

‘imagining a Japanese world completely different from mine’ (Ortega 2012: 138) (3.15) con envolturas hecho [sic] de arroz with wrappers made of rice ‘with wrappers made of rice’ (Ortega 2012: 138) Ortega (2012: 138) concludes that “[…] complexification strategies other than subordination can be important resources for writers, for example, phrasal modification”. This allows the conclusion that in addition to the length- and subordination-based complexity measures, gains in NP complexity play a bigger part in the development of L2 skills than implied by much of the literature on SLA. The analysis of complexity differences of New English varieties on the basis of NP modification patterns, thus, is in keeping with these insights.

3.1.3 Summary: Language contact and simplification in the field of SLA In this section, the phenomena of language contact and simplification have been analysed from the point of view of individual L2 acquisition. The upshot from this discussion of SLA results for the study of NPs in New Englishes will be summarised in the following. – With both language contact and simplification having been shown convincingly to play a role in individual SLA, it can be assumed safely that present-day speakers of New Englishes, too, will make use of these strategies to some extent as long as they use English as an L2 rather than an L1 (cf. 2.1.5). In addition, processes of SLA will have played a role in the early phases of the emergence of New Englishes; these are particularly formative stages of development which are supposed to have ramifications to the present day.

46

3 Modelling language change in New Englishes

– Furthermore, in the field of SLA, a whole range of studies have found NP syntax to be susceptible to both L1 influence and simplification, which reinforces the basic hypotheses of the present study. – Syntactic transfer should be possible even between highly dissimilar languages if the structures in question are as general and unspecific to particular languages as NPs. The substrates in postcolonial contexts, thus, can be assumed to have had an impact on English in this area of variation despite their typological distance from English. – From the point of view of SLA, it seems plausible for speakers of highly advanced New Englishes to use more complex structures than speakers of less developed varieties, as there are clear correlations between simplificatory tendencies and skill levels in English.

3.2 An evolutionary model of language change for New Englishes In the previous section, it has been demonstrated that the phenomena of L1 transfer and simplification are widely attested in L2 production and can be considered typical effects of individual language acquisition (cf. 3.1). Processes occurring at the level of individual speakers, however, fail to provide a sufficient explanation for the features of present-day varieties of New Englishes. What is needed for a full understanding of the emergence of New Englishes is an account of the processes by which SLA-induced features spread from individual speakers to the community and become part of the communal linguistic norm. The objective of this section is the development of an evolutionary model of language change which mediates between individual language use and large-scale language change. As for its basic premises, it will be based on a model devised by Croft (2000), which uses concepts adapted from evolutionary theory to account for language change. It will elaborate considerably on the basic account of the evolution of New Englishes given in 2.1.3 by invoking a fully-fledged, rigorous theory of language change. I will lay the groundwork for Croft’s (2000) theory by briefly sketching the general tenets of an evolutionary approach to language change. In what follows, I will present Croft’s theory and briefly review his assumptions. Then, I will outline the key processes of language change which have led to the emergence of New Englishes in accordance with Croft’s theory.

3.2 An evolutionary model of language change for New Englishes

47

3.2.1 Evolutionary approaches to language change The use of evolutionary ideas in linguistics, although dating back to the 19th century (cf. Sampson 1994: 21), has started in earnest since the 1980s (cf. Rosenbach 2008: 25). Evolutionary concepts have been employed, on the one hand, in order to account for the emergence of languages at large, e.g. with a focus on linguistic diversity (cf. Dixon 1997) or grammatical universals (cf. Kirby 1999). What is most relevant from the point of view of the present study, however, are approaches which apply evolutionary theory to phenomena of language change, e.g. in the approaches championed by Keller (1994), Lass (1997: 370–90), Haspelmath (1999), Mufwene (2001, 2008) or Croft (2000). Croft (2000: 10–2) (along with Lass 1990) views biological evolution and language change as two instantiations of the same general process of evolution, which can be defined by the following criteria: (a) variation, i.e. the “continuing abundance of different elements” (Dennett 1995: 343); (b) selection or “differential fitness” (Dennett 1995: 343), depending on the interaction with the environment in question; (c) replication, i.e. the capability of the elements in question to “create copies or replicas of themselves” (Dennett 1995: 343). These factors are to be found in biological evolution and language change alike. While represented by mutations, procreation/reproduction and the differential survival rates of organisms in biology, they feature in language in the form of linguistic variation, the transmission of language from generation to generation and in the differential chances of linguistic innovations to spread in a given speech community (cf. Rosenbach 2008: 24; 26; 32).

3.2.2 The key assumptions in Croft’s theory of language change 3.2.2.1 Theoretical prerequisites There are two aspects of Croft’s understanding of language and language change which are worth noting at the outset. First, his theory is distinctly cognitivelinguistic and usage-based; second, he views language as a “fundamentally social phenomenon” (Croft 2000: 87). A cognitive, utterance-based theory of change — In line with the stance taken in the present study (cf. 1.3.2.2), Croft’s theory is distinctly cognitive and usage-based in assuming that language structure emerges from language use (cf. 1.3.2.2). As a corollary of this, he views the basic linguistic units undergoing change as utterances or “linguemes” (Croft 2000: 29), which are defined as “a particular, actual occurrence of the product of human behavior in communicative interaction […], as

48

3 Modelling language change in New Englishes

it is pronounced, grammatically structured, and semantically and pragmatically interpreted in its context” (Croft 2000: 26; cf. also Croft and Cruse 2007: 291–327). Linguemes, which are coextensive with the notion of constructions in construction grammar (cf. 1.3.2.2), may be “schematic”, i.e. represented by structural linguistic templates which can be filled variably (e.g. constituent orders like SVO or SOV, or NP structures such as N-Gen or Adj-N), or “substantive”, i.e. having the shape of lexically filled expressions (such as kick the bucket; cf. Croft 2000: 201; Fillmore, Kay, and O’Connor 1988: 505). This aspect renders Croft’s (2000) theory particularly helpful in the present context, as it allows for the modelling of cross-linguistic transfer of abstract syntactic structures. As another consequence of the usage-based nature of his approach, Croft stresses that rather than undergoing change due to self-propelled internal processes, languages change by virtue of speakers changing them. This sets his framework apart from earlier child-based theories of language change, which put language change down to a large extent to imperfect acquisition by children (cf. Paul 1886: 31). A thoroughly social theory of language change — Croft explicitly takes on board modern sociolinguistics by viewing language as a “fundamentally social phenomenon” (Croft 2000: 87). This leads him to posit an intimate relationship between linguistic codes and their associated domains and speech communities (cf. Croft 2000: 92): “[A] linguistic code belongs to a (speech) community, and a community is defined by a domain”. As a corollary, with codes carrying social meaning (cf. Croft 2000: 91), every lingueme is strictly defined as a “threecornered sign-in-a-community” (Croft 2000: 92), i.e. with regard to its form, its meaning and in its relation to speech communities, individual social domains and the codes associated with them. From this point of view, even the choice between the linguemes needed in one language e.g. for informal conversation or formal academic essays, is a social one, by virtue of the two codes belonging to different domains or sub-domains, and their association with distinct overlapping speech communities. The goal of reaching social success (which, according to Croft (2000: 88), is the prime goal of language at large), is tied to the knowledge of the convention of domain-specific codes.

3.2.2.2 The components of Croft’s (2000) evolutionary theory Croft (2000) capitalises on Hull’s (1988: 397–431) generalised theory of evolution, which provides a generalised evolutionary underpinning for any phenomenon which can be described in evolutionary terms (cf. 3.2.1). Hull’s theory revolves around the notion of a “replicator”, which is defined as “an entity that passes on

3.2 An evolutionary model of language change for New Englishes

49

its structure largely intact in successive replications” (Hull 1988: 408). Depending on the field of inquiry, replicators may be genes in biology, which are passed on from organism to organism, or, in science, elements of scientific thinking such as “beliefs about the goals of science, proper ways about realizing these goals, […] [or] accumulated data” (Hull 1988: 434). In any of these scenarios, there are three distinct ways in which replicators can be passed on in the process of evolution (cf. Croft 2006: 94). Normal replication is the process of creating an identical copy of a replicator. Altered replication, or “innovation”, refers to the change of individual replicators under replication. Differential replication, or “propagation”, refers to proportional changes in the overall set of replicators. In the course of many copying processes, certain replicators are preferred over others, which leads to a change in their overall proportions. As these stages of the evolutionary process characterise any kind of evolutionary phenomenon (cf. Croft 2006: 96), they lend themselves to an adaptation for the purpose of capturing language change. For Croft (2000), the equivalents of Hull’s (1988) replicators are linguistic utterances, or “linguemes” (cf. 3.2.2.1). In the following, it will be laid out how linguemes can be expected to undergo the three evolutionary processes of normal, altered and differential replication.

3.2.2.2.1 Normal replication in language evolution Normal replication in language evolution is defined as “conformity to convention” (Croft 2000: 31) — the use of utterances fully in line with the linguistic norms valid in the speech community in question. Normal replication is the component of Croft’s theory which allows for the diachronic stability of languages. The tendency of speakers to conform to the convention is attributed to two distinct motivations. First, speakers generally pursue the overarching aim of being understood by their interlocutors, which is most likely to succeed when a given utterance conforms to the linguistic convention known in the speech community (cf. 3.2.2.1). Second, speakers may, involuntarily and unintentionally, follow cognitive routines entrenched in their minds, which are in line with the conventional structure of language. Croft (2000: 72), for instance, reports that he cannot help using American English vocabulary like kleenex even when in a British context.

50

3 Modelling language change in New Englishes

3.2.2.2.2 Innovation in language evolution Innovation in language evolution, in turn, lies at the heart of language change. It is brought about by speakers using innovative structures, which, rather than conforming to the linguistic convention of a speech community, involve novel “remappings between form and function” (Croft 2000: 166). Once more, Croft seeks to root this process in a speaker’s pragmatic and functional requirements in concrete linguistic exchanges. Motivations which cause speakers to innovate include (a) expressiveness, which can be put down to the goal of being noticed by interlocutors, or by the aim of raising one’s own social esteem by being playful and entertaining (cf. Keller 1994: 97; cf. also 3.2.2.1); (b) a desire to avoid misunderstanding; (c) a tendency towards economy, i.e. the “goal of using as little time as possible in achieving […] interactional goals” (Croft 2000: 75) . As can be seen, innovation is construed as a functional process because innovation is assumed to occur when linguistic features are adapted in order to optimise their function in the given speech situation (cf. Croft 2000: 8).

3.2.2.2.3 Propagation in language evolution Propagation, in turn, takes place whenever the proportions of linguemes in the lingueme pool change due to speakers preferring particular linguemes already present in the speech community over others. Croft (2000: 73) identifies two motivations which may cause speakers to prefer certain linguemes over others, with (a) representing an intentional process, while (b) is involuntary and unintentional. (a) When adopting a specific lingueme from the lingueme pool, language users follow Keller’s (1994: 96) maxim “Talk like the others talk”, which makes them accommodate their speech to the speech of interlocutors in the specific domain or sub-domain of society in which they operate. In doing so, they seek to blend in socially, and, by implication, achieve social success by influencing others. (b) The other factor which governs speakers’ choices of linguemes is unintentional. Propagation may also be triggered by the mere cognitive entrenchment of linguemes. Croft (2000: 74), for instance, finds himself involuntarily relying on British English forms in conversations with speakers of American English, which he puts down to exposure to British English during a stint in the United Kingdom. It is important to note that “[…] the selection process [i.e. propagation, T.B.] is essentially a social one” (Croft 2000: 32). Keller’s (1994: 96) maxim “Talk like the others talk” mentioned above, which has been identified as the speakers’ main

3.2 An evolutionary model of language change for New Englishes

51

rationale for differential replication, after all, is based on a desire to fit in socially in the respective domain or subdomain. According to Croft (2000: 32), [t]he variants in a linguistic variable have social values associated with them. Speakers select variants to use — that is, to replicate — in particular utterances on the basis of their social values: overt or covert prestige, the social relation of the speaker to the interlocutor, etc.

According to this, the functional factors which govern innovation do not play a significant part in propagation (cf. 3.2.2.2.2).

3.2.3 Problematising Croft’s (2000) evolutionary approach In this section, Croft’s evolutionary approach to language change will be discussed briefly.

3.2.3.1 Arguments for Croft’s (2000) evolutionary approach From the point of view of the present study, the theory is particularly advantageous due to its cognitive orientation, which offers a highly useful theoretical backdrop for a corpus-linguistic study of New Englishes (cf. 1.3.2). Its constructivist notion of linguemes is helpful as it allows for the cross-linguistic transfer both of lexically filled (substantive) and purely structural (schematic) linguemes (cf. 1.3.2.2). The NP structures under discussion in the present study can be modelled as schematic linguemes, which may be subject to cognitive entrenchment much in the same way as actual lexical items. Furthermore, unlikeKeller (1994) or Haspelmath (1999), Croft (2000: 145–8) allows for language contact as a factor in language change and explicitly considers the role of second languages. The model can also be credited with a number of plus points on a theoretical plane. – Croft (2000: 4) is adamant that “[l]anguages don’t change; people change language through their actions” (cf. 3.2.2.1). Consequently, all assumptions made about language change are traced back to the pragmatic needs of concrete speakers in concrete situations. This commitment to “methodological individualism” (Keller 1994: 104) prevents him from reifying linguistic systems in an undue way and, incidentally, is in line with a long-standing, highly plausible discourse in historical linguistics, which has emphasised the significance of individual language users in analysing language change (cf. von Humboldt 1836: 41; Coşeriu 1974. [1958]: 38; Mufwene 2008: 182–3).

52

3 Modelling language change in New Englishes

– Croft’s theory can also be credited with circumventing the “Teleological Fallacy” (Haspelmath 1999: 188), i.e. the tendency of historical linguists to posit teleological mechanisms as explanations for phenomena of language change. Williams (1987: 174), for instance, identifies an “impulse toward regularization” as a reason for speakers of New Englishes leaving out the third-person singular s in verbal morphology. This amounts, in effect, to ascribing a phenomenon of language change which can be observed at the level of the linguistic system (i.e. simplification) to the conscious, intentional and goal-directed action of speakers (cf. Croft 2000: 66). Teleological explanations are hardly convincing, though, as there are clear cases where individual speakers’ intentions and the effects of large-scale language change are at variance: As demonstrated by Keller (1994: 74), for instance, the pejorisation of the Middle High German word vrouwe ‘noble woman’ in modern standard German, where it has developed into the default word for ‘woman’ (Frau) cannot possibly be ascribed to a collective aversion to women. In contrast, it must be due to the “game of gallantry” (Keller 1994: 74) played in the courtly medieval German society, in which individual speakers again and again followed the maxim “‘in case of doubt choose an expression that is a notch too high rather than too low’” (Keller 1994: 73), which, in the long run, has led to the semantic bleaching of expressions referring to women (cf. also Winter-Froemel 2011: 161). Croft’s (2000) solution to the teleology problem rests on his overall conception of language change, which proceeds at two distinct levels: the level of individual speakers (where linguemes are innovated by innovation) and the level of the speech community (where innovations are integrated into the communal linguistic system through propagation). The intentions of individual speakers at the level of innovation, however, need not necessarily coincide with the large-scale direction of change at the level of the linguistic system (cf. Winter-Froemel 2011: 160; cf. Coşeriu 1974. [1958]: 194 and Keller 1994: 77). – Finally, Croft’s (2000) theory is one of the very few fully-fledged evolutionary theories of language change (cf. Rosenbach 2008: 33) to have had a wide appeal in linguistics (cf. Haspelmath 1999; Seiler 2006; Rosenbach 2008; WinterFroemel 2008; Ansaldo 2009; van Rooy 2010; van Rooy 2011; Winter- Froemel 2011).

3.2.3.2 Points of criticism of Croft (2000) The key point of criticism voiced in the literature concerns Croft’s (2000) sole focus on social factors in propagation (cf. 3.2.2.1), which turns out to be an idiosyncratic and radical decision if one considers other approaches in the field. Functional

3.2 An evolutionary model of language change for New Englishes

53

factors in the propagation of linguistic structures are part and parcel of Haspelmath’s (1999: 190) evolutionary theory of language change, which states that selection in linguistic evolution depends on “the usefulness (or ‘user optimality’) of linguistic structures”; they are likewise assumed to play a role in the theories advanced by Mufwene (2008: 150) and Thomason (2005: 60). Seiler (2006) demonstrates convincingly that in the propagation of a rare form of prepositional dative marking in Alemannic and Bavarian dialects of German, functional factors must have played a role (cf. Seiler 2006: 177). I would argue that, considering the evidence in favour of and the absence of unanimous evidence against the relevance of functional factors, it is not advisable to exclude the possibility of functional influence in propagation. It seems more than justified to open up Croft’s (2000) model and allow for functional factors at the social level. This is in line with suggestions made by Rosenbach (2008: 44), who argues that “the evidence does not speak for the exclusive role of social factors in the selection process”, and WinterFroemel (2008: 211), who likewise opts for a wide conception of selection, which does not exclude functional factors categorically. This modified notion of selection/propagation will be taken into account in 3.2.4 In addition, Winter-Froemel (2011) points out that Croft’s (2000) notion of propagation conflates two distinct dimensions of change. It refers, for one thing, to the mere adoption of innovations by individual language users, independent of large-scale ramifications in the speech community (cf. Croft 2000: 73). At the same time, though, it signifies the spread of a lingueme in the speech community at large: “The propagation of a novel linguistic variant is essentially the adoption of a new linguistic convention by the community” (Croft 2000: 174). Rather than breaking up the notion of propagation into two sub-processes, as suggested by Winter-Froemel (2011: 207), however, in the present study, it will be merely taken into account that propagation may be seen in these two perspectives.

3.2.3.3 A revised outline of Croft’s (2000) model of language change Figure 3.2 shows an outline of Croft’s (2000) model of language change which takes into account the changes suggested above. Language stability is guaranteed by normal replication, i.e. replication of linguemes which are in line with the communal norm. Innovation, i.e. the altered replication of linguemes by individual speakers, is driven by functional considerations. Propagation, in turn, which is due to speakers preferring specific linguemes present in the lingueme pool over others, is governed by three distinct factors: (a) speakers’ intentional decisions to adopt given linguemes by virtue of their knowledge of their social values, following Keller’s (1994: 96) maxim “Talk like the others talk”. The social value of a lingueme also encompasses the codes (or registers) used by a single

54

3 Modelling language change in New Englishes

social, functional and non-intentional factors

functional factors

language change

innovation

normal replication

propagation

(adoption at the level of individual speakers)

(spread at the level of the communal norm)

Fig. 3.2. An altered version of Croft’s (2000) model of language change

speakers (even if monolingual) in different domains (cf. 3.2.2.1); (b) speakers’ unintentional, involuntary replication of certain linguemes due to their degree of entrenchment; (c) speakers’ intentional replication of certain linguemes due to functional factors.

3.2.4 Modelling linguistic processes in the emergence of New Englishes in the light of Croft’s (2000) theory of language change On the basis of the altered version of Croft’s (2000) model (cf. figure 3.2), language contact and simplification can now be explained fully. Croft’s (2000) model accounts both for the tendency of individual language learners to simplify and to take over features of local languages and explains the selection of such features at the communal level.

3.2.4.1 Language contact in the light of Croft’s (2000) innovation Language contact starts out as an innovation process in individuals. A bilingual speaker using two distinct languages on a regular basis will notice similarities between the languages, which leads to “interlingual identification” (Odlin 2003: 442), i.e. the “establishment […] of a cognitive link between the corresponding linguemes of the two languages with respect to their identity in substance” (Croft 2000: 146). Subsequently, the “counterpart forms in the two languages are rendered variants of a single variable, that is, two alternative forms for a single function for that speaker” (Croft 2000: 147). From the point of view of the speaker, the

3.2 An evolutionary model of language change for New Englishes

55

result of this process is a novel pairing of meaning and form. A form belonging to another linguistic code available in the speaker’s repertoire can now be used to express a concept for which there already is an expression in the speaker’s native language (cf. Croft 2000: 147).

3.2.4.2 Simplification in the light of Croft’s (2000) innovation Simplification likewise starts out as an innovation process. However, it is given short shrift in Croft’s account. In particular, he never explicitly relates it to any maxim of innovation which makes it plausible for speakers to simplify their L2 in concrete communicative situations. I would argue, though, that such a maxim is readily available in his framework. As mentioned above, speakers follow the “principle of economy”, which can be broken down into two concrete functional goals (cf. 3.2.2.2.2): (a) Keller’s (1994: 98) maxim “Talk in such a way that you do not expend superfluous energy.”, i.e. a processing factor restricting the processing cost on the part of the speaker, and (b) the speakers’ goal of spending as little time as possible on a communicative interaction (cf. Croft 2000: 75). These maxims, while intended for language use in general, take on added significance in L2 use. With a second languages (like English) in postcolonial contexts being used more rarely than the local native languages and in restricted domains only, both the cognitive cost of using L2 linguemes and the time required for a given utterance must be expected to be higher than in a speaker’s L1. This “principle of economy” provides a concrete, functionally and pragmatically plausible reason for individual speakers’ tendency to simplify the structure of their L2 in concrete interactions (cf. Williams 1987: 169).

3.2.4.3 Language contact and simplification in a propagation perspective The accounts of both language contact and simplification given in the previous sections need to be complemented by corresponding processes of propagation. First, language contact phenomena may be propagated intentionally owing to their social values, which encompasses both their alignment with specific speaker groups and even their aptitude at the level of individual domains or codes within one speech community (cf. 3.2.2.1). Second, speakers may also adopt and propagate contact-induced linguemes unintentionally, without wanting to express their group alignment by their choice, purely by virtue of the degree of cognitive entrenchment of a particular variant of a lingueme in their minds (cf. 3.2.3.3). As regards simplification, three distinct mechanisms of propagation can be differentiated: (a) There may be a functional preference for simpler structures by virtue of the principles of economy mentioned above (cf. 3.2.4.2), which, at a

56

3 Modelling language change in New Englishes

societal level, too, lead speakers to opt for those L2 features which require little cognitive effort.(b) Speakers are led to prefer simpler structures over more complex structures on an unintentional basis because in non-native varieties, simple linguemes are more entrenched than complex ones due to a higher overall usage frequency (cf. 3.2.2.2.3). (c) In principle, it is conceivable that intentional social factor also have a role to play with regard to simplification. Simpler structures may acquire an association with a particular (social, and linguistic) domain, where they are propagated as markers of identification with the norms of the domain.

3.3 Summary: Modelling language contact and simplification in New Englishes In this chapter, an explanatory model for the two processes of language contact and simplification in new English varieties has been developed. In 3.1, both processes were illuminated from the point of view of SLA, as the field of SLA provides an exceptionally rich body of results on the linguistic effects of individual L2 acquisition. There is ample evidence that both syntactic structures in general and, more specifically, NP structures, are subject to simplification and cross-linguistic influence in SLA usage. In 3.2, in turn, Croft’s (2000) model of language change was adapted for the analysis of New Englishes with the aim of accounting for the factors which bring about the spread of individual speakers’ innovations to the level of the speech community. Using this model, effects of both language contact and simplification in the present-day norm of a variety can be explained holistically, based on the motivations of individual speakers in concrete communicative situations, which subsequently translate to a large-scale change of the communal norms of language use. The insights from SLA gained in the previous section about L1 transfer and simplification in individual L2 learners (cf. 3.1) were used to corroborate the assumptions about language contact and simplification made in Croft’s (2000) theory. This chapter, thus, prepares the ground for a description of cross-linguistic influence and simplification in the syntax of NPs in New Englishes. It also responds to calls in the literature for a stronger consideration of the results of SLA in the study of New Englishes (cf. 1.3.1). At the same time, rather than ascribing language change in New Englishes to a diffuse notion of “structural nativisation” operating at the abstract level of whole varieties (cf. 2.1.5), the chapter invokes a fully-fledged theory of language change which provides plausible functional and pragmatic motivations for language change.

4 Kenyan and Singaporean English In this chapter, the varieties of Singaporean English and Kenyan English will be characterised with regard to the context of their emergence, their stages of linguistic evolution and their present status.

4.1 Singaporean English Singaporean English will be briefly characterised in general terms; subsequently, the typological properties of its multilingual background will be addressed in detail.

4.1.1 Singaporean English as a variety 4.1.1.1 The history and evolution of Singaporean English In the following, the historical background of British involvement in Singapore will be sketched briefly along the lines of Schneider’s (2007) Dynamic Model (cf. 2.1.2; Turnbull 1980: 94–5 and Abshire 2011: 38 for a general history of Singapore).

(a) Phase 1 (1819–1867): Foundation — Phase 1 of the evolution of English in Singapore starts with the foundation of Singapore in 1819, and it is assumed to last until 1867, when Singapore was made a British Crown Colony (cf. Schneider 2007: 153). Due to an influx of settlers from a variety of national and ethnic backgrounds, including Malays, Chinese and Arabs, English was but one of many languages in use in the first decades, and was largely restricted to settlers of European descent (cf. Gupta 1998: 108–9). The dominant lingua francas of this early period of Singapore would have been Bazaar Malay and, in particular, Southern Chinese varieties such as Teochew, Hokkien or Cantonese. English, in comparison, was a minor player in a highly heterogeneous linguistic matrix. As it was restricted to the stl strand, i.e. the small English minority in Singapore, bilingualism was rare, and significant cross-linguistic influence is unlikely to have happened. (b) Phase 2 (1867–1942): Exonormative stabilisation — The complex pattern of multilingualism remained largely intact after Singapore had become a British Crown Colony in 1867, with the contact varieties of Malay and, to a lesser extent,

58

4 Kenyan and Singaporean English

Southern Chinese languages, serving as the main lingua francas (cf. L. Lim 2007: 452). Despite that, it can be assumed that the role of English was enhanced and stabilised, which marked the beginning of phase 2. In this phase, language teaching started to play a role. According to Gupta (1998: 110), since the late 1850s, the British started pursuing educational policies which involved English-medium teaching. While instruction in English was originally intended for Europeans/Eurasians, the percentage of non-European attendants of English-medium schools had risen to about 50% by the end of the 19th century, and grew by substantial numbers until the 20th century. By the 1910s, Chinese children “thronged into the English schools” (Gupta 1998: 114). Moreover, from the 1900s onwards, more and more Chinese entered the teaching profession, whereas in the 19th century, the majority of English teachers had been Eurasian (cf. L. Lim 2007: 452). It was a natural consequence of English-medium schooling that, although still having a somewhat “elitist touch” (Schneider 2007: 154), bilingualism started to spread slowly in this phase. According to a 1921 census, in the urban areas of the Straits settlement, 6% of the Chinese, 3% of the Malay and 8% of the Indian population had a command of English (cf. Gupta 1998: 115). As is typical of phase 2, a number of lexical borrowings are documented for this period, especially for place names, flora and fauna, including Jurong, Katong, Tekong Kechil, brinjal ‘aubergine’, taugeh ‘beansprout’ or lalang ‘tropical grass’ (cf. Schneider 2007: 155). Despite that, however, it is safe to assume that the linguistic orientation at large was broadly exonormative, which resonated with the overall cultural orientation towards the British (cf. Schneider 2007: 154). (c) Phase 3 (1945–1970s): Nativisation — The traumatic three years of Japanese occupation from 1942–45, which cast serious doubt on the protection offered by the British colonial power (cf. Schneider 2007: 155), brought about the psychological detachment among the British settlers typical of phase 3 . After the British takeover in 1945, independence campaigners became more and more vocal in their criticisms, which contributed to an overall awareness that “the colonial tradition was broken” (Schneider 2007: 155). Despite that, there was a further rise in the numbers of children attending English-medium schools in this period, especially in the Malay ethnic group (cf. Gupta 1998: 115). After Singapore’s independence in 1965, even more efforts were made to make English-medium education available to all, with English being made a compulsory school language in 1965 (cf. L. Lim 2007: 452). The main reasons for the success of English after independence were the shift in identity since World War II and the government’s promotion of English-medium

4.1 Singaporean English

59

schooling, along with the ethnic neutrality of English and the lack of mutual intelligibility of local Singaporean languages (cf. Schneider 2007: 156). The rising levels of English-knowing bilingualism in this period led to widespread language contact, which allowed for a process of structural nativisation typical of phase 3 (cf. 2.1.2) . (d) Phase 4 (1970s–): Endonormative stabilisation — There is widespread agreement in the literature, however, that since the 1970s, Singaporean English has moved on even further. In the same way as the collective awareness of Singapore as a city state has been growing since phase 3, the linguistic identities of Singaporeans have crystallised and stabilised. Singaporean English has moved on to a status of endonormativity, where deviations from the British standard are no longer perceived as errors but have been amalgamated into a stable, distinctive, new variety of English. L. Lim (2001) even discusses first indications of phase 5. According to Schneider’s (2007) account Singaporean English is a highly advanced variety on the verge of becoming a full-blown variety of English. This view, however, must be supported by actual linguistic evidence. In the following sections, key pieces of linguistic and sociolinguistic evidence will be collated which allow for an assessment of the status of Singaporean English.

4.1.1.2 Singaporean English and Singlish Singaporean English is far from being a unitary phenomenon. Rather, the variety represents a gradient between distinct strata, with the majority of analysts recognising two extreme poles, i.e. a colloquial variety called Colloquial Singaporean English or Singlish, and a variety of Standard Singaporean English. A typical (probably stereotypical) example of the former is exemplified in (4.1); the latter is illustrated in (4.2). (4.1)

(4.2)

Eh, better do properly, lah. Anyhow do, wait kena scolding. And then, you always ask her for favour, and still don’t want to do properly. Must lah. Like that do cannot. Do again. Come, I help you. Two people do, very fast finish. One person do, not so fast. You see, almost finish. Wah, she see this, she will be very happy. Then we get big angpow for sure this New Year! (Alsagoff and C. L. Ho 1998: 129) You had better do this properly. If you don’t, you may get told off. And since you are always asking her for favours, you should at least do this properly

60

4 Kenyan and Singaporean English

for her. You should! You cannot do it like this. Do it again. Come, let me help you. Two people can finish this job very quickly. One person will not be able to do it as fast. You see, we’re almost done. Wow, when she sees this, she’ll be very happy. We will definitely get a very big present from her this New Year. (Alsagoff and C. L. Ho 1998: 129) It is obvious from (4.1) that Singlish is marked by quite severe deviations from the British standard, which are largely due to “substantial substrate-influenced grammatical restructuring” (Bao and Hong 2006: 105), primarily from Sinitic languages, while Standard Singaporean English “resembles very much the formal register of Englishes used in other parts of the world” (Alsagoff and C. L. Ho 1998: 129–30)¹. It is important to note that the present study will focus on the standard end of the style continuum, which is represented by the corpora of the ICE family (cf. 6.1.1). The key arguments for this choice are the following: First, the differences between the two strata of language are not necessarily differentiated as clearly as in (4.1)–(4.2). (4.3) shows a passage of Singaporean English taken from a transcript of a discussion between Singaporeans recorded by Leimgruber (2012:9). The portions in bold print indicate compliance with Standard English norms, while the underlined passages are distinctly Singlish (the labels 3f, 4m etc. identify individual speakers and flag them as male or female; asterisks (*) indicate hypercorrection, cf. Leimgruber (2012:9)). (4.3)

3f Fifteen minute ah, is it? Aiyoh, don’t know what to say ah. 4m Eh after this ø finish ah? 2f No. One of you have to go walk around the school, meet your friends and talk. 4m Are you sure ah? 2f Because he want to see how we all talk, normally… 3f (***) stupidity. 2f That’s what him say to us just now. 4m Serious? So dumb. 3f Do you realise that he’s actually quite nervous? 4m I know!

1 Attempts at capturing the sociolinguistic relationship between the two strata of language in Singapore abound (cf. John T. Platt 1975; Gupta 1989; Gupta 1994; Pakir 1991; Alsagoff 2010; Leimgruber 2013 for the key approaches; Leimgruber 2012: 4–8 for an overview); dealing with them in detail, however, would go beyond the domain of this study.

4.1 Singaporean English

2f 1m 4m 2f 3f 2f

61

Yeah. He will edit it lah, I think. So confirmation right, we will be Bangkok. (***) Yeah. [laugh] Ok lah, confirm ah, Bangkok. Let’s have a try ah

It is obvious that this conversation is ambiguous. Rather than belonging clearly to either Singlish or Standard Singapore English, it oscillates between the strata, as the speakers use expressions conforming to the British Standard (such as dosupport in don’t know or do you realise, an overt pronominal subject in he will edit it) or localised expressions at various levels of formality (including e.g. discourse particles such as lah and ah, zero verbal endings as in want etc.), depending on context and communicative goals. Therefore, Singlish and Standard Singaporean English should not be viewed as sharply delimited concepts. Rather, they form two poles on a continuum from more standard to more colloquial forms of English, which, in actual discourse, inevitably intersect and overlap (cf. also Wee 2008b: 259). Second, while it is surely true that a comparison of the standard end of Singaporean English with Standard Englishes worldwide will yield “few [of the] categorical differences” (Gupta 2010: 69) to be found between typical strongly vernacular varieties, one can safely assume that even in standard varieties, “there are regionally patterned differences in preference” (Gupta 2010: 69). Along these lines, it will be assumed that texts and conversations in Standard Singaporean English partake of the same style continuum as Singlish, which means that they, too, are open to effects such as contact influence and simplification, just as the lower stylistic levels exemplified above. They, too, exhibit unique regional characteristics, even though these characteristics may be less conspicuous than in the more vernacular strata of variation, e.g. by appearing in the shape of statistical preferences in the choice of constructions only (cf. 1.3.1).

4.1.1.3 Linguistic features of Singaporean English The most salient grammatical features of Singaporean English can be summarised as follows. (a) Phonetics and phonology — Typical phonological features include a lack of aspiration in voiceless consonants like /p,t/, the realisation of /θ, ð/ as /t,d/ (wordinitially, /tin/ for thin) and /f/ (word-finally, /bref/ for breath) (cf. Wee 2008b: 266–

62

4 Kenyan and Singaporean English

8), a marked tendency towards consonant cluster reduction (/teks/ for texts, cf. also Gut 2007) or the use of a schwa epenthesis instead of syllabic consonants, with e.g. /ˈbʌtn̩/ realised as /ˈbɑtən/ (cf. Wee 2008b: 269). (b) Morphology and Syntax — Verb forms tend to lack inflectional markers, with both time reference and aspect marking frequently effected by analytic markers such as yesterday (He eat here yesterday), always, still, already (habitual: She always borrow money from me — perfective: My son ride bicycle already — progressive: Late already, you still eat.). This system of analytic aspect markers has been analysed in terms of cross-linguistic influence from local varieties of Chinese (cf. Bao 2005: 239–50). Noun phrases tend to lack articles (I don’t have ticket; cf. Wee and Ansaldo 2004: 60–1); in addition, the Standard English distinction between count and non-count nouns may be obliterated (luggages, equipments; Wee 2008a: 596). Singlish has been documented to have a characteristic relativisation strategy using an invariant relative marker one (That boy pinch my sister one very naughty. ‘That boy who pinched my sister is very naughty’; cf. (2.5), cf. also Wee 2008a: 597; Gil 1998). Alsagoff and C. L. Ho (1998: 145–7) analyse such relative clauses as results of a complex amalgamation between English and Chinese structures. They view the position of one after the relative clause as a reflex from Chinese, where the relative pronoun likewise follows the clause; the position of the relative clause after the NP head, however, runs counter to Chinese norms, while it follows the English template (cf. Alsagoff and C. L. Ho 1998: 146–7)². At the clause level, speakers of Singlish have a tendency not to spell out verbal complements in full, with e.g. the sentence (4.4) being fully grammatical, even though both subject (“pro-drop”) and object are omitted and need to be reconstructed contextually (cf. Alsagoff and C. L. Ho 1998: 147). (4.4)

Every year, øsubj must buy øobj for Chinese New Year (Alsagoff and C. L. Ho 1998: 147)..

Singlish also exhibits topicalisation structures which appear marked from a Standard English point of view, as in (4.5), which can be attributed to the topiccomment structure typical of varieties of Chinese (cf. 3.1.1.2).

2 There is no agreement in the literature about the status of relative one (cf. Wee and Ansaldo 2004: 66–71 for an overview and Gil 1998 for an alternative conception, where one is analysed as a “reifier”)

4.1 Singaporean English

(4.5)

63

Certain medicines we don’t stock in our dispensary. (Alsagoff and C. L. Ho 1998: 148)

Likewise, there are un-inverted wh-interrogatives as in She took what, which must also be due to contact with Sinitic languages (cf. Alsagoff and C. L. Ho 1998: 149). Apart from the default standard English passive construction, there is a contactinduced “kena passive”, as in John kena scold ‘John was scolded’ (from Malay). Another much-cited feature is invariant tag questions, as in They give him a medal, is it? (cf. Alsagoff and C. L. Ho 1998: 150). In line with Sinitic languages spoken in the region, Singlish exhibits grammaticalised reduplication phenomena such as hypocoristic boy-boy ‘boy-friend’, with the reduplication adding an undertone of familiarity and closeness, or sweet-sweet, where reduplication intensifies the meaning of the adjective (cf. Wee 2008a: 601–2). One of the most striking features of Singlish, finally, is a set of discourse particles originally borrowed from Chinese, which have come to be fully integrated into the grammatical system of the variety (cf. L. Lim 2007 for a detailed account of their emergence). Well-known examples include lah, which “indicates [the] speaker’s mood[…] and appeals to the addressee to accommodate this mood[…]” (Wee 2008a: 604), ah, which emphasises that a question requires a response or what, which marks information as being obvious (cf. Wee 2004: 117–26). There is also a range of studies which have demonstrated that Singaporean English differs from other New Englishes in terms of more complex, statistical patterns of usage. Gut (2007: 355), for instance, finds a high rate of consonant cluster reduction. She attributes this to a “tacit endonormative standard”, which allows the transfer of the preference for simple consonant clusters in Sinitic languages to English, while in Nigerian English, an exonormative orientation bars substrate influence on English, which leads to high rates of consonant cluster retention. In Mukherjee and Gries’s (2009: 47–8) study of verb complementation in New Englishes, Singaporean English stands out as the variety with the highest number of verbs whose preferred complementation patterns differ from those found in British English (cf. 2.1.2). Both Hong Kong and Indian English, both of which are in lower stages of evolution, are closer to their input variety with regard to this criterion, which, for Mukherjee and Gries (2009: 48), confirms the high evolutionary stage of Singaporean English. This shows that Singaporean English has also developed localised grammatical preferences in the areas of non-categorical, probabilistic grammatical variation mentioned in 1.3.1, which will take centre stage in the present study.

64

4 Kenyan and Singaporean English

This overview of grammatical innovations shows that a wide range of grammatical features set apart Singaporean English from its input variety. This is a clear confirmation of the high degree of nativisation assumed.

4.1.1.4 The sociolinguistics of Singaporean English However, another prerequisite for endocentricity is a high degree of sociolinguistic entrenchment. In what follows, the key sociolinguistic characteristics of Singaporean English will be sketched briefly.

4.1.1.4.1 Patterns of bi- and multilingualism Table 4.1 shows the proportions of ethnic groups in Singapore according to the 2010 census of population (cf. Department of Statistics Singapore 2010: Table 1). As can be seen, almost 75% of Singaporeans are ethnically Chinese; Malays and Indians are represented by about 13% and 9% respectively, while the remaining 3% of Singaporeans belong to various other ethnic groups. It does not come as a surprise that this high degree of multi-ethnicity should be reflected in widespread multilingualism among Singaporeans. Table 4.2 shows the “languages literate in” of Singaporeans according to the 2010 census³. 16.31% and 11.06% of Singaporeans are monolingually Chinese or English respectively. By far the most frequent multilingual literacy pattern is Chinese-English bilingualism, which is attested for

Table 4.1. Population by ethnic group according to the 2010 census of population in Singapore (Department of Statistics Singapore 2010: Statistical Tables — Table 1) Ethnic group

N

%

Chinese Malays Indians Others

2,793,980 503,868 348,119 125,754

74.08 13.36 9.23 3.33

Sum

3,771,721

100.00

3 According to the census, the overall number of literate Singaporeans is 2,977,088 (cf. Department of Statistics Singapore 2010: Table 39). The sum of the individual speaker numbers given in the census (which have been replicated faithfully in table 4.2), however, amounts to 2,977,090. It should be noted that all census data have been extracted automatically from Excel files provided on the website of the Department of Statistics in order to guarantee the correct representation of all results.

4.1 Singaporean English

65

Table 4.2. “Languages literate in” according to the 2010 census of population in Singapore (Department of Statistics Singapore 2010: Table 39) Language pattern

N

%

Chinese only English only Malay only Tamil only Non-Official Language only English & Chinese only English & Malay only English & Tamil only English & Non-Official Language only Other Two Languages only English, Chinese & Malay only English, Chinese & Tamil only English, Malay & Tamil only Other Three or More Languages

485,511 329,194 47,278 10,939 5,292 1,305,705 390,124 104,570 82,972 12,898 96,660 377 16,423 89,147

16.31 11.06 1.59 0.37 0.18 43.86 13.10 3.51 2.79 0.43 3.25 0.01 0.55 2.99

Sum

2,977,090

100.00

43.86% of the population, followed by 13.1% who have a command of English and Malay. The other patterns are less significant, as each has a share of less than 4% of the population. In this context, two further sociolinguistic census results are worth noting. It is interesting, for one thing, that “English-knowing bilingualism” (Kachru 1982: 42) is stratified in a quite homogeneous way among the ethnic groups of Singapore. The majority of ethnically Chinese, Malays, Indians and “Others”, in their majority have a pattern of multilingualism which involves English; only a minute proportion do no have English in their repertoire (cf. Department of Statistics Singapore 2010: Table 45). However, the census also documents that the knowledge of English “is still closely correlated with social class” (Schneider 2007: 157). The rate of English-knowing bi- and trilingualism turns out to be lowest for Singaporeans having no qualification and gradually rises to an overall rate of more than 70% among Singaporeans having a secondary educational qualification. Conversely, Chinese (and Malay) monolingualism is most frequent in unskilled Singaporeans, while primary, lower secondary and secondary levels of qualification are associated with a marked decrease of such patterns (cf. Department of Statistics Singapore 2010: Table 45).

66

4 Kenyan and Singaporean English

4.1.1.4.2 The social stratification of Singaporean English The degree to which a New English variety has made inroads as a home language is an important indicator of its degree of entrenchment in a society. Table 4.3 shows the most frequent languages spoken at home among residents older than five years in Singapore according to the 2010 census of population⁴. It is immediately evident from this that English is, in fact, used quite widely in people’s private lives, with more than 32% of Singaporean households stating it to be the primary home language. It is only surpassed in this role by Mandarin, which has a share of almost 36%. The other major home languages are Malay (12.19%) and Hokkien (7.03%), while the remainder of languages are of minor importance in this respect, with proportions of less than 4%. Once more, the census is also informative with regard to ethnic group and social class. The strong position of English in the population at large is replicated best among the ethnically Chinese and the Indians, where the rate of English oscillates around a third, and for the “Others” group, where the share of English exceeds 60%. Malays, in contrast, stand out in this respect, as a mere 17% of them speak English at home, while the vast majority (82.66%) opt for Malay (cf. Department of Statistics Singapore 2010: Table 55; L. Lim, Pakir, and Wee 2010: 8; cf. Chong and Seilhamer 2014). Once more, it turns out that the use of English as a home language is likewise correlated with high educational levels, with only the

Table 4.3. The languages spoken most frequently at home according to the 2010 census of population in Singapore (Department of Statistics Singapore 2010: Table 47) Language

N

%

Mandarin English Malay Hokkien Cantonese Tamil Teochew Other Indian languages Others Other Chinese dialects

1,211,505 1,097,443 414,475 238,843 121,136 110,667 94,302 40,334 37,600 32,750

35.64 32.29 12.19 7.03 3.56 3.26 2.77 1.19 1.11 0.96

Sum

3,399,055

100.00

4 The sum of 3,399,055 shown in table 4.3 corresponds to the individual speaker numbers, while the census has 3,399,054.

4.1 Singaporean English

67

group having secondary education, in fact, reaching 30% of households which use English as a home language, while its proportion bottoms out in the lowest educational levels, with roughly 14% in the lower secondary group, 6.6% among people with primary education and a mere 2.7% among families with the lowest level of education, i.e. no qualification (cf. Department of Statistics Singapore 2010: Table 55).

4.1.1.4.3 Language policy With the spread of English in Singapore well-known to have been “a result of deliberate government planning and policy” (Bokhorst-Heng 1998: 303), it is important to take into account language-policy in Singapore. Before the mid-1950s, i.e. under British colonial rule, the Mandarin, Malay, Tamil and English speech communities devised their own programmes of language education (cf. BokhorstHeng 1998: 288), and “English was in no way to be taught en masse” (BokhorstHeng 1998: 288). Active language policies promoting English were implemented from the mid-fifties onwards in order to address the high degree of linguistic diversity (cf. Bokhorst-Heng 1998: 292); in addition, there were hopes that the knowledge of an international language would bring economic advantages, and the ethnic neutrality of English played a role, too (cf. L. Lim 2010: 29). However, English was to be embedded into an “English-knowing bilingual system” (L. Lim 2010: 30). In addition to the international language English, Singaporeans should be taught one of the three official languages, Mandarin, Malay and Tamil (“mother tongues”) “in order to maintain Asian values” (L. Lim 2010: 30). Other vernaculars were deliberately suppressed through the so-called “Speak Mandarin Campaign”, a long-standing government-run campaign against Chinese vernaculars other than Mandarin (such as Hokkien and Cantonese). The long-term effects of this “aggressive” (L. Lim 2010: 30) language policy led to a marked increase of Mandarin as a home language from 10% in 1980 to 35.64% in 2010, at the expense of Hokkien and Bazaar Malay, and, crucially, to a spread of English as main lingua franca cf. L. Lim 2010: 30–1; cf. table 4.3. This two-tiered approach can be considered the foremost characteristic of Singaporean language policy (cf. L. Lim 2010: 38) to the present day. Governmental language policy in Singapore, however, can also be viewed as a hindrance to the development of a distinct variety of English in Singapore. The emergence of Singlish since the mid-1970s (cf. 4.1.1.2) has been looked on with a wary eye by the government, amid fears that Singaporean English may turn unintelligible for non-Singaporeans, which, in the eyes of the government, would jeopardise its role as a link to the Western world (cf. Bokhorst-Heng 1998: 303). For this reason, the “Speak Good English Movement” was launched in 2000, with the

68

4 Kenyan and Singaporean English

purpose of advocating and promoting the use of standard English among Singaporeans (cf. Bokhorst-Heng 2005: 185). Measures in the wake of the “Speak Good English Movement” include the enforcement of Standard English in language education, speeches, events, printed information sheets and billboards (cf. Bruthiaux 2010: 92). Early materials, for example, featured recommendations that instead of asking You want to try? in service encounters, it would be more appropriate to spell out the question in full as Would you like to try that (cited by Gupta 2010: 72). In the context of the “Speak Good English Movement”, however, there have also been vociferous debates in Singaporean society, with large swaths of Singaporean society defending the use of Singlish in, e.g., TV and cinemas (cf. Schneider 2007: 158). This aspect of the linguistic situation in Singapore attests to the fact that despite its strong degree of entrenchment in the population, the official level in Singapore has not yet converged on a common view and acceptance of English.

4.1.1.4.4 Language attitudes It is interesting, however, that despite qualms on the part of the government, the attitude towards Singaporean English in the population of Singapore appears to be highly positive. P. K. W. Tan and D. K. H. Tan (2008) use a matched-guise technique to elicit Singaporean pupils’ (aged 15–16) views on utterances in American English, Standard English with a Singaporean accent and a distinctly Singlish passage (cf. P. K. W. Tan and D. K. H. Tan 2008: 470). The texts are read out by a Singaporean who, thanks to an extended stint in California, is able to “switch convincingly between Singaporean and (Californian) American accents” (P. K. W. Tan and D. K. H. Tan 2008: 470). The results show that acrolectal Standard English with a Singaporean accent receives by far the most favourable ratings for all items, outstripping both American English and Singlish on all counts (cf. P. K. W. Tan and D. K. H. Tan 2008: 474). In addition, with regard to Singlish, it turns out that while the pupils may not view the colloquial Singaporean variety as cool and trendy, they have come to appreciate it as an identity marker and use it in order “to interact effectively in the community” (P. K. W. Tan and D. K. H. Tan 2008: 476). The study provides evidence of the high degree of entrenchment and the complementary relationship of both Standard Singaporean English and Singlish.

4.1.1.5 Summary: Assessing the degree of endo- and exonormativity in Singapore Considering the characteristics of Singaporean English addressed in the previous sections, what is the evidence of the linguistics independence, self-sufficiency

4.1 Singaporean English

69

and endonormativity of Singapore claimed in the literature (cf. 4.1.1.1)? The linguistic, feature-based evidence for endonormativity is strong (cf. 4.1.1.3), which attests to the fact that Singaporean English “has gone through a vibrant process of structural nativisation” (Schneider 2007: 158). From a sociolinguistic point of view, there is, for one thing, the exceptional degree of its spread in society, and, in fact, down the social scale. It was demonstrated above that almost 80% of Singaporeans speak a pattern of languages which involves English; more than a third of households use it as a home language, and, in fact, as an L1 or “co-L1” (Schneider 2007: 157). These factors can be considered clear indications for the stability and self-sufficiency typical of Schneider’s (2007) phase 4. The only area which does not allow for wholly unanimous conclusions with regard to endonormativity is represented by the attitudes of Singaporeans towards their own language variety. While, at the level of the Singaporean population the evidence of a stable internal linguistic norm is overwhelming (cf. 4.1.1.4.4), there are clear indications at the level of official language policy that “the issue of norm selection is still under discussion” (Schneider 2007: 160), as has been demonstrated on the basis of the “Speak Good English Movement” (cf. 4.1.1.4.3: Schneider 2007: 160). These remnants from earlier evolutionary phases allow the conclusion that at an official level, Singaporean society has not fully converged on a common norm of English, which would be expected in a settled, endonormative variety. Taken together, however, the evidence for Singaporean English being a phase-4 variety is strong.

4.1.2 The local languages in Singapore In order to prepare the ground for the analysis of substrate influence in Singaporean English, this section will deal with the indigenous languages of Singapore and their NP structures.

4.1.2.1 The indigenous languages of Singapore: past and present The aim of this section is to provide an overview of the main substrate languages of Singaporean English and to assess their importance for the development of the variety, both at present and in the past. Table 4.4 shows a comprehensive overview of the languages spoken in today’s Singapore according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013). Judging from table 4.4, by far the most frequent local language is Mandarin Chinese, which is spoken by 34.77% of Singaporeans. The other Chinese varieties

70

4 Kenyan and Singaporean English

Table 4.4. Languages of Singapore according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013). The source specified as “2010 census” refers to Department of Statistics Singapore (2010). The percentages to not add up to 100% due to rounding. Language

Speakers

%

Chinese, Mandarin English Malay Chinese, Mĭn Nan (Hokkien, Teochew) Chinese, Yuè (Cantonese) Tamil Chinese, Hakka Chinese, Min Dong Malayalam Chinese, Pu-Xian Hindi Malay, Baba Panjabi, Eastern Gujarati Chinese, Min Bei Sindhi Sinhala Madura Orang Seletar Javanese Bengali Telugu Malay, Standard

1,210,000 1,100,000 414,000 333,000 121,000 111,000 69,000 34,200 26,300 14,100 13,100 10,000 5,670 4,120 4,000 3,970 3,140 900 880 800 600 600 0

34.77 31.61 11.90 9.57 3.48 3.19 1.98 0.98 0.76 0.41 0.38 0.29 0.16 0.12 0.11 0.11 0.09 0.03 0.03 0.02 0.02 0.02 0.00

Sum

3,480,380

100.03

Source

Phylum

2010 census 2010 census 2010 census 2010 census 2010 census 2010 census 1980 2000 2010 census 2000 2010 census Pakir (1986) 2010 census 2010 census 1985 2010 census 2010 census 1985 2000 1985 1985 2000 —

Sino-Tibetan Indo-European Austronesian Sino-Tibetan Sino-Tibetan Dravidian Sino-Tibetan Sino-Tibetan Dravidian Sino-Tibetan Indo-European Creole Indo-European Indo-European Sino-Tibetan Indo-European Indo-European Austronesian Austronesian Austronesian Indo-European Dravidian Austronesian

are considerably less frequent, with Mĭn Nan varieties, i.e. mostly Hokkien/Teochew, at 9.57%, Yuè varieties of Chinese, i.e. Cantonese, at 3.48%, and all other Chinese varieties used by very low percentages of the population. Malay and Tamil, the two official languages, have proportions of 11.9% and 3.19% respectively. All other languages, including further Indian varieties, are likewise represented only by tiny fractions of the population. This set of languages spoken in today’s Singapore, however, is markedly different from the languages spoken earlier in its history. It is well-known, though, that according to Mufwene’s (2001: 28–9) “founder effect”, linguistic innovations from the earliest periods of a language contact scenario stand a particularly high chance of persisting in a speech community.

4.1 Singaporean English

71

As can be seen in figure 4.1, with regard to the relative importance of contact languages, the linguistic history of Singapore can be divided into three major phases (cf. L. Lim 2007: 453–9). English

Bazaar Malay

Hokkien

Mandarin

Cantonese Age I

Age II

Age III

1800s1970s

mid1970s–late 1980s

late 1980s– 2000

Fig. 4.1. L. Lim’s (2007: 458) overview of the historical relevance of indigenous languages in Singapore. — language is dominant; — language is present; — language is waning in prominence.

(a) Age I: 1800s–mid-1970s — From the inception of Singapore, the main lingua francas in the region, across ethnic boundaries, were contact varieties of Malay (cf. L. Lim 2007: 453). Baba Malay, a variety of Malay restructured through language contact with Sinitic languages (cf. 4.1.2.2.1), served as the vernacular language of the Straits-born Chinese (or “Babas”) of mixed Chinese-Malay ancestry, who were economically successful and had a high social standing. The spread of English among them was particularly rapid, which makes this group an important agent of language contact (cf. L. Lim 2010: 24). Furthermore, Chinese immigrants in Singapore, who, for the most part, came from Southern China (i.e. the provinces of Chaochow, Fujian, and Guangdong), spoke Southern Chinese varieties like Teochew, Hokkien and Cantonese. Due to the high number of ethnically Chinese in early Singapore (according to L. Lim 2007: 454, 45.0% of Singaporeans in 1836), these languages would have played a major role (cf. L. Lim 2007: 454). Hokkien was, in fact, the most widely understood

72

4 Kenyan and Singaporean English

Chinese variety until the 1970s, and Cantonese, too, was of major importance (cf. L. Lim 2010: 24). Mandarin Chinese, in contrast, did not play a role at the earliest stages but started to spread only in the 1920s (cf. L. Lim 2010: 24). Likewise, both Standard Malay, today’s National Language, and Indian languages like Tamil, Telugu, Malayalam, Punjabi or Bengali had a very limited impact, as they never served as languages of wider communication but were restricted to their respective communities (cf. L. Lim 2007: 453–5). (b) Age II: mid-1970s–late 1980s — In this phase, both Bazaar Malay and Baba Malay — although restricted to the Peranakans living in the Straits of Singapore anyway — lost ground due to the spread of English, which was due to governmental language policy (cf. L. Lim 2010: 30–1). For the same reason, Mandarin made enormous inroads in this period, with its proportion rising from 10% in 1980 to 30.1% in 1990 (cf. L. Lim 2010: 30). Conversely, Hokkien and Cantonese, despite being still in use in the Chinese community, lost their earlier significance (cf. L. Lim 2007: 456). Varieties of Indian languages, in contrast, remained minority languages (cf. L. Lim 2010: 31). (c) Age III: late 1980s–2000 — In this phase, the linguistic situation of the previous phase remained largely in place. If anything, the increasing cultural dominance of Hong Kong in the region, along with film and music, increased the prominence of Cantonese. This historical perspective alters the relevance of the set of local languages spoken in present-day Singapore for the present study. The main players in the contact ecology of Singaporean English can be ranked as follows. Historically speaking, contact varieties of Malay, i.e. Baba Malay and, in particular, Bazaar Malay, are of the utmost significance due to their spread as lingua francas in the region, despite the fact that in present-day Singapore, they are no longer significant (cf. figure 4.1). Hokkien, Cantonese and Teochew, too, played a major role in early Singapore and, to an extent, continue to be used (cf. figure 4.1). Mandarin, despite its low significance in early Singapore, must be considered a significant factor due to the inroads it has made by now, following the government’s “Speak Mandarin Campaign”. Importantly, Standard Malay, despite its status as a National Language today, is of minor importance, as is Tamil and the other varieties of Indian languages; neither language will be taken into account in the typological survey below.

4.1 Singaporean English

73

4.1.2.2 NPs in the local languages of Singapore The analysis of NPs in the contact ecology of Singapore will focus on the main players in the contact ecology identified in the previous section.

4.1.2.2.1 NPs in Bazaar Malay and Baba Malay Both Bazaar Malay and Baba Malay are vernacular contact varieties of Malay, which resemble each other linguistically and are mutually understandable. Despite that, there are some points of difference: Bazaar Malay is an “interethnic lingua franca”, which emerged from a trade context as a prime lingua franca in the whole of the Malay peninsula (cf. Aye 2005: 1). It has, however, receded greatly due to the spread of English since the 1970s, with fewer than 10,000 speakers left today (cf. Aye 2013: 86). Bazaar Malay has never had native speakers and is not firmly associated with any ethnic group, which renders it a somewhat diffuse phenomenon which “show[s] different degrees of focusing […] depending on the speech community” (Ansaldo 2009: 61). Bazaar Malay has been classified as a pidgin (cf. Aye 2013: 87), whose lexifier is clearly (colloquial) Malay, while the main substrate is Hokkien, along with further varieties of (Southern) Chinese and various Indian languages (cf. Aye 2005: 37); its lexis is marked by a Malay basis and numerous borrowings from Hokkien (cf. Aye 2005: 296). Similarly, its phonology is influenced by Hokkien and further local languages (cf. Aye 2013: 88; cf. also Aye 2005: 296), and its morphosyntax, too, is indicative of “the competition of Malay and Hokkien” (Aye 2005: 297), with Hokkien-induced features including, for instance, isolating morphology, topic-comment structures and punya as a marker of possessive (cf. below), attributive, relative or locative relationships (cf. Adelaar 2005: 212–5). Baba Malay, in contrast, is the mother tongue of the Peranakan, a population group descended from the first generation of Chinese who, possibly as early as 1400, migrated to the Malay Peninsula, i.e. to Melaka, Penang and Singapore, and intermarried with Malay women. Baba Malay was transplanted to Singapore as many of the Straits-born Chinese, especially from Melaka, migrated to Singapore during the 19th century (cf. Ansaldo 2009: 66; cf. S. Lim 1988: 8–10). Baba Malay, too, is the result of language contact between Malay and Southern Chinese varieties belonging to the Mĭn group, especially Hokkien, but, to some extent, also Teochew, which has been variably seen as a “corrupted form of Malay”, a dialect of Malay (cf. Pakir 1986: 33) or a creole (cf. Pakir 1986: 3). The phonology of Baba Malay is highly similar to Malay (cf. S. Lim 1988: 14; Ansaldo 2009: 163); its lexicon and grammar alike, too, display a Malay substrate (including e.g. existential constructions using the Malay verb ada) with considerable admixture from Hokkien, such as a strongly isolating morphology or topic-comment structures (cf.

74

4 Kenyan and Singaporean English

Ansaldo 2009: 163; 166–74; 177). The majority of these features are shared with Bazaar Malay (cf. Ansaldo 2009: 166). It does not come as a surprise, then, that the syntax of NPs in both Bazaar Malay and Baba Malay is likewise marked by Sinitic influences, while being based on Standard Malay. With regard to NP structures, this entails a competition between Sinitic prenominal structures and Malay-derived postnominal structures (cf. Tadmor 2009: 800). Relevant phrases taken from Standard Malay can be seen in the following (all taken from Tadmor 2009: 800). (4.6)

rumah besar house big

(4.7)

‘a/the big house’ rumah guru house teacher

(4.8)

‘a/the teacher’s house’ rumah ini house dem-prox ‘this house’

It is fully plausible that the corresponding structures in Bazaar Malay vary between pre- and postnominal⁵: Possessives in present-day Bazaar Malay occur variably before or after the head (cf. (4.9)–(4.10)), as do adjectives (cf. (4.11)–(4.12)), with the positional variants not differing significantly in terms of their meaning or pragmatics (cf. Aye 2005: 108). Numerals invariably stand before their heads; prepositional phrases follow the head. Finally, demonstratives can occur pre- or postnominally (cf. Aye 2005: 107–9). (4.9)

Dia jatuh ah ini anjing lagi gigit dia tangan. 3sg fall.down dem dog more bite 3sg hand

‘He fell down (and what’s) more, the dog bit his hand.’ (Aye 2005: 107) (4.10) Dia dua-dua nama dua tempat ah. 3sg two-two name two place ‘Both are the names of two places.’ (Aye 2005: 107)

5 Aye’s (2005: 104–17) study is based on present-day Baba Malay, even though, of course, earlier data would be desirable (cf. 4.1.2.1). For want of a better option, however, it can be assumed that present-day Singapore Bazaar Malay is “as close as we can get to a specific BZM [Bazaar Malay] variety” (Ansaldo 2009: 162).

4.1 Singaporean English

75

berat, pegan baru petir, lu mesti tahu berapa 2sg must know how.much heavy 2sg hold new box tahu. know

(4.11) Lu

‘If you hold the new box, you certainly know how heavy it is, you know.’ (Aye 2005: 108) (4.12) Ini budak kecil kena pukul. dem child small pass beat ‘The little child was beaten.’ (Aye 2005: 108) Another important structure is the punya construction, which precedes NP heads. The origin of punya is a Malay verb denoting possession, which, in Bazaar Malay, has been restructured into a multi-functional particle used to prefix NPs with phrase types such as NPs, APs, PPs, quantifiers or clauses (cf. Bao and Aye 2010: 157). The punya construction can express a wide variety of semantic relationships, including possession (cf. (4.13)) or part-whole relationships (cf. (4.14); cf. Aye 2005: 115). It may represent a calque of Hokkien e (cf. Aye 2005: 256), which has a very similar syntactic distribution (cf. Aye 2005: 256–64). (4.13) Dia punya badan bagus bah

3sg punya body

good

‘His body is very good.’ (Aye 2005: 113) periuk punya tutup ah kasi terbang ah, turun dem pot punya lid cause fly go.down bawah. under

(4.14) Itu

‘The lid of the pot flew and dropped down.’ (Aye 2005: 113–4) The yang construction, finally, a Malay-derived relative construction, is categorically post-nominal (cf. (4.15); Aye 2005: 116). (4.15) So saya tahu

so 1sg

already ada orang yang sudah Minum know already exist person rel perf drink

‘So, I knew there was someone who drank (it).’ (Aye 2005: 116) Table 4.5 provides a summary of the NP structures available in Bazaar Malay, along with the NP structures of the relevant input languages Malay and Hokkien. Upon comparison, the choice available between pre- and postnominal NPs in Bazaar Malay turns out to be a plausible result of language contact between Standard Malay and Hokkien.

76

4 Kenyan and Singaporean English

Table 4.5. NP structures in Bazaar Malay, Standard Malay and Hokkien according to Aye (2005: 104–26). Prenominal patterns have been underlined.

Possessives

Adjective phrases

Numerals

Prepositional phrases Demonstratives Relative constructions

Bazaar Malay

Standard Malay

Hokkien

Poss. — punya — N Poss. — N N — Poss. AP — punya — N AP — N N — AP Num — N

N — Poss.

Poss. — e — N Poss. — N

N — AP N — yang — AP

AP — N AP —e— N

Num —cl— N N — (yang) — Num — cl N — PP

Num —cl— N

PP — punya — N N — PP Dem — N N — Dem. N — Dem. Rel. cl. — punya — N N — yang — Rel. cl. N — yang — Rel. cl.

PP — e — N Dem. — cl. — N Rel. cl. — e — N

In addition, it is highly interesting that in actual speech, there are clear preferences with regard to these structural choices. Aye (2005: 241) finds that “the Hokkien-like constructions, that is, the punya construction and premodifers [sic] are most frequently used”; in addition, it is particularly educated speakers trained in Standard Malay who use postnominal possessives (cf. Aye 2005: 243). Finally, the pre-nominal relative construction using punya outstrips the Malay-derived postnominal yang construction (in Aye’s (2005: 249) data, there are 260 yang relatives but a mere 16 yang constructions). Consequently, despite the presence of post-nominal (head-initial) tendencies derived from Standard Malay, NPs in present day Bazaar Malay are dominantly modified prenominally. The NP structures of Baba Malay, too, must be seen as the result of a historical development from the head-initial structures of Standard Malay (cf. above) to the largely head-final nature of the modern variety. In this case, the historical development can be traced directly, thanks to Thurgood’s (1998) study of 19th century Malay, which is based on the scrutiny of a wide variety of 19th century texts, including narrative, the first Baba Malay newspaper, called Bintang Timor, which appeared from 1884–5 and 19th century grammars of Malay (cf. Thurgood 1998: 10). In the following, each structure will be described first in its Old Baba Malay form taken from Thurgood’s (1998) study, and, subsequently, in the modern Baba Malay form studied by Pakir (1986).

4.1 Singaporean English

77

Demonstratives in Old Baba Malay can be used both pre- and postnominally (cf. Thurgood 2001: 486). In present day Bazaar Malay, in contrast, the preferred position of demonstratives is prenominal (cf. Pakir 1986: 133; cf. also S. Lim 1988: 26): (4.16) ini

buku this book ‘this book’ (Pakir 1986: 133)

However, they may also be used in the traditional postnominal position, which, according to a study of taped casual conversations with nine speakers of Baba Malay by S. Lim (1988: 1; 47), applies in about 18 % of cases (no relevant examples are provided). Possessives in Old Baba Malay can be expressed both prenominally (using the Hokkien-derived puña construction, which is equivalent to the punya construction analysed in the section on Bazaar Malay (cf. above)⁶, and postnominally (cf. Thurgood 1998: 137). The very same applies to present-day Baba Malay (cf. (4.17), where puña is reduced to mia, and (4.18) for the postnominal variant; cf. also S. Lim 1988: 47). (4.17) si-gayoŋ

mia anak nmp-name pm child ‘Gayoŋ’s child (Pakir 1986: 141) gua house 1sg

(4.18) ruma

‘my house’ (Pakir 1986: 142) Adjective phrases in Old Baba Malay occur postnominally, in line with traditional Malay (cf. Thurgood 1998: 113). In present-day Baba Malay, both the pre- and postnominal positions are possible for adjectives (cf. Pakir 1986: 150). By means of the puña particle, adjective phrases can be used prenominally, as can be seen in (4.19); (4.20), shows a postnominal adjective phrase (cf. S. Lim 1988: 47)⁷.

6 I adopt Pakir’s (1986: 139) spelling for the Baba Malay form instead of Aye’s (2005: 104– 17) , which, in the present study, is used in order to refer to the corresponding expression in Bazaar Malay. 7 According to Pakir (1986: 151), this alternation between pre- and postnominal adjective phrases creates differences in pragmatic focus. Pendek mia baŋku in (4.19) puts a contrastive focus on the modifier pendek ‘low’ (implying: ‘a stool that is low, as opposed to another stool that is high’),

78

4 Kenyan and Singaporean English

(4.19) pendek mia

short

baŋku mpm stool

‘a stool that is low’ (Pakir 1986: 150) (4.20) baŋku pendek

stool

short

‘a low stool’ (Pakir 1986: 150) A relative clause, in Old Baba Malay, “follows the noun it modifies” (Thurgood 1998: 142). According to Pakir (1986), this postnominal relative yang construction is still possible in present-day Baba Malay, which, however, has also developed a prenominal variant: (4.21) gua bəli mia

I

baraŋ buy mpm thing

‘thing that I bought’ (Pakir 1986: 149) A summary of the data on NPs in Baba Malay can be seen in table 4.6. Once more, the data of two input languages of Standard Malay and Hokkien have been added for comparison. Again, the alternation between pre- and postnominal NPs fits in with Baba Malay’s nature as a contact variety. By way of language contact, the head-initial substrate of Standard Malay has undergone change through the head-final structures contributed by Hokkien. Old Baba Malay can be viewed as an intermediate stage, since only some head-final patterns have made inroads. Once more, however, in actual usage, pre- and postnominal patterns are not on a par in Baba Malay from a usage-based perspective. Both Pakir (1986: 142) and S. Lim (1988: 46–7) find “many more examples of the [Hokkien-like] possessive construction than the [Standard Malay] possessive construction”⁸. Consequently, present-day Baba Malay, too, can be viewed as a largely head-final language. Both Bazaar Malay and Baba Malay, despite their head-initial beginnings, have long been characterised by the dynamic competition of Malay (i.e. postnominal) and Hokkien/Sinitic (i.e. prenominal) structures, with the latter vastly dominant in present-day usage, which implies that contact influence has brought about

while baŋku pendek in (4.20) represents the unmarked order, which may be paraphrased as ‘a low stool’. 8 Pakir’s (1986: 142) assessment is based on random spot checks in every 20th of her 101 tape recordings (cf. Pakir 1986: 15). No further details are provided about the exact sampling procedure.

4.1 Singaporean English

79

Table 4.6. NP structures in Baba Malay, Old Baba Malay, Standard Malay and Hokkien according to Pakir (1986), S. Lim (1988) and Thurgood (1998). Prenominal patterns have been underlined.

Demonstratives

Possessives

Adjective Phrases

Relative constructions

Baba Malay

Old Baba Malay

Standard Malay

Hokkien

Dem — N

Dem — N

N — Dem.

Dem. — cl. — N

N — Dem Poss. — puña — N N — Poss. AP — puña — N N — AP

N — Dem Poss. — puña — N N — Poss. N — AP

N — Poss.

N — AP

Poss. — e — N Poss. — N AP — N

Rel. cl. — puña — N N — yang — Rel. cl.

N — yang — Rel. cl.

N — yang — AP N — yang — Rel. cl.

AP —e— N Rel. cl. — e — N

a large-scale typological transformation of NP structures in these varieties. This has two ramifications for the present study. First, among speakers who are bilingual in English and one of the Malay contact varieties, one can expect head-final contact influence from Bazaar Malay or Baba Malay on their English. Likewise, historically speaking, it can be expected that since the late 19th century, both Malay varieties have been turning more and more head-final, which may have had an impact on earlier stages of Singaporean English. Second, the case of Bazaar Malay or Baba Malay shows that two varieties have been influenced heavily by Sinitic languages in terms of the preferences of their speakers for either headinitial or head-final NPs. This allows the conclusion that the idea of a similar degree of influence of English is not unfounded.

4.1.2.2.2 NPs in varieties of Chinese In comparison to those of Bazaar Malay and Baba Malay, the NP structures of varieties of Chinese are strongly uniform and allow for safe typological generalisations, not least because of one of the most characteristic typological properties of Chinese languages being the tendency to “place modifying elements before the modified element” (Dryer 2003: 52).

80

4 Kenyan and Singaporean English

In Cantonese, “the noun phrase is rather strictly head-final” (R. S. Bauer and Matthews 2003: 146) , with possessive constructions, demonstratives, attributive adjectives (which are “normally followed by the linking particle ge” (Matthews and Yip 2011: 180)), as well as relative clauses all occurring in prenominal position. Matthews and Yip (2011: 477) mention but one exception from this general pattern: Numerals may occur post-nominally in a construction “used especially for commodities, [which] emphasizes the number, for example in advertising”): (4.22) gó

sāam go sailouhjái ge fuhmóuh those three cl children lp parents

‘those three children’s parents’ (Matthews and Yip 2011: 104) (4.23) lī go behngyàhn this cl patient ‘this patient’ (Matthews and Yip 2011: 107) haih go hóu hou-haak ge yàhn. your mummy is cl very hospitable lp person

(4.24) Léih māmih

‘Your mum’s a really hospitable person.’ (Matthews and Yip 2011: 180) (4.25) jā-chē fāan hohk ge hohksāang drive-car return school lp student ‘students who drive to school’ (Matthews and Yip 2011: 105) NPs in Hokkien, too, conform to this head-final pattern, which extends, e.g., to adjective phrases (cf. (4.26)), numerals (cf. Bodman 1955: 42), possessives (cf. Bodman 1955: 324) and relative constructions (cf. (4.27)): (4.26) cīn

hoù ē lû-tiăm very good e hotel ‘very good hotels’ (Bodman 1955: 68; glosses by Aye 2010: 3.2)

(4.27) Tān ī-siêng khiā ē chŭ

Tan doctor live

e house

‘the house that Dr. Tan lives in’ (Bodman 1955: 130; glossed and slightly abridged by Aye 2005: 256)

4.1 Singaporean English

81

Teochew, too, can be expected to be head-final on typological grounds (cf. Dryer 2003: 52). This is confirmed by Yeo’s (2011) account⁹. For instance, both Teochew possessives and relative clauses are pre-nominal: (4.28) a

dzik²-kə³³ thɔɪ̃ ⁵³ tsɛ¹¹ …a ʊaŋ³³laɪ ⁵⁵ si³³ intj enter-comp.dir see a while intj actually be ŋɪaʊ ³⁵tshə¹¹-kaɪ¹¹ khɪa¹¹kɛ ³³ mouse-gen house

‘Ah, (let us) enter and look, it actually is the mice’s home.’ (Yeo 2011: 54) (4.29) a naŋ⁵³ thɔɪ̃ ⁵³ hi³³kai¹¹ tshɪʊ¹¹ aʊ³³bʊɛʔ ⁵ u¹¹ tseɪk² kaɪ ³³ intj 1pl see dem tree behind have one cl tseɪŋ³³-tʊa¹¹ tsaŋ³³ kaɪ¹¹ tshɪʊ¹¹ int-be.big cl nom tree ‘Ah, we see (that) behind that tree there is a tree that is very big.’ (Yeo 2011: 55) Unsurprisingly, in Mandarin, too, “modifiers must precede their heads” (Li and Thompson 1981: 19). This extends, for instance, to so-called “associative phrases” (Li and Thompson 1981: 113), i.e. prenominal modifiers attached to a head using the particle de, which cover a range of meanings comparable to of -genitives in English (cf. (4.30)), adjective phrases (cf. (4.31)) and relative clauses (cf. Li and Thompson 1981: 116). (4.30) tùzi

de ĕrduō rabbit gen ear

‘rabbit’s ear’ (Li and Thompson 1981: 113) (4.31) tā shi yi ge hăo rén 3sg be one cl good person ‘He/she is a good person.’ (Li and Thompson 1981: 117)

9 Yeo’s (2011) Bachelor’s thesis written at Nanyang Technological University in Singapore is one of the few descriptions of Teochew through the medium of English. The supervisor of Yeo’s (2011) BA thesis, Dr Alexander R. Coupe, Associate Professor at Nanyang Technological University, considers her results reliable as “[h]er heritage is Teochew, and […] she used members of her family as consultants, […] [and] she at least has a passive understanding of Teochew if she isn’t a native speaker” (Coupe, Alexander R., p.c. June 5, 2014).

82

4 Kenyan and Singaporean English

4.1.2.2.3 Compounding in the languages of Singapore Although the focus of the present study is on syntax rather than morphology, a brief cursory look at Chinese morphology suggests itself, considering that in NP structures, the boundaries between morphology and syntax can be described as blurry (cf. 5.1.4.1). Nominal compounding in varieties of Chinese is described as “the most productive means of word formation (approximately 80% of Chinese words are compounds […])” (Basciano, Kula, and Melloni 2011: 209). Likewise, Halliday (1967: 9) identifies nominal compounding in Chinese as the prime strategy of vocabulary extension, whereas the borrowing of terms from other languages plays a minor role. Relevant Mandarin Chinese compounds are shown in (4.32)–(4.33). (4.32) fēi-jī

fly-machine ‘airplane’ (Post 2007: 152) (4.33) mù-dì eye-target ‘purpose; aim; goal’ (Post 2007: 152) Quite obviously these compounds exhibit the same kind of head-final structure as the NPs discussed above. This shows that head-final structures are not only present in Singapore in the shape of prenominal modification in the NP, but also with regard to nominal compounding. From a cognitive, usage-based point of view (cf. 1.3.2.2), it is not implausible that such schematic structures would carry over from morphology to syntax (cf. 5.1.4.1)¹⁰.

4.1.2.2.4 Weighing the typological evidence: NPs in the languages of Singapore Summing up, the evidence is strong that NPs in Singaporean English have been under primarily head-final influence for a long time. In the variety’s early, formative stages (i.e. Age I in figure 4.1), it came from Hokkien, Cantonese and Teochew; early head-initial influence from Bazaar Malay and Baba Malay soon yielded to Sinitic pressure as these languages developed head-final structures themselves. In phases II and III (cf. figure 4.1) i.e. as of the 1970s, however, with the demise of Bazaar Malay and Baba Malay and the rise of Mandarin, Sinitic languages have

10 This structure, however, has not carried over to Bazaar Malay and Baba Malay, both of which have head-initial compounds (cf. Aye 2005: 92).

4.2 Kenyan English

83

been by far dominant, which means that, once more, the majority of NPs contributed to the feature pool by local languages were head-final.

4.2 Kenyan English 4.2.1 Kenyan English as a variety 4.2.1.1 The history and evolution of Kenyan English In this section, the evolution of Kenyan English will be traced on the basis of the phases of development suggested in Schneider’s (2007: 189–97) Dynamic Model (cf. 2.1.2; for the history of Kenya, cf. Bennett and A. Smith 1976, Hemphill 1963, Marsh and Kingsnorth 1972 or Mwanzi 1985).

(a) Phase 1 (1860s–1920): Foundation — It is assumed that the foundational phase 1 of the evolution of Kenyan English started in 1860 and ended with the establishment of the colony in 1920 (cf. Schneider 2007: 189). As the English speech community was extremely small in these earliest phases (which was marked, especially, by missionary activities), language contact between the stl and the idg strand must have been rare and casual. Early mission schools established by orders used mainly local languages (cf. Gorman 1974:404). The stl strand, however, grew sizeably during the 19th century, as the number of British settlers rose to 9,000 by 1914 — in line with the initial conceptualisation of Kenya as a settlement colony (cf. Schneider 2013: 137). Towards the end of the period, restricted bilingualism must have arisen, which is the prerequisite for any form of language contact (cf. Schneider 2007: 191). (b) Phase 2 (1920–late 1940s): Exonormative stabilisation — The establishment of the colony in 1920 helped to stabilise both the situation of the settlers and the variety of English in Kenya, which was now used regularly in many new official domains such as business, law and colonial administration (cf. Schneider 2007: 191). Such forms of interaction between the stl and the idg strands must have lead to the slow emergence of bilingualism (cf. Schneider 2007: 38). This development, however, has to be viewed against the backdrop of British language policy. Owing to the theory of indirect rule developed by Lugard, which advocated respect for indigenous cultures, English was never forced upon Kenyans; instead, the British government adopted a policy of respect towards local languages (cf. A. M. Mazrui and A. A. Mazrui 1996: 273). Kiswahili was used widely even by British officials who regularly interacted with non-English speak-

84

4 Kenyan and Singaporean English

ing members of the public. In addition to these philanthropic aims, the caution was due to fears that teaching the language of the rulers to the African population at large would foster resistance movements (cf. A. M. Mazrui and A. A. Mazrui 1996: 274–5). Consequently, “the presence of a strong British settler community was initially a curse rather than a blessing to the spread of English” (A. M. Mazrui and A. A. Mazrui 1996: 275). Kenya is quite unusual in terms of the fact that for most of the time, more members of the stl strand must have been bilingual than members of the idg strand (cf. Schneider 2007: 191–2); among Kenyans, bilingualism was restricted to the upper echelons of society for an extended period of time. In spite of that, a range of lexical borrowings hail from this period, which, quite typically, stand for objects of local significance, including baobab “[a] tree (Adansonia digitata) with a stem of enormous thickness” (Simpson and Weiner 1991: 105), simba ‘lion’, or kuan ‘boiled cornmeal’ (cf. Schneider 2007: 192). These loan words attest to the beginning of structural nativisation in this period. (c) Phase 3 (late 1940s–today): Nativisation — The cut-off point for phase 3 may have been the late 1940s (cf. Schneider 2007: 192). The step into this phase can be put down to three major forces. First, since the 1940s, the British had a general sense of the end of the colonial period. In addition, events like the Mau Mau rebellion, a lengthy bloody altercation between the British administration and the underground Mau Mau movement between 1952 and 1956 (cf. Bennett and A. Smith 1976: 135), and, of course, Kenya’s political independence in 1963 (cf. Schneider 2007: 192) caused “the identities of the country’s two major populations to crystallize and their relationship to polarize” (Schneider 2007: 192), which is a crucial prerequisite of phase 3. Second, British language policy ensured that the use of English was not obliterated in the process. After World War II, for one thing, it was decided to abandon the practice of teaching three languages at schools and, instead, focus on the combination of English and a local vernacular (cf. Nabea 2009: 124). This made Kiswahili lose its key role, which, in turn, led to English making even more inroads (cf. Githiora 2008: 241). The British boosted their efforts to spread English and targeted much wider swaths of Kenyan society, in an attempt to create a local élite in the face of Kenya’s imminent independence. By the early 1950s, English had been firmly established both in primary and intermediate schools (cf. A. M. Mazrui and A. A. Mazrui 1996: 276). After Kenya’s independence in 1963, the report of the Ominde commission (1964) recommended the continuation of English language teaching from primary school onwards, while Kiswahili was marginalised further (cf. Muthwii 2004: 36). Reasons for Kenya’s

4.2 Kenyan English

85

commitment to English included its lack of ethnic associations and the lack of teaching materials for indigenous language (cf. Muthwii 2004: 36). At the government level, English was declared an official language to be used for documentary purposes (cf. Nabea 2009: 125), in upper courts and in parliament (cf. A. M. Mazrui and A. A. Mazrui 1996: 287). An attempt to make Kiswahili the sole parliamentary language in 1974 failed due to practical problems in implementing multilingualism, and English was re-introduced in 1979 (cf. A. M. Mazrui and A. A. Mazrui 1996: 288–9). Third, the spread of English was due to the favourable reaction of the population (cf. Gorman 1974: 430). Once more, a short overview of the linguistic features of Kenyan English will help to put its analysis as a nativised variety in phase 3 to the test.

4.2.1.2 Linguistic features of Kenyan English (a) Phonetics and phonology — The vowel system is marked e.g. by the Bantuinduced levelling of vowel contrasts, which leads to a merger of the fleece/kit and the goose/foot lexical sets and the avoidance of diphthongs (cf. Schmied 2004b: 927). From a phonotactic point of view, Kenyan English shares with many other New Englishes a tendency for consonant cluster reduction and syllabletimed rhythm patterns (cf. Schmied 2004b: 929–20). (b) Lexis — A wide range of Kenyan coinings has been documented, such as joblessly, pedestrate ‘walk’, overlisten, young husband ‘son’ etc. Skandera (2003: 159–61). Skandera (2003: 159–61) has unearthed idioms specific to Kenyan English, including talk nicely ‘bribe’ (159) or youth-winger ‘young party member’ (160). Mwangi (2003: 235) detects idiosyncrasies in the usage of prepositions. (c) Grammar — The most salient grammatical features found include the extension of progressive forms to stative meanings, as in is smelling, an arbitrary choice between infinitive and -ing-form as in tried to walk/tried walking, the omission of articles, resumptive pronouns as in As for me and my house, we declared war on poverty, invariant question tags, and a degree of flexibility in word order (cf. Schmied 2004a: 930–6). Hudson-Ettle (1998: 173–4) finds an unusually high frequency of nominal that clauses in Kenyan English. (d) Register — Van Rooy, Terblanche, et al. (2010) analyse register differentiation in East African English. Using the results of Biber’s (1988) register studies of British English as a benchmark, they find, for one thing, a similar extent of register

86

4 Kenyan and Singaporean English

differentiation as in the native mother variety. Simultaneously, they note a “general undertone of greater formality” (van Rooy, Terblanche, et al. 2010: 346) across Biber’s (1988: 128–55) dimensions of register variation, which they attribute to the restriction of English to formal domains of language use, while informal situations are dominated by local vernaculars. Judging from this, it seems fair to say that in Kenyan English, “structural nativisation has been in full swing” (Schneider 2007: 195), as a wide range of features can be detected which deviate from the British norm. This, however, does not mean that the grammatical innovations have already brought about a stable, coherent new variety. In the following, a discussion of the sociolinguistic distribution and attitudes towards the variety will help to gauge the sociolinguistic status of linguistic features of Kenyan English more precisely.

4.2.1.3 The sociolinguistics of Kenyan English In this section, the sociolinguistic profile of Kenyan English will be analysed by addressing the patterns of multilingualism found in Kenya and the social distribution of Kenyan English, as well as language attitudes.

4.2.1.3.1 Multilingualism in Kenya In addition to Kiswahili and English, which have a status as “official languages” (cf. National Council for Law Reporting 2010: 14), the language ecology of Kenya involves more than 60 local vernacular languages belonging to different African language phyla. In what follows, the role of these languages will be characterised briefly.

(a) The local vernaculars, which belong to four African language phyla, are the languages learnt most frequently as L1s in childhood. The results of a 2006 survey shown in table 4.7 indicate that for 90.4% of respondents, the language learnt first during childhood was a local vernacular. To the present day, these languages dominate informal contexts (cf. also Heine and Möhlig 1980: 67). This is corroborated by table 4.8, which shows the home languages of Kenya according to the 2006 Kenya National Adults Literacy Survey. Quite obviously, the “Mother tongues” lead the pack with a share of 80.8%, while no other language has made major inroads in this respect.

4.2 Kenyan English

87

Table 4.7. Language first learned during childhood and still used according to the 2006 Kenya National Adults Literacy Survey (Kenya National Bureau of Statistics 2006) Language

%

English Kiswahili Mother tongue Others No answer

1.1 6.5 90.4 0.3 1.7

Sum

100.0

Table 4.8. Language spoken at home most of the time according to the 2006 Kenya National Adults Literacy Survey (Kenya National Bureau of Statistics 2006) Language

%

English Kiswahili Mother tongue Others No answer

1.8 17.1 80.8 0.3 0.0

Sum

100.0

(b) Kiswahili, in contrast, is not usually spoken as L1, but plays a major role as a nationwide vehicular language. According to Table 4.9, which shows the patterns of L2 knowledge according to Heine and Möhlig (1980: 61), back in 1968–70, Kiswahili was the most frequent second language, with 42% of Kenyans having only a command of Kiswahili and more than 25% of Kenyans knowing some combination of Kiswahili with English and/or local vernaculars¹¹. Kiswahili is used widely, both in informal and in formal domains. According to Githiora’s (2002: 177) survey of 294 respondents in Nairobi, it is by far the preferred language in everyday informal interactions in the urban context of Nairobi; neither Sheng, a mixed code drawing on Kiswahili and English (cf. 4.2.1.3.1) nor English plays a similarly prominent role. Kiswahili is also used as “the universal lingua franca in small-scale trade and media” (Githiora 2008: 245). At the same time, in formal domains of language use, Kiswahili, being an African language, plays a special role as a carrier of national identity (cf. Githiora 2008: 245); it is laid down in Kenya’s

11 More recent results than Heine and Möhlig (1980) could not be found in the literature reviewed.

88

4 Kenyan and Singaporean English

Table 4.9. Heine and Möhlig’s (1980: 61) overview of patterns of second-language knowledge in Kenya; based on surveys conducted 1968–70 Second languages known

%

none Vernacular (second languages) Kiswahili English Kiswahili & vernacular English & vernacular English & Kiswahili English & Kiswahili & Vernacular

33.7 0.5 42.0 0.3 7.7 0.2 10.3 5.3

Sum

100.0

constitution as both the “national” and the “official language”, and, along with English (and Kenyan Sign Language), serves as the language of parliament (cf. National Council for Law Reporting 2010). (c) Sheng and Engsh — Recent decades have seen the emergence of two local language varieties in Kenya, which go under the label of “Sheng” and “Engsh”. Sheng, which has emerged in Nairobi City since Kenya’s independence in 1963 (cf. Githiora 2002: 159) and has spread “to most […] urban areas of Kenya” (Meierkord 2009: 5), relies on Kiswahili in terms of its syntax, but is interspersed with loan influences from English, Dholuo and Maasai (cf. Githiora 2002: 159); it has been variably characterised as a code-switching variety, a pidgin, a slang (cf. Meierkord 2009: 5), a peer language, a dialect of Kiswahili or even a creole (cf. Githiora 2002). Engsh, which originates from the more affluent Westlands and Parklands areas of Nairobi and is still associated with higher classes (cf. Abdulaziz and Osinde 1997: 50), is more obviously based on English than Sheng, but likewise draws heavily on local languages (cf. Abdulaziz and Osinde 1997: 50). According to Schneider (2007: 48), such codes frequently accompany the identity constructions in phase 3 of the emergence of New Englishes; they are particularly likely to occur in situations where the idg strand still strongly relies on local languages, while the access to English is limited, as is the case in Kenya. The focus of this study, however, is the standard variety of Kenyan English identified above; neither Sheng nor Engsh will be discussed any further.

4.2 Kenyan English

89

(d) English — Assumptions made in the literature about the number of Kenyans having a command of English vary considerably. Kembo Sure (1991: 246) claims 5%, without, however, specifying his source, while Skandera (2003: 19) assumes 15%–20%, based on a rough estimation on the basis of previous studies. Heine and Möhlig (1980: 63), in turn, on the basis of their own sociolinguistic surveys (which are highly dated) suggest roughly 17%. The Centre for Immigration Studies finally, claims 80% (cf. Center for Immigration Studies 1996), which is, however, only based on a statement of a spokesperson from the Kenyan Embassy in Washington. These estimations, apart from varying widely, are problematic in being either speculative, anecdotal or out of date. There are no further empirical studies; in the 1999 and 2009 censuses in Kenya (cf. Kenya National Bureau of Statistics 2009b), no data on language knowledge or use were collected¹². Considering the overall impression of the vigorous spread of English in Kenya (cf. Schneider 2007: 193–4), excessively low estimations are probably wrong; at the same time, 80%, as cited above, seems overstated. The sociolinguistic status of English in Kenya is clearly that of an L2. As documented in table 4.7, for over 90% of respondents of the 2006 Kenya National Adults Literacy Survey, the L1 was one of Kenya’s local languages, while a mere 1.1 % of Kenyans grew up with English as their mother tongue (cf. Githiora 2008: 242; Skandera 2003: 18). Table 4.9 shows that 16.1% of Kenyans report a pattern of L2 knowledge involving English¹³.

4.2.1.3.2 The social distribution of English (a) Domains of use of Kenyan English — What is the functional role occupied by English in Kenya’s multilingual ecology? The domains of use established in the literature include education, where English plays a crucial role as a medium of instruction from class four onwards and is the primary language of examination (cf. Muthwii 2004: 36). In the Kenya Certificate of Primary Education ending the

12 An e-mail inquiry made at the Kenya National Bureau of Statistics (p.c. 20 March 2014) confirms that no studies on language use or knowledge in Kenya have been conducted in recent years. The 2006 Kenya National Adults Literacy Survey (cf. tables 4.7–4.8) represents the most topical data set on language use in Kenya available today. 13 Whiteley (1974b: 43;49) comes to similar conclusions with regard to patterns of language knowledge in Kenya. His data, which were collected between 1968 and 1969 (cf. Whiteley 1974a: 36), however, are not more up-to-date than Heine and Möhlig’s (1980).

90

4 Kenyan and Singaporean English

eight years of primary education (KCPE), all subjects (except for Kiswahili) are examined in English (cf. Gacheche 2010: 17). English is also used widely by radio stations (cf. Michieka 2005: 177–8), by the majority of TV stations and a proportion of newspapers and magazines (cf. A. M. Mazrui and A. A. Mazrui 1996: 289), and for purposes of advertising (cf. Michieka 2005: 179). In business, English has a strong position; “[q]ualifications for most jobs include proficiency in English” (Muthwii 2004: 37). Finally, English is both the major parliamentary language and the language for bills and political statements. According to the constitution, Kiswahili could be used for this purpose, too (cf. National Council for Law Reporting 2010: 71), but several consecutive governments have preferred English for their communication (cf. Githiora 2008: 250), while only few members of parliament opt for Kiswahili (cf. Githiora 2008: 247). In sum, English is used in all sorts of official and semi-official domains. Unlike Singaporean English, though, it does not seem to have made major inroads in people’s private lives, with a mere 1.8% of households using English on a day-to-day basis (cf. table 4.8). English in Kenya has spread down the social hierarchy only to a limited extent. (b) Social, regional and demographic tendencies — Apart from domains of use, other well-known determinants of English use and knowledge in Kenya are social, regional and demographic in nature. Age effects have been found in all relevant studies. Heine and Köhler (1981: 38) establish that back in 1969–70, the knowledge of English was particularly widespread in the younger generation. Among their respondents older than 39, less than 5% spoke English. These dated results are confirmed by Fink (2005: 53). Even more recently, SACMEQ II and SACMEQ III, two consecutive studies of the Kenyan school system based on data collected from 2000–2005 and 2005–2011 respectively, likewise attest to the importance of English in the youngest generation. As can be seen in table 4.10, an astounding 86.4% of pupils at Standard 6 between 2000 and 2005, and an even greater proportion, 90.9%, of pupils between 2005 and 2011, reported that they used English outside school on a regular basis (i.e. ‘sometimes’, ‘most of the time’ or ‘all the time’)¹⁴. The literacy rates in English also vary from region to region. However, contrary to Heine and Möhlig’s (1980) studies back in 1969–70, who find a massive decline in rural areas of Kenya (where rates are down to 3.2%), more recent stud-

14 The SACMEQ results shown in table 4.10 should be taken with a grain of salt, as it is not clear what proportions of the respondents reported using English ‘sometimes’, ‘most of the time’ or ‘all the time’. Furthermore, speakers tend to be unreliable in self-reports on their own choice of linguistic codes, particularly when asked about prestige languages (cf. Milroy and Gordon 2003: 210–1).

4.2 Kenyan English

91

Table 4.10. Proportions of pupils at Standard 6 reporting that they speak English language outside school ‘sometimes’, ‘most of the time’ or ‘all the time’ according to SACMEQ II (2000– 2005) and SACMEQ III (2005–2011)

Central Coast Eastern Nairobi North-Eastern Nyanza Rift Valley Western Kenya

% 2000–2005

% 2005–2011

80.9 95.8 85.1 93.2 96.4 89.8 84.4 88.1 86.4

88.0 93.1 88.6 95.3 97.8 94.8 89.5 90.9 90.9

ies state that over the last decades, English has been spreading vigorously even in such regions (cf. Skandera 2003: 19). The 2005–2011 survey data from the SACMEQ III project shown in table 4.10 likewise document that a sizeable proportion (way above 80%) of Kenyan pupils report that they use English regularly outside school, no matter what region they come from (cf., however, the problems noted in Footnote 14). This indicates that by now, the question of geographical region may actually be less significant with regard to language use than at the time of Heine and Möhlig’s survey, even though all estimations in this respect have to remain somewhat speculative (cf. A. M. Mazrui and A. A. Mazrui 1996: 284). There are also tendencies in terms of social class and ethnic group. Githiora (2008: 242), for instance, confirms that “the higher status groups [in Kenya] are more proficient in English”. Schneider (2007: 193) identifies the upper and middle classes as the focus of English use and knowledge. Heine and Möhlig’s (1980: 63) dated survey also indicates an ethnic stratification, with the knowledge of English most widespread among the Gikuyu, the Embu and the Kamba (due to high levels of schooling), which is confirmed by Kioko and Muthwii (2003: 133–74), who find that English is most likely to be used in the workplace among the Dholuo, the Oluluyia, the Gikuyu and the Kamba. It is clear that English in Kenya is restricted in its spread and entrenchment in several respects: It is a language of the young, which is spoken primarily in urban

92

4 Kenyan and Singaporean English

regions, it has associations with higher classes and, consequently, it is also restricted to certain ethnic groups.

4.2.1.3.3 Language attitudes The ambivalent nature of English in Kenya is summarised aptly by Kembo Sure (1991: 246), who states that “[i]n terms of numbers English appears to be a minority language, but its functions provide it with a unique status, which in other contexts can only be acquired by a (numerically) majority language”. It is hardly surprising that such a language brings about ambivalent attitudes. On the one hand, attitudes towards English are markedly positive in some respects. Kembo Sure (1991: 245) finds a generally positive stance among Kenyan pupils towards both Kiswahili and English, with clear indications that English is viewed instrumentally, while Kiswahili has an integrative function, much like a “symbol of Kenyan nationhood”. Likewise, Fink (2005: 63), in a questionnaire survey on language attitudes in Kenya, asks 119 Nairobi residents for a set of assessments, including, for instance, “the preferred language in radio news”, “the most trustworthy language” or “the most important language in 30 years’ time”. In seven out of ten cases, English is rated more positively than the local vernaculars, Sheng and Kiswahili; it is surpassed by the local vernaculars and/or Kiswahili only with regard to trustworthiness and friendliness and with regard to the language which the president and local leaders should use¹⁵. At the same time, there is also no shortage of criticism of English. One problem identified by many authors is its social divisiveness, which perpetuates the situation of the early days of colonialism, when “English education was given to a select few” (Kembo Sure 2003: 251). It is argued, for instance, that restricted access to English-medium newspapers and radio stations may lead to actual inequalities in Kenyan society (cf. Kembo Sure 2003: 252). Furthermore, Muthwii (2004: 37) identifies numerous problems in English education, such as a tendency for the denigration of the local vernaculars, conflicts between localised and external British norms as well as poor English language skills among pupils, which, in her eyes, create an inner-Kenyan “‘third world’ that languishes in semi-literacy and illiteracy” (Muthwii 2004: 39). Such and similar criticisms of Kenyan language policy, which have been voiced by numerous further authors (cf. A. M. Mazrui and A. A. Mazrui 1996; Kiarie 2004; Nabea 2009; Gacheche 2010; Orwenjo 2012) show that despite the positive attitudes noted above, the status of English in Kenya is

15 It should be noted that Fink’s (2005) study is based on data from Nairobi. It is likely that in rural regions, the preference for English will be less pronounced.

4.2 Kenyan English

93

by no means settled, and that attitudes, in perspective, must be characterised as “partly schizophrenic” (Schneider 2007: 194). Attitudes towards actual localised linguistic variants are likewise ambivalent. While some reports suggest that the awareness for speaking a separate variety of English is quite low (cf. Skandera 2003: 34), Zuengler (1982: 123), for instance, reports that “Kenyan English is not yet popularly accepted as a variety on its own” (cf. also Muthwii 2004: 38). Schmied (2004b: 924) points out that this does not seem to apply for pronunciation, which is confirmed by Mwangi’s (2003: 16) observation that teachers clinging to the British or American standard are an object of ridicule, while “many Kenyans would not hesitate to say that such people [using an authentic Kenyan accent] speak ‘good English’” (Mwangi 2003: 16). With regard to pronunciation, thus, it would seem that Kenyan English is developing first signs of endonormativity. There is also some anecdotal evidence with regard to the acceptance of local norms in grammar and syntax, with Skandera (2003: 37–8), e.g., citing a Kenyan informant who prefers Kenyan invariant tag questions over the British forms (cf. 4.2.1.2) or Mwangi (2003: 16) reporting that lexical innovations like best maid are likely to be accepted. Buregeya (2006), in a questionnaire survey among 188 Nairobi university students found that nine of 20 morphological, syntactic and lexical features (such as second born, secretarial for ‘secretarial work’ or anyhowly) were endorsed as correct by at least 60% of respondents (cf. Buregeya 2006: 213). The attitudes of Kenyans to localised linguistic variants, just like their views of English at large, oscillate between the poles of aversion and acceptance, which allows the conclusion that the variety is firmly established but not yet fully settled.

4.2.1.4 Summary: Assessing the degree of endo- and exonormativity in Kenya The foregoing discussion allows for a holistic assessment of the degrees of endoand exonormativity of Kenyan English, which, according to Schneider’s (2007) Dynamic Model has been in phase 3 since the late 1940s. There are two key properties of Kenyan English which show it to be firmly rooted in phase 3, and which confirm that it has not yet progressed very far towards endonormativity. First, there are its low speaker numbers in comparison to Kiswahili and its restriction to broadly formal and official domains of usage as well as its association with upper and educated classes, with one of the consequences of this unequal distribution of English being the emergence of mixed codes such as Sheng and Engsh. The opportunities for the development of the linguistic homogeneity typical of phase 4, thus, seem quite restricted. A second major argument for placing Kenya at the conservative end of phase 3 are the widespread reports about the lack of acceptance of Kenyan norms of English, the continued orientation towards an external norm

94

4 Kenyan and Singaporean English

(cf. 4.2.1.3.3) and “community-internal discussions of the adequacy of linguistic usage” (Schneider 2007: 43). While it is undeniable that there is a large number of palpable, recurring linguistic deviations from the British standard (cf. 4.2.1.2), they still sit uneasily with the ever-present British norm, and, at times, are still perceived, if implicitly, as deviations or even errors. Simultaneously, there are some indications pointing towards nativisation, including the considerable number of innovations documented at various levels of linguistic structure, which, in addition, seem to have been gaining acceptance slowly. Furthermore, from a language attitudes point of view, a range of studies describe the overall view of Kenyan English as markedly positive; in addition, it is enshrined in the constitution as an official language and firmly entrenched in the linguistic landscape of the country. Recent evidence shows that it may be starting to make inroads as a home language, and that the restriction of English to urban areas assumed in the traditional literature may be no longer valid (cf. 4.2.1.3.1). In sum, Kenyan English clearly lacks the homogeneity and stability of phase 4; Schneider’s (2007) evaluation of Kenyan English as a variety in phase 3 turns out to be fully warranted.

4.2.2 The local languages in Kenya The linguistic substrate of Kenyan English comprises more than 60 indigenous languages belonging to four African language phyla. With these languages being the prototypical L1s in Kenya, which are learnt first in childhood and, in later life, are vastly dominant in the informal domain (cf. 4.2.1.3.1), they can be expected to have a major influence on the structure of Kenyan English. In the following, I will provide a short overview of the basic properties, the typological classification, and the use of Kenya’s African languages and by a documentation of the structure of their NPs .

4.2.2.1 Kenya’s African languages: Speaker numbers, social, geographical and ethnic distribution Kenya, which has an estimated population of 45,925,301 (cf. CIA 2016, estimate for July 2015), is marked by an exceptionally high degree of linguistic diversity (cf. Heine and Nurse 2000: 4). Estimations of the number of indigenous languages spoken in Kenya range from 34 to 42 or 50 to 67 (cf. Whiteley 1974b: 27; Kiarie 2004: 56; Githiora 2008: 236; Lewis, Simons, and Fennig 2013). Being the most recent and comprehensive overview, the Ethnologue’s estimation of 67 languages will be used as a baseline in the present study, even though any such assessment

4.2 Kenyan English

95

is beset by the notorious problem of delineating languages and dialects (cf. Githiora 2008: 237)¹⁶. Out of the set of four African language phyla posited by Greenberg (1963), three are represented in Kenya: Niger-Congo, Afro-Asiatic and Nilo-Saharan, while there are no languages belonging to the Khoisan phylum; in addition, there is a minute set of Indo-European languages including English and a creole. Table 4.11 shows the affiliations of the individual Kenyan languages with these phyla along with speaker numbers and relevant percentages (computed on the basis of the 2009 census) according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013). The breakdown of speaker numbers of these languages indicates that Niger-Congo languages, all of which are from the Bantu subbranch of the large phylum, are vastly dominant in terms of speaker numbers, as they are spoken by more than 60% of Kenyans; Nilo-Saharan languages, in contrast, are used by 30.91%, and a mere 7.69% of Kenyans have a command of Afro-Asiatic languages, which, in the case of Kenya, are exclusively Cushitic, with the exception of one Arabic dialect, which belongs to the semitic subbranch (cf. table 4.11). Indo-European languages, including English, Gujarati, Konkani and Panjabi, and Nubi, a creole, clearly play a negligible role as local vernaculars (cf. table 4.11)¹⁷. Therefore, in a typological investigation of Kenyan English, the major crosslinguistic influences to be expected are from the languages of the Niger-Congo and the Nilo-Saharan phyla, with Afro-Asiatic languages playing a minor role. In what follows, these general tendencies will be elaborated on using social, geographical and ethnic criteria of individual languages. The speaker proportions of the Kenyan vernaculars gleaned from the Ethnologue (cf. Lewis, Simons, and Fennig 2013) are represented in table 4.12¹⁸.

16 The Ethnologue has been criticised on two counts: First, for a tendency of positing an excessively large number of languages because of using simplistic criteria of mutual intelligibility for the separation of languages and dialects, and, second, because of the use of out-of-date source material (cf. Paolillo 2006: Section 5). Despite these shortcomings, it is considered “good enough to be useful” (Paolillo 2006: Section 5) for the present purpose. 17 These proportions are largely in accordance with the figures provided by Heine and Möhlig (1980: 10). According to them, Niger-Congo (→ Bantu) languages are spoken by 65% of Kenyans; Nilo-Saharan (→ Nilotic) languages by 30%, and Afro-Asiatic (→ Cushitic) languages by 3%. 18 The speaker numbers in the Ethnologue are largely based on the 2009 census of Kenya (cf. Kenya National Bureau of Statistics 2009b; cf. also table 4.12). As the census did not include any explicit questions on language knowledge, it is likely that the Ethnologue’s figures are based on the census results on ethnic affiliation (cf. Kenya National Bureau of Statistics 2009a). With regard to the remaining data sources specified in table 4.12, “Larsen, SIL (1994)” refers to unpublished data collected by Iver A. Larsen, who is affiliated with the Summer Institute of Linguistics (SIL; cf. Larsen 2014). The data marked as “SIL”, “United Bible Societies” and “Bible

96

4 Kenyan and Singaporean English

Table 4.11. Kenyan languages by language phyla, with speaker numbers and proportions based on the Ethnologue (cf. Lewis, Simons, and Fennig 2013) Phylum

Corresponding Kenyan languages

Speaker number

Niger-Congo (Bantu)

Chichonyi-Chidzihana-Chikauma, Chidigo, Chiduruma, Dawida, Ekegusii, Gichuka, Gikuyu, Kamba, Kiembu, Kigiryama, Kimîîru, Kipfokomo, Kiswahili, Kitharaka, Kiwilwana, Kuria, Lubukusu, LuidakhoLuisukha-Lutirichi, Lukabaras, Lulogooli, Lutachoni, Mwimbi-Muthambi, Nyala, Olukhayo, Olumarachi, Olumarama, Olunyole, Olusamia, Olushisa, Olutsotso, Oluwanga, Sagalla, Singa, Suba, Taveta (35 languages) Dholuo, Keiyo, Kipsigis, Maasai, Markweeta, Nandi, Okiek, Omotik, Pökoot, Sabaot, Samburu, Terik, Teso, Tugen, Turkana (15 languages) Arabic Omani, Aweer, Borana, Burji, Daasanach, Dahalo, El Molo, Orma, Rendille, Somali, Waata, Yaaku (12 languages) English, Gujarati, Konkani/Goan, Panjabi/Eastern (4 languages) Nubi (1 language)

22,747,132

61.12

11,504,050

30.91

2,862,580

7.69

88,200

0.24

15,500

0.04

Nilo-Saharan (Nilotic)

Afro-Asiatic (Cushitic)

Indo-European Creole

%

Table 4.12. Speaker numbers of Kenyan languages according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013) and their percentages of the overall number of speakers. The percentages do not add up to 100 due to rounding. KNBS refers to the 2009 census in Kenya (cf. Kenya National Bureau of Statistics 2009b). Language Gikuyu Dholuo Kamba

Speaker number 6,623,000 4,044,000 3,893,000

Source

% 17.80 10.87 10.46

KNBS KNBS KNBS

Translation&Literacy” appear to be unpublished data collected by the respective organisations. If only a year is given, the Ethnologue does not specify any source. No more recent data on language use in Kenya could be found in the literature; other analyses (e.g. Owino 1999) use highly outdated sources.

4.2 Kenyan English

Table 4.12. Speaker numbers of Kenyan languages according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013) and their percentages of the overall number of speakers. The percentages do not add up to 100 due to rounding. KNBS refers to the 2009 census in Kenya (cf. Kenya National Bureau of Statistics 2009b). Language Somali Ekegusii Kipsigis Kimîîru Lubukusu Turkana Nandi Kigiryama Maasai P�koot Lulogooli Terik Luidakho-Luisukha-Lutirichi Kiembu Chiduruma Teso Keiyo Olunyole Oluwanga Borana Nyala Kuria Dawida Lutachoni Lukabaras Sabaot Samburu Chidigo Chichonyi-Chidzihana-Chikauma Markweeta Kitharaka Olumarachi Olumarama Tugen Suba Olushisa Kiswahili Olukhayo Olusamia

Speaker number 2,386,222 2,205,000 1,916,000 1,658,000 1,433,000 989,000 949,000 944,000 842,000 633,000 618,000 601,000 598,000 429,000 397,000 339,000 314,000 311,000 309,000 277,800 273,000 260,000 254,000 253,000 253,000 241,000 237,000 217,000 184,000 180,000 175,905 155,000 152,427 140,000 139,000 137,000 131,000 125,000 125,000

% 6.41 5.92 5.15 4.45 3.85 2.66 2.55 2.54 2.26 1.70 1.66 1.61 1.61 1.15 1.07 0.91 0.84 0.84 0.83 0.75 0.73 0.70 0.68 0.68 0.68 0.65 0.64 0.58 0.49 0.48 0.47 0.42 0.41 0.38 0.37 0.37 0.35 0.34 0.34

Source KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS Larsen, SIL (1994) KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS Larsen, SIL (1994) KNBS KNBS KNBS KNBS KNBS KNBS KNBS KNBS no source given KNBS KNBS

97

98

4 Kenyan and Singaporean English

Table 4.12. Speaker numbers of Kenyan languages according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013) and their percentages of the overall number of speakers. The percentages do not add up to 100 due to rounding. KNBS refers to the 2009 census in Kenya (cf. Kenya National Bureau of Statistics 2009b). Language Olutsotso Sagalla Kipfokomo Okiek Mwimbi-Muthambi Gichuka Orma Rendille Gujarati English Burji Taveta Kiwilwana Nubi Arabic, Omani Spoken Waata Daasanach Panjabi, Eastern Aweer Konkani, Goan Dahalo Yaaku Omotik El Molo Singa Sum

Speaker number

%

Source

122,000 100,000 95,000 79,000 70,000 70,000 66,300 60,000 50,000 24,300 23,700 21,000 16,800 15,500 15,000 12,600 12,500 10,000 8,000

0.33 0.27 0.26 0.21 0.19 0.19 0.18 0.16 0.13 0.07 0.06 0.06 0.05 0.04 0.04 0.03 0.03 0.03 0.02

3,900 400 50 50 8 0

0.01 0.00 0.00 0.00 0.00 0.00

KNBS United Bible Societies (1992) KNBS KNBS SIL (1980 ) SIL (1980) KNBS KNBS SIL (1995) 2006 KNBS KNBS KNBS KNBS 1995 KNBS KNBS SIL (1995) Bible Translation & Literacy 2014 1987 Brenzinger (1992) 1983 1980 Larsen, SIL (1994) no source given

37,217,462

100.01

As can be seen, the distribution of these languages is strongly skewed towards Gikuyu (17.80% — Niger-Congo), Dholuo (10.87% — Nilo-Saharan), Kamba (10.46% — Niger-Congo), Somali (6.41% — Afro-Asiatic), Ekegusii (5.92% — Niger-Congo) and Kipsigis (5.15% — Nilo-Saharan), which are the only languages spoken by more than 5% of the population. This confirms the dominance of the Niger-Congo and Nilo-Saharan phyla. The other languages, in contrast, are considerably less

4.2 Kenyan English

99

frequent, with 49 of them, for instance, having a share of less than 1% of the population¹⁹. Furthermore, the knowledge and use of these indigenous languages strongly depends on speakers’ affiliations with ethnic groups, and, by implication, on speakers’ associations with geographical regions of Kenya — Chiduruma is spoken by the Duruma, Ateso is the language used by the Ateso, and Kitharaka is the code of the Atharaka (cf. Lewis, Simons, and Fennig 2013), to give just a few examples; similar associations hold for each of the vernaculars. With regard to geographical factors, a (once more, highly dated) regional survey conducted by Heine and Möhlig (1980) indicates that languages belonging to the Nilo-Saharan phylum (i.e. Nilotic languages) are primarily spoken in the West, while more than the whole Eastern half is dominated by Cushitic (Afro-Asiatic) languages, and two small swaths in the south are dominated by Bantu (i.e. Niger-Congo) languages.

4.2.2.2 NPs in the languages of Kenya 4.2.2.2.1 Generalisations about NPs in African languages A first important clue about NP structures in Kenya comes from typological generalisations about the languages of Africa. According to the World Atlas of Language Structures (World Atlas of Language Structures; (WALS; Dryer and Haspelmath 2013)), the vast majority of African languages have postnominal adjectives, while languages with prenominal adjectives are few and far between. Most of them are restricted to a narrow belt spanning Cameroon, the Central African Republic, Sudan and Ethiopia. Heine’s (1976: 23) typological analysis of 300 African languages shows that this applies to all major NP dependents, including adjectives, demonstratives, numerals, genitives and relative clauses (cf. Leyew and Heine 2008: 22 for a similar conclusion). Languages in Kenya, thus, are spoken in a linguistic area whose hallmark is head-initial NPs. As this tendency is not equally distributed among African languages and language phyla (cf. Dryer 2011: 298), however, in the following, the NP structures in the individual language phyla spoken in Kenya (cf. 4.2.2.1) will be considered individually.

19 The overall number of speakers resulting from summing up all speaker numbers given in the Ethnologue amounts to 37,217,462. This sum is lower than the population of Kenya determined in the 2009 census (cf. Kenya National Bureau of Statistics 2009b), which is 38,610,097 and likewise considerably lower than the present-day estimation of Kenya’s population size, which is 45,010,056 (cf. 4.2.2.1).

100

4 Kenyan and Singaporean English

4.2.2.2.2 NPs in the Niger-Congo languages of Kenya Table 4.13 shows an overview of the NP structures of the Niger-Congo languages spoken in Kenya according to the Ethnologue, with a focus on a range of NP structures. It is clear at first sight that there is a strong tendency towards head-initial order. Adjectives, demonstratives, and, slightly less obviously (due to gaps in the literature), genitives, numerals and relative clauses occur post-nominally. Interestingly, this tendency is very pronounced in Gikuyu, Kamba and Ekegusii, which are spoken by more than 34% of Kenyans (cf. 4.2.2.1). Examples of head-initial NPs taken from Gikuyu involving adjectives, demonstratives, genitives and relative clauses are shown in (4.34)–(4.37). (4.34) nyũngũ

nene cls9pot cls9big ‘the big pot’ (Mugane 1997: 35) ĩno cls9pot cls9-dem

(4.35) nyũngũ

‘this pot’ (Mugane 1997: 34) (4.36) nyũngũ ya cũcũ cls9pot cls9-ass cls1grandmother ‘grandmother’s pot’ (Mugane 1997: 35) iria nene cls10pot cls10-rl cls10big

(4.37) nyũngũ

‘pots that are big’ (Mugane 1997: 35) NPs taken from Ekegusii can be seen in (4.38)–(4.40). All of them qualify as headinitial. (4.38) omo-gaambi omo-be

cls1-leader

cl1-bad

‘a/the bad leader’ (Oyori Ogechi 2002: 60) (4.39) e-karamu eye cls9-pen this ‘this pen’ (Oyori Ogechi 2002: 60) (4.40) omosani bwa Agoki

friend

of

Agoki

‘friend of Agoki’ (Oyori Ogechi 2002: 61) Another highly frequent Niger-Congo language, Kamba, likewise fits the headinitial pattern. Brutzer (2012. [1906]: 6), for instance, in his early Kamba grammar,

4.2 Kenyan English

101

Table 4.13. NP structures of the Niger-Congo languages of Kenya according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013); underlining indicates that the structure in question is head-initial;“NA” means that the respective piece of information is not available. Cf. also Brutzer (2012. [1906]) for Kamba, Oyori Ogechi (2002: 60–2) for Ekegusii, Appleby (1961: 12) for Lulogooli and Polomé (1967: 142–4) for Kiswahili. Language

%

Kamba 10.46 Ekegusii 5.92 Kimîîru 4.45 Lubukusu 3.85 Kigiryama 2.54 Lulogooli 1.66 Luidakho- 1.61 LuisukhaLutirichi Kiembu 1.15 Chiduruma 1.07 Olunyole 0.84 Oluwanga 0.83 Nyala 0.73 Kuria 0.70 Dawida 0.68 Lukabaras 0.68 Lutachoni 0.68 Chidigo 0.58 Kitharaka 0.47 Olumarachi 0.42 Olumarama 0.41 0.37 Olushisa 0.37 Suba Kiswahili 0.35 Olukhayo 0.34 Olusamia 0.34 Olutsotso 0.33 0.27 Sagalla Kipfokomo 0.26 Gichuka 0.19 Mwimbi0.19 Muthambi Taveta 0.06 Kiwilwana 0.05 Singa 0.00 Chichonyi- 0.49 ChidzihanaChikauma

N — ADJ

N — GEN

N — DEM

N — NUM

N — REL

N — ADPOS

N — ADJ N — ADJ NA NA NA N — ADJ N — ADJ

N — GEN N — GEN NA NA NA NA NA

N — DEM N — DEM NA NA NA N — DEM N — DEM

NA NA NA NA NA NA NA

N — REL NA NA NA NA NA NA

NA NA NA NA NA NA NA

NA NA N — ADJ N — ADJ N — ADJ NA NA N — ADJ N — ADJ NA NA N — ADJ N — ADJ N — ADJ NA N — ADJ N — ADJ NA N — ADJ NA NA NA NA

NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA N — GEN NA NA NA NA NA NA NA

NA NA N — DEM N — DEM N — DEM NA NA N — DEM N — DEM NA NA N — DEM N — DEM N — DEM NA N — DEM N — DEM NA N — DEM NA NA NA NA

NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA N — NUM NA NA NA NA NA NA NA

NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA N — REL NA NA NA NA NA NA NA

NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA prepositions NA NA NA NA NA NA NA

NA NA NA NA

NA NA NA NA

NA NA NA NA

NA NA NA NA

NA NA NA NA

NA NA NA NA

102

4 Kenyan and Singaporean English

states unanimously: “Das Adjectiv [sic] steht immer hinter dem regierenden Substantiv” (‘adjectives invariably follow the noun by which they are governed’). Two of his examples are shown in (4.41)–(4.42). The same applies for demonstrative markers, genitives and possessive markers, as well as relative clauses (cf. Brutzer 2012. [1906]: 8; 32; 33; 36–7). (4.41) múndu munéne

man

tall

‘a tall man’ (Brutzer 2012. [1906]: 6) ịnéne wound large

(4.42) ịtáụ

‘large wounds’ (Brutzer 2012. [1906]: 6) Furthermore, it is highly significant that the head-initial tendency is also strongly pronounced in Kiswahili, which is extremely widespread as a lingua franca (cf. 4.2.2.1). (4.43) shows the postnominal position of adnominal adjectives; (4.44) illustrates numerals. (4.45) exemplifies the default postnominal position of demonstratives, which, however, may also occur pre-nominally (cf. Schadeberg 1992: 16), which is typical of “direct style” (Polomé 1967: 143; cf. also Perrott 1976: 41). (4.46), finally, shows Kiswahili relative constructions (cf. Lipps 2011: 16–20). (4.43) maji

machungu water bitter ‘bitter water’ (Schadeberg 1992: 16) watatu children three

(4.44) watoto

‘three children’ (Schadeberg 1992: 16) (4.45) mtu yule man this ‘this man’ (Perrott 1976: 41) ambaye anakula person amba eat

(4.46) mtu

‘a person who is eating’ (Lipps 2011: 17) The remaining Niger-Congo languages shown in table 4.13 are not documented to a sufficient extent in the medium of English to allow conclusions about NP structures; it seems highly plausible from a typological point of view, though, that they will be largely head-initial, too (cf. also Dryer 2011: 298–307

4.2 Kenyan English

103

for a typological survey which confirms this tendency for the whole Niger-Congo phylum).

4.2.2.2.3 NPs in the Nilo-Saharan languages of Kenya Table 4.14 shows the structure of NPs of Nilo-Saharan (Nilotic) languages in Kenya according to the Ethnologue. Once more, head-initial structures are widespread. With the exception of demonstratives in Maasai, head-initial NPs are attested for all of the most frequent languages like e.g. Dhouluo, Turkana, Nandi, Maasai or Pökoot. In Dhouluo, for instance, numerals follow their heads (cf. (4.47)), as do adjectives (cf. (4.48)), possessives (cf. Okoth Okombo 1997: 59) and relative clauses (cf. Okoth Okombo 1997: 26; cf. Dryer 2011: 298–307 for a similar overview). (4.47) Orito

ndigni adek. He is guarding bicycles three ‘He is guarding three bicycles.’ (Stafford 1967: 10)

Table 4.14. NP structures of the Nilo-Saharan languages of Kenya according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013); underlining indicates that the structure in question is head-initial;“NA” means that the respective piece of information is not available. Cf. also Stafford (1967: 8;10;52); Okoth Okombo (1997) for Dholuo; Barton (1921: 45) for Turkana Language

%

Dholuo 10.87 Kipsigis 5.15 Turkana 2.66 Nandi 2.55 Maasai 2.26 Pökoot 1.70 Terik 1.61 Teso 0.91 Keiyo 0.84 Sabaot 0.65 Samburu 0.64 Markweeta 0.48 Tugen 0.38 Okiek 0.21 Omotik 0.00

N — ADJ

N — GEN

N — DEM

N — NUM

N — REL

N — ADPOS

N — ADJ NA N — ADJ N — ADJ N — ADJ N — ADJ NA NA NA NA NA NA NA NA NA

N — GEN NA N — GEN N — GEN N — GEN N — GEN NA NA NA NA NA NA NA NA NA

N — DEM NA N — DEM Suffix DEM — N Suffix NA NA NA NA NA NA NA NA NA

N — NUM NA N — NUM N — NUM N — NUM N — NUM NA NA NA NA NA NA NA NA NA

N — REL NA N — REL N — REL N — REL N — REL NA NA NA NA NA NA NA NA NA

prepositions NA prepositions prepositions prepositions prepositions NA NA NA NA NA NA NA NA NA

104

4 Kenyan and Singaporean English

(4.48) Dhiang’ ma - ber

cow

rel-good

‘a good cow’ (Okoth Okombo 1997: 36)²⁰ Despite the fact that gaps in the literature do not allow for statements about the remaining set of languages, the head-final tendency is likewise strong in the NiloSaharan phylum (cf. also Dryer’s (2011: 298–307) typological survey, which comes to the same result).

4.2.2.2.4 NPs in the Afro-Asiatic languages of Kenya Table 4.15, finally, shows the NP structures in the Afro-Asiatic (i.e. Cushitic) languages of Kenya, which are considerably less widespread in Kenya than NigerCongo and Nilo-Saharan languages (cf. 4.2.1.3.1). Quite obviously, there is again evidence for head-initial patterns, even though there are two counterexamples, as

Table 4.15. NP structures in the Afro-Asiatic languages of Kenya according to the Ethnologue (cf. Lewis, Simons, and Fennig 2013); underlining indicates that the structure in question is headinitial;“NA” means that the respective piece of information is not available. Cf. also Kirk (1905: 35;125); Saeed (1993: 121; 246) for Somali. Language

%

N — ADJ

N — GEN

N — DEM

N — NUM

N — REL

N — ADPOS

Somali Borana Orma Rendille Burji Arabic Omani Spoken Daasanach Waata Aweer Dahalo El Molo Yaaku

6.41 0.75 0.18 0.16 0.06 0.04

N — ADJ NA NA NA NA NA

NA NA NA NA NA NA

Suffix NA NA NA NA NA

NUM — N NA NA NA NA NA

N — REL NA NA NA NA NA

NA NA NA NA NA NA

0.03 0.03 0.02 0.00 0.00 0.00

N — ADJ N — ADJ NA NA NA NA

GEN — N N — GEN NA NA NA NA

N — DEM N — DEM NA NA NA NA

N — NUM N — NUM NA NA NA NA

NA N — REL NA NA NA NA

NA postpositions NA NA NA NA

20 Attributive adjective constructions like (4.48) can be viewed as “minimal one-word relative clauses, in which, […] the clitic ma functions as relativizer” (Okoth Okombo 1997: 36). This implies that truly attributive adjectives, as in English, do not exist in Luo.

4.2 Kenyan English

105

both Somali numerals and Daasanach genitives are pre-nominal. It is significant, though, that Somali, the most frequent Cushitic language, qualifies as clearly head-initial, as shown in (4.49)–(4.50). (4.49) dálká

wèyn country-def big ‘the big country’ (Saeed 1993: 121) (4.50) labá naagóod two women ‘two women’ (Saeed 1993: 182) (4.49) shows the postnominal position of adjectives. As can be seen in (4.50), however, Somali numerals are pre-nominal (cf. also Dryer’s (2011: 298–307) typological survey, which confirms that the head-initial tendency in Afro-Asiatic is less pronounced than in the other phyla). With their speaker proportions very small across the board (cf. table 4.11), the remaining Cushitic languages mentioned in table 4.15 can be considered less significant in terms of their typological influence on Kenyan English.

4.2.2.2.5 Compounding in African languages As stated in 4.1.2.2.3, it makes sense to look briefly into nominal compounding in this typological survey, as nominal compounds have been frequently argued to blend into syntactic constructions (cf. 5.1.4.1). In (4.51)–(4.53), examples of Swahili compounds, which are invariably head-initial, are shown. (4.51) mwanachama

child party ‘party member’ (Schadeberg 1992: 12) (4.52) msumeno knife teeth ‘saw’ (Schadeberg 1992: 12) (4.53) gari (la) moshi car (of) smoke ‘railway train’ (Schadeberg 1992: 12) Such constructions, however, are quite marginal in several respects. For one thing, they may actually be “shortened nominal possessives” (Schadeberg 1992: 12), i.e. a result of the omission of the possessive marker la shown in parentheses in (4.53). Furthermore, in “most Bantu languages […] this pattern of compounding

106

4 Kenyan and Singaporean English

tends to be no longer productive and the attested compounds might have rather idiomatic meaning” (Basciano, Kula, and Melloni 2011: 208). Bantu languages, quite generally, “prefer[…] other strategies of word formation (affixation or reduplication) over root compounding” (Basciano, Kula, and Melloni 2011: 204–5). In sum, N+N compounds in Bantu languages cannot be expected to have a major influence on English²¹.

4.2.2.2.6 Weighing the typological evidence: Head-initial NPs in Kenya’s languages In the preceding discussion, it was demonstrated that a wide variety of languages from three language phyla contribute head-initial NP structures to the feature pool in Kenya. Against the background of the insight that the whole of Africa prefers head-initial NPs (cf. 4.2.2.2), there is a great deal of evidence for the view that the lingueme pool in Kenya is dominated by head-initial NPs, in sharp contrast to the substrates of Singaporean English.

4.3 Singaporean English and Kenyan English in perspective The key properties of the varieties under study established in this chapter can be summarised in table 4.16. Table 4.16. Key properties of Singaporean English and Kenyan English

Dominant NP structures Evolutionary stage

Singaporean English

Kenyan English

head-final Phase 4 (nativisation)

head-initial Phase 3 (endonormative stabilisation)

As can be seen, Singaporean English and Kenyan English differ, for one thing, in terms of the dominant NP structures in their substrates. This difference will be capitalised on in the study of language contact phenomena in the varieties under study. Furthermore, there is a clear difference in terms of evolutionary stages, which is expected to lead to differences in the degree of simplification. As was

21 Compounds involving verbal elements, however, are not infrequent in Bantu languages. Basciano, Kula, and Melloni (2011: 221), for instance, cite Kiswahili m-pita njia ’passer-by’, which consists of a deverbal agentive noun m-pita ‘passer’ and njia ‘street’.

4.3 Singaporean English and Kenyan English in perspective

107

demonstrated above (cf. 4.1.1.4.2), the ubiquity of English in Singapore across domains of life, classes and social structures offers excellent chances for the acquisition and use of complex structures; in less advanced varieties, in which English is confined to certain domains, and where the educational system is less effective in providing language skills, simplification is likely to play more of a role.

5 The English NP — structure and variation This chapter addresses the structure and the variability of the English NP. At the beginning, a syntactic model of the NP will be developed. The following section will address the variability of NP structures in language use. Then, I will develop a model of NP complexity. The last section delineates the variable context to be addressed in the analysis of NPs in the present study.

5.1 The structure of the English NP Structural models of the English NP do not necessarily have intrinsic cognitive reality from Croft’s (2000) or Langacker’s (1987) cognitive-linguistic points of view, but are only valid by virtue of their degree of entrenchment (cf. 1.3.2.2). Despite that, a corpus-based study requires a degree of formalisation of the structure in question. For this reason, the development of a syntactic template is indispensable.

5.1.1 The model by Bache (2000) The approach used in the present study will be based on a functionalist model suggested by Bache (2000), which is shown in figure 5.1. It consists of four functional slots. The Determination zone serves to “signal the kind of reference involved in the expression of a nominal” (Bache 2000: 170). Bache (2000: 171) divides the Determination zone into three sub-zones which are defined mainly in terms of their relative positions: Central determiners are, as a rule, realised by articles, further possessive, indefinite or demonstrative determiners, or genitive constructions. They may be preceded by predeterminers such as such and what or by postdeterminers such as every in her every wish (cf. figure 5.1). The Modification zone is likewise divided into three sub-functions. Mod. I, such as usual, own or same, “help single out or quantify the referent of the construction in relation to some context” (Bache 2000: 235). Further examples include ordinals (first, next), cardinal numbers and quantifiers (two, many), comparative forms (older, smaller) and “Others” such as only, own, same, other, subsequent, chief, general. The slot called “Mod. II” is occupied by central descriptive adjectives such as deep, quiet, disdainful. Mod. III, in turn, “subcategorize the head they modify” (Bache 2000: 235). They can be realised by a variety of word classes, including “deverbal adjectives” (leading, internalized), colour adjectives (green, yellow) and many

such what

pre-

post-

his such

central-

the her the the my a a this any

Determination

last

best

usual own same

Specification (Mod. I) sound handsome beautiful little

Description (Mod. II)

Modification

English naval French

Classification (Mod. III)

Table 5.1. The structure of the NP according to Bache (2000: 160; 171; 238–9)

stock officer actress girl student, fool fool day luck

Categorisation

with the shy smile who left school early

(multi-functional)

5.1 The structure of the English NP

109

110

5 The English NP — structure and variation

denominal adjectives (industrial, presidential), as well as NPs (metal, silk). The Categorisation zone introduces the referent of the NP. Postmodifiers, finally, fulfil a variety of functions comprising determination (e.g. the sudden death of my father), modification (the little girl with the shy smile), categorisation (no additional staffing, academic or secretarial) and complementation (this very dull visit to her parents).

5.1.2 Evaluating Bache’s (2000) model Bache’s (2000) model recommends itself in a number of respects. First, he imposes a strictly functional interpretation onto the elements of the NP. Unlike e.g. Quirk et al. (1985), whose well-known model is mainly based on “word classes and semantic classes of premodifiers” (Mazaud 2004: 43), with the category of descriptive modifiers, for instance, defined by their realisation through “central gradable adjectives […], which satisfy all […] criteria of adjective status” (Quirk et al. 1985: 1338), Bache uses clear functional labels such as “description”, “categorisation”, etc., which relate the syntactic category in question to its typical communicative purpose. He also assumes that the order of prenominal elements from left to right stands in an iconic relationship to our experience of situations. Determiners, for instance, are put first because of their immediate relevance to the disambiguation of referents (cf. Bache 2000: 239). Second, the model has a complex hierarchical structure involving zones and sub-zones, which allows for a more fine-grained analysis of NP elements than e.g. the “flat” structure assumed in Halliday’s (1989: 159–62) functional model, where determiner-like elements such as the “numerative” and the “deictic” are situated at the very same level as modifying elements (cf. Breban 2010a: 18). One problem of Bache’s (2000) study is the close association assumed between the functional slots within the NP and particular phrasal classes. For instance, “[c]ategorization […] [is] associated with the head noun, determination with articles, pronouns and genitive constructions, and modification with adjectives” (Breban 2010a: 18). Such associations are, however, frequently overridden in actual usage. The central adjective small, for instance, which would qualify as a descriptive modifier (Mod. II) in Bache’s grammar, is used as a classifier (Mod. III) in (5.1), as, in this case, it does not primarily add a semantic feature ‘small’ but singles out an ad-hoc subclass of varieties of fuchsia (cf. Breban 2010a: 32–3). (5.1)

Although fuchsias on the whole look very similar to each other, each variety needs to be treated differently. Some varieties such as ‘Marily Olsen Nellie Nuttal’ and ‘Minirose’ need small, pot plant treatment while others such as

5.1 The structure of the English NP

111

‘Greenpeace Space Shuttle’ and many of the species are best grown in large pots or in the greenhouse border. The small varieties mentioned make excellent show plants where the pot size is restricted to 31/2 in diameter. (Breban 2010a: 33; from the Collins Cobuild Corpus.) Therefore, the interpretation of these slots should be “opened up” (Breban 2010a: 33) in order to allow for such unusual formal realisations. Secondly, the concept of Mod. I assumed in Bache’s (2000) model seems overly inclusive, as it comprises ordinal numbers and cardinal numbers, quantifiers, comparatives and superlatives of adjectives as well as “Others” such as only, own, same, other, subsequent, chief and general. In the present study, ordinal numbers, cardinal numbers and quantifiers such as two or many will be assigned to the determination zone, as they clearly have “secondary identificational values” (Breban 2010a: 28;; cf. also Davidse 2004; Langacker 1991: 81– 2). Comparatives and superlatives (better farming methods) will be kept in the slot intended for central adjectives, as, in my view, their descriptive function is dominant (cf. Breban 2010b: 258). The only elements which will be accepted in Mod. I are what Bache (2000) refers to as “Others”, i.e. a set of (marginal) adjectives, including only, own, same, other, subsequent, chief, general. Finally, Bache’s (2000) model lacks two specific slots. For one thing, he fails to accommodate the type of emphasiser shown below. (5.2) (5.3)

the total disappearance of the urban episcopate a sort of apprenticeship

Such emphasisers either boost or down-tone meanings from the perspective of the speaker; they should not be yoked together with the secondary determiners in Bache’s (2000) Mod. I as they have a clearly modifying function and lack the referential effect of secondary determiners. Therefore, a separate slot for emphasisers and downtoners will be assumed (cf. also Quirk et al. 1985: 1338; Ghesquière 2009: 314). Objective descriptors, another marginal subtype of premodifier, which can be seen in (5.4)–(5.6), are likewise not covered by Bache’s (2000) model. (5.4) (5.5) (5.6)

a large growing family (Feist 2012: 46) immensely gifted disabled Irish writers (Feist 2012: 46) little black iron fences (Feist 2012: 46)

What these elements share with Mod. II is their descriptive characteristic; they ascribe descriptive properties to the referent of the head noun. What they lack, however, is comparability: Neither *a large very growing family nor *very disabled

112

5 The English NP — structure and variation

Irish writers can be viewed as felicitous paraphrases of (5.4)–(5.5); the phrase little very black iron fences is grammatical, but instantiates a different adjective black, which does not refer to the mere visual property but to different shades of a colour. The non-gradability of such adjectives and participles can be attributed to their “perceptual meaning” (Feist 2012: 48), which makes them more objective than typical Mod. II adjectives, which encode more strongly subjective characterisations (cf. e.g. hopeless or ugly). In the way of explanation, Feist (2012: 49) suggests that, when talking about basic perceptions, “we do not construe the quality in degrees; we simply construe the quality (for example red, pregnant, depressed) as being present, not absent”. The prototypical elements realising such objective descriptors are colour adjectives, and adjectives referring directly to visual perception such as black, short or Byzantine; however, they also “grade off to more conceptual words” (Feist 2012: 47) such as old. Furthermore, they comprise ingand ed- participles, as shown in (5.4)–(5.5), as they, too, can be related to perception. By virtue of these differences to other modifiers, I will assign such elements to a separate slot and use the term “objective descriptor” (cf. Ghesquière 2009: 314).

5.1.3 An adapted model of the NP In what follows, I am going to present a structural template of the NP which uses Bache’s (2000) model as its starting point and takes on board the alterations developed in the previous section. Figure 5.2 shows the outline of the model, which consists of a zone of Determination and Premodification, the Head and a Postmodification zone. They will be discussed in turn.

5.1.3.1 The determination zone For the analysis of the Determination zone, I largely take over the classes of determiners suggested by Bache (2000) (cf. 5.1.1) and Quirk et al. (1985: 257). Predeterminers are defined positionally as the determiners occurring in the first position of the determiner zone. They include such elements as all, both, half, double, twice, three times, such, and what (cf. Bache 2000: 171; Quirk et al. 1985: 257). Central determiners are realised by definite and indefinite, possessive and demonstrative determiners, as well as genitives (cf. 5.1.1). Numeratives, finally, are defined in accordance with Quirk et al. (1985: 261). They comprise cardinal numerals (two, three), ordinal numerals (second, third), so-called “general ordinals” (Quirk et al. 1985: 261) (first, last), closed-class quantifiers (few) and open-class quantifiers (a lot of, a large number of ).

Premodification

Head

Postmodification

predet. central det. num. Emph. Mod. I Mod. II Objective descriptor Mod. III such large climatic changes in the Pleistocene

all this activity

a particularly pure form of amnesia […]

these profound words

the total disappearance of the urban episcopate

mankind’s sheer inability to communicate with each other

the same legal status

the particular learning deficits in the verbal and visuo-spatial domains the moving train

the red background

those withered thighs

the two array options

Determination

Table 5.2. A model of the NP geared towards the present study, based on Bache (2000) and the discussion in 5.1.2

5.1 The structure of the English NP

113

114

5 The English NP — structure and variation

5.1.3.2 The premodification zone Emphasisers such as total or veritable will be assigned to the leftmost modification zone (cf. 5.1.2). Mod. I, in turn, is defined as the class of “secondary determiners” exemplified in (5.7)–(5.8), which serve to support the determiner and help to identify the referent of the NP in a precise way (i.e. Bache’s “Others”; cf. 5.1.2)¹. (5.7) (5.8)

additional, certain, chief, entire, existing, following, further, main, present, previous, principal, remaining, other, only, opposite (Sinclair 1990) other, same, identical, complete, entire, whole, above, aforementioned, certain, customary, expected, famous, given, habitual, necessary, normal, notorious, obvious, odd, ordinary, original, particular, possible, probable, regular, respective, special, typical, usual, various, well-known (Halliday 1989)

Many of the adjectives shown in (5.7)–(5.8) also have descriptive uses. It can only be determined in context whether they function as Mod. I or e.g. Mod. II (cf. 5.1.2). Following Breban’s (2010) call for flexibility (cf. 5.1.2), I will also allow into this slot any other type of phrase used in this function, such as e.g. this type of (cf. Davidse, Brems, and de Smedt 2008: 150–1; cf. also 5.1.4.3): (5.9)

This type of dislocation caused much unemployment in the affected countries which could not compete as well as the Japanese.

Mod. II is prototypically realised by adjective phrases headed by central, gradable adjectives, which characterise the referent of the NP in a subjective way (cf. Quirk et al. 1985: 1338): (5.10) a powerful urge (5.11) a very rich network of knowledge (5.12) rich laymen within the Pictish society

1 This definition of Mod. I is equivalent to the slot called “post-deictic” in Halliday’s (1989: 159) functional analysis of the English NP.

5.1 The structure of the English NP

115

Their functional centrality in realising “modification at its purest” (Bache 2000: 239) is reflected by the central position which they take in the premodification zone (cf. 5.1.1)². Objective descriptors, a marginal and rare class of modifiers realised primarily by adjectives referring to direct visual impression such as blue, but also by participles such as growing or disabled, have a syntactic slot of their own (cf. 5.1.2). The Mod. III slot, finally, comprises NP constituents with a classifying rather than descriptive function. In (5.13), for instance, medical does not assign any descriptive property to the model referred to (in Bolinger’s (1967: 15) words, “referent modification”), but rather classifies it as a special type of model (“reference modification”; Bolinger 1967). Apart from adjective phrases (as in (5.13)), whose heads are frequently denominal, it is usually noun phrases (cf. (5.14)) and, occasionally, ing-forms (cf. (5.15)) that realise Mod. III (cf. Bache 2000: 243)³. (5.13) the medical model (5.14) village problems (5.15) running races (Feist 2012: 47)

5.1.3.3 The postmodification zone Postmodifiers may serve a range of functions, including determination, modification, categorisation and complementation (cf. 5.1.1). According to Quirk et al. (1985: 1239; 1244–60), the major types of postmodification include prepositional phrases, non-finite and finite clauses. Rarely, adverb phrases or postposed adjectives may be used. Postmodifiers can be stacked, which is particularly frequent in formal texts. In principle, two basic structures involving long, stacked postmodification are possible. The NP in (5.16) is postmodified by one relative clause only, into which, however, further postmodified NPs are embedded at successively deeper levels of embedding.

2 Cf. Wulff (2003), Feist (2009, 2012) and Bache (1978) for the ordering principles which apply when several descriptive adjective phrases are combined in Mod. II position (e.g. a long intelligent book). 3 I make a distinction between ing-forms such as running in (5.15) and ing-participles with true participle meaning, such as growing in a large growing family shown in (5.4) (p. 111), which will be classified as objective descriptors (cf. 5.1.2; cf. also Feist 2012: 47).

116

5 The English NP — structure and variation

(5.16) the epidemic which hit the Maasai community who had been a stumbling block to the traders who were flying the interior of the continent The type of structure in (5.16), however, needs to be distinguished from the case shown in (5.17), where several postmodifiers follow the NP head, which are independent of each other. (5.17) this guy at Embankment tube station who stood there in in skintight jeans

Such stacked postmodifiers are interpreted in a recursive way. The prepositional phrase at Embankment tube station refers to the head this guy, while the relative clause who stood there in in skintight jeans takes as its antecedent the whole syntagm guy at Embankment tube station (cf. Quirk et al. 1985: 1296–7; Dik 1978: 55–6). In the analyses in 7, such stacked postmodifiers will be referred to as Postmod. 1, Postmod. 2, Postmod. 3 etc. This numbering, however, is not meant to suggest a similar functional differentiation as the numerals I, II, III in Mod. I–III; instead, the numbers only refer to the sequential position of postmodifiers.

5.1.4 Syntactic phenomena and problems in the NP In this section, I am going to address a small set of much-discussed problems in NP syntax which require separate comment, including the difference between modification and compounding in N+N compounds, unusual formal realisations in the Premodification zone and so-called SKT-constructions involving nouns such as sort, kind and type.

5.1.4.1 Compounding vs. syntax in premodification It is a well-known fact that the sequences of NPs shown in (5.18)–(5.19) are ambiguous between a syntactic interpretation (with the first NP viewed as a modifier; a typical example interpreted in this way is shown in (5.18)) and morphological readings (i.e. compounds, with (5.19) showing typical examples). (5.18) hitchhiker chap (5.19) (a) steel bridge (Giegerich 2004: 1–2) (b) boat train (L. Bauer 1998: 66)

5.1 The structure of the English NP

117

(c) watch-maker (Giegerich 2004: 1–2) (d) girlfriend (L. Bauer 1998: 66) Linguistic criteria which can be used to differentiate syntactic and morphological cases include, among others, stress patterns, with compounds assumed to have fore-stress like ˈapple cake, while in syntactic constructions, the second component is stressed, like apple ˈpie (cf. L. Bauer 1998: 70) and the possibility to coordinate the modifying NP, which is permissible with syntactic constructions only, as in iron and steel bars (cf. L. Bauer 1998: 74; cf. also Giegerich 2004; Plag 2006; Plag et al. 2008; Pastor Gómez 2011). L. Bauer (1998: 67–77), however, debunks each of the criteria used, showing that neither orthography nor stress assignment nor any of the syntactic tests mentioned above allow a consistent judgement about the type of a given N+N construction. This, for him, presents a strong case for the “position that there is no clear division between these two sets of constructions” (L. Bauer 1998: 66) . For this reason, I will follow a practice adopted by several previous studies on NP structures which “accept spelling as a guide” (Jucker 1992: 68). Both Both Leech et al. (2009: 215) and Jucker (1992: 68), “as a matter of convenience” (Leech et al. 2009: 215), view as modifiers all pre-head elements which are separated from the head by a space .

5.1.4.2 “Lexical phrases” in the Premodification zone In the premodification slot, one also finds complex realisations which cannot be described in terms of “orthodox” phrase structures: (5.20) within-generation events (5.21) The supplementary information-gaining processes (5.22) Singapore’s ’learn from Japan policy’ (5.23) “true to scale” representations of data (5.24) a more semantics-based grammar Frequent structural templates of such constructions, which have been dubbed  “lexical phrases” (Fischer 1998: 21–2)⁴, include PPs (cf. (5.20)), participial con-

4 Fischer’s (1998: 21–2) term “lexical phrase” is derived from Nattinger and DeCarrico (1992: 1), although their concept of “lexical phrase” is not restricted to modifiers within NPs; “lexical phrases”, in their sense, comprise a wide range of fixed expressions and phraseological chunks such as that goes without saying, as it were and the like.

118

5 The English NP — structure and variation

structions (cf. (5.21); (5.24)), whole sentences (cf. (5.22)) or APs with complementation (cf. (5.23); cf. Fischer 1998: 48–9; Mazaud 2004: 38). The sense of unorthodox constructions is underscored by hyphens (as in (5.20)–(5.21) or (5.24)) or quotation marks as in (5.22)–(5.23); univerbation, in contrast, is rare (cf. Mazaud 2004: 26–7). In the literature, “lexical phrases” have been captured as constructions lying halfway between the poles of prototypical words and prototypical syntactic phrases (cf. Fischer 1998: 22), or as “(temporary) conversion” (Mazaud 2004: 26). Clearly, from a cognitive point of view (cf. 1.3.2.2), the most adequate view of such structures would involve prototype categories covering “orthodox” as well as such unusual premodification structures, with “lexical phrases” epitomising the cognitive view that the lexicon blends into syntax (cf. 1.3.2.2; cf. also Hilpert 2014: 79), even though a comprehensive analysis along such lines would go beyond the domain of this study. In the corpus analysis in 7, however, the ambiguity of lexical phrases will be taken into account by assigning them to a separate formal category of modifiers (cf. 5.2.2.2).

5.1.4.3 SKT-constructions (5.25)–(5.27) show so-called “SKT-constructions” (Keizer 2007: 152). (5.25) No they I mean if even in that sort of a situation if the police run in

(5.26) I like all this type of nonsense (5.27) the western kind of look The syntactic discussion about these constructions in the literature (cf. Denison 2006; Keizer 2007: 152–88; Davidse 2009; Brems and Davidse 2010) has revolved around the ambiguous status of the elements sort, kind or type. The following cases can be distinguished (cf. Davidse, Brems, and de Smedt 2008: 143–53): Lexical head uses, to start with, have number concord with the head (cf. Davidse, Brems, and de Smedt 2008: 143): (5.28) I can’t stand that kind of dog. (Davidse, Brems, and de Smedt 2008: 143) Uses as Mod. I, in turn, are frequently marked by a lack of number congruence with the head. They instantiate “more intricate deictic meanings and phoric relations than simple determiners” (Davidse, Brems, and de Smedt 2008: 151; cf. 5.29).

5.2 Variation in the English NP

119

(5.29) I mean I don’t associate you with uh you know one of these sort of skills like driving. (Davidse, Brems, and de Smedt 2008: 150) Uses as Mod. II involve a far greater range of modifying adjectives than lexical head uses and are characterised by referring to instances rather than subtypes (cf. Davidse, Brems, and de Smedt 2008: 147): (5.30) It’s a cool quirky kind of song. (Davidse, Brems, and de Smedt 2008: 147) Cases of Mod. III (cf. (5.31)) follow a classifying AP or NP (cf. Davidse, Brems, and de Smedt 2008: 149). (5.31) like with the European sorta look (Davidse, Brems, and de Smedt 2008: 149) In the corpus analyses in 7, SKT-constructions will be assigned to a separate formal category, and their functions will be determined according to the classification shown above.

5.2 Variation in the English NP The focus of the present study is the variability in the use of English NPs in two new English varieties influenced by divergent typological environments. An important prerequisite for any variationist study of a linguistic feature of this kind is a precise definition of the envelope of variation of the phenomenon in question. This section will address the variability of English NPs from two perspectives: first, by dealing with the formal realisation and complexity of NP modification, and, second, by addressing general factors which have been found to influence the choice between pre- and postmodification.

5.2.1 Factors influencing the form and complexity of modification 5.2.1.1 Language-internal factors In what follows, three major language-internal factors impacting on the complexity of NP usage will be addressed.

(a) Word class of the head noun — The structure of NPs is strongly governed by the word class of the noun realising the head. Typical heads which are frequently

120

5 The English NP — structure and variation

pre- and postmodified are common nouns (cf. Biber, Johansson, et al. 1999: 581), whereas pronouns of all kinds (personal pronouns, most demonstratives) have, on average, a quite weak tendency towards modification, as they refer to entities “known by the listener/reader” (Biber, Johansson, et al. 1999: 581); few pronouns are used more or less regularly with postmodifiers, such as indefinite pronouns (one) or those, while premodification is virtually excluded for pronouns with the exception of one or all. Proper nouns, too, are rarely modified, as they “refer[…] to a specific person, place, or institution” (Biber, Johansson, et al. 1999: 583). (b) Syntactic position — Another powerful determinant of NP structure is their syntactic position. F. G. A. M. Aarts (1971: 289) finds, first and foremost, that the subject position is strongly associated with structurally ‘light’ items, i.e. pronouns, names and simple NPs, with e.g. 73.70% of subjects (in F. G. A. M. Aarts’s (1971) corpus) realised by pronouns or names, while modified patterns are extremely rare (i.e. between 3.58% and 5.77%). The positions typically realised by modified NPs (with percentages ranging from 13.36% to 19.61%) are objects and adverbials (cf. also de Haan 1993: 86, who confirms F. G. A. M. Aarts’s result on the basis of more refined statistical techniques). Consequently, a categorical difference must be made between subjects and non-subjects. Similarly, Schilk and Schaub (2016: 68–9) find that an NP’s syntactic position is an excellent predictor of its complexity, with subjecthood making structurally simple NPs very likely. This tendency is closely associated with Behaghel’s (1930) Law of Increasing Constituents, which captures the cross-linguistic observation that long constituents tend to come after short ones. It can also be related to highly basic functional principles uncovered by Hawkins (1994: 17). He argues that this order is ultimately due to the processing capability of the human parser. When parsing a sentence, speakers seek to minimise the number of words which need to be processed before the immediate constituent structure of the clause is recognised. In doing so, they seek to keep low the number of “elements that must be held in working memory and that are relevant to a given parsing decision” (Hawkins 1994: 60). In (5.32), for instance, which, quite obviously, is at variance with Behaghel’s (1930) Law of Increasing Constituents, a total of eleven words have to be processed before the last constituent is reached, i.e. before the structure of the clause can be decoded. Sentence (5.33), in contrast, is much more felicitous as a mere four words need to be processed to arrive at the last immediate constituent.

5.2 Variation in the English NP

(5.32) I

VP [gave NP [the

1 find] 10 (5.33) I

121

PP [to

valuable book that was extremely difficult to 2 3 4 5 6 7 8 9 Mary]]. (Hawkins 1994: 57)

11

VP [gave

PP [to Mary] NP [the valuable book that was 1 2 3 4 extremely difficult to find]]. (Hawkins 1994: 57)

Speakers’ actual preference of the order represented in (5.33) over the order in (5.32) is borne out by a corpus analysis conducted by Hawkins (1994: 182). In 90.2% of 480 cases taken from a written corpus made up of novels which allow both structures, the EIC principle predicts the order of NP and PP correctly. In sum, Hawkins’s EIC principle provides a cognitively and functionally plausible explanation for the observation that long constituents follow short ones, and, by implication, for F. G. A. M. Aarts’s (1971) and others’ observation of the shortness of NPs in subject position. In addition, Hawkins’s (1994) theory makes it clear that the short-before-long prediction made by the EIC principle is peculiar to broadly head-initial languages, such as English (which may be viewed as a case of a “less consistent” (Primus 2001: 856) head-initial language). In consistently head-final languages such as Japanese, the opposite applies. Long constituents stand before shorter ones, which is likewise helpful for the human parser (cf. Hawkins 1994: 105). The typological twist given to this well-known word order preference is highly noteworthy with regard to the present study. If Singaporean English is primarily influenced by head-final languages, it may well be that the EIC principle will play less of a role than in Kenyan English, which is dominated by largely head-initial languages (cf. 4.2.2.2.6), (c) Semantic class of the head noun — One of the few studies to have addressed the influence of the semantic class of the head noun is de Haan (1989). He looks, for instance, into the proportions of NPs having premodifiers by the semantic class of the head noun, finding that NPs without any pre-head slot are significantly overrepresented among nouns referring to human beings (cf. de Haan 1989: 151). Abstract nouns, in contrast, have pre-head functions in the majority of cases, with a percentage of 62.8%. This appears plausible considering that abstract nouns are frequently used in formal contexts, where precise expression and high explicitness are called for. With regard to the present study, the conclusion must be that the semantic class of the head noun has an impact on modification patterns, with

122

5 The English NP — structure and variation

a clear line dividing concrete and animate nouns, which have generally low modification rates, from abstract nouns, which are frequently pre- and postmodified. 5.2.1.2 Language-external factors Apart from the language-internal, i.e. syntactic and semantic factors discussed in 5.2.1.1, there is a range of language-external, broadly sociolinguistic factors affecting syntactic variation in NPs.

(a) Modality — A variety of studies has found a fundamental difference between NPs in different modalities, i.e. in spoken and written discourse. Miller and Weinert (1998), for instance, compare NP structures in an oral narrative, an extract from spontaneous conversation and a written corpus consisting of readers’ letters (cf. Miller and Weinert 1998: 139). In both of the spoken registers, there is an excessively high percentage of personal pronouns (44.9% and 48.9% of all NPs respectively), while all types of modified NPs are few and far between, especially NPs involving PPs (6.0% and 5.6%) or relative clauses (3.2% and 3.8%). Readers’ letters, in contrast, are marked by a much lower proportion of personal pronouns, while the proportions of modified NPs, in particular, cases of adjectival premodification (19.7%) and modifying PPs (18.8%), are much higher. In the way of interpretation of such differences, Halliday (1987: 66) states that “[s]poken and written language […] tend to display different kinds of complexity”. Written texts are “marked by complexity in the nominal group” (Halliday 1987: 73), i.e. they frequently aim to compress large amounts of information into NPs by embedding other NPs. Spoken language, in contrast, typically relies on simple NPs, and its specific kind of complexity lies “in the clause complex” (Halliday 1987: 73). These differences can be put down, for instance, to a “continuum from most spontaneous to most self-monitored language” (Halliday 1987: 69), which ascribes the strongly condensed, NP-prone nature of written discourse to a high degree of self-monitoring by writers, while the preference for clauses in spoken language, in contrast, is rooted in a much more casual stance of speakers towards their own language production. Halliday’s (1987) considerations are confirmed by Schäpers (2009) on the basis of a corpus study of ICE-GB, with regard to the complexity of written NPs (cf. Schäpers 2009: 123); his idea of a complexity tradeoff, however, is debunked by Schäpers’s study, as in her corpus, the clausal syntax of written texts outdoes the complexity found in spoken language (cf. Schäpers 2009: 153).

5.2 Variation in the English NP

123

(b) Register — F. G. A. M. Aarts (1971) studies modification patterns of NPs in four different registers, viz. “light fiction”, scientific writing, informal speech and formal English (comprising both spoken and written data). His results boil down to a clear association of complex NPs with formal registers. Both scientific writing and formal English have a high rate of modified NPs (cf. F. G. A. M. Aarts 1971: 289), with e.g. 41.0% of subjects in scientific writing having a postmodifier, while formal registers avoid pronouns. De Haan (1993: 85) adds to this result that it is, in particular, the tendency to modify subjects which marks out non-fictional (formal) texts as against fictional (informal), where subjects tend to be simple. These tendencies are largely confirmed by Schilk and Schaub’s (2016: 69) study on the basis of multifactorial statistical models. In a study on British newspaper language, Jucker (1992) finds that NP modification depends on even more specific stylistic factors. In his corpus, the proportion of modified NPs varies markedly between down-market papers (where it turns out to be low) to up-market papers, where it is exceptionally high (cf. Jucker 1992: 115), and from newspaper section to section, with e.g. the sports section in upmarket papers having a particularly low number of modified NPs (cf. Jucker 1992: 129). In addition, even individual subtypes of pre- and postmodifiers co-vary with newspaper sections: High proportions of descriptive premodifiers (such as grim and gloomy) are associated with the sports and the arts sections, as their function is “descriptive and characterising” (Jucker 1992: 150; cf. Jucker 1992: 151), which fits in particularly well with these registers (cf. Jucker 1992: 153–4); articles on foreign news, home news and business, in contrast, are marked by a high incidence of classifying premodifiers (a Scottish prison), which are “non-emotive” (Jucker 1992: 149) and do not require personal judgement (cf. Jucker 1992: 154). Classifiers (“Mod. III” in 5.1.3) also show characteristic formal variation. According to Jucker, adjectival classifiers like renal in renal disease tend to evoke a more formal, educated register than nominal ones such as kidney in kidney disease; This fits in with the high share of nominal classifiers in down-market papers in comparison to the adjectival realisations in up-market ones, which aim at a more educated readership. The strong dependence of NP structure on register (and even thematic) preferences, thus, shows that it is crucial to control for register when studying NPs. This will be taken into account in the corpus study below.

124

5 The English NP — structure and variation

5.2.2 Factors influencing the choice between pre- and postmodification This section is dedicated to factors which have been found to influence speakers’ and writers’ choice between pre- and postmodification. Language-internal and language external factors will be considered in turn.

5.2.2.1 Language-internal factors impacting on the choice between pre- and postmodification The interchangeability between pre- and postmodification in English is strongly limited as many post-head structures lack a pre-head variant. For instance, relative clauses, PPs or that-clauses will only ever occur in post-head position in standard English, which can be put down to their length (cf. 5.2.1.1); the pre-head position, hence, is clearly only available for relatively short elements such as short NPs, adjective phrases etc. With regard to the whole range of NP modification patterns, there are hardly any other general factors which determine the choice between pre- and postmodification. It is only for a range of highly specific subtypes of NPs that such factors have been made out, including e.g. the alternation between a nominal pre-head classifier and the corresponding of -phrase shown in (5.34). (5.34) (a) the terror balance (b) the balance of terror (Sørensen 1980: 78) With regard to such cases, determinants have been made out like the “reduced explicitness” (Quirk et al. 1985: 1330–1) of premodifiers (the Lincoln road allows e.g. directional meanings ‘the road to Lincoln’, but does not cover the local meaning of by in the tree by a stream; cf. also also Sørensen 1980: 79–80), the “relative impermanence” (Quirk et al. 1985: 1331) conveyed by the premodified structures (the man in the corner cannot be paraphrased by the corner man), and information structure (response probability is more likely than the probability of a response if the response has already been mentioned). Further relevant factors are known to play a role in all kinds of syntactic alternations, such as rhythm, with a preference for an “ideal weak-strong alternation” (Rosenbach 2014: 232) or persistence, which implies a preference for a given construction over its alternant if the construction has occurred in the preceding context (cf. Rosenbach 2014: 232). These factors, however, are only helpful for the analysis of specific subtypes of NPs, while they do not play the same role for NPs in general. In the corpus study in 7, thus, it will be important to take into account that factors which govern the choice between pre- and postmodification at the level of the whole NP cannot be

5.2 Variation in the English NP

125

controlled for comprehensively and that the interchangeability of pre- and postmodifiers is limited.

5.2.2.2 Language-external factors impacting on the choice between pre- and postmodification The language-external factors known to influence the choice between pre- and postmodification are, in the main, register, modality and variety. With regard to register and modality, Biber, Johansson, et al. (1999) find, for one thing, that, across registers, premodification is somewhat more frequent than postmodification; at the same time, it emerges that this difference is more pronounced in conversation and fiction than in the most formal written register, academic prose, where the relationship is roughly 25% vs. 20% (cf. Biber, Johansson, et al. 1999: 578). Furthermore, multiple postmodifiers turn out to be a hallmark of written, and, especially, academic written discourse, where they have a frequency of more than 5000 per million words, while they amount to roughly 500 in conversation and roughly 2800 and 3200 respectively in fiction and news (cf. Biber, Johansson, et al. 1999: 642; cf. also Schäpers 2009: 114). Apart from register, a range of studies have also addressed NP structures with regard to their variation between regional varieties of English. Mazaud (2004) focuses on so-called “complex premodifiers”, i.e. “constructions consisting of two or more words that are used in the slot between determiner and head noun of a noun phrase” (Mazaud 2004: 24): (5.35) the once-highflying Nasdaq (5.36) her two university age children (5.37) commando style course of exercise Her corpus comprises newspaper articles from Great Britain, Ireland, the US, Canada, Jamaica, Australia, New Zealand and Singapore. Table 5.3 shows the overall frequencies of complex premodifiers in these varieties. Apart from the fact that British newspapers have a lower frequency of complex premodifiers than US newspapers (cf. Mazaud 2004: 243; 247), it is particularly relevant for the present study that the variety which leads the pack in terms of complex premodification is clearly Singaporean English. Not only does it have the highest token frequency of complex premodifiers of all varieties involved (cf. table 5.3), some of which are exemplified in (5.38)–(5.40).

126

5 The English NP — structure and variation

Table 5.3. The token frequencies of complex premodifiers across newspapers (cf. Mazaud 2004: 242) — frequencies normalised per 100,000 words Newspaper

N

region

Scotsman 148.55 Great Britain Mirror 156.25 Great Britain Jamaica Gleaner 167.67 Jamaica Times 182.32 Great Britain Sydney Morning Herald 198.81 Australia Irish Times 203.90 Ireland New York Times 219.93 US New Zealand Herald 235.95 New Zealand Age 283.23 Australia Globe & Mail 287.03 Canada USA Today 292.63 US Straits Times 308.24 Singapore

(5.38) feel-good stories (Mazaud 2004: 332) (5.39) ‘brick-and-mortar’ branches (Mazaud 2004: 332) (5.40) shared-use programme (Mazaud 2004: 336) Singaporean English also stands out in terms of a high number of original formations (cf. Mazaud 2004: 255), many full sentences in premodifier position (cf. (5.41)–(5.43)) and, incidentally, by featuring the longest premodifier in the sample, which is shown in (5.44). (5.41) her ‘just-too-bad-isn’t-it’ throwaway shrug (Mazaud 2004: 340) (5.42) his never-say-die spirit (Mazaud 2004: 340) (5.43) its ‘I love this game’ advertisements and half-time shows (Mazaud 2004: 340) (5.44) The I-have-conquered-the-world-and-I-am-so-clever-because-I-canplay-golf swagger (Mazaud 2004: 340) Mazaud (2004: 267–8) interprets the high proportion of complex premodifiers in two ways. For one thing, she suggests that it may be due to the high degree of endonormativity reached by the variety of English spoken in Singapore. In addition, she intimates that it “may be caused by some substratum, e.g. Chinese” (Mazaud 2004: 256), without, however, pursuing this hunch any further. The guid-

5.3 NP complexity

127

ing hypothesis of the present study, of course, is that Sinitic substrate influence may, in fact, be the reason for her findings (cf. 4.1.2.2). Another study in which Singaporean English stands out with regard to NP structures is Gut and Coronel’s (2012) account of relativisation strategies in the Nigerian, Jamaican, Philippine and Singaporean components of the International Corpus of English. They find that the strongest differences in their dataset hold between Nigerian English and Singaporean English, with the incidence of relative clause highest at 9,394 per million words in the Nigerian corpus component, and the Singaporean component scoring lowest at a mere 7,280 relative clauses per million words (𝑝 ≤ 0.001) Although this result is not interpreted along such lines by Gut and Coronel (2012: 237), it seems plausible from the contact-linguistic position taken in the present study to ascribe the preference for postnominal relative clauses in Nigerian English to the pan-African preference for head-initial NPs (cf. 4.2.2.2). The lower frequency noted for Singaporean English, in turn, could be put down to the head-final typology of the local languages of Singapore (cf. ??). Hudson-Ettle and Nilsson (2002), finally, look into differences in adjectival modification between British and East African English on the basis of the East African component of the International Corpus of English (ICE), and British English data, on the one hand from the Uppsala Press Corpus and, on the other hand, from the Uppsala Corpus of Travel Texts. They expect a low share of attributive adjectives in East African English, as they consider this variety more “oral” than British English⁵. However, they find only minute differences between the Kenyan articles taken from ICE-EA and the general feature articles from the UPC (23 vs. 22 NPs with adjectival premodification per 1000 words). It is only the British “travel feature articles” from UCTT that stand out as having a considerably higher incidence of premodifying adjectives. This, however, is a mere register effect which does not reveal anything about variety-specific tendencies (cf. Hudson-Ettle and Nilsson 2002: 51–2).

5.3 NP complexity One of the key parts of the present study is dedicated to the investigation of simplification tendencies in NP structures in new varieties of English. For this pur-

5 This surprising view is rooted in an earlier study by Hudson-Ettle (1998: 173–4), where an unusually high proportion of that clauses in ICE-EA is interpreted as an “oral” feature (cf. HudsonEttle and Nilsson 2002: 35). Hudson-Ettle and Nilsson (2002: 35) acknowledge, however, that this assumption is at variance with other accounts, in which Kenyan English is described as a distinctly formal, bookish variety (cf. e.g. Schmied 1991: 54).

128

5 The English NP — structure and variation

pose, it is paramount to develop a valid method of measuring the complexity of NPs. First, the general notion of linguistic complexity will be problematised and discussed. Then, the main metrics of NP complexity used in the literature will be reviewed and discussed with regard to the needs of the present study. Finally, I will seek to arrive at an operationalisation of NP complexity geared towards the needs of the present study.

5.3.1 Problematising linguistic complexity The notions of simplicity and complexity, which have already been addressed with respect to varieties of New Englishes (cf. 2.2.2), have been the subject of lively debate in the recent literature. Two salient dichotomies will be discussed below. First, there is a difference between the global complexity of linguistic systems at large and the local complexity of individual areas of grammar (cf. Miestamo 2008: 29). Many authors in the recent literature have converged on the view that the assessment of the overall complexity of a linguistic system is a “very ambitious and […] probably hopeless endeavour” (Kortmann and Szmrecsanyi 2012: 8) or, indeed, a “wild goose chase” (Deutscher 2009: 243). It would be very hard, for one thing, to find a representative and yet manageable set of linguistic features for this purpose (cf. Deutscher 2009: 249); furthermore, it is hardly possible to ensure that these features are cross-linguistically comparable (cf. Miestamo 2008: 30). In contrast to that, the analysis of the complexity of individual, well-defined and narrowly circumscribed areas of grammar, such as the complexity of concrete phonological, morphological, syntactic or semantic phenomena is generally considered feasible (cf. Kortmann and Szmrecsanyi 2012: 9) and has been adopted widely (cf. Miestamo, Sinnemäki, and Karlsson 2008). Hence, this study, too, will restrict its focus to the complexity of NPs only. A second important dichotomy in complexity research is the distinction between absolute and relative complexity (cf. Miestamo 2006: 346; Miestamo 2008: 24). Absolute complexity refers to the complexity of linguistic features from a “theory-oriented point of view” (Miestamo 2006: 346). It is a deductive concept based on a preconceived theory of language structure, which can be exemplified e.g. by Kolmogorov complexity, according to which the complexity of an object 𝑦 is assessed with reference to the “minimal length 𝑙(𝑝) of the ‘program’ 𝑝 for obtaining 𝑦” (Kolmogorov 1965: 5; cf. also Juola 2008: 91). The complexity of a linguistic subsystem at large, thus, depends, for instance, on the number of its constituents. A phoneme inventory consisting of 34 phonemes requires a longer ‘program’ for its complete description than a phoneme system having fewer phonemes (cf. Miestamo 2008: 24). This kind of complexity, thus, is measured

5.3 NP complexity

129

without reference to language users. Relative complexity, in contrast, is defined with respect to language users as it depends on the processing or production cost of a speaker or listener in concrete communicative situations (cf. Miestamo 2006: 346). Kusters (2003: 6), for instance, defines the complexity of verbal paradigms as relative to an idealised “outsider”, i.e. a “second language learner, who is not acquainted with the speech community of which s/he is learning the language, and who wants to use the language in order to transmit meaningful messages”. From such a learner’s point of view, the difficulty of verbal paradigms depends on the number of semantic categories expressed morphologically, on the transparency of the form-meaning relationships and on the degree of isomorphy between semantics and linguistic form (cf. Kusters 2003: 21). In addition, there are, in fact, areas of transition between the two notions of complexity. Miestamo (2008: 32–3) demonstrates that the relative criteria for complexity used by Kusters (2003) bear a surprising similarity to the absolute ones used by McWhorter (2008). A preference for absolute metrics is usually found in studies on language typology (cf. e.g. Miestamo 2008: 26), while sociolinguists and psycholinguists alike (cf. e.g. Kusters 2008: 5), due to their usage-based construal of language, have tended to opt for the relative approach (cf. Kortmann and Szmrecsanyi 2012: 10). Some authors, however, have also used composite metrics, which fits in well with the idea that there is no clear line dividing these two approaches. Kortmann and Szmrecsanyi (2009b: 269), for instance, take an abstract linguistic notion as their starting point, such as the use of she/her for inanimate referents (which is considered “ornamentally complex”), or the levelling of the difference between past and past participle forms (e.g. He had went), which, from an absolute point of view, is considered a feature simplifying the linguistic system in comparison to the standard (cf. Kortmann and Szmrecsanyi 2009b: 273; cf. 2.2.2.2). At the same time, however, there is a clear implication that the complexity of relevant linguistic items depends on the difficulty which it poses for individual speakers in actual discourse (cf. Kortmann and Szmrecsanyi 2012: 11). Their approach, thus, can be viewed as absolute and relative at the same time. In the present study, too, the main complexity metrics used for NPs will be composite absolute-relative metrics. NP complexity will be gauged on the basis of abstract linguistic criteria involving e.g. the counts of syntactic nodes, which qualifies as an absolute measure of complexity (cf. 5.3.2). In order to ensure the plausibility of this metric, however, all assumptions will be corroborated by means of psycholinguistic studies attesting to the actual relevance of these complexity differences for speakers (cf. 5.3.2.2). In addition, it is worth noting that in the context of research into linguistic complexity, the notion of complexity required for the study of NPs is somewhat special in being geared towards one particular, quite narrow hierarchical

130

5 The English NP — structure and variation

syntactic structure. The debates about complexity in the recent literature, in contrast, have largely focused on the complexity of whole linguistic systems or subsystems, either in terms of the presence or absence of particular linguistic structures (cf. e.g. Kortmann and Szmrecsanyi 2009b) or e.g. with regard to the morphosyntactic complexity of inflectional paradigms (cf. e.g. Kusters 2003). The narrow focus on the complexity of NPs, however, is justified by recent research in the field of SLA (cf. 3.1.2.2), which has shown that such hierarchical measures, in particular complex NPs, are powerful predictors of the competence level of learners.

5.3.2 The major metrics of NP complexity 5.3.2.1 NP complexity measures: an overview In this section, the three most well-established ways of measuring NP complexity found in the literature will be discussed in turn.

(a) Measures of hierarchical complexity — Measures of hierarchical complexity are based on the counting of syntactic nodes of NPs, which requires formalisation on the basis of a syntactic theory. Figures 5.1–5.2, for instance, show Berlage’s

±ê 6

±

˥Țǘ

ƭʂʂɈ˒

Fig. 5.1. Berlage’s (2014: 36) analysis of the NP the books ±ê 6 Ɔ

±ʂɩ ± ɩƆɱ

êê êʾǘʳ

±ê

ʂȄ

±ʂɩ ǃɀ

±

ȇʾǘƆ˥

Țʂɱʂ˺ʾ

Fig. 5.2. Berlage’s (2014: 5) analysis of the NP a man of great honour

5.3 NP complexity

131

(2014: 5; 36) analysis of two English NPs, the books and a man of honour, along the lines of Huddleston and Pullum (2002). Hierarchical complexity differences between these NPs can be measured by counting all syntactic nodes in the tree structures. According to this, the NP in figure 5.1 has three nodes, while the one in figure 5.2 has a total of ten nodes. Alternatively, only phrasal nodes can be counted, which amounts to one for figure 5.1 and to three in figure 5.2 (cf. Berlage 2014: 37; Wasow 1997: 85). No matter which approach is used, figure 5.1 invariably turns out to be less complex than figure 5.2. (b) Measures of length, too, could be employed in order to gauge complexity differences between NPs. The NP in figure 5.1 is also less complex than figure 5.2 in being shorter than the former. Length can be operationalised both in terms of word, character and syllable counts, which are correlated extremely strongly with correlation coefficients 𝑟 around 0.98 (cf. Wolk et al. 2013: 12). Word counts are a natural choice for practical reasons, however, as they can be established automatically using e.g. an R script. (c) Further, more specific measures of complexity — A range of authors have used specific measures of complexity aiming at particular features of syntactic structure. Szmrecsanyi (2004: 1034), for instance, computes an “Index of Syntactic Complexity”, which involves counting, first, all NPs and VPs in a given clause, and, second, all markers of subordination such as conjunctions (because, since, when…) or wh-pronouns such as who, whose or which. The complexity index of a given clause is computed by inserting the counts for these elements into a formula which weights subordination markers more heavily than VPs and NPs (cf. Szmrecsanyi 2004: 1034). Similarly, Berlage’s (2014: 6) “degree of sententiality” measures whether a given “NP involves a clause (a VP or clause in a tree diagram) as part of its structure” (Berlage 2014: 6). Finally, Ravid and Berman (2010: 11), in a study in an FLA framework, look into the number of different phrase types embedded into NPs. From this point of view, an NP which involves both an adjectival premodifier and a clausal postmodifiers would be ranked higher than an NP with a clausal postmodifier only. 5.3.2.2 Previous studies confirming the relevance of complexity metrics This section briefly summarises major studies which have confirmed the cognitive relevance of the type of NP complexity metrics presented in the previous section.

132

5 The English NP — structure and variation

(a) Berlage (2014) measures NP complexity on the basis of three variables: hierarchical NP complexity, the length (in words) of NPs and the degree of “sententiality” of NPs, i.e. the question of whether they contain clausal structures, such as relative clauses, non-finite clauses, etc. (cf. Berlage 2014: 250). On the basis of these criteria, she arrives at the complexity hierarchy represented in table 5.4. Type 1, which lacks any postmodifier, is assumed to be less complex than types 2–9 (cf. Berlage 2014: 44–5) as it has very few syntactic nodes; types 2–4, which clearly go beyond the complexity of Type 1 but do not involve sentential elements, in turn, are assumed to be more complex than type 1, and types 5–9 are most complex in containing sentential elements. The two key criteria underlying this classification, thus, are counts of syntactic nodes and sententiality. Table 5.4 also shows that in Berlage’s view, hierarchical complexity in NPs is created by postmodification only, while the premodifying string only counts towards the length measure and does not increment an NP’s hierarchical complexity (cf. 5.3.3). Table 5.4. Berlage’s (2014: 44) NP complexity hierarchy for NPs involving up to one postmodifier; cf. Berlage (2014: 45–58) for the example sentences (5.45)–(5.53). Type Modification 1 2 3 4 5 6 7 8 9

Non-postmodified NPs Coordinated NPs NP + PP NP + AP Gerundial constructions NP + non-finite clause NP + finite relative clause free wh-clause NP + that-clause

Example (5.45) (5.46) (5.47) (5.48) (5.49) (5.50) (5.51) (5.52) (5.53)

the vital reports the houses and the gardens varieties with a long maturation period the day suitable his constant questioning of my motives the person to do the job the person who read the book what happened to her the rumour that he is leaving

In Berlage’s (2014) study, the complexity of these modification types is tested on the basis of a large set of corpora made up of British and American newspaper texts and the Brown and Frown corpus pair, with a focus on the following environments of syntactic variation: (a) the realisation of a final VP in constructions involving as far as, which is more likely with short and simple NPs than with long and complex NPs: (5.54) As far as going on holiday is concerned/goes/Ø (Berlage 2014: 81) (b) positional variants with light verb constructions, where long and complex NPs as complements tend to be realised clause-finally:

5.3 NP complexity

133

(5.55) (a) The gunmen entered the settlement, taking the residents hostage. (Berlage 2014: 120) (b) He added that the Serbs had taken hostage 300 women and children. (Berlage 2014: 120) (c) the position of notwithstanding, with the prenominal position more likely for complex NPs: (5.56) (a) Notwithstanding the logic, virtual organisations are virtually absent. (Berlage 2014: 162) (b) This book notwithstanding, he eludes us still. (Berlage 2014: 162) (d) the use of the to-infinitive after help, which is more likely in (5.57)(b) than in (5.57)(a) due to the length and complexity of the postverbal NP (cf. 5.2.1.1): (5.57) (a) It is the fastest way to help those people gain a bigger share of the world economy. (b) ‘We were brought into being with the express intention of helping those who have been on the margins of society to move to the centre stage’. (Berlage 2014: 203) Having analysed these variational contexts using mono- and multifactorial statistical techniques, Berlage arrives at two complexity hierarchies for NP modification patterns, one for cases involving up to one postmodifier, which is shown in table 5.5 and the other one for cases involving more than one postmodifier, which is represented in table 5.6. Table 5.5. A complexity hierarchy of NP modification patterns involving up to one postmodifier, from Berlage (2014: 255) Complexity rank Modification pattern (1) (2) (3) (4)

non-postmodified NPs NP+PP — NP+AP — coordinated NPs NP+non-finite clause — NP+gerundial constructions NP+finite relative clause

As can be seen, the complexity hierarchies of modification patterns confirm the influence of length, hierarchical complexity and sententiality (cf. table 5.4). As for length, it is obvious that the least complex NPs are invariably

134

5 The English NP — structure and variation

Table 5.6. A complexity hierarchy of NP modification patterns involving more than one postmodifier, from Berlage (2014: 255) Complexity rank Modification pattern (1) (2) (3) (4) (5)

non-postmodified NPs NPs involving coordinations/PPs/APs NPs involving (non-finite) supplements NPs involving gerundial constructions and/or non-finite clauses NPs involving finite clauses (e.g. finite relative clauses, complement clauses introduced by that, or free whclauses)

non-postmodified patterns. Moreover, in both tables, long, clausal patterns turn out to be more complex than the nominal patterns involving PPs and APs. In terms of hierarchical complexity, large numbers of phrasal nodes correlate with high complexity rankings. NPs belonging to Type (1), in both tables, have fewer nodes than the other types involving postmodification; among the postmodified types, in turn, finite clauses, which have the highest number of nodes, turn out to be most complex. With regard to sententiality, sentential modifier types are found to be most complex. This clearly indicates that “the syntactic complexity of NPs is not sufficiently captured by the number of phrasal nodes they contain” (Berlage 2014: 257), but that the degree of sententiality plays a major role. In sum, Berlage’s (2014) study is highly relevant in documenting the cognitive relevance of complexity levels of NPs. In addition, due to her usage-based approach, it can be assumed that her NP complexity levels, although partly deriving from a formal syntactic theory and determined on the basis of L1 English users only, reflect general usage tendencies.

(b) The relevance of length and hierarchical complexity has also been ascertained from a psycholinguistic point of view. A study by Ferreira (1991: 216) involves asking test subjects to memorise sentences involving NPs of different degrees of length and structural complexity (determined on the basis of the number of syntactic nodes), which are shown on a computer screen, and instructing them to repeat each sentence once it has disappeared. Examples of the NPs include the river (short), the large and raging river (long, low syntactic complexity), the river near their city (long, medium syntactic complexity) or the river that stopped flooding (long, high syntactic complexity).

5.3 NP complexity

135

The study demonstrates that both length and complexity affect initiation times. Test subjects require 518 milliseconds on average in order to repeat sentences involving short subjects, while they need 613 milliseconds for the three long types of subjects (cf. Ferreira 1991: 215); likewise, initiation times for the long-low complexity condition differ significantly from those of the medium and high complexity levels (cf. Ferreira 1991: 216). Ferreira’s (1991) study, thus, documents that hierarchical complexity (operationalised as syntactic node counts) and length (in words) can be reliably mapped onto the initiation times of test subjects. Hence, the idea that these factors should influence speakers’ processing load is plausible. Nodal complexity is not merely an abstract measure based on formalist considerations but can be accepted as “cognitively […] ‘real”’ (Szmrecsanyi 2004: 1031).

5.3.3 Discussion of the main metrics of syntactic complexity In the following, two problems surrounding the role of NP length and hierarchical complexity as complexity metrics will be discussed briefly.

(a) The degree of correlation of length and hierarchical complexity is a moot point. Wasow (1997: 93) finds correlation coefficients of 0.94 to 0.99 between the two (cf. also Szmrecsanyi 2004: 1037); Berlage (2014: 250), in contrast, states that “both NP-length and NP-structure successfully predict the variation”⁶. In the present study, in the absence of unanimous evidence, it will be assumed that “there is no 1:1 correlation between NP-length and NP-structure” (Berlage 2014: 14) and that an independent analysis of length and hierarchical complexity is worthwile (cf. 5.3.4). (b) Inclusive vs. exclusive operationalisation of hierarchical NP complexity — Authors also differ with regard to the question of what types of syntactic elements can raise NP complexity. Hawkins (1994: 59) opts for an inclusive approach by counting “[A]ll the terminal and non-terminal nodes” (Hawkins 1994: 59), unlike e.g. Ferreira (1991: 214), who only accepts phrasal nodes; likewise, he considers both pre- and postmodification “psycholinguistically significant” (Hawkins 1994:

6 She also finds that for some of her variables, length is more important than hierarchical complexity (e.g. for the take hostage construction; cf. 5.3.2.2), while other constructions (e.g. the “as far as variable”, cf. 5.3.2.2) are more strongly influenced by hierarchical complexity.

136

5 The English NP — structure and variation

59). Berlage (2014), in contrast, presupposes that the only elements which raise the complexity levels of NPs are postmodifiers (cf. 5.3.2.2), while premodifiers only count towards the length measure of the NP. Her reasons for doing so are based on the view of Huddleston and Pullum (2002); furthermore, she cites Altenberg’s (1982: 78) conclusion that “right branching expansions tend to create structural problems of a different order than left-branching ones”, which is based mainly on his observation of the relative shortness of premodifiers as against postmodifiers (cf. Altenberg 1982: 77–8). These arguments, both of which are, to a degree, arbitrary and theory internal, are never questioned by Berlage (2014) but taken over in a quite uncritical way. In my view, it can hardly be doubted that premodifiers do add to the complexity of NPs, especially if one considers cases like the long, hierarchically complex classifiers cited in (5.22). This does not detract from the fact that the prototypical long modifiers are postmodifiers. However, under the assumption that grammar is a gradual phenomenon eschewing categorical definitions (cf. Langacker 1987: 14), any type of phrasal embedding into an NP increases processing and production costs. Consequently I will resort to an inclusive approach of NP complexity and accept that both pre- and postmodification can raise the degree of complexity of NPs. 5.3.4 How NP complexity will be operationalised In this section, I will lay out how NP complexity measures will be operationalised for the purpose of the present study. The notion of complexity will be used in two distinct ways: For one thing, I will address the complexity of whole NPs, like Berlage (2014), which will allow conclusions about simplification tendencies in New Englishes (cf. 2.2.2). In addition to that, I will also gauge the individual complexities of the pre- and postmodifying string, which may allow conclusions about L1 influence. It is assumed that the typological factors outlined above (cf. 4.1.2– 4.2.2) may lead to the complexification of premodifiers or postmodifiers in Singaporean English and Kenyan English respectively.

5.3.4.1 The complexity of modifiers 5.3.4.1.1 The complexity of premodifiers In accordance with the conclusions drawn above (cf. 5.3.2), both length and hierarchical complexity will be used as independent metrics of premodifier complexity.

5.3 NP complexity

137

(a) Length of premodifiers — The type of long premodifiers shown in (5.58) will be considered more complex than short premodifiers. (5.58) the postwar international political economy (b) Hierarchical complexity of premodifiers — The hierarchical complexity of premodifiers will be assessed on the basis of counting syntactic nodes. Rather than relying on own node counts, however, the complexity hierarchies developed by Berlage (2014) will be capitalised on (cf. 5.3.2.2). Premodifiers will be ranked according to the criteria shown in figure 5.3. (1Prem ): simple phrases < (2Prem ): mod. or coord. non-clausal phrases < (3Prem ): clausal structures Fig. 5.3. Complexity hierarchy of premodifiers

Type (1Prem ), the simplest form of premodification, is exemplified in (5.59)–(5.60). (5.59) Beautiful area (5.60) the newspaper clippings

Type (2Prem ), in turn, which is illustrated in (5.61)–(5.62), will be considered more complex due to the premodifiers being modified themselves, or by virtue of being coordinated, as in (5.63). In addition, I presuppose that the simple, non-clausal instances of lexical phrases (cf. 5.1.4.2) shown in (5.64)–(5.66), too, are at this intermediate complexity level. (5.61) (5.62) (5.63) (5.64) (5.65)

a very loose sense infant mortality rate a tertiary or upper secondary education within-generation events The supplementary information-gaining processes (5.66) a more semantics-based grammar

Examples of Type (3Prem ) can be seen in (5.67)–(5.69). There is no doubt that the somewhat unwieldy clauses functioning as lexical phrases (cf. 5.1.4.2) can be viewed as cognitively most demanding.

138

5 The English NP — structure and variation

(5.67) Singapore’s ’learn from Japan policy’ (5.68) The Malaysia’s look east’ policy (5.69) an entirely separate “insert before current entry” function

Beyond the analysis of premodifiers in terms of the three types (1Prem )–(3Prem ), the hierarchical complexity of premodifiers will be ranked by the following two criteria, both of which are related to the hierarchical complexity of phrases but cut across the criteria mentioned so far. – Premodifiers will be classified according to their individual formal realisations. This criterion will help to uncover idiosyncratic preferences in varieties such as the Singaporean penchant for sentential premodifiers documented above (cf. 5.2.2.2). – It is assumed that the hierarchical complexity of the premodifying string as a whole can rise through the stacking of individual functional types of premodifiers. NPs in which only one of these is realised, as (5.70), are hierarchically simpler than cases where multiple slots are occupied, as in (5.71). (5.70) a very sadMod. II story (5.71) Aik Leng has very goodMod. II ballMod. III sense.

– It is assumed that the hierarchical complexity of the premodifying string will rise by the embedding of further NPs (cf. (5.72); cf. also Jucker 1992: 112–4). (5.72) thirty SIAI-Marchetti trainer and light attack aircraft

5.3.4.1.2 The complexity of postmodifiers The key measures to be used for postmodifiers are likewise length and hierarchical complexity.

(a) Lengths of postmodifiers will be used as a measure of complexity. (b) Hierarchical/structural complexity of postmodifiers — Postmodifiers can be ranked in terms of their hierarchical complexity as shown in figure 5.4, with the ranking largely based on the results gained by Berlage (2014) (cf. 5.3.2.2).

5.3 NP complexity

139

(1Pom ): PPs(, APs) < (2Pom ): non-finite clausal structures < (3Pom ): finite clausal structures Fig. 5.4. Complexity hierarchy of postmodifiers

Type (1Pom ) comprises PPs (and APs, which occur only rarely) (cf. (5.73)–(5.74)), which contribute least to the NPs’ complexity. (5.73) the need for Compression of Information to limited space

(5.74) the cashbox full of money Type (2Pom ), in turn, is shown in (5.75)–(5.77). It involves non-finite clauses. (5.75) a process defined as simplification or pidginization

(5.76) his attempt to bring some order amongst the gathering

(5.77) his sister, having arrived at Hungary to have an eye on him

Importantly, non-finite clauses will be taken into account even if they are modifiers of NPs embedded at higher levels of embedding. In (5.78), for instance, the non-finite clause does not modify the NP head directly but is embedded into a PP. (5.78) the attainment of order and safety as needed by the individuals

Type (3Pom ), which is exemplified in (5.79)–(5.80), involves finite clauses; such postmodifiers are assumed to be most complex. (5.79) the people who developed it (5.80) the belief that its failure or eradication may evoke a curse from the ancestors In addition, the complexity of postmodifiers will be gauged by the same set of additional criteria as premodifiers:

140

5 The English NP — structure and variation

– Once more, beyond the formal types (1Pom )–(3Pom ), the individual formal realisations of postmodifiers will be analysed, with the aim of revealing specific preferences e.g. for certain subtypes of phrases. – It is assumed that the use of two or more postmodifiers in a row renders the postmodifying portion of a given NP more complex. An example of two stacked postmodifiers can be seen in (5.81), where the PP is followed by a non-finite clause. (5.81) the dialect of KiswahiliPostmod. I used in ZanzibarPostmod. II

– The hierarchical complexity of the postmodifying string rises by the embedding of further NPs (cf. (5.82)). (5.82) NINDEF’s follow-up to the establishment of the SDU within the Ministry of Finance

5.3.4.2 The complexity of whole NPs In addition to the complexity of pre- and postmodifiers, the complexity of whole NPs will play a role in the analyses in 7. Once more, Berlage’s (2014) corpus-based rankings of structures will serve as a basis of argumentation. In the following, length measures and hierarchical complexity differences will be dealt with in turn. 5.3.4.2.1 Length of NPs Counts of orthographic words will serve as a basic complexity metric for the complexity of whole NPs. The NP in (5.83), thus, which is 14 words long, would be considered more complex than an NP modified by a short three-word prepositional phrase such as (5.84). (5.83) groups of petty men who offer plans and tricks to gain the enemy’s confidence (5.84) their appeal to the peasants Length measures will also be considered separately for subjects and nonsubjects, as, according to Hawkins’s (1994) “EIC principle” (cf. 5.2.1.1), short (and simple) subjects are functionally advantageous from a parsing perspective. In a variety which simplifies, it can be expected that these advantages are being capitalised on more strongly than in a very advanced or native variety.

5.3 NP complexity

141

5.3.4.2.2 Hierarchical complexity of whole NPs In order to gauge the complexity of a given NP as a whole, the complexity levels of pre- and postmodifiers found above (cf. 5.3.4.1) need to be related systematically to the complexity of NPs at large. It is important to note, though, that, despite the view that they both add to NP complexity (cf. 5.3.3), pre- and postmodifiers cannot be measured by quite the same yardstick. Extreme cases of phrasal premodification shown in (5.67)–(5.69), despite being structurally licit, are clearly exceptional. Generally speaking, though, the premodifier zone does not allow for the same degree of structural elaboration as postmodifiers (cf. 5.3.2.2). For this reason, postmodifiers must be assumed to have a stronger bearing on a given NP’s overall complexity than premodifiers. Table 5.7 shows a complexity ranking of whole NPs which takes into account this difference between pre- and postmodifiers. As can be seen, the basic distinction lying at the heart of this model of complexity is the number of modifiers. Level (1NP ) is unmodified; levels (2NP )–(7NP ) are either pre- or postmodified. In order to accommodate the complexity differences between different pre- and postmodifier types established above (cf. 5.3.4.1), levels (2NP )–(7NP ) are ranked in order. NPs with premodifiers, which are assumed to be principally simpler than postmodified ones, are considered less complex than NPs with postmodifiers. In addition, both

Table 5.7. The complexity levels of whole NPs Complexity level Structure (8NP )

(7NP ) (6NP ) (5NP ) (4NP ) (3NP ) (2NP ) (1NP )

Example

(5.85) most significant and invaluable international political roles for Japan postmod. level (3Pom ) (5.86) guys who were perhaps “less qualified” than they postmod. level (2Pom ) (5.87) tools to achieve her goals postmod. level (1Pom ) (5.88) support among the people prem. level (3Prem ) (5.89) ’learn from Japan policy’ prem. level (2Prem ) (5.90) a social class continuum prem. level (1Prem ) (5.91) a staple income unmodified NP (5.92) superficiality pre- and postmod.

142

5 The English NP — structure and variation

pre- and postmodified NPs can be ranked in terms of the three complexity levels of their modifiers (1Prem )–(3Prem ) and (1Pom )–(3Pom ) (cf. 5.3.4.1.1–5.3.4.1.2). Level (8NP ), finally, which is considered most complex, has both a pre- and a postmodifier. In principle, NPs at this level, too, would have to be subcategorised further according to the three complexity levels of pre- and postmodifiers respectively. For the sake of clarity and legibility, however, the six possible permutations of the three pre- and the three postmodifier types will not be included into the model. Furthermore, three further measures will be employed in order to assess the hierarchical complexity of NPs, which cut across the complexity types shown in table 5.7: – For one thing, the numbers of pre- and postmodifiers used per NP will serve as an additional hierarchical complexity measure. NPs featuring a range of functionally distinct premodifier types, as in (5.93), will be viewed as more complex than NPs with single premodifiers; in addition, multiple, stacked postmodifiers will be assumed to raise the complexity of NPs, as exemplified in (5.94). (5.93) quite patchy autobiographical ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ memories ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Mod. II

Mod. III

(5.94) his campaign ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ against Brittany ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ in 799 Postmod. 1

Postmod. 2

It stands to reason that several pre- and/or several postmodifiers will complexify the hierarchical structure of a given NP (cf. Schäpers 2009: 110–1 for a similar measure). On the basis of the observation that postmodifiers in English can, in principle, be expanded more easily than premodifiers (cf. 5.3.2.2), NPs will be ranked as follows according to their pre- and postmodification patterns: premodified NPs < multiply premodified NPs < postmodified NPs < multiply postmodified NPs < pre- and postmodified NPs < multiply pre- and postmodified NPs. – NP complexity will also be measured according to “syntactic depth” (cf. Jucker 1992: 113; Ravid and Berman 2010: 11–2; cf. 5.3.2.2). In the NP shown in (5.95), a point, being at the top of the embedding hierarchy, would be assigned the level 0; this growth would be assigned to level 1 in being a part of a modifier of a level-0 NP; the remaining NPs interest, CD ROM data and CD ROM, in turn, are embedded at successively deeper levels of embedding. (5.95) It does raise a point on this growth of interest ROM data. ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ in CD ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ level 2 level 4 ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ level 0 level 1 level 3

– With Hawkins (1994) having demonstrated that simple NPs in early sentence positions are advantageous from a parsing point of view (cf. 5.2.1.1), the degree

5.3 NP complexity

143

to which the varieties make use of this advantage will serve as a measure of simplification. More precisely, the question will be to what extent they rely on very simple subjects such as (5.96), and to what degree they tolerate complex subjects such as (5.97). (5.96) the government still owns the land (5.97) The rate of consumption is accelerated by a foreign object in front of me. The decision to focus on the difference between subjects and non-subjects only is based on the fact that F. G. A. M. Aarts (1971), de Haan (1993) and Jucker (1992) consider this the most effective distinction of clause positions (cf. 5.2.1.1). Jucker (1992: 259), in fact, states explicitly that a more fine-grained distinction into syntactic functions has not proved effective in his study. It has to be acknowledged that some of the complexity levels shown in table 5.7 are contestable. One may think, for instance, of an NP with a sentential premodifier as in (5.84) (p. 140) which, in fact, outdoes another NP with a very simple postmodifier, as in (5.89) (p. 141) in terms of overall complexity. This would be at variance with the stipulation that premodifiers are, as a rule, simpler than postmodifiers. However, due to the rarity of complex, sentential premodifiers, this problem is not a major one. A further problem worth noting has already been mentioned with regard to length (cf. 5.3.4.2.1). In the model of complexity, postmodification contributes to complexity more than does premodification. Given the structural properties of the English NP, this is unavoidable: A variety favouring postmodification will always outdo a variety which prefers premodification in terms of overall NP complexity. For this reason, this measure of complexity, too, cannot be used in order to differentiate ICE-SING and ICE-EA clearly in terms of complexity if any of the hypotheses with regard to preferences for pre- and postmodified NPs in these varieties bear out. It will, however, serve as a metric for complexity differences between ICE-GB and the two new varieties under scrutiny. The criteria lying at the heart of this classification of NP complexity, thus, likewise correspond to Berlage’s (2014: 44–5) approach (cf. 5.3.2.2). In a wider perspective, they also fit in with the notion of simplification developed along the lines of Croft’s (2000) model of language change (cf. 3.2.4.2). According to him, the underlying principle of simplification is the speaker’s desire to “minimiz[e] the time involved in achieving the extralinguistic goals” (Croft 2000: 75). I have argued above that the economy principle, although at work in any linguistic interaction, can be expected to assume particular significance in L2 situations where speakers use a language which they know imperfectly (cf. 3.2.4.2). Language production

144

5 The English NP — structure and variation

in an L2 is inevitably more time-consuming than language production in an L1. As Ferreira’s (1991) results have shown, NP complexity as described in this section is directly correlated with the time needed for the production of a given NP (cf. 5.3.2.2).

5.4 Summary: Delineating the variable context in NPs In this chapter, the English NP has been approached from three angles: The discussion in 5.1 addressed its syntactic structure and helped to delineate the field of inquiry of the present study. 5.2, in turn, provided an overview of previous results on the variation of NPs, with reference to their syntactic position, the semantic class of the head noun, register and language variety. 5.3, finally, addressed the notion of NP complexity and developed criteria for its assessment, which likewise will play a major part in the analysis of NPs in Singaporean English and Kenyan English.

5.4.1 Methodological conclusions from the discussion of NP variation The discussion of the factors governing variation in the English NP above (cf. 5.2) allows the following set of conclusions. The two factors which are at stake in the corpus study below are substrate influence/language contact and SLA-based simplification. In what follows, it will be pointed out how the remaining factors identified so far will be addressed. (a) The influence of the word class of the head noun will be taken into account through the following decision: Only lexical NPs such as activities, speed, wheelchair or bodies will be taken into account. Both proper names (Mike Heafy, London, Stanmore etc.; cf. ) and any pronouns (it, who, each, whatever etc.) will be excluded, due to their extremely low modification rates (cf. 5.2.1.1). In addition, only the modification zone of NPs will be taken into account, i.e. the elements defined as Emphasiser, Mod. I, Mod. II, objective descriptor and Mod. III (cf. 5.1.3.2); determiner choices are governed by grammatical regularities, which are at the speaker’s disposal to a lesser degree than the choice of pre- or postmodifiers (cf. 5.1.3.1). (b) The syntactic position of the NP (cf. 5.2.1.1) and the semantic class of the head noun (cf. 5.2.1.1) will be addressed by way of individual cross-tabulations and statistical tests of the syntactic and semantic distribution of NP structures (cf. 7.3.1.2.2, 7.3.2.2.2 and B.3). Furthermore, the use of random samples and large token numbers will ensure that no particular context is favoured.

5.4 Summary: Delineating the variable context in NPs

145

(c) The influence of modality and register (cf. 5.2.1.2–5.2.1.2) will be addressed at the corpus level by the use of two specific text registers (private dialogue and student essays, cf. 6.2.1) at two opposite ends of the formality scale. The key inquiry will be conducted separately for the spoken and written parts of the sample in order to tease out contact effects, with the separate discussions, however, complemented by comparative analyses.

5.4.2 The variable contexts in the NP Following the analysis of NP structure in the previous sections, it will now be possible to delineate the variable contexts for the corpus analyses in 7 with the aim of investigating effects of language contact and simplification in the varieties at hand. From the point of view of a cognitivist, usage-based theory of language (cf. 1.3.2.2; 3.2.4.1), the NP structures covered in above along traditional descriptivist lines (cf. 5.1.3), which are deemed necessary despite cognitive-linguistic qualms (cf. 5.3.3), have to be viewed as a complex taxonomic network of schematic linguemes, which is illustrated in figure 5.5. Different NP modification patterns such as premodification, postmodification, simultaneous pre- and postmodification as well as all of the more specific sub-patterns, such as Mod. II — Head, Mod. II — Mod. III — Head, etc. are represented in the speaker’s minds in the form of schematic linguemes (cf. 3.2.2.1), with individuals differing in the degree of entrenchment of the NP linguemes. From the point of view of the present study, the NPs in figure 5.5 can be differentiated along two dimensions: First, with regard to whether they are pre- or postmodified; second, with regard to their degree of complexity (i.e. length and intricacy of hierarchical structure). In addition, all individual modifiers used with NPs, such as Mod. I, Mod. II and Mod. III, in their various

Schematic NP linguemes

unmodified NPs

(...)

[Mod. II - Head]

Premodified patterns

[Mod. II - Mod. III - Head]

Postmodified patterns

(...)

[Head - Postmod. 1]

Pre- and postmodified patterns

[Head - Postmod. 1 - Postmod. 1]

(...)

[Mod. I - Mod. III - Head - Postmod. 1]

Fig. 5.5. NP structures as schematic linguemes (exemplary overview; non-exhaustive)

146

5 The English NP — structure and variation

Modifier linguemes

Mod. I

(...)

AP with marginal adj.

Mod. II

SKT-construction

(...)

AP with central adj.

Mod. III

(...)

AP with classifying adj.

Postmodifier

NP

PP

fin. cl.

nonfin. cl.

AP

Fig. 5.6. NP modifiers as schematic linguemes (exemplary overview; non-exhaustive)

formal realisations (cf. 5.1.3), can likewise be thought of as schematic linguemes, which stand in the kind of network relationships illustrated in figure 5.6, with Mod. I–III and postmodifiers representing higher-ranking, unspecific linguemes, whereas the more specific linguemes at the bottom level of figure 5.6 correspond to the different formal realisations as AP, NP etc. Once more, the underlying assumption must be that speakers will differ in terms of the degree of entrenchment of these modifier linguemes, as their frequency of use differs from speech community to speech community (cf. 1.3.2.2). The statistical patterns of such NPs found in the corpus texts at hand, consequently, can be interpreted as the result of the selective adoption and propagation of NPs present in the lingueme pool (cf. 3.2). In order to elucidate the effects of language contact and simplification which are at stake in the present study, it makes sense to focus on the following broad areas of variation in the English NP: (a) variation in the use of pre- and postmodified NP linguemes, which will allow conclusions about contact-induced preferences in the adoption/propagation of head-final or head-initial NP structures; (b) variation between structurally complex and structurally simple NP linguemes (cf. 5.3). This will provide a handle on the understanding of simplification tendencies in the varieties at hand, which have led to the propagation of simpler linguemes than in the input variety British English. In what follows, these two perspectives will be elaborated on.

(a) The choice between pre- and postmodification will be exploited analytically in two major ways: With a focus on the broad choice between pre- and postmodification, and with a focus on the complexity of pre- and postmodifiers. – Focus on the choice between pre- and postmodified linguemes — The full set of NP linguemes sketched in figure 5.5, can, for one thing, be viewed in binary terms, as speakers have a broad choice between pre- and postmodifying

5.4 Summary: Delineating the variable context in NPs

147

linguemes. This area of variation, thus, will allow conclusions about contactinduced preferences for pre- or postmodification. The problem about this view of NP variation, however, is its degree of abstraction. Furthermore, with the exception of very few patterns, there is no full exchangeability between pre- and postmodification in English (cf. 5.2.2.1). This issue will be addressed in three ways: (a) For one thing, all analyses will be based on random samples from large populations, which will ensure that no particular context of NP use is favoured; (b) the analyses of NP structures will be based on large token numbers in order to ensure a sufficient degree of generalisation across individual NPs and their contexts of use; (c) the analysis of typological preferences will not rely on the mere counts of overall frequencies of pre- and postmodified NPs but will also address differences in the complexity of pre- and postmodifiers, which allow conclusions about contact-induced preference patterns unhampered by the problem of non-interchangeability, as will be seen in the following. – Focus on the complexity of pre- and postmodification — It will be assumed L1 influence will also lead to greater formal complexity of pre- and postmodified linguemes, both in terms of length and hierarchical complexity (cf. types (1Prem )–(3Prem) ) and (1Pom )–(3Pom ) in the model developed in 5.3). This hypothesis is rooted in the view that a structure favoured by substrate languages will be used more fully and more elaborately by speakers, which is e.g. suggested by the complexity of premodifiers in Singaporean English identified by Mazaud (2004) (cf. 5.2.2.2). This area of variation, too, will help to reveal the patterns of contact-induced adoption of pre- or postmodification. Singaporean English can be expected to have complex pre- and simple postmodifiers; Kenyan English is likely to exhibit the opposite tendency.

(b) In order to explore the choice between simple and complex NP linguemes in the varieties at hand, the whole NP linguemes sketched in figure 5.5 will be taken into account, which differ in terms of their complexity (cf. 5.3.4). Using the complexity ranking of NP linguemes developed above (cf. 5.3.4), each NP found can be gauged with regard to its hierarchical complexity level and its length of NPs. Language contact, thus, will be analysed in terms of two complementary perspectives, while simplification will be approached with regard to the structural complexity of whole NPs only.

Part B: NPs in Kenyan English and Singaporean English

6 Methodology, corpus handling and statistics In this chapter, I will address the methodology used in the present study.

6.1 General methodology: Corpus, software and statistics I will start by dealing with the structure and design of the International Corpus of English; in turn, both the software employed in the present study and statistical methods will be addressed

6.1.1 The International Corpus of English The International Corpus of English (ICE) has been created in order to “provide the resources for comparative studies of the English used in countries where it is either a majority first language […] or an official additional language” (Greenbaum 1996: 3; cf. also Greenbaum 1988). To date, sub-corpora for Canada, East Africa, Great Britain, Hong Kong, India, Ireland, Jamaica, New Zealand, Nigeria, The Philippines, Singapore, Sri Lanka (written) and USA (written) have been published (cf. The ICE Project 2014). In order to ensure comparability, all national components follow a parallel design, shown in table 6.1 (cf. Nelson 1996: 27). Each of them consists of 500 texts comprising 2000 words each, 300 of which represent spoken English, while the remaining 200 ones have been compiled from written sources. All in all, each ICE corpus component amounts to 1 million words (cf. Nelson 1996: 27) and consists of a rich set of text types associated with a variety of domains of language use (cf. table 6.1; Nelson 1996: 28). The texts are supposed to “date from 1990 to 1994 inclusive” (Nelson 1996: 28), which is true for ICE-GB and ICE-SING; the data contained in ICE-EA (which was published in 1999) date from 1995 to 1996, which, however, is a minute difference during which diachronic language is unlikely to have occurred (cf. Schmied and Hudson-Ettle 1999: 16). Writers and speakers sampled in the corpora are supposed to be 18 years old and must either be natives of the respective country or have “moved there at an early age” (Nelson 1996: 28); in addition, they need to have finished secondary education through the medium of English (cf. Nelson 1996: 28). Among the three ICE components to be used in the present study, ICE-GB (cf. Nelson, Wallis, and B. Aarts 2002: 3–4) and ICE-SING (cf. Nelson 2002: 4-5) are largely in accordance with the typical structural template (cf. table 6.1), apart from the fact that sociolinguistic markup detailing speakers’ and writers’ ages,

152

6 Methodology, corpus handling and statistics

Table 6.1. ICE texts and text categories, from Nelson (2002: 4–5; 29–30). Text category

ICE tags

N

private

Direct Conversations Telephone Calls

S1A-001 to S1A-090 S1A-091 to S1A-100

90 10

180,000 20,000

public

Class Lessons Broadcast Discussions Broadcast Interviews Parliamentary Debates Legal Cross-examinations Business Transactions

S1B-001 S1B-021 S1B-041 S1B-051 S1B-061 S1B-071

to to to to to to

S1B-020 S1B-040 S1B-050 S1B-060 S1B-070 S1B-080

20 20 10 10 10 10

40,000 40,000 20,000 20,000 20,000 20,000

unscripted

Spontaneous Commentaries Unscripted Speeches Demonstrations Legal Presentations

S2A-001 S2A-021 S2A-051 S2A-061

to to to to

S2A-020 S2A-050 S2A-060 S2A-070

20 30 10 10

40,000 60,000 20,000 20,000

scripted

Broadcast News Broadcast Talks Non-broadcast Talks

S2B-001 to S2B-020 S2B-021 to S2B-040 S2B-041 to S2B-050

20 20 10

40,000 40,000 20,000

non-professional writing

Student Essays Examination Scripts

W1A-001 to W1A-010 W1A-011 to W1A-020

10 10

20,000 20,000

correspondence

Social Letters Business Letters

W1B-001 to W1B-015 W1B-016 to W1B-030

15 15

30,000 30,000

academic writing

Humanities Social Sciences Natural Sciences Technology

W2A-001 W2A-011 W2A-021 W2A-031

to to to to

W2A-010 W2A-020 W2A-030 W2A-040

10 10 10 10

20,000 20,000 20,000 20,000

non-academic writing

Humanities Social Sciences Natural Sciences Technology

W2B-001 W2B-011 W2B-021 W2B-031

to to to to

W2B-010 W2B-020 W2B-030 W2B-040

10 10 10 10

20,000 20,000 20,000 20,000

reportage

Press News Reports

W2C-001 to W2C-020

20

40,000

instructional writing

Administrative Writing Skills & Hobbies

W2D-001 to W2D-010 W2D-011 to W2D-020

10 10

20,000 20,000

persuasive writing

Press Editorials

W2F-001 to W2F-020

20

40,000

creative writing

Novels & Stories

W2E-001 to W2E-010

10

20,000

printed

Written

non-printed

monologue

Spoken

dialogue

Type

Sum

size

500 1,000,000

their professions and employment, the conversation participants, and the date and place of recording are only provided in ICE-GB. ICE-EA, in contrast, does not conform exactly to the required ICE structure, as the compilers had to make a range of adjustments owing to the specific sociolinguistic situation in East Africa.

6.1 General methodology: Corpus, software and statistics

153

Key problems were the impossibility of placing Kenyan and Tanzanian varieties of written English in one category (due to sociopolitical differences made plausible in Schmied and Hudson-Ettle 1999), the difficulty of obtaining data for some of the ICE categories in the Kenyan setting and the rarity of texts with a length of 2000 words in East Africa, due to the restricted usage conditions of Kenyan English (cf. Schmied and Hudson-Ettle 1999: 4). One consequence of this is the separation of the corpus into a Kenyan and a Tanzanian component. Further idiosyncrasies of ICE-EA concern the spoken section, which is smaller than in the other national components, with e.g. the category “private dialogue”, which is represented by 90 texts in the British component, comprising a mere 30 texts, most of which were sampled in Kenya¹. The written part of ICE-EA, despite likewise subject to minor changes, is more closely in line with the common ICE design (cf. Schmied and Hudson-Ettle 1999: 8–10 for details). From the point of view of the present study, the use of ICE corpora has clear advantages. They are unique in providing a large set of corpora from all over the world following a parallel design. This is especially important for studies taking into account more than one variety, as manual sampling of data in several geographically distant regions of the world would be excessively time-consuming and costly. In addition, they offer a rich variety of text types, including, especially, spoken data. One of the problems surrounding ICE corpora is their size of 1 million words. Corpus linguists generalise from a corpus to the speech community as a whole, which requires a reasonably large corpus. The required size, however, is dependent on the frequency of the phenomenon under study. Biber (1993), for instance, shows that the cumulative frequency distribution of frequent phenomena such as prepositions tends to be linear, i.e. their frequency rises linearly across adjacent 200-word segments of a given corpus, while the frequency of a rare feature such as relative clauses fluctuates strongly from segment to segment (cf. Biber 1993: 250– 2). The size of the ICE corpora, thus, can be viewed as appropriate for phenomena with a reasonably high text frequency (cf. Mukherjee and S. Hoffmann 2006: 160), such as e.g. aspectual forms (cf. Yao and P. Collins 2012), preposition placement (cf. T. Hoffmann 2011) or particle verbs (cf. Zipp and Bernaisch 2012). There is no doubt that this includes NPs, which are exceptionally frequent. For the analysis of rare phenomena, however, e.g. the investigation of idioms, rare grammatical structures such as cleft sentences or studies focusing on individual lexical items 1 Due to the scarcity of spoken dialogue in a Kenyan and Tanzanian context, the compilers of ICE-EA resort to a novel text category written as spoken, which includes direct transcripts of speech, such as parliamentary Hansards and written records of cross-examinations (cf. Schmied and Hudson-Ettle 1999: 7–8).

154

6 Methodology, corpus handling and statistics

etc., “a token-size of one million is simply not enough” (Schilk 2006: 313), and considerably larger corpora are called for. Another problem concerns style and style stratification. Being restricted to speakers and authors who have completed secondary schooling (cf. 6.1.1), the ICEcorpora can be expected to contain, by and large, an educated form of their variety. Considering the well-known “pyramid-shape relationship between linguistic variability […] and style stratification” (Schneider 2007: 82), however, this is unfortunate. Finally, there is a range of basic sampling problems. To give just one example, almost 75% of the “direct conversations” sampled for ICE-EA, which will be used in the corpus analysis in 7, are based on conversations between students after class (cf. 6.1.1). The parallel sections in ICE-GB and ICE-SING, by contrast, have been sampled in a diverse set of situational contexts. There is no doubt that such divergences cast some doubt on the comparability of ICE corpora. I would argue that these stylistic and sampling problems, while precluding inquiries addressing sociolinguistic aspects of language use, are less of a problem for a study having a broad typological focus, where the sociolinguistic stratification within varieties or languages tends to play a minor role and generalisation over registers is the norm.

6.1.2 Software One of the key software tools used in the present study is the “R” environment (cf. R Development Core Team 2013), which provides both a comprehensive set of statistical functions and text mining functions, which facilitate the extraction of NPs from corpus texts (cf. Gries 2008). A second important tool is the CLAWS tagger (version 4 for Windows), which is used to tag the corpora for parts of speech (cf. Garside 1996; Garside and N. Smith 1997). CLAWS uses two main techniques, namely a probabilistic method, which involves the statistical assessment of the most likely tag for a given word on the basis of its immediate context² and a rule-based approach, which uses a set of fixed grammatical rules and patterns (cf. Garside and N. Smith 1997: 102; cf. Garside and N. Smith 1997: 111–3 for details of the tagging process). According to Garside and N. Smith (1997: 119), the degree of accuracy attained by CLAWS ranges

2 The assessment is based, first, on the probability of one tag following another (for instance, the word following a determiner tag is a great deal more likely to be a noun than a verb), and, second, on the probability “of a word being assigned a particular tag from the list of all possible tags for the word” (Garside and N. Smith 1997: 104).

6.1 General methodology: Corpus, software and statistics

155

from 96%–97%, which has been confirmed in independent studies; van Rooy and Schäfer (2003: 836), for instance, show that even in non-native learner English produced by South African L2 learners of English, the accuracy rate reached by CLAWS is 96.26% (which turns out to be considerably better than the other two taggers tested the Brill tagger and the TOSCA-ICLE tagger).

6.1.3 Statistical methods Ratio data, to start with, will be dealt with in standard ways. They will be compared on the basis of their arithmetic means; in order to represent their dispersion, I will use the standard deviation (cf. Gries 2009: 116) and, in addition, due to the standard deviation’s dependence on the arithmetic mean, the so-called variation coefficient, which is computed by dividing the standard deviation by the mean (cf. Gries 2009: 117). The dispersion of ratio variables will also be assessed by visualising the data using standard box-plots (cf. Gries 2009: 277), with, however, an extra horizontal dashed lines representing the overall arithmetic mean. Following standard procedures, the influence of single independent categorical variables (e.g. variety or syntactic position) on dependent ratio variables (such as NP length or modifier length) will be tested for significance by using monofactorial ANOVAs (cf. Baayen 2008: 103–5 and Gries 2009: 275–83) along with Tukey’s Honestly Significant Difference (Tukey HSD) as a post-hoc test (cf. Gries 2009: 279–80; Baayen 2008: 106–7). If any of the prerequisites for ANOVAs are not met, i.e. variance homogeneity, which can be established on the basis of the so-called Bartlett-test (cf. Gries 2009: 275), and a normal distribution of the residuals (cf. Gries 2009: 275), which can be determined through the Shapiro-Wilk test of normality (cf. Gries 2009: 281), the Kruskal-Wallis rank sum test will be used instead of ANOVAs (cf. Gries 2009: 283; Baayen 2008: 108), with multiple pairwise 𝑈-tests (cf. Hirsig 2000: 6.119) serving as post-hoc procedures³. Multifactorial test designs involving ratio variables, which target e.g. the influence of both variety and register on the lengths of NPs or modifiers, will be addressed by means of two-way ANOVAS (cf. Gries 2009: 284). This method can likewise only be applied under the assumption of variance homogeneity (cf. A. P. Field, Miles, and Z. Field 2012: 518). If this condition is not met, “robust ANOVAs” (cf. Wilcox 2003: 344; cf. also Wilcox and Schönbrodt’s (2014) R package “WRS” and Welch 1951) are a useful alternative, as they yield reliable results even with unequal variances by using trimmed mean values, i.e. mean values based on data 3 Post-hoc procedures for the Kruskal-Wallis test are not standardised to the same degree as Tukey HSD tests are for ANOVAs (cf. above; cf. also Rasch et al. 2006: 166).

156

6 Methodology, corpus handling and statistics

100

with a range such as the top 20% and the bottom 20% removed (cf. Wilcox 2003: 305–6)⁴. With regard to categorical data, the key descriptive strategies include twoand three-dimensional contingency tables. Percentages will be visualised by means of stacked bar charts like those in figure 6.1. The greater-than and lessthan signs followed by asterisks refer to the results of a so-called “Hierarchical Configural Frequency Analysis” (HCFA), which is the key method used to analyse differences in the distribution of categorical data. It is applied by means of an R script implemented by Gries (2004). HCFAs were originally developed in medicine and psychology, where they help to unearth syndromes and psychological types in an objective way by establishing the sets of symptoms which co-occur most frequently in the patients in a sample (cf. Krauth and Lienert 1973). In a distribution of linguistic features like the one shown in figure 6.1, the HCFA procedure can demonstrate whether any combination of the varieties (i.e. ICE-GB, ICE-SING and ICE-EA) and any of the modification types (i.e. premodified or postmodified) is particularly frequent (so-called “types”) or infrequent (so-called “antitypes”; cf. Gries 2009: 244). The HCFA procedure involves establishing all possible interactions of features, with each of them represented by an individual contingency table, including the frequencies of individual variables (e.g. premodified NPs) and interactions

>**

60

80

ICE-SING > ICE-EA (cf. 5.3.4.1.1). (7.11) (7.12)

ice core samples the concept of social class

These considerations will now be summarised in hypotheses 1–3.

Hypothesis 1. In Kenyan English, head-initial NPs will be more frequent than in British English, while in Singaporean English, head-final NPs will be more frequent than in the input variety.

Hypothesis 2. Singaporean English will have longer premodifiers than British English; Kenyan English will have longer postmodifiers than British English.

Hypothesis 3. Premodifiers in Singaporean English will be hierarchically more complex than in British English; Kenyan English will have postmodifiers which are hierarchically more complex than in British English. The hierarchical complexity of pre- and postmodifiers will be assessed on the basis of… (a) the individual formal realisations of pre- and postmodifiers (cf. 5.3.4.1.1; 5.3.4.1.2). (b) the criteria developed above for premodifiers (cf. 5.3.4.1.1), which classify them into unmodified, modified or clausal, i.e. levels (1prem )–(3prem ), and the criteria developed for postmodifiers (cf. 5.3.4.1.2), which classify them into phrasal, clausal (non-finite) or clausal (finite), i.e. levels (1pom )–(3pom ). (c) the numbers of pre- and postmodifying slots occupied (cf. 5.3.4.1.1). (d) the numbers of NPs embedded in either pre- or postmodifying function (cf. 5.3.4.1.1).

168

7 Studying NP modification in Singaporean English and Kenyan English

7.1.2 Hypotheses related to simplification Complexity differences can be expected to appear, first, with regard to length differences and, second, in terms of the hierarchical structure of NPs (cf. 5.3). The variable of NP length will be operationalised in a straightforward way on the basis of word counts. As was pointed out above (cf. 5.3.4.2.1), length differences will be considered with regard to the corpus at large, but also analysed separately for subjects and non-subjects in order to find out to what extent speakers of Singaporean English and Kenyan English rely on the functional advantages of simple subjects (cf. 5.2.1.1). Hierarchical complexity, in turn, will be assessed on the basis of the eightmembered complexity scale (1NP )–(8NP ) developed above (cf. 5.3.4.2.2) as well as the numbers of pre- and postmodifiers used per NP, and the numbers of NPs embedded into other NPs (cf. 5.3.4.2.2). The expected hierarchy of complexity across varieties for all of these distributions is ICE-GB > ICE-SING > ICE-EA. The native variety can be expected to be most complex; Kenyan English, which is the least advanced non-native variety in evolutionary terms, is likely to show the strongest effects of SLA-induced simplification, while Singaporean English will arguably cover the middle ground (cf. 4.3). In addition, the distribution of the levels of NP complexity across subjects and non-subjects will be analysed with the underlying assumption that the less advanced a variety, the more it will capitalise on the parsing and processing advantages offered by simple constituents at the beginning of the clause, i.e. in subject position (cf. 5.2.1.1). As mentioned earlier, however, one of the problems of this hierarchical complexity measure is the asymmetry between pre- and postmodifiers in terms of syntactic expandability (cf. 5.3.4.2.2). While British English, being the only native variety, can be plausibly expected to be more complex than the two new varieties, the complexity relationship of ICE-SING and ICE-EA will depend on the effects of Hypothesis 1. If ICE-SING, in fact, turns out to underuse postmodification, it will necessarily emerge as less complex on the scale between (1NP )–(8NP ), as postmodifiers contribute more strongly to overall complexity than premodifiers. Consequently, in the hypotheses below, the question of the complexity relationship between ICE-SING and ICE-EA will deliberately be left open. The considerations with regard to complexity, thus, can be summarised in the following set of hypotheses.

7.1 Hypotheses on NP structures in Singaporean English and Kenyan English

169

Hypothesis 4. (a) NPs in Singaporean English and Kenyan English will be shorter than NPs in British English. (b) Speakers of Singaporean English and Kenyan English will rely more strongly on short (and, hence, simple) subjects than speakers of British English.

Hypothesis 5. NPs in ICE-SING will be less complex hierarchically than NPs in the input variety ICE-GB; NPs in ICE-EA, in turn, will have simpler hierarchies than NPs in ICE-SING. The following criteria will be used to measure hierarchical complexity²: (a) the eight levels of NP complexity (1NP )–(8NP ) developed above (cf. 5.3.4.2.2) (b) the distribution of the levels of NP complexity across subjects and non-subjects (c) the numbers of pre- and postmodifiers used per NP (cf. 5.3.4.2.2) (d) the numbers of NPs embedded into other NPs (cf. 5.3.4.2.2)

Both language contact and simplification phenomena will be investigated in both S1A, i.e. informal speech, and W1A, which contains student essays (cf. 7.1.1; 6.1.1).

7.1.3 Register-specific areas of interest The comparison of informal spoken (S1A) and highly formal written texts (W1A) will allow a number of conclusions, even though the main focus of the present study is on the effects from cross-linguistic influence and simplification (cf. 7.1.3). The following questions have been identified as particularly relevant areas of investigation (cf. 7.1.3 for details).

(a) Implications for varieties — Are there variety-specific differences in register differentiation? Do informal registers in New Englishes turn out to be more formal

2 Hierarchical complexity, too, is in principle dependent on the proportion of pre- and postmodifiers, since postmodifiers can be expanded syntactically much more easily than premodifiers. However, it is not dependent to the same degree as length, because even long postmodifiers can be hierarchically simple (e.g. by being non-clausal).

170

7 Studying NP modification in Singaporean English and Kenyan English

than in native varieties, as noted in previous analyses (cf. 2.1.3)? This question will be pursued, in particular, in the sections dealing with simplification (cf. 7.3). (b) Implications for registers — Can any differential effects of language contact or simplification be explained via the properties of registers? (c) Implications for NP features — What features have been propagated in the informal domain (i.e. S1A), what features have also been propagated in the formal domain (i.e. W1A)? What may be the reason for the register-specific way of propagation of particular features? None of these questions, however, will take centre stage; rather, they will be discussed alongside the main hypotheses.

7.2 Testing hypotheses about language contact phenomena In this section, the hypotheses on the effects of language contact in Singaporean English and Kenyan English will be tested. The two registers under scrutiny will be addressed in turn (cf. 5.2.1.2). Subsequently, the results for S1A and W1A will be compared; the last section discusses and interprets the results.

7.2.1 Language contact in private dialogue () 7.2.1.1 The frequencies of pre- and postmodified NPs in S1A (Hypothesis 1) In the following, a general perspective on the overall frequencies of pre- and postmodified NPs in ICE-GB, ICE-SING and ICE-EA will be complemented by a range of follow-up analyses.

7.2.1.1.1 The overall frequencies of pre- and postmodified NPs in S1A Table 7.1 and figure 7.1 show the frequencies of pre- and postmodified NPs in ICE-GB, ICE-SING and ICE-EA. The relationship for ICE-GB is 57.62% vs. 42.38% ; cf. (7.13)–(7.18) for examples from ICE-GB. (7.13) (7.14) (7.15)

ready mixed cement   a much less popular route   educational establishments  

7.2 Testing hypotheses about language contact phenomena

171

Table 7.1. Premodification vs. postmodification in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Type of mod.

%

N

ICE-SING %

N

ICE-EA %

N

All

100

postmodified 42.38 189 32.99 128 53.49 207 524 premodified 57.62 257 67.01 260 46.51 180 697 All 100.00 446 100.00 388 100.00 387 1,221

>**

60

80

***

>*

0

* >***

Mod. I − Head

ICE−GB ICE−SING ICE−EA

Fig. 7.3. The co-occurrence patterns of all pre- and all postmodifier types in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A (cf. table A.3, p. 322, for exact HCFA results)

0

10

20

30

40

50

Head − Postmod. 1 Mod. III − Head Mod. II − Head

Mod. II − Head − Postmod. 1 Mod. III − Head − Postmod. 1 Descriptor − Head Mod. I − Head − Postmod. 1 Mod. II − Mod. III − Head Head − Postmod. 1 − Postmod. 2 Emph. − Head − Postmod. 1 Emph. − Mod. III − Head Emph. − Head Mod. I − Mod. III − Head Emph. − Head − Postmod. 1 − Postmod. 2 Mod. I − Mod. III − Head − Postmod. 1 Mod. II − Descriptor − Head Mod. II − Head − Postmod. 1 − Postmod. 2 Emph. − Mod. I − Head Emph. − Mod. I − Head − Postmod. 1 − Postmod. 2 Emph. − Mod. II − Head Emph. − Mod. II − Head − Postm. 1 − Postm. 2 Emph. − Mod. III − Head − Postmod. 1 Mod. I − Head − Postmod. 1 − Postmod. 2 Mod. I − Mod. II − Head − Postmod. 1 Mod. II − Mod. III − Head − Postmod. 1 Descriptor − Mod. III − Head Head − Postmod. 1 − Postmod. 2 − Postmod. 3 Mod. I − Mod. II − Head

7.2 Testing hypotheses about language contact phenomena

177

178

7 Studying NP modification in Singaporean English and Kenyan English

– Mod. II — Descr. — Head: (7.41)

that large cat ⏟⏟⏟⏟⏟⏟⏟⏟⏟ ginger ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ ⏟⏟⏟⏟⏟ Mod. II Obj. descr. Head

– Mod. I — Mod. III — Head — Postmod. 1: (7.42) the only kind people the firm ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ in ⏟⏟⏟⏟⏟⏟⏟ really ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Mod. I

Mod. II

Head

Postmod. 1

Judging from this, it seems that in the native variety, the NP is marked by a somewhat higher variability of modification patterns than in the non-native varieties under study, which leads to the greater number of syntagmatic co-occurrence patterns. This, however, cannot be ascribed to contact influence in a plausible way. Instead, an explanation in terms of simplification tendencies due to general processes of SLA seems more likely. This aspect of variation will be addressed again in the analyses of simplification below (cf. 7.3.1.3).

7.2.1.2 Complexity differences in pre- and postmodifiers in S1A (Hypotheses 2–3) Further effects of language contact to be expected in ICE-SING and ICE-EA are complexity differences specific to either the pre- or the postmodifying portion of the NP (cf. 7.1.2). This section sets out to detect such effects. Length differences between modifiers and different degrees of hierarchical complexity will be dealt with in turn.

7.2.1.2.1 Length differences specific to pre- or postmodification in S1A (Hypothesis 2) In what follows, I will address the lengths of the whole premodifying string as well as the lengths of Mod. II and Mod. III; cases of Mod. I will be disregarded due to their rarity. Figure 7.4, to begin with, shows a box-and-whisker plot (cf. 6.1.3) of the lengths of the whole premodifier string of each NP in ICE-GB, ICE-SING and ICE-EA; a summary of the corresponding core indicators of the length differences is given in table 7.4⁵.

5 The sums shown in table 7.4 comprise both NPs which are premodified only and NPs which have a pre- and a postmodifier.







3







2

Length (in words)

4

5

7.2 Testing hypotheses about language contact phenomena



+

+

1

+

ICE−GB ICE−SING

179

ICE−EA

Variety

Fig. 7.4. The lengths of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EAS1A. Horizontal dashed line: arithmetic mean of all length values; -signs: arithmetic means for individual varieties (cf. 6.1.3)

+

Table 7.4. The lengths of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICEEA-S1A ICE-GB ICE-SING ICE-EA Number of premodifying strings Arithmetic mean SD Variation coefficient

316 1.37 0.68 0.49

278 1.35 0.70 0.52

213 1.23 0.61 0.49

Monofactorial ANOVA test

F = 2.8899, df = 2, p = 0.05616

The lack of a box for ICE-EA shows immediately that this variety stands out with short premodifying strings, while ICE-SING patterns together with ICE-GB. As can be seen in table 7.4, an average premodifying string in ICE-GB and ICE-SING consists of 1.37 and 1.35 words respectively; in ICE-EA, in contrast, the mean value is down to 1.23 words. Hierarchically complex premodifiers, as shown in (7.43)– (7.45), many of which involve SKT-constructions (cf. 5.1.4.3), thus, are more frequent in British English and Singaporean English than in Kenyan English.

180

7 Studying NP modification in Singaporean English and Kenyan English

(7.43) a four month kind of a attachment (7.44) our Elijah musical sort of music (7.45) a slightly sort of cynical attitude Kenyan English, in contrast, relies more strongly on short premodifiers of the type shown in (7.46)–(7.48). (7.46) a philosophical treatise (7.47) a very fussy person (7.48) the adult literacy classes Even though this fits in very well with Hypothesis 2, this result has to be taken with a grain of salt since, according to a monofactorial ANOVA⁶, this difference is not significant, with the 𝑝 value, however, just below the significance threshold (𝑝 = 0.0587; cf. table 7.4). In addition, the post-hoc TukeyHSD test shown in table 7.5 reveals that none of the differences between the individual varieties is significant. Table 7.5. A TukeyHSD post hoc test of the mean lengths of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Varieties ICE-SING ICE-GB ICE-GB

diff vs. vs. vs.

ICE-EA ICE-EA ICE-SING

0.117,776,2 0.135,511,38 0.017,735,18

lwr

upr

-0.025,177,721 0.260,730,1 -0.003,663,874 0.274,686,6 -0.111,355,479 0.146,825,8

p-value 0.129,698 0.058,303 0.944,257,3

Table 7.6, in turn, shows the mean lengths of descriptive Mod. II in ICE-GBS1A, ICE-SING-S1A and ICE-EA-S1A⁷. With regard to the arithmetic means, NPs in ICE-SING are slightly longer than in ICE-GB (1.40 words vs. 1.34 words), while the average length for ICE-EA (1.20 words) is markedly lower. The cases of Mod. II which are longer than one word are primarily realised by APs involving degree modification through adverb phrases, as in (7.49)–(7.52), which are rarely longer than two words, as in (7.52). Only one of them exceeds the length of three words

6 A Bartlett test reveals the variances of the length differences to be homogeneous (𝐾2 = 5.265, 𝑑𝑓 = 2, 𝑝 = 0.0719). For this reason, an ANOVA procedure is permissible (cf. 6.1.3). 7 Due to the excessively low length measures for Mod. II, a boxplot visualisation is not possible; no boxes are rendered by the R function.

7.2 Testing hypotheses about language contact phenomena

181

Table 7.6. The lengths of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A ICE-GB ICE-SING ICE-EA Number of Mod. II Arithmetic mean SD Variation coefficient

92 1.34 0.56 0.42

47 1.40 0.58 0.41

54 1.20 0.63 0.52

Monofactorial ANOVA test F = 1.4111, df = 2, p = 0.2464

(cf. (7.53)); this modifier, however, counter to Hypothesis 2, occurs in the Kenyan sample. (7.49) (7.50) (7.51) (7.52)

a very subtle change a particularly good singer the most realistic portrayal Very talented brilliant people who have travelled a lot very widely

(7.53) a very important but very difficult question

A unanimous confirmation of Hypothesis 2, thus, is not possible on the basis of Mod. II either, even though, as in the case of the whole premodifying string (cf. figure 7.4), p. 179, the pattern of deviation found in both ICE-SING and ICE-EA accords well with expectations. Table 7.7 lists the length differences between the classifying Mod. III in the varieties under study⁸.

Table 7.7. The lengths of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A ICE-GB ICE-SING ICE-EA Number of Mod. III Arithmetic mean SD Variation coefficient

176 1.19 0.50 0.42

194 1.29 0.69 0.53

128 1.10 0.33 0.30

Kruskal-Wallis rank sum test H = 7.234, df = 2; p = 0.02686*

8 Once more, the low degree of length differences of Mod. III does not allow for a boxplot visualisation; no boxes are drawn by the algorithm.

182

7 Studying NP modification in Singaporean English and Kenyan English

Table 7.8. Pairwise Wilcoxon rank sum test of the lengths of Mod. III in ICE-GB-S1A, ICE-SINGS1A and ICE-EA-S1A ICE-GB

ICE-SING

ICE-SING p = 0.227 — ICE-EA p = 0.227 p = 0.023*

The lengths of Mod. III exhibit a stronger degree of variation between the varieties than the length distributions considered so far: ICE-SING clearly leads the pack, ICE-GB covers the middle ground, while ICE-EA has the lowest value, which confirms Hypothesis 2. A Kruskal-Wallis rank sum test⁹ demonstrates that these differences are statistically significant at 𝑝 = 0.027∗∗ (cf. table 7.7). A pairwise comparison of the individual varieties by means of a repeated Wilcoxon rank sum test (cf. 6.1.3) shown in table 7.8, in turn, demonstrates that it is only the interaction between ICE-SING and ICE-EA that is significant (𝑝 = 0.023∗ ); the two new varieties, in contrast, do not differ significantly from their common input variety. One might be led to conclude that, metaphorically speaking, they have moved in different directions from their common starting point British English due to divergent contact influence. From a formal point of view, the cases of Mod. III longer than one word primarily consist of modified phrases used as Mod. III, as in (7.54) or stacked Mod. III, as in (7.55). (7.54) the research scholarship schemes ICE-SIN:S1A-068#93:1:D> (7.55) American card holders Rare long cases, such as (7.56) (three words) or (7.57) (four words) frequently involve SKT-constructions (cf. 5.1.4.3). (7.56) a counselling kind of setup (7.57) a four month kind of a attachment Figure 7.5, finally, shows two boxplot representations of the lengths of the whole postmodifying strings in ICE-GB, ICE-SING and ICE-EA; Figure 7.5a represents all values including outliers; in the zoom version in figure 7.5b, all outliers have been omitted, which brings out the main differences in greater detail.

9 The lack of homogeneity of variances bars the use of the ANOVA procedure; Bartlett’s 𝐾2 = 76.0278, 𝑑𝑓 = 2, 𝑝 = 2.2 ⋅ 10−16 .

7.2 Testing hypotheses about language contact phenomena

(b) Without outliers (zoom view)

10

(a) Including all values

183

● ● ● ● ● ● ● ●

● ● ● ● ● ● ●

+

6

+

+

● ● ●

+

+

0

+

● ● ●

2

20



4

30



10

Length (in words)



Length (in words)

8

40



ICE−GB ICE−SING

ICE−EA

ICE−GB ICE−SING

Variety

ICE−EA

Variety

Fig. 7.5. The lengths of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EAS1A. Horizontal dashed line: arithmetic mean of all length values; -signs: arithmetic means for individual varieties (cf. 6.1.3)

+

Unsurprisingly, postmodifiers exhibit a far greater range of lengths than premodifiers (cf. figure 7.4 on p. 179), which is immediately evident from the boxplots. The mean length of postmodifiers in ICE-GB is 4.29 words. The value for ICE-SING of 3.87 words turns out to be markedly lower than that, and the ICE-EA mean of 4.60 words clearly exceeds that of ICE-GB (cf. table 7.9). According to a KruskalWallis rank sum test¹⁰, this difference is significant at 𝑝 = 0.02194∗ (cf. table 7.9). Table 7.10 shows the results of a pairwise 𝑈-test (cf. 6.1.3), which teases out the differences between the individual corpus parts. As can be seen, the only significant difference lies between ICE-SING and ICE-EA (𝑝 = 0.016∗∗ ). Neither the differences between ICE-GB and ICE-SING nor those between ICE-GB and ICE-EA reach the required level of significance. As above, this allows the conclusion that the two non-native varieties, as it were, have moved into different directions. Their respective “distance” from their common input variety is still minor, but they have 10 As can be shown by the Bartlett test of homogeneity of variances (𝐾2 = 28.4528, 𝑑𝑓 = 2, 𝑝 = 6.631 ⋅ 10−7 ), the variances of the three length measures are not equal, which bars ANOVAs.

184

7 Studying NP modification in Singaporean English and Kenyan English

Table 7.9. The lengths of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICEEA-S1A ICE-GB ICE-SING ICE-EA Number of postmodifiers Arithmetic mean SD Variation Coefficient

248 4.29 3.44 0.80

146 3.87 3.14 0.81

240 4.60 4.49 0.98

Kruskal-Wallis rank sum test H= 7.639, df = 2; p = 0.02194*

Table 7.10. Pairwise Wilcoxon rank sum test of the mean lengths of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A ICE-GB

ICE-SING

— ICE-SING p = 0.144 ICE-EA p = 0.27 p = 0.016*

come to be significantly different from each other, owing to typological influence. The kinds of long postmodifiers like (7.58) (45 words), (7.59) (33 words) or (7.60) (27 words), thus, are more likely to occur in ICE-EA than in ICE-SING: (7.58) a way in which we can also persuade them that it is beneficial to them as members of our society and as individuals as well and as fathers and so forth if the position of women can be changed so that she is no longer an object (7.59) in a culture where the man is supposed to be the head and the woman the neck or there down and it is not possible for an educated woman to accept the position of the neck (7.60) a situation whereby I got so much interested that in her assignment in the story I would miss doing anything else but not that not to get good marks As can be seen, formally speaking, such excessively long cases are based on elaborate clausal postmodification of abstract nouns such as way, culture or situation (cf. (7.58)–(7.60)), which are low in meaning themselves (cf. 5.2.1.1); furthermore, they frequently involve the coordination of constituents within the postmodifiers (cf. 7.2.1.2.2 for a detailed analysis of the form of postmodifiers).

7.2 Testing hypotheses about language contact phenomena

185

Short postmodifiers as in (7.61)–(7.63), in contrast, while underrepresented in ICEEA (cf. the missing lower portion of the box for ICE-EA in figure 7.5b), are more typical of ICE-SING. (7.61) the wedding itself (7.62) a tourist of Singapore (7.63) survey forms not returned In sum, if length is accepted as a proxy of syntactic complexity, broadly speaking, ICE-SING has complex premodifiers and simple postmodifiers. Conversely, ICE-EA is marked by simple premodifiers and complex postmodifiers. 7.2.1.2.2 Hierarchical complexity differences between pre- and postmodifiers in S1A (Hypothesis 3) In this section, the complexity of modifiers will be measured in terms of hierarchical syntactic structure (cf. 5.3.4.1). In the following, a short general overview of the formal realisations of pre- and postmodifiers will be followed by a complexity assessment based on the numbers of pre- and postmodifying slots occupied in NPs and the preferred position of NP embedding in ICE-GB, ICE-SING and ICE-EA (cf. 5.3.4.1). The following investigation of the formal categories used as pre- and postmodifiers will provide preliminary insights into complexity differences (cf. 5.3.4.1.1) without being geared directly to any of the hypotheses. With regard to premodification, I will focus on Mod. II and Mod. III due to their being the most frequent modifier types (cf. 7.2.1.1.2; cf. also 7.2.1.2.1). They will be addressed in 7.2.1.2.2– 7.2.1.2.2; in 7.2.1.2.2, the formal realisations of postmodification will be discussed.

(a) Formal realisations of Mod. II — Table 7.11 and figure 7.6 show the formal realisations of Mod. II in ICE-GB, ICE-SING and ICE-EA. As expected, the vast majority of Mod. II are realised by APs (cf. (7.64)–(7.66)). (7.64) a much less popular route (7.65) his very busy day (7.66) some of the aggressive customs A very small percentage of Mod. II only are occupied by elements other than APs. (7.67) shows one of the rare cases of several coordinated APs used in a row. ICE-GB has one SKT-construction in this position, which is shown in (7.68); (7.69), which is taken from ICE-SING, represents a rare case of a lexical phrase used as a Mod. II.

186

7 Studying NP modification in Singaporean English and Kenyan English

Table 7.11. The formal realisations of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Formal realisation of Mod. II

0.00 1.09 2.17 2.17 94.57 100.00

N

%

N

ICE-EA %

N All

0 2.13 1 0.00 0 1 1 0.00 0 0.00 0 1 2 0.00 0 0.00 0 2 2 2.13 1 1.85 1 4 87 95.74 45 98.15 53 185 92 100.00 47 100.00 54 193

80

100

lexical phrase sort/type/kind-constr NP several mods. AP All

%

ICE-SING

0

20

40

60

lexical phrase SKT−constr. NP several mods. AP

ICE−GB

ICE−SING

ICE−EA

Fig. 7.6. The formal realisations of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A

(7.67) a vivid bigger imagination (7.68) a huge sort of lake on our doorstep (7.69) the most uh high-class place

The frequencies of the non-adjectival, complex cases of Mod. II in figure 7.6 decrease gently from the native variety via ICE-SING to ICE-EA, which, however, is a far cry from statistical significance.

7.2 Testing hypotheses about language contact phenomena

187

(b) Formal realisations of Mod. III are shown in table 7.12 and figure 7.7¹¹. As can be seen, in contrast to Mod. II, the varieties exhibit clear differences in terms of formal realisations. Table 7.12. The formal realisations of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Formal realisation of Mod. III

N

%

N

ICE-EA %

N

All

0.00 0 0.52 1 0.00 0 1 0.00 0 1.55 3 0.00 0 3 0.57 1 0.00 0 0.00 0 1 0.57 1 0.52 1 2.34 3 5 1.14 2 4.12 8 0.78 1 11 1.70 3 6.19 12 1.56 2 17 2.27 4 1.55 3 2.34 3 10 2.27 4 0.00 0 0.00 0 4 2.84 5 6.70 13 0.00 0 18 17.61 31 12.89 25 34.38 44 100 71.02 125 65.98 128 58.59 75 328 100.00 176 100.00 194 100.00 128 498

80

100

VP present part. quant. gen. sort/type/kind-constr. ing-form several mods. past part. lexical phrase AP NP All

%

ICE-SING

0

20

40

60

>*

VP present part. quant. gen. SKT−constr. i ng−form several mods. past part. lexical phrase AP NP

ICE−GB

ICE−SING

ICE−EA

Fig. 7.7. The formal realisations of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A (cf. table A.4, p. 323, for exact HCFA results) 11 Cf. Footnote 3; p. 115 for the understanding of the terms ing-form and “present part.” used in the table.

188

7 Studying NP modification in Singaporean English and Kenyan English

Most notably, the relative proportions of NPs (cf. (7.70)) and APs (cf. (7.71)– (7.72)) differ considerably from variety to variety. (7.70) background resistance by women (7.71) diplomatic relations (7.72) a physiological need

The proportion of NPs decreases gradually from ICE-GB (71.02%) via ICE-SING (65.98%) to a low point of 58.59% in ICE-EA. APs, in turn, are least frequent in this position in ICE-SING (12.89%), slightly more frequent in ICE-GB (17.61%), and their frequency of 34.38% in ICE-EA is significantly higher than in the other varieties according to the HCFA procedure. Adjectival realisation of Mod. III, mostly by means of APs headed by denominal adjectives, like (7.73)–(7.89), thus, appears to be a particularly typical feature of Kenyan English. (7.73) (7.74) (7.75) (7.76) (7.77) (7.78) (7.79) (7.80) (7.81) (7.82) (7.83) (7.84) (7.85) (7.86) (7.87) (7.88) (7.89)

basic need Kenyan writer African girl personal notes commercial colleges Western government tribal wars African woman immaterial culture psychological point of view foreign affairs creative work literary criticism economic bit creative writing sexual harassment domestic issues

There are two possible reasons for the high Kenyan frequency of APs, both of which are not typological in nature but draw on other considerations. For one thing, adjectival classifiers are associated with higher registers than nominal ones (cf. 5.2.1.2). The high share of APs in Kenyan English, thus, may be a reflection of

7.2 Testing hypotheses about language contact phenomena

189

a certain “bookishness” of spoken registers in ICE-EA, which is due to the strong association of English with the educational sphere, while everyday situations are dominated by local vernaculars (cf. 4.2.1.3.2; cf. also van Rooy, Terblanche, et al. 2010: 346). In fact, many of the adjectives in question are formal and technical in nature (cf. (7.73)–(7.89)). In addition, the ICE section “private dialogue” in ICE-EA consists primarily of conversations with students, which may also be the cause of a more formal register of conversation. A second reason could, in fact, be related to complexity differences. Nominal premodification is marked by a degree of implicitness by allowing multiple interpretations, as was shown above (cf. 5.2.2.1); the Kenyan preference for APs in ICE-EA, thus, may be the effect of a tendency towards more explicit and, hence, less complex structures. The formal realisations of Mod. III by structures other than NP and AP turn out to be highly infrequent across the board (cf. figure 7.7). It is unusual for a Mod. III to be realised by a present participle, (7.90) or a genitive, (7.91). (7.90) visiting professors (7.91) Ladies’ finger

Despite their low occurrence rates, it is obvious that they are used more readily in ICE-SING than in ICE-EA. The bar for ICE-SING in figure 7.7 indicates that ingforms (cf. Footnote 3; p. 115), SKT-constructions (cf. 5.1.4.3), genitives, and present participles, if infrequently, all appear in ICE-SING, while they are either rare or wholly absent from ICE-EA. Despite the lack of statistically sound evidence, thus, ICE-SING can be said to have a slightly higher formal diversity in Mod. III position than ICE-EA and to use more of the modifiers with low type frequency, which, I would argue, is another proxy of prehead complexity and thus supports Hypothesis 3. (c) Formal realisations of postmodifiers — Table 7.13 and figure 7.8 provide an overview of the realisations of Postmod. 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A (Postmod. 2 and all further postmodifying slots are disregarded due to their low frequencies). Across varieties, prepositional phrases lead the pack (72.98%— 70.55%— 63.33%; cf. (7.92)–(7.93)), while relative clauses are the second most frequent option (14.11%— 15.07%— 23.33%; cf. (7.94)–(7.95)). (7.92) the weather in San Francisco (7.93) the other blacks in all those spheres

190

7 Studying NP modification in Singaporean English and Kenyan English

Table 7.13. The formal realisations of Postmod. 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Formal realisation of Postmod. 1

0.00 0.40 0.00 0.00 1.21 3.63 2.02 5.65 14.11 72.98 100.00

N

%

N

ICE-EA %

0 0.00 0 0.42 1 0.00 0 0.00 0 0.00 0 0.83 0 1.37 2 0.42 3 3.42 5 1.67 9 4.11 6 0.00 5 1.37 2 3.33 14 4.11 6 6.67 35 15.07 22 23.33 181 70.55 103 63.33 248 100.00 146 100.00

N

All

1 0 2 1 4 0 8 16 56 152 240

1 1 2 3 12 15 15 36 113 436 634

100

idiom quant. num pronoun NP AdvP fin. cl. nonfin. cl. rel. cl. PP All

%

ICE-SING

0

20

40

60

80

idiom quant. num. pronoun NP AdvP fin. cl. nonfin. cl. rel. cl. PP

ICE−GB

ICE−SING

ICE−EA

Fig. 7.8. The formal realisations of Postmod. 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A

(7.94) the people who come with no data and have to actually go and start doing their survey from scratch (7.95) a question I’ve always been asked even by the Kenyan media

7.2 Testing hypotheses about language contact phenomena

191

Even though none of the frequency deviations reaches the level of statistical significance, there is one noteworthy observation: While ICE-GB and ICE-SING pattern together in terms of the relationship between PPs and relative clauses, ICE-EA has a markedly higher incidence of relative clauses (23.33%) than the other varieties under study, at the expense of PPs (63.33%, while both ICE-GB and ICE-SING have shares of PPs beyond 70%). Considering Berlage’s (2014: 255) emphasis on finite modifying clauses as a prime factor in NP complexity (cf. 5.3.2.2), this may be viewed as a tentative indication of a contact-induced rise in complexity levels specific to the postmodifying zone in Kenyan English, even though the evidence is not fully convincing due to the lack of statistical significance. The remaining forms of NP modification do not show marked differences between the varieties under study. Non-finite clauses as in (7.96) (5.65%— 4.11%— 6.67%), finite clauses other than relative clauses as in (7.97) (2.02%— 1.37%— 3.33%) or pronouns, exemplified in (7.98) (primarily reflexive pronouns in emphatic use; 0.00%— 1.37%— 0.42%), for instance, are similarly infrequent in the varieties under scrutiny. (7.96) any barrier hindering you from that kind of a marriage

(7.97) the fact that I’ll be totally responsible for my own diet

(7.98) the wedding itself

Summing up, this analysis has yielded a range of preliminary indications that Singaporean English and Kenyan English differ in the preferred formal realisation of the pre- and postmodifying string of NPs. Some of them may be due to complexity differences. Furthermore, the main deviations concern premodifiers in ICE-SING and postmodifiers in ICE-EA. At least in an indirect way, thus, they may be viewed as evidence in favour of Hypothesis 3. The key question, however, is to what extent these differences in formal realisation correlate with complexity differences in the sense of the definition of complexity developed above (cf. 5.3.4.1). The following steps are dedicated to the comparison of the hierarchical complexity levels of pre- and postmodifiers between the varieties under study. 7.2.1.2.2–7.2.1.2.2 will provide analyses of the complexity levels of Mod. II and Mod. III respectively; Emphasisers, Mod. I and objective descriptors are too infrequent across the board to warrant such an analysis. Point 7.2.1.2.2, in turn, takes into view the premodifying string as a whole by generalising across all functional subzones. 7.2.1.2.2 will target the complexity levels of postmodifiers.

192

7 Studying NP modification in Singaporean English and Kenyan English

Table 7.14. The complexity levels of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Complexity level of Mod. II

%

N

ICE-SING %

N

ICE-EA %

N All

60

80

100

(2Prem ) — modified/coordinated phrases 27.17 25 36.17 17 14.81 8 50 (1Prem ) — simple phrases 72.83 67 63.83 30 85.19 46 143 All 100.00 92 100.00 47 100.00 54 193

0

20

40

level (2Prem) − modified/coord. p. level (1Prem) − simple phrases

ICE−GB

ICE−SING

ICE−EA

Fig. 7.9. The complexity levels of Mod. II in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A

(a) The complexity of Mod. II — Table 7.14 and figure 7.9 show the proportions of these modifier types in ICE-GB, ICE-SING and ICE-EA¹² The modified type (2Prem ), as against (1Prem ), patterns in a way which is consonant with Hypothesis 3. It is more frequent in ICE-SING than in ICE-GB , and less frequent in ICE-EA than in the input variety, even thought there is no statistical significance. (7.99)–(7.101) show representative modified cases classified as (2Prem ) from ICE-SING.

12 It should be noted that non-sentential lexical phrases, such as sell-by in sell-by date were counted as “modified”. Modifiers containing SKT-constructions (cf. 5.1.4.3) such as sort of a were filed as “modified”.

7.2 Testing hypotheses about language contact phenomena

193

(7.99) his very busy day (7.100) quite a good course (7.101) a very practical example Practically all of the modifiers of APs realising Mod. II in ICE-SING are amplifiers such as very or extremely or downtoners such as quite (cf. Quirk et al. 1985: 589– 90). Cases of coordinated Mod. II, as in (7.102), in contrast, are utterly infrequent: (7.102) a proper and accurate appraisal of oneself Sentential Mod. II, i.e. (3Prem ), finally, do not feature at all in the sample. The evidence for substrate influence in terms of the complexity of Mod. II, thus, cannot be confirmed with certainty from a statistical point of view due to the low token numbers involved; I would argue, however, that it does not seem wholly implausible. (b) The complexity of Mod. III — Table 7.15 shows a contingency table of the complexity levels of the realisations of Mod. III, which is represented graphically in figure 7.10. Across varieties, the vast majority of them are simple phrases, i.e. cases of (1Prem ) like (7.103)–(7.104). Table 7.15. The complexity levels of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Complexity level of Mod. III

%

N

ICE-SING %

N

ICE-EA %

N

All

(2Prem ) — modified/coordinated phrases 9.66 17 18.56 36 6.25 8 61 (1Prem ) — simple phrases 90.34 159 81.44 158 93.75 120 437 All 100.00 176 100.00 194 100.00 128 498

(7.103) an African girl (7.104) a Western buffet As regards the proportions of modified or coordinated premodifiers, i.e. (2Prem ), interestingly, it is once more ICE-SING that stands out with the top score of 18.56%.

7 Studying NP modification in Singaporean English and Kenyan English

60

80

100

194

0

20

40

level (2Prem) − modified/coord. p. level (1Prem) − simple phrases

ICE−GB

ICE−SING

ICE−EA

Fig. 7.10. The complexity levels of Mod. III in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A (cf. table A.5, p. 323, for HCFA results)

ICE-GB is way below that at 9.66%, and ICE-EA turns out to be even lower, with a share of merely 6.25%. Relevant cases are shown in (7.105)–(7.109). (7.105) (7.106) (7.107) (7.108) (7.109)

the research scholarship schemes our Elijah musical sort of music the wife inheritance thing property development exercise the western kind of look

For the most part, such modifiers have a modifier themselves, or represent SKTconstructions. Coordinated phrases, as in (7.110), are extremely rare across varieties, and the highest complexity level (3Prem ), sentential realisation, does not figure at all in the sample. (7.110) economic and uh political stress from from the outside world

Once more, however, the HCFA procedure does not yield any significant configuration, even though the frequency of modified Mod. III in ICE-SING is classified as

195

7.2 Testing hypotheses about language contact phenomena

“marginally significant” (𝑝 = 0.07) by the algorithm (cf. table A.5 in Appendix A, p. 323), which indicates that it approximates the necessary significance level¹³. (c) The complexity of the whole premodifying string — This part of the analysis focuses on the complexity of the premodifying string as a whole. The pertinent result is shown in table 7.16 and visualised in figure 7.11. It emerges that ICE-EA has a markedly lower frequency of complex premodifiers (12.21%), which is a far cry from the 20.25% and 22.30% yielded for ICE-GB and ICE-SING respectively. This

Table 7.16. The complexity levels of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Complexity level of premodifying string

%

N

ICE-SING %

N

ICE-EA %

N

All

60

80

100

level (2Prem ) — modified/coordinated phrases 20.25 64 22.30 62 12.21 26 152 level (1Prem ) — simple phrases 79.75 252 77.70 216 87.79 187 655 All 100.00 316 100.00 278 100.00 213 807

0

20

40

level (2Prem) − modified/coord. p. level (1Prem) − simple phrases

ICE−GB

ICE−SING

ICE−EA

Fig. 7.11. The complexity level of the whole premodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A

13 The label “marginal significant” means that a given 𝑝-values lies between 0.05 and 0.1 (cf. Gries 2009: 32; cf. also 6.1.3).

196

7 Studying NP modification in Singaporean English and Kenyan English

pattern, too, is in line with Hypothesis 3; once more, however, the interpretation is hampered by the lack of statistical significance. (d) The complexity of postmodifiers — The relevant distribution according to the criteria developed above (cf. 5.3.4.1.2) is shown in table 7.17 and visualised in figure 7.12. The most frequent postmodifier types (68.55%— 76.03%— 60.83%) are invariably phrases devoid of clausal structures, as illustrated in (7.111)–(7.114); (7.112) shows a rare case of an appositional NP as a postmodifier.

Table 7.17. Complexity levels of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Complexity level of postmodifying string

22.18 9.27 68.55 100.00

N

%

N

ICE-EA %

55 19.18 28 30.83 23 4.79 7 8.33 170 76.03 111 60.83 248 100.00 146 100.00

N

All

74 20 146 240

157 50 427 634

60

80

100

level (3Pom ) — clausal (finite) level (2Pom ) — clausal (non-finite) level (1Pom ) — phrasal All

%

ICE-SING

0

20

40

level (3Pom) − clausal (finite) level (2Pom) − clausal (nonfinite) level (1Pom) − phrasal

ICE−GB

ICE−SING

ICE−EA

Fig. 7.12. The complexity levels of the whole postmodifying string in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A

7.2 Testing hypotheses about language contact phenomena

197

(7.111) my friend from Hong Kong (7.112) the bus Greyhound from uh Chicago to Indiana (7.113) the strength of my English language competency

(7.114) a lady in mini-skirt With regard to non-simple patterns, however, a glance at figure 7.8 reveals a noteworthy pattern. Compared to ICE-GB, ICE-EA stands out with a marked rise of postmodifiers at complexity level 3 (finite clauses — 30.83% vs. 22.18%)¹⁴, which resonates with Hypothesis 3. Examples involving a relative clause, an appositive that-clause and a finite wh-clause as part of a prepositional phrase are provided in (7.115)–(7.120)¹⁵. (7.115) the legal skills that are required for reasoning and analytical work

(7.116) sense that maybe I I mean come out with little

(7.117) some aspect of how I how I exercise my thinking

(7.118) this issue how religion came into the society (7.119) We are not like guys have no shame (7.120) a time when his wife left for some three weeks and he he didn’t know why she had left ICE-SING, in contrast, has a particularly low frequency of complexity level (2Pom ) in comparison to both ICE-GB and ICE-EA (non-finite clauses — 4.79% in ICE-SING as against 9.27% in ICE-GB and 8.33% in ICE-EA). Examples of non-finite clauses can be seen in (7.121)–(7.124). (7.121) his clearance to do something (7.122) the better person to tell him what who who has been working in the library committee

14 Postmodifiers were assigned to this highest complexity level regardless of the syntactic level at which a finite clause occurred: In (7.117), the formal class of the postmodifier is a PP, but into the PP, a finite nominal wh-clause (how I how I exercise my thinking) is embedded. 15 The Kenyan English example (7.118) appears to be a non-standard case of a relative clause lacking a subject relative pronoun, which, however, has not so far been described as a systematic feature of this variety (cf. Schmied 2004a: 932–5).

198

7 Studying NP modification in Singaporean English and Kenyan English

(7.123) a responsibility to generate resources for the family (7.124) a matter of uh expense of importing spare parts

Taken together, both ICE-SING and ICE-EA show deviations which, in principle, appear interpretable from the point of view of Hypothesis 3, which, however, are not convincing due to the lack of statistical evidence. The complexity of the premodifying string as a whole also rises by the simultaneous use of several modifier types (cf. 5.3.4.1.1). In the following, 7.2.1.2.2 will address the number of different premodifying slots occupied in the varieties under study, while 7.2.1.2.2 will deal with postmodifiers.

(a) Number of premodifying slots occupied — As can be seen in table 7.18 and figure 7.13, the vast majority of NPs in the sample have one premodifier only, as in (7.125)–(7.126). An NP with more than two of the premodifying slots occupied cannot be found in any of the varieties. Table 7.18. The number of premodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Number of premods. 2 1 All

%

N

ICE-SING %

N

ICE-EA %

N

All

8.54 27 3.24 9 4.69 10 46 91.46 289 96.76 269 95.31 203 761 100.00 316 100.00 278 100.00 213 807

(7.125) a very sadMod. II story (7.126) sexualMod. III harassment The frequency of NPs having two functionally distinct modifiers, like (7.127)– (7.128), varies gently: It is highest in ICE-GB, lowest in ICE-SING, and reaches an intermediate rate in ICE-EA. (7.127) Aik Leng has very goodMod. II ballMod. III sense (7.128) a certainMod. I traditionalMod. III god

199

60

80

100

7.2 Testing hypotheses about language contact phenomena

0

20

40

2 prems. 1 prem.

ICE−GB

ICE−SING

ICE−EA

Fig. 7.13. The number of premodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A

The relationship between ICE-SING and ICE-EA, at first sight, may seem to contradict Hypothesis 3. With the frequency differences so small and the token rates, especially in the two non-native corpora, so low, however, it will not be assumed that this distribution reflects any valid facts about contact effects in the varieties under study. (b) Number of postmodifying slots occupied — Much the same applies to the number of postmodifier slots occupied in the corpora under scrutiny, which are tabulated in table 7.19 and visualised in figure 7.14. The vast majority of NPs have one

Table 7.19. The number of postmodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Number of postmods. 3 2 1 All

%

N

ICE-SING %

N

ICE-EA %

N

All

0.00 0 0.68 1 0.00 0 1 6.05 15 1.37 2 2.08 5 22 93.95 233 97.95 143 97.92 235 611 100.00 248 100.00 146 100.00 240 634

7 Studying NP modification in Singaporean English and Kenyan English

60

80

100

200

0

20

40

3 poms. 2 poms. 1 pom.

ICE−GB

ICE−SING

ICE−EA

Fig. 7.14. The number of postmodifying slots occupied per NP in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A

modifier only. A vanishingly low proportion have two, as illustrated in (7.129), and one single Singaporean NP shown in (7.130) has three postmodifiers . (7.129) his own experiences as a young manPostmod. 1 the in the tribal wars between the Kikuyus and the MaasaisPostmod. 2 (7.130) this book Myths and Legends in of SingaporePostmod. 1 published by VJ TimesPostmod. 2 who got things like got chapters like Pulau HantuPostmod. 3 ¹⁶ As before, no hard-and-fast conclusions can be drawn from the cases of multiple postmodification due to excessively low token numbers. If at all, the slightly higher rate of double postmodification may be attributed to British English being a native variety. No plausible effects from contact influence, however, can be made out (cf. 7.3.1.3 for an interpretation of this distribution in terms of simplification). Finally, Hypothesis 3c predicts that typological influence may lead to differences in the numbers of NPs embedded as either pre- or postmodifiers in the varieties under study. Table 7.20 shows the relevant proportions of pre- and posthead functions of level-1 NPs in the sample, which are visualised in figure 7.15. Additionally,

16 In (7.130), the relative pronoun who is used in a non-standard way to refer to the referent of book.

7.2 Testing hypotheses about language contact phenomena

201

Table 7.20. The position of NPs embedded at level 1 in ICE-GB-S1A, ICE-SING-S1A and ICE-EAS1A Variety ICE-GB Position

%

N

ICE-SING %

N

ICE-EA %

N

All

100

pre-head 40.06 137 55.79 159 30.43 77 373 post-head 59.94 205 44.21 126 69.57 176 507 All 100.00 342 100.00 285 100.00 253 880

80

**

40

pre−head post−head

>*

0

20

**

2 prem. − 2 pom. 1 prem. − 2 pom. 2 prem. − 1 pom. 1 prem. − 1 pom. 0 prem. − 3 pom. 0 prem. − 2 pom. 0 prem. − 1 pom. 2 prem. − 0 pom. 1 prem. − 0 pom.

ICE−GB

ICE−SING

ICE−EA

Fig. 7.37. The numbers of pre- and postmodifiers in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A (cf. table A.10, p. 325, for exact HCFA results)

260

7 Studying NP modification in Singaporean English and Kenyan English

involving one pre- or one postmodifier reflect the familiar set of preferences in ICE-SING and ICE-GB: The pattern involving one premodifier only is significantly more frequent in ICE-SING, while the postmodified pattern turns out to be more frequent in ICE-EA, which, however, is more plausibly interpreted in terms of typology than in terms of simplification (cf. 7.2.1.1). More relevant than that, the distribution of the most complex types involving both pre- and postmodification reveals an interesting pattern. The most frequent type is the one involving one premodifier and one postmodifier (cf. (7.278)), which turns out to be roughly equally frequent in ICE-GB and ICE-EA (9.31% — 7.38%), while being rare in ICE-SING (4.43%). (7.278) some diplomatic with Russia ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ relations ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Postmod. 1

Mod. III

The patterns ranked as even more complex, which involve several pre- and/or several postmodifiers, have small but tangible frequencies in ICE-GB and ICE-EA, while being completely absent from ICE-SING; they include “2 prem. — 1 pom.” (exemplified in (7.279)), “2 prem. — 2 pom.” (cf. (7.280)), “1 prem. — 2 pom.” (cf. (7.281)) (cf. table 7.57). about books (7.279) the sort of nicest ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ thing ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Emph.

Mod. II

Postmod. 1

they list here of aid ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (7.280) the ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ sort of uhm typical ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ progress ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Emph.

Postmod. 1

Mod. I

Postmod. 2

³³ (7.281) his own as a young man ⏟⏟⏟⏟⏟⏟⏟⏟⏟ experiences ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Mod. I

Postmod. 1

the in the tribal wars between the Kikuyus and the Maasais ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Postmod. 2

The conclusion must be that with regard to these complex pre- and postmodified types, the (highly tentative) complexity ranking to be derived is ICE-GB > ICE-EA > ICE-SING, even though the low frequencies certainly do not allow far-reaching conclusions.

33 The AP typical in (7.280) is viewed as a case of Mod. I because it relates the referent of progress to a pre-established type. As both the predicative alternation (the sort of progress is typical) and modification by very are possible, however, it may also be viewed as Mod. II (cf. 5.1.3.2).

7.3 Testing hypotheses about simplification

261

7.3.1.2.4 Levels of NP embedding in S1A (Hypothesis 5d) The distribution of NPs embedded into other NPs (cf. 5.3.4.2.2) in the sample at hand is shown in table 7.58 and visualised in figure 7.38. Unsurprisingly, NPs in S1A are generally marked by “shallow” levels of embedding. The vast majority of NPs occur at the clause level. With regard to NPs at level one, which function as modifiers of an NP at level 0 (cf. (7.282)—(7.285)), figure 7.38 reveals an important tendency with respect to ICE-EA.

Table 7.58. Number of NPs embedded into other NPs per variety (cf. 5.3.4.2.2) in ICE-GB-S1A, ICE-SING-S1A and ICE-EA-S1A Variety ICE-GB Level of embedding

N

%

N

ICE-EA %

N

All

0.07 1 0.00 0 0.00 0 1 0.22 3 0.00 0 0.07 1 4 2.75 37 2.36 28 2.15 29 94 25.45 342 23.97 285 18.73 253 880 71.51 961 73.68 876 79.05 1,068 2,905 100.00 1,344 100.00 1,189 100.00 1,351 3,884

100

Level 4 Level 3 Level 2 Level 1 Level 0 All

%

ICE-SING

80

ICE-SING & ICE-EA ICE-SING and ICE-EA ICE-GB & ICE-SING > ICE-EA (ICE-EA)

Hierarchical complexity levels of whole NPs Complexity of subjects Hypothesis 5 Numbers of pre- and postmodifiers Levels of NP embedding (Results from typological analyses)

ICE-GB > ICE-SING > ICE-EA ICE-GB & ICE-SING > ICE-EA ICE-GB >? ICE-EA > ICE-SING ICE-GB & ICE-SING > ICE-EA ICE-GB > ICE-SING & ICE-EA

Simplification in…

ICE-SING and ICE-EA ICE-EA (ICE-SING and ICE-EA) ICE-EA ICE-SING and ICE-EA

Hypothesis 4, to begin with, predicts differences in terms of the length of whole NPs between ICE-GB and the new varieties at hand (cf. Hypothesis 4a). As demonstrated above (cf. 7.3.1.1.1), both ICE-SING and ICE-EA have significantly shorter NPs than the native variety ICE-GB, with ICE-SING exhibiting the lowest values. This fits in well with Hypothesis 4. Another result, the extreme shortness of Singaporean NPs, however, can not be unanimously attributed to simplification but must be simultaneously seen as an effect of contact influence; with postmodifiers naturally being the main factor in rendering an NP long, ICE-EA must inevitably outdo ICE-SING in terms of overall NP length. This shows that in some cases, contact influence and simplification are impossible to tease apart (cf. 7.3.1.1.1). The test of Hypothesis 4b, in turn, showed that ICE-EA relies on somewhat shorter subjects than the other varieties, which, although pointing in the direction of simplification, does not allow convincing conclusions to be drawn due to the lack of statistical significance. Hypothesis 5a predicts complexity differences between the NPs in ICE-GB and the non-native varieties. The analysis of NPs in the sample yielded marked,

7.3 Testing hypotheses about simplification

265

statistically significant differences between the varieties at hand in terms of the simplest NP pattern involving no modifier at all (cf. 7.3.1.2.1). Unmodified NPs are least frequent in ICE-GB, most frequent in ICE-EA, while ICE-SING neatly covers the middle ground. The remaining patterns (2NP )–(7NP ), conversely, decrease in their cumulative frequency from ICE-GB via ICE-SING to a minimum in ICE-EA. This result points towards a plausible complexity ranking, with the native variety most complex and the least advanced variety least complex, which resonates well with Hypothesis 5a (cf. 7.3.1.2.1). Hypothesis 5b — The findings with regard to the distribution of complexity types across subjects and non-subjects in the varieties at hand fit in with the results in the previous paragraph. ICE-EA has a particularly strong preference for simple subjects, while ICE-GB and ICE-SING are slightly more tolerant towards modified, complex NPs in this function (cf. 7.3.1.2.2). Under the assumption that the principle of end-weight benefits speakers and listeners functionally in terms of both parsing and producing (cf. 5.2.1.1), this accords with the idea of a particular preference for functional simplicity in Kenyan English. Hypothesis 5c predicts complexity differences in NPs with regard to the numbers of pre- and postmodifiers used. The analysis above (cf. 7.3.1.2.3), however, has yielded very fine-grained tendencies only. If the small and insignificant frequency differences in terms of the complex modification types are accepted as interpretable in any way, it would seem that ICE-GB is most complex, ICE-EA covers the middle ground and ICE-SING is least complex, because in the Singaporean part of the sample, the complex types involving multiple pre- and/or postmodification do not occur at all. This ranking, however, must be viewed as tentative. Hypothesis 5d, which predicts differences in the depth of NP embedding in the varieties at hand, was confirmed in terms of a significantly lower frequency of NPs embedded at level 1 in ICE-EA (cf. 7.3.1.2.4). Singaporean English, in contrast, clearly patterns with British English in this respect. Taken together, the analysis of S1A has yielded stable evidence in support of the hypotheses pertaining to simplification. NPs in the two new varieties at hand have been found to be simpler than in British English in terms of both length and hierarchical structure. Table 7.59 also shows that simplification is a highly pervasive phenomenon; significant statistical effects pointing towards simplificatory tendencies in comparison to British English can be found in both of the new varieties, with the exception of the levels of NP embedding and the distribution of complexity levels

266

7 Studying NP modification in Singaporean English and Kenyan English

across syntactic functions, both of which turn out to be lower in ICE-EA only. Language users in both varieties, thus, adopt simpler NP linguemes than speakers of the input variety British English. It is also worth considering the patterns of variation of the individual varieties to be seen in table 7.59 ignoring significances: In seven out of the seven cases, ICE-EA either turns out to be least complex or patterns with the less complex of the remaining two varieties. It never reaches the top complexity level, even though it once covers the middle ground between ICEGB and ICE-SING. ICE-SING, in contrast, patterns variably with ICE-EA (twice), with ICE-GB (three times) and ends up halfway between the other two varieties (once). A single time, it is outdone by ICE-EA in terms of complexity. These observations corroborate the impression that the tendency for simplification is more pronounced in Kenyan English, while ICE-SING is more variable and tends towards the complex pole. In spite of being a weak and tenuous effect, this fits in with the evolutionary stages of the varieties (cf. 4.3).

7.3.2 The complexity of NPs in student essays () In the following, the complexity of NPs in W1A will first be analysed with regard to NP lengths (cf. 7.3.2.1). Subsequently, I will explore differences in the hierarchical complexity of NPs (cf. 7.3.2.2) and present selected results on simplification which have emerged in the analysis of language contact phenomena (cf. 7.3.2.3). The final section will summarise the main points of the chapter.

7.3.2.1 Length differences between NPs in W1A (Hypothesis 4) Once more, an analysis of the overall length differences between NPs in W1A will be followed by an inquiry into the correlation of NP length and syntactic position.

7.3.2.1.1 Length differences between whole NPs in W1A (Hypothesis 4a) Table 7.60 shows the mean values, the SDs and the variation coefficients of the length measures in question. Figure 7.39 is a boxplot visualisation of the length values, once more with the left-hand pane (figure 7.39a) showing a version with all outliers, and the right-hand pane (figure 7.39b), a zoom version excluding the outliers. The mean values reveal a familiar pattern of variation. The -signs plotted into figure 7.39b show a clear ranking in terms of average NP length from British English via the phase-4/5 variety, Singaporean English, to Kenyan English. The average NP consists of 4.94 words in ICE-GB, 4.63 words in ICE-SING and a mere 4.19

+

7.3 Testing hypotheses about simplification

267

Table 7.60. The lengths of NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A ICE-GB ICE-SING ICE-EA Number of NPs Arithmetic mean SD Variation coefficient

1544 4.94 5.16 1.04

1430 4.63 4.42 0.95

1615 4.19 4.80 1.15

Kruskal-Wallis rank sum test H = 47.7421, df = 2, p = 4.295 · 10-¹¹***

(b) Without outliers (zoom view)

12

(a) Including all values

40

● ●

10

+

8

+

● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

6

● ●

+ 4

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Length (in words)

30 20



+

10

Length (in words)

● ●

0

2

+

+

ICE−GB ICE−SING

ICE−EA

ICE−GB ICE−SING

Variety

ICE−EA

Variety

Fig. 7.39. The lengths of NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A. Horizontal dashed line: arithmetic mean of all length values; -signs: arithmetic means for individual varieties (cf. 6.1.3)

+

words in ICE-EA. This constitutes a highly significant difference according to a Kruskal-Wallis test summarised in table 7.60, with 𝑝 far below 0.001∗∗∗ . The exact difference between varieties can be determined by means of a pairwise Wilcoxon test; its results are listed in table 7.61. What emerges from them is that with respect to mean NP lengths, it is ICE-EA that stands out in particular; ICE-SING is different from ICE-EA but not from ICE-GB; ICE-EA, in contrast, differs significantly from

268

7 Studying NP modification in Singaporean English and Kenyan English

Table 7.61. Pairwise Wilcoxon rank sum test of the mean lengths of NPs in ICE-GB-W1A, ICESING-W1A and ICE-EA-W1A ICE-GB

ICE-SING

ICE-SING p = 0.9 — ICE-EA p = 4.2 · 10-⁹*** p = 1.00 · 10-⁸***

both varieties at hand, both from the native input variety and from the more advanced Singaporean English.

7.3.2.1.2 Length differences between whole NPs and syntactic position in W1A (Hypothesis 4b) The question arises to what extent the varieties at hand make use of the parsing (and possibly producing, cf. 5.2.1.1) advantages which are offered by short (and simple) subjects in accordance with Hawkins’s (1994) EIC principle (cf. 5.2.1.1). Table 7.62 and figure 7.40 show the lengths of subjects and non-subjects in W1A; table 7.63 details the numbers of subjects and non-subjects analysed. With regard to NPs in subject position, ICE-SING clearly patterns with ICE-GB, with its NPs only slightly longer than in the native variety (4.10 words — 4.24 words). Long subjects shown in (7.289)–(7.290), thus, are not too uncommon in these varieties. (7.289) Non-agriculturists who owned land increased from 18% in 1906–07 to 31% in 1929–30. (7.290) Their historical background of being “conquered people” made them regard the government reforms as interfering in their lives.

Table 7.62. The lengths of whole NPs by syntactic position and variety in ICE-GB-W1A, ICE-SINGW1A and ICE-EA-W1A Subject

Non-subject

Mean Var. coeff. Mean Var. coeff. ICE-GB 4.10 0.975,802,9 5.28 1.046,343 ICE-SING 4.24 0.947,739,3 4.82 0.952,589,3 ICE-EA 3.01 1.182,320,3 4.59 1.109,163,9

4.5

5.0

269

4.0

Overall mean

3.5

ICE−GB ICE−SING ICE−EA

3.0

Mean of overall NP length

7.3 Testing hypotheses about simplification

subject

nonsubject

Fig. 7.40. The correlation of syntactic position and NP length in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A

Table 7.63. The numbers of subjects and non-subjects in ICE-GB-W1A, ICE-SING-W1A and ICEEA-W1A ICE-GB ICE-SING ICE-EA Sum subject 439 463 non-subject 1,105 967 Sum 1,544 1,430

409 1,311 1,206 3,278 1,615 4,589

In ICE-EA, in contrast, subjects comprise a mere 3.01 words on average, which means that short subjects exemplified in (7.291)–(7.293) are particularly frequent. (7.291) Language is not realised in the abstract. (7.292) In such cases the contacts is too specialised […] (7.293) To him, it is in this sense that all human societies have a “culture” and a “civilization”. Like in the parallel analysis of S1A (cf. 7.3.1.1.2), it turns out that ICE-EA exhibits the greatest length difference between subjects and non-subjects, whereas, in ICE-GB and ICE-SING, the lengths are largely on a par. Again, the slope of the line representing ICE-EA in figure 7.40 seems steeper than for ICE-SING, which is indicative of a higher degree of difference in the less advanced variety. The question, however, is whether this visual impression can be corroborated on a statistical basis. With the variances clearly inhomogeneous (Length by Variety:

270

7 Studying NP modification in Singaporean English and Kenyan English

Bartlett’s 𝐾-squared = 35.3474, df = 2, 𝑝-value < 2.111e-08; Length by Function: Bartlett’s 𝐾-squared = 122.9794, df = 1, 𝑝-value < 2.2e-16), a default ANOVA is precluded; instead, a robust ANOVA will be conducted (cf. Wilcox 2003). Table 7.64 shows the results of the procedure. Both the effects of Function and Variety are Table 7.64. The impact of Function and Variety on the mean lengths of the whole premodifying string according to a robust two-way ANOVA (cf. Wilcox 2003: 347) Variable(s)

Q

p-value

Function 48.786,82 0.001 *** Variety 67.789,57 0.001 *** Register × Variety 86.490,07 0.013 **

significant. However, it is particularly interesting that the interaction Function × Variety, too, proves significant, which confirms the impression gained from figure 7.40 that there is a difference in slope between the varieties at hand. An important follow-up question, however, is which varieties interact with each other. For this purpose, Wilcox’s (2003) post-hoc test for robust ANOVAS can be used; its results are shown in table 7.65. As can be seen, the main effect of Function proves significant (as is obvious from the last column labelled significance𝑝 adj. ), which confirms the result of the robust ANOVA³⁵. The effect of Variety, too, is borne out, with the post-hoc test, however, showing that only the differences between ICE-GB and ICE-EA and between the two new varieties are significant; ICE-GB and ICE-SING, in contrast, are too similar to yield a significant difference (cf. figure 7.40). The key point, however, is the interaction between variables (i.e. the last three lines of table 7.65). It is interesting that while insignificant for the pairs ICE-GB and ICE-SING as well as ICE-GB and ICE-EA, the interaction proves significant for the difference between ICE-SING and ICE-EA. This allows the conclusion that the differences in slope found for Singaporean English and Kenyan English in figure 7.40, in fact, are significant. Syntactic position has differential effects on NP length the two new varieties at hand. 35 As pointed out above (cf. 6.1.3), the two criteria used to assess a significance in such post-hoc tests are whether the “test” value exceeds the “crit” value, and, furthermore, whether the confidence intervals labelled as “ci.lower” and “ci.upper” in table 7.65 cross zero (a given difference is significant if the intervals do not cross zero). In the interpretation of the main effect of Function in table 7.65, the former criterion fails (the test values are below the crit value by way of being negative), but the confidence intervals do not cross zero. For this case, A. P. Field, Miles, and Z. Field (2012: 539) suggest that one may ignore the minus value of the negative test value and thus interpret the configuration as significant.

Function × Variety

Function Variety

subject vs. non-subject ICE-GB vs. ICE-SING ICE-GB vs. ICE-EA ICE-SING vs. ICE-EA ICE-GB ICE-SING ICE-GB ICE-EA ICE-SING ICE-EA

1.961,236 2.39 2.39 2.39 2.39 2.39 2.39

6.242,895,9 7.301,198,8 -1.268,029 1.459,925 2.948,487

crit

test -6.984,756 -0.762,533,8 0.323,482,3 0.284,287,6 0.257,205,6 0.249,614,7 0.284,287,6 0.257,205,6 0.249,614,7

se 1,866.612 1,271.367 1,236.635 1,270.542 1,271.367 1,236.635 1,270.542

df

1.605,707,6 1.822,486,5 -0.360,484,9 0.375,500,9 0.735,985,8

-2.259,445 -0.216,778,9

psihat

0.990,986,3 1.225,907,4 -1.039,932,2 -0.239,220,5 0.139,406,7

-2.893,87 -0.896,226,2

ci.lower 0.462,668,4 2.220,428,9 2.419,065,6 0.318,962,4 0.990,222,2 1.332,564,9

-1.625,02

ci.upper

3.953,726 . 10-¹² 0.445,882,9 5.892,435 . 10-¹⁰ 5.013,767 . 10-¹³ 0.205,019,71 0.144,564,656 0.003,251,707

p.value * n.s. * * n.s. n.s. *

Table 7.65. Post-hoc test (cf. Wilcox 2003: 436) of the interaction of the length of the whole premodifying string with Function and Variety (cf. 6.1.3); “𝑝 (not) adj.” ‘𝑝 (not) adjusted for multiple testing’ (cf. 6.1.3)

7.3 Testing hypotheses about simplification

271

272

7 Studying NP modification in Singaporean English and Kenyan English

This buttresses the prediction that Kenyan English will rely on short subjects in a particular way. In Hawkins’s (1994) terms, thus, the implication is that, in W1A, in line with the hypotheses under discussion, speakers of Kenyan English capitalise on the functional advantages of short subjects to a stronger degree than speakers of Singaporean English, which can be attributed to the general simplification tendencies in these varieties noted elsewhere (cf. 2.2.2). At the same time, this result also allows for the additional conclusion that in Kenyan English, which is primarily affected by head-initial languages, the EIC principle plays a bigger role, as it is only in this set of languages where, according to Hawkins (1994), the ordering “short before long” offers advantages in parsing (cf. 5.2.1.1); this may carry over to Kenyan English. The weak effect of syntactic position on NP length in Singaporean English, in contrast, may be due to a lower importance of the EIC principle, since in head-final languages, the ordering of short before long is not assumed to be advantageous by Hawkins (1994) (cf. 5.2.1.1). It may be that the difference between Singaporean English and Kenyan English is due to both Kenyans capitalising on the advantages of the EIC principle due to a stronger degree of simplification, and likewise due to the effect of contact with primarily head-initial or head-final languages. On the basis of the present corpus data, however, it can hardly be decided which of these factors is stronger.

7.3.2.2 Hierarchical complexity in W1A (Hypothesis 5) In this section, the hierarchical structure of NPs will be analysed on the basis of the following set of variables: the frequencies of the complexity levels of NPs (cf. Hypothesis 5a; 7.3.2.2.1); the simplicity of subjects as against nonsubjects (cf. Hypothesis 5b; 7.3.2.2.2); the numbers of pre- and postmodifiers per NP (cf. Hypothesis 5c; 7.3.2.2.3); and the levels of NP embedding (cf. Hypothesis 5d; 7.3.2.2.4).

7.3.2.2.1 Hierarchical complexity levels in W1A (Hypothesis 5a) Table 7.66 and figure 7.41 show the frequencies of the complexity levels by variety. The HCFA results are indicative of a range of marked differences, especially with regard to the least and the most complex NP types. It is striking, for one thing, that type (1NP ), i.e. unmodified NPs, should be significantly underrepresented in ICE-GB with a proportion of 24.55%, while ICE-SING has a highly similar share of 25.25% (which, according to the HCFA procedure, is likewise underrepresented in a “marginally significant” way). In ICE-EA, in contrast, such unmodified NPs have a much higher incidence of 35.98%, which constitutes a highly significant difference according to the HCFA test. Examples can be seen in (7.294)–(7.295).

7.3 Testing hypotheses about simplification

273

Table 7.66. The complexity level of whole NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A (cf. 5.3.4.2 for the definition of complexity levels) Variety ICE-GB Complexity level of whole NP

100

level (8NP ) level (7NP ) level (6NP ) level (5NP ) level (4NP ) level (3NP ) level (2NP ) level (1NP ) All

%

19.51 279 4.69 67 3.78 54 24.69 353 0.00 0 4.27 61 17.83 255 25.25 361 100.00 1,430

%

N

All

11.83 191 721 5.20 84 254 4.03 65 189 24.09 389 1,115 0.00 0 1 2.11 34 148 16.78 271 840 35.98 581 1,321 100.00 1,615 4,589

40

60

***

**

80

%

ICE-SING

ICE−GB

ICE−SING

ICE−EA

Fig. 7.41. The complexity level of whole NPs in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A (cf. 5.3.4.2 for the definition of complexity levels; cf. table A.12, p. 326, for exact HCFA results)

(7.294) its vocabulary (7.295) redundancy With regard to such unmodified NPs, thus, ICE-SING patterns with ICE-GB, while ICE-EA stands out as clearly different. This is partly in accordance with Hypothesis 5a: Although there is no neat complexity ranking as in the corresponding analysis

274

7 Studying NP modification in Singaporean English and Kenyan English

for S1A (cf. figure 7.35, p. 254), the least advanced variety turns out to have the largest proportion of utterly simple NPs. The second set of differences unearthed by the HCFA procedure, in turn, refers to the most complex type (8NP ), i.e. NPs which are both pre- and postmodified. They are illustrated in (7.296)–(7.301). (7.296) a clear distinction between its understanding/generative capabilities which rely strictly on linguistic knowledge and those which rely on domain knowledge (7.297) a meaningful socio-linguistic conceptualisation of Singapore English

(7.298) the most important factor preventing them from seeking employment

(7.299) other modern countries where social status is achieved rather than ascribed (7.300) the social and cultural dimension of progress (7.301) an amusement aspect, which in most cases, is used in artistic work such as novels Type (8NP ), too, discriminates well between the varieties at hand. The most complex NPs in the sample are strongly overrepresented in ICE-SING, whereas they are highly significantly underrepresented in ICE-EA. NPs such as (7.300)–(7.301), thus, are not as likely to occur in Kenyan student essays as in British or Singaporean ones. The proportion for ICE-GB, which covers the middle road between ICE-SING and ICE-EA, is unremarkable from a statistical point of view. As regards the most complex NP type, thus, Hypothesis 5a is also partially confirmed, as it is plausible for the least advanced variety to have the lowest proportion of highly complex NPs. At the same time, it comes as a surprise that Singaporean English should turn out to be more complex than the input variety British English (cf. its high proportion of complex, pre- and postmodified NPs). In order to reconcile this with the predictions made by Hypothesis 5a, one may assume that due to the association of English with the sphere of education, Singaporean writers are somewhat over-zealous in their aim to write formal academic prose and thus are even more prone than British students to encode information compactly in the form of NPs. Being less inhibited by SLA-induced limitations than Kenyan writers, they are led to use structures which are even more complex than those of British student writers. With regard to the remaining levels of complexity, no statistically significant frequency differences are uncovered by the HCFA test. The only palpable effect concerns a gentle decrease of the types (2NP ) and (3NP ) in ICE-EA (16.78% and

7.3 Testing hypotheses about simplification

275

2.11%) as against ICE-GB and ICE-SING, which may be related to the typological tendencies against premodification found above (cf. 7.2.2.1.1). Examples of types (2NP ), premodification by simple phrases, and (3NP ), premodification by phrases modified themselves or coordinated, can be seen in (7.302)–(7.306). (7.302) (7.303) (7.304) (7.305) (7.306)

international recognition dynamic social planning the slightly smaller difference the daily newspaper editors a tertiary or upper secondary education

Type (4NP ), i.e. NPs premodified by a clausal premodifier, are exceptionally rare, with one single occurrence hailing from ICE-GB (cf. (7.307)); quite characteristically, the lexical phrase insert before current entry is enclosed in quotation marks, which are used frequently to flag the ad-hoc character of such constructions (cf. 5.1.4.2). (7.307) an entirely separate “insert before current entry” function

The new varieties do not make use of such constructions, though. In particular, ICE-SING, where typological influence may be expected to favour prenominal clauses (cf. 4.1.2.2), clearly does not rely on lexical phrases. This is somewhat surprising in view of the results of Mazaud’s (2004) study of newspaper language, where Singaporean writers showed a strong penchant for such constructions (cf. 5.2.2.2). It seems likely that register effects play a role here: Student essays represent a plain, academic text type which, despite its strong overall reliance on nominal constructions, allows playful or experimental use of language only to a very limited extent, while newspaper language is known to push linguistic boundaries regularly for creative and innovative purposes (cf. Mazaud 2004: 3). The remaining complexity types, interestingly, have highly similar probabilities of occurrence in the three samples at hand. Level (5NP ), which represents NPs postmodified by simple phrasal postmodifiers, has stable frequencies of around 24% in all three varieties (cf. (7.308)–(7.309)). (7.308) jobs outside their homes (7.309) a lot of noise in the room Level (6NP ), i.e. NPs postmodified by non-finite clauses, which are illustrated in (7.310)–(7.311), have similar shares in the corpora at hand of 4.53%, 3.78% and

276

7 Studying NP modification in Singaporean English and Kenyan English

4.03% respectively. NPs at level (7NP ), in turn, likewise have unremarkable proportions of 6.67%, 4.69% and 5.20%. They are exemplified in (7.312)–(7.313). (7.310) the factors accounting for the differential ethnic population growth in Peninsular Malaysia since independence (7.311) the need to secure markets previously acquired through exports and to avoid protectionism (7.312) people living in the Island of Pate (7.313) The Hungarians, burning with racism Taken together, thus, it is striking that it is the NP structures making up the endpoints of the complexity scale which show the strongest frequency differences, while the intermediate cases are almost equally frequent across corpora. Speakers of New Englishes seem to be particularly sensitive to such extremely complex or extremely simple NPs. The overall complexity ranking to be derived from this distribution, thus, is ICE-SING > ICE-GB > ICE-EA.

7.3.2.2.2 Hierarchical complexity levels and syntactic position in W1A (Hypothesis 5b) Table 7.67 and figure 7.42 show the correlation of NP complexity and syntactic position in ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A. The result is very similar to

100 80 60

100

***

0

0

0

Nonsubject

level (8NP) level (7NP) level (6NP) level (5NP) level (4NP) level (3NP) level (2NP) level (1NP)

ICE-EA ICE-GB=ICE-SING=ICE-EA ICE-SING >³⁸ICE-GB > ICE-EA ICE-GB > ICE-SING & ICE-EA

Hypothesis 5a

Simplification in…

ICE-EA ICE-EA — ICE-EA ICE-SING and ICE-EA

37 The ranking of ICE-SING as higher in complexity levels than ICE-GB is based on the significantly higher frequency of the most complex type (8NP ); otherwise, the distributions in ICE-GB and ICE-SING are highly similar (cf. figure 7.41; p. 273).

7.3 Testing hypotheses about simplification

285

Hypothesis 4 predicts a complexity ranking of British English, Singaporean English and Kenyan English in terms of length differences between NPs (cf. Hypothesis 4a). The analysis conducted above (cf. 7.3.2.1.1) has shown that in terms of NP lengths, ICE-SING patterns with ICE-GB. NPs in ICE-EA, in contrast, are significantly shorter than in either of the two other varieties at hand. This fits in with the predictions of the Dynamic Model in that with regard to complexity, Singaporean English, which represents a highly advanced variety, approximates British English, while ICE-EA, being in phase 3 only, turns out to be considerably less complex (cf. 7.3.2.1.1)³⁹ . As regards Hypothesis 4b, which predicts variety-specific degrees of preference for short subjects, it turned out that length differences between subjects and non-subjects in ICE-EA were significantly more pronounced than in ICE-SING, which allows the conclusion that, due to simplification, speakers of Kenyan English are more reliant on short subjects than speakers of Singaporean English. Hypothesis 5a predicts lower levels of NP complexity in both ICE-SING and ICEEA. With regard to this criterion, too, marked differences between the varieties could be made out above (cf. 7.3.2.2.1), with the most notable results pertaining to the frequencies of the simplest level (1NP ) and the most complex level (8NP ). ICEEA clearly turns out to be least complex, with a strong overrepresentation of (1NP ) and a palpable underrepresentation of (8NP ). This is, once more, fully in line with the predictions of Hypothesis 5a: The least advanced variety exhibits unmistakable preferences for very simple NPs, while showing an aversion to highly complex structures. The relationship between ICE-GB and ICE-SING, in turn, comes as a surprise, as with respect to (8NP ), ICE-SING turns out to be slightly more complex than ICE-GB, because the proportion of this most intricate NP type is significantly higher in ICE-SING than in the other varieties (cf. figure 7.41, p. 273). It has been suggested that this may be due to Singaporean students seeking to render their English particularly formal in an educational context, while the register aimed at by British students is somewhat more relaxed (cf. 7.3.2.2.1 ). On the whole, thus, Hypothesis 5a has been largely confirmed: Differences in NP complexity levels in

38 Again, the ranking of ICE-SING as more complex than ICE-GB is not unanimous, as the frequencies are very close. The significantly low proportion of level-0 NPs, however, in my view, warrants a classification of ICE-SING as slightly more complex than ICE-GB. 39 It should be noted that the problem concerning the intersection of typological and complexity differences noted with regard to length differences in S1A (cf. 7.3.1.1.1) does not play a role here, since in W1A, there is hardly any difference in the frequencies of pre- and postmodified NPs between the varieties under study (cf. 7.2.2.1.1).

286

7 Studying NP modification in Singaporean English and Kenyan English

ICE-GB-W1A, ICE-SING-W1A and ICE-EA-W1A can be related plausibly to the status of the varieties under study (cf. 7.3.2.2.1). Hypothesis 5b — The correlation between syntactic position and complexity levels confirms a result known from S1A: ICE-EA has a much stronger tendency to realise subjects by unmodified, simple NPs belonging to (1NP ) than do the other varieties. This likewise fits the evolutionary statuses assumed for the varieties at hand, since Singaporean English patterns with ICE-GB in allowing more complex NPs in subject position, which, of course, comes with a functional cost (cf. 5.2.1.1). Hypothesis 5c predicts lower complexity levels in the new varieties with regard to the number of pre- and postmodifiers used in NPs. However, no meaningful complexity differences could be made out between the varieties with regard to this criterion in any of the registers at hand (cf. 7.3.2.2.3). Hypothesis 5d predicts differences in terms of the numbers of NPs embedded into other NPs as modifiers. The analyses, first and foremost, confirm ICE-EA to be lower in complexity owing to a reduced frequency of NPs embedded at level 2. The comparison of ICE-GB and ICE-SING, in turn, once more shows that ICE-SING outdoes ICE-GB by a narrow margin, as Singaporean level-0 NPs are underrepresented in comparison to higher levels of embedding. The conclusions, thus, are highly similar to those drawn for Hypothesis 5a: In line with the Dynamic Model, ICE-EA proves least complex. Once again, ICE-SING appears to be slightly more on the complex side in comparison to ICE-GB as the proportion of level-0 NPs is particularly low, which, again, may be due to a degree of “overcorrection” by Singaporean students, who aim at a particularly high level of complexity due to English being associated with the sphere of education (cf. 7.3.2.2.4). A range of results from typological analyses have been re-interpreted in terms of simplicity as they attest to a lower degree of complexity in both ICE-SING and ICE-EA in comparison to ICE-GB, with regard to the lengths of Mod. II, the modification rates of Mod. II or the structural variability of the NP as measured by co-occurrence patterns of modifier types (cf. 7.3.2.3). In sum, in W1A, too, Hypothesis 4 and Hypothesis 5 pertaining to SLA-induced simplification have been confirmed. However, in contrast to S1A (cf. 7.3.1.4), in five out of seven distributions, only ICE-EA proves markedly simpler than the other varieties. It is only the results from typological analyses that show both varieties simultaneously to be simpler than ICE-GB. Quite clearly, in W1A, the adoption of simpler NP linguemes is primarily a preference of language users in Kenya;

287

7.3 Testing hypotheses about simplification

student writers from Singapore, in contrast, are comparable to British writers in terms of NP complexity. Table 7.70 also confirms the impression of ICE-EA being more prone to simplification than ICE-SING. In all six of the six interpretable cases, ICE-EA is either least complex or patterns with the less complex variety. ICE-SING, once more, patterns variably with ICE-GB, ICE-EA or, in fact, twice exceeds the complexity of ICE-GB. These patterns confirm the evolutionary statuses assumed for the two varieties; in contrast to speakers of Kenyan English, Singaporean English writers, quite obviously, are uninhibited by SLA-induced simplification and, in formal texts, even aim at a higher level of NP complexity than native speakers.

7.3.3 Simplification in S1A and W1A — a comparison The discussion in the following will be based on the results on simplification gained in the separate analyses of S1A and W1A in 7.3.1–7.3.2 , which are summarised in table 7.71 (cf. also Footnote 29; p. 245). Table 7.71. The varieties with the most obvious effects of simplification by register/modality. ‘effect in one variety’ ; ‘effect in both varieties’ Criterion

Length differences between whole NPs Hypothesis 4 Lengths of subjects

Hierarchical complexity levels of whole NPs

Complexity of subjects

Hypothesis 5

Numbers of pre- and postmodifiers

Levels of NP embedding

(Results from typological analyses)

S1A

W1A

Contact/simpl. in…

Contact/simpl. in…

ICE-GB > ICE-SING & ICE-EA ICE-GB & ICE-SING > ICE-EA ICE-SING and ICE-EA

ICE-EA

ICE-GB & ICE-SING > ICE-EA ICE-GB & ICE-SING > ICE-EA (ICE-EA)

ICE-EA

ICE-GB > ICE-SING > ICE-EA ICE-SING >? ICE-GB > ICE-EA ICE-SING and ICE-EA

ICE-EA

ICE-GB & ICE-SING > ICE-EA ICE-GB & ICE-SING > ICE-EA ICE-EA

ICE-EA

ICE-GB >? ICE-EA > ICE-SING ICE-GB=ICE-SING=ICE-EA (ICE-SING and ICE-EA)



ICE-GB & ICE-SING > ICE-EA ICE-SING >? ICE-GB > ICE-EA ICE-EA

ICE-EA

ICE-GB > ICE-SING & ICE-EA ICE-GB > ICE-SING & ICE-EA ICE-SING and ICE-EA

ICE-SING and ICE-EA

The table presents a very clear variety-specific pattern of variation. Both varieties exhibit simplification in S1A, with four of the seven cases based on

288

7 Studying NP modification in Singaporean English and Kenyan English

distributions in which both Singaporean English and Kenyan English are affected (despite the very slight predominance of Kenyan English noted above in 7.2.2.1.1). In the written register W1A, in contrast, ICE-EA clearly takes precedence, with five out of the seven distributions referring to Kenyan English only, and Singaporean English appearing only once. While speakers of Singaporean English, thus, largely shed their preference for adopting simpler NPs than those in the native variety when writing formal scholarly texts (in fact, they even occasionally use more complex linguemes than in British English; cf. 7.3.2.4), the tendency remains in place in Kenyan English. This comparison of simplification tendencies also allows conclusions about the degree of register differentiation of the two varieties under study (cf. 2.1.3), or, more precisely, with regard to the question of excessive formality of registers in new varieties (cf. 2.1.3). However, there were only few pieces of evidence of informal registers having a particular tinge of formality. As shown in figure 7.7 (p. 187), ICE-EA has a strong preference for APs in Mod. III position, which can be viewed as an effect of a higher degree of formality (cf. 7.2.1.3). Apart from this, however, NP syntax is not indicative of excessive formality of informal texts in the two varieties.

7.3.4 Discussion and interpretation: Simplification in NPs in Singaporean English and Kenyan English In the previous sections, NP modification patterns in Singaporean English and Kenyan English have been analysed for simplification tendencies. A fair share of cases was found where simplification can be claimed, both on the basis of length differences and on the basis of the hierarchical complexity of linguemes (cf. 7.3.1.4; 7.3.2.4). The prediction was for both new varieties to be simpler than ICE-GB, but for ICE-EA to turn out to be even simpler than ICE-SING, due to its lower evolutionary stage. When viewed across corpora, this predicted pattern of complexity is borne out for ICE-EA: In table 7.71 (p. 287), ICE-EA can be seen to partake of eleven out of the eleven plausible frequency distributions (including one non-significant one) in both of the registers and ICE-SING appearing in five out of eleven cases only. It seems that speakers of varieties in low evolutionary stages, in fact, systematically adopt simpler NP linguemes than do speakers of more highly advanced varieties. In addition, Kenyan English, for the most part, represents the least complex variety, or patterns with the less complex one; it covers the middle ground only once and never goes beyond the input variety British English (cf. table 7.71). Singaporean English likewise conforms to expectations by patterning variably with either the native variety or Kenyan English, which confirms

7.3 Testing hypotheses about simplification

289

its highly advanced status as a phase-4 variety. However, it also goes twice beyond the complexity of British English, which was attributed to hypercorrection (cf. 7.3.2.4). From the point of view of Croft’s (2000) theory of language change, the simplification tendencies, too, must be ascribed to the process of propagation. It is due to speakers adopting (or having adopted) simpler linguemes available in the respective feature pool. The comparison of the formal and the informal registers above, however, with Singaporean English using simplification strategies in S1A only, while Kenyan English exhibits simplified NPs in both registers (cf. 7.3.3), allows further conclusions. In the way of explanation, it is, first and foremost, worth mentioning that the Singaporean pattern is more strongly in line with standard English register norms than the Kenyan one, as informal conversation is, for clear functional reasons, dominated by simple NPs, while academic language, in native discourse, is rife with highly complex NPs, likewise with a clear functional logic (cf. 5.2.1.2). Against this background, the question arises whether the simple reason for the register-specific pattern of variation may be that Kenyans aim at a novel norm of academic texts which is marked by simpler NPs than the native British norm. This answer, however, seems hardly convincing. There is no doubt that Kenyans have access to academic writing which follows the standard English norm, and I would argue that it is unlikely for speakers of a phase-3 variety, which is clearly still in the process of developing an internal norm, to abolish NP complexity as a property of formal texts, which is one of the hallmarks of this register. The interpretation offered in the following, thus, presupposes that Kenyans, in principle, aim at the same level of complexity as native speakers when producing formal academic language. In order to account for the pattern, it is worth remembering that the motivation of individual speakers for the adoption/propagation of simplified linguemes was put down to a set of three possible factors (cf. 3.2.4.3): (a) a functional desire for communicative economy at the level of the speech community due to restrictions imposed by the use of an L1; (b) socially-driven preferences for simple linguemes; (c) the involuntary use of simpler linguemes due to their cognitive entrenchment.

I would argue that the pattern of variation in Kenyan English, in conjunction with the knowledge about the evolutionary stage of the variety, allows the conclusion that speakers of Kenyan English adopt simpler linguemes in both S1A and W1A primarily due to the functional factor mentioned in (a). With English having the status of an L1 in Kenya, which, in everyday life, is eclipsed by local vernaculars and thus used considerably less frequently than in Singapore, speakers of Kenyan English feel restrictions typical of SLA in language production and resort to NPs

290

7 Studying NP modification in Singaporean English and Kenyan English

which offer a greater economy of production (and possibly processing; cf. 5.2.1.1). In Singaporean English, in contrast, which is more advanced, such restrictions are no longer in place to the same extent, as English is used widely, has spread down the social hierarchy to a greater extent than in Kenya and has even made inroads as an L1. Consequently, speakers are able to use complexity levels at will, which allows them to do full justice to the register-specific complexity of NPs in the academic register W1A. They even go beyond the British complexity levels in a number of cases, which has been ascribed to an effect of hyper-correction (cf. 7.3.2.2.1). The question arises, however, why it is that, despite this, Singaporeans should use NPs in S1A which are simpler than in the native variety, if, as stated above, there are indications that SLA-induced restrictions no longer work as a limiting factor in this variety, not even in the formal register, where excessively high levels of complexity are necessary. I would argue tentatively that Singaporeans’ adherence to simple NPs in S1A, in fact, may be due to the association which, during decades of propagation, simple linguemes have acquired for these speakers with the informal domain, as markers e.g. of casualness. This would mean that beyond the functional motivation for the adoption assumed for Kenyan English, Singaporeans might propagate simple linguemes intentionally, because they have a casual ring to them, which fits the informal domain, and, because they have turned into what Croft (2000: 92) has called a “three-cornered sign-in-a-community” (cf. 3.2.2.1). This interpretation entails that in varieties in earlier evolutionary stages which have not yet been fully subject to endonormative stabilisation, simplification in NPs as a strategy of language production is primarily driven by the functional need to employ simpler linguemes, which is not at the language users’ disposal. In advanced varieties, in contrast, simplification has gained social salience and can be used by speakers at will, depending on its adequacy in given domains or registers. This argumentation would also imply that it is only in Singaporean English that the long-term propagation of simplified linguemes since the inception of the variety has brought about changes in the linguistic norm. The functional restrictions in Kenyan English, in contrast, would seem to be more of a transient usage phenomenon which is due to today’s speakers merely adopting (rather than propagating, cf. 3.2.3.2) simpler linguemes, and which, rather than having an impact on the emergent norm, in the course of the future development of the variety, is likely to subside. The third motivation for the propagation of simple linguemes mentioned above (cf. (c)), viz. the quasi-automatic, involuntary use of simpler linguemes due to their high token frequency in the speech community, and, hence, their high degree of cognitive entrenchment, may likewise play a role in the registers and varieties where simplification has been found; to what extent, however, the use

7.4 Summary and discussion: Language contact and simplification

291

of simpler NPs has become automatised or is due to functional restrictions in the individual varieties is hard to pinpoint.

7.4 Summary and discussion: Language contact and simplification across varieties and registers Table 7.72 lists all results with regard to simplification and language contact in S1A and W1A, based on the separate analyses above (cf. 7.3.1). The results shown in the table will not be reiterated in full as they were spelt out and interpreted in detail earlier (cf. 7.2.1.3 for the results on language contact in S1A; cf. 7.2.2.3 for the results on language contact in W1A; cf. 7.2.3 and 7.2.4 for their comparison and interpretation; cf. 7.3.1.4 for the results on simplification in S1A; cf. 7.3.2.4 for the results on simplification in W1A; cf. 7.3.3 and 7.3.4 for their comparison and interpretation). In the bigger picture, the results indicate that the modification patterns of NPs in Singaporean English and Kenyan English are influenced by both a variety-specific factor, i.e. language contact, and by simplification, which plays a universal role across varieties due to their specific contexts of use and emergence as L2s (cf. 2.2.1.2). The comparative methodology adopted, which takes into account both the input variety and two new varieties with divergent typological backgrounds, created an opportunity to disentangle these effects⁴⁰. It is worth noting, furthermore, that the notion of simplification developed in the present study is novel in being geared towards hierarchical structure, while the default notion of complexity in the field focuses on morphosyntactic structure or the overall absence/presence of linguistic properties (cf. 2.2.2.2). The exact relationship between variety-specific phenomena of language contact and universal simplification tendencies can be pinpointed even more clearly when considering the overall pattern of variation in the varieties across registers. In what follows, the core implications of this pattern will be reiterated briefly (cf. 7.2.4 and 7.3.4 for details). With regard to language contact, no variety-specific tendencies could be made out above (cf. 7.2.3). It seems that the best predictor of contact effects in NP syntax is, in fact, register: Language contact is clearly noticeable in S1A across varieties, while it is rather weak in W1A (although still noticeable), regardless of variety (cf. table 7.73). The following conclusion has been derived from this pattern: The preferences for pre- or postmodification are primarily due to unintentional propagation owing to the high degree of cognitive

40 This does not apply to differences between NP lengths in S1A, which could not be unanimously ascribed to either language contact or simplification (cf. 7.3.1.1.1).

292

7 Studying NP modification in Singaporean English and Kenyan English

Table 7.72. Language contact and simplification by register/modality and variety. Brackets as in “(ICE-SING)” mean: ‘plausible but insignificant statistical deviation’; boldface and no brackets, as in “ICE-SING” mean: plausible and significant frequency deviation; “—” ‘no plausible effect’; ‘effect in one variety’ ; ‘effect in both varieties’ Criterion

Frequencies of pre- and postmodified NPs Hypothesis 1 Modifier subtype

Language contact

Hypothesis 2

S1A

W1A

Contact/simpl. in…

Contact/simpl. in…

ICE-SING and ICE-EA

ICE-SING and ICE-EA

(ICE-EA) ICE-EA

Lengths of the whole premodifying string

(ICE-EA)

Lengths of Mod. II

(ICE-SING and ICE-EA)

Lengths of Mod. III

ICE-SING and (ICE-EA)

Formal realisations of Mod. II

— ICE-EA (ICE-SING)



Formal realisations of Mod. III

(ICE-SING) and ICE-EA

— (ICE-SING)

Formal realisations of postmodifiers

(ICE-EA)

Complexity levels of Mod. II

(ICE-SING and ICE-EA)

Complexity levels of Mod. III

(ICE-SING)

(ICE-SING)

Complexity levels of premodifying string

(ICE-EA)

(ICE-SING)

Complexity levels of postmodifiers

(ICE-SING and ICE-EA)



Number of premodifying slots



Number of postmodifying slots



Number of embedded NPs pre-/postnom.

Length differences between whole NPs Hypothesis 4 Lengths of subjects Simplification



Co-occurrence of pre- and postmodifiers

Lengths of the whole postmodifying string ICE-SING and (ICE-EA)

Hypothesis 3

(ICE-EA)

ICE-SING and ICE-EA

Hierarchical complexity levels of whole NPs

Complexity of subjects Hypothesis 5 Numbers of pre- and postmodifiers

Levels of NP embedding

(Results from typological analyses)

ICE-SING and ICE-EA



(ICE-SING) (ICE-EA) — ICE-SING and ICE-EA

ICE-GB > ICE-SING & ICE-EA ICE-GB & ICE-SING > ICE-EA ICE-SING and ICE-EA

ICE-EA

ICE-GB & ICE-SING > ICE-EA ICE-GB & ICE-SING > ICE-EA (ICE-EA)

ICE-EA

ICE-GB > ICE-SING > ICE-EA ICE-SING >? ICE-GB > ICE-EA ICE-SING and ICE-EA

ICE-EA

ICE-GB & ICE-SING > ICE-EA ICE-GB & ICE-SING > ICE-EA ICE-EA

ICE-EA

ICE-GB >? ICE-EA > ICE-SING ICE-GB=ICE-SING=ICE-EA (ICE-SING and ICE-EA)



ICE-GB & ICE-SING > ICE-EA ICE-SING >? ICE-GB > ICE-EA ICE-EA

ICE-EA

ICE-GB > ICE-SING & ICE-EA ICE-GB > ICE-SING & ICE-EA ICE-SING and ICE-EA

ICE-SING and ICE-EA

7.4 Summary and discussion: Language contact and simplification

293

Table 7.73. The susceptibility of varieties to language contact and simplification in S1A and W1A S1A language contact both glish glish simplification both glish glish

W1A Singaporean En- neither in Singaporean and Kenyan En- English nor in Kenyan English Singaporean En- only in Kenyan English and Kenyan En-

entrenchment of one NP type. With the register-specificity of language contact so strong in both varieties, however, there is at least a potential for the intentional propagation of pre- or postmodified NP linguemes due to their local and, hence, informal associations (cf. 7.2.4). With regard to simplification, in turn, S1A showed relevant effects in both varieties at hand. In W1A, however, simplification tendencies subside in ICE-SING, while they continue to appear in ICE-EA (cf. 7.3.3; table 7.73). This allowed for the following conclusions: – The pattern confirms the prediction made in Hypothesis 5 that Kenyan English should simplify English NPs more strongly than Singaporean English, due to its lower evolutionary stage, as language contact affects both registers in Kenyan English but only the informal one in Singaporean English. This fits in with the picture of Kenyan English drawn above (cf. 4.3), including its comparatively low speaker numbers, its confinement to certain official or semi-official domains, and, as a result, lower overall skill levels. – It was suggested tentatively that simplification in Kenyan English may be due to functional factors without having a major impact on the communal linguistic norm, while in Singaporean English, it is likely to have gained an association with the informal domain, as speakers appear to adopt simple NP linguemes intentionally, independent of SLA-based restrictions (cf. 7.3.4). Furthermore, this pattern shows that the validity of Mair’s (2007: 84) hypothesis of intervarietal convergence in writing (cf. 2.1.3) depends on the type of phenomenon under scrutiny. It holds true for the variety-specific phenomenon of language contact but falls short of the data in terms of the universal phenomenon of simplification. All in all, the implication from these tendencies is that language contact

294

7 Studying NP modification in Singaporean English and Kenyan English

and simplification differ in terms of their cognitive status: In the analysis of NP modification patterns in this study, the variety-specific factors appear to have established themselves in a stabler, more homogeneous way (i.e. by a similar kind of propagation and similar effects on usage norms in both varieties), while the universal factor is more transient in depending on the evolutionary stage of the variety at hand.

8 Conclusion The goal of the present inquiry has been the analysis of NP structures in Singaporean English and Kenyan English with the aim of uncovering effects of language contact and simplification, two factors which have been shown to be key to the linguistic evolution of New Englishes (cf. 1.1). Below, I will first sketch the outline of the study and provide an overview of the key results of the corpus analysis. Then, the implications of the present study in the wider field of New English scholarship will be considered.

8.1 The outline of the study Part A laid the groundwork for the corpus analysis by dealing with the basic concepts presupposed by the research question. – Chapter 2 provided an introduction to the field of New Englishes on the basis of a definition of these languages and by presenting an outline of Schneider’s (2007) Dynamic Model. Language contact and simplification, which can be expected to affect NPs, were pinpointed as two major formative processes in the evolution of New Englishes, out of the wide range of linguistic innovations found therein (cf. 1.2). – Chapter 3 was dedicated to the development of a theory of language change and its adaptation for the purpose of the present study. Initially, results of research into language contact and simplification from the field of SLA were discussed, as they allow valuable insights into these processes from the point of view of individual language users. The discussion arrived at the following key conclusions: First, both language contact and simplification are highly common effects in the acquisition of an L2; second, it is well-known that they affect NP structures; third, simplification tendencies can be expected to subside as individual speakers gain in language skills. Considering that New Englishes are largely acquired and used as L2s, these results help to render the main hypothesis of the study plausible (cf. 2.1.5). Subsequently, Croft’s (2000) model of language change was adapted for the present study. It construes language change as an evolutionary two-step process involving innovations at the level of individuals, which lead to novel pairings of meaning and form; these innovations may subsequently transfer to the communal language norm by way of being adopted and propagated in the speech community. The model allows for a holistic explanation of language contact and simplification in New Englishes.

296

8 Conclusion

– Chapter 4 provided an account of Singaporean and Kenyan English, which led to the insight that, in accordance with the analyses provided in the literature, Singaporean English must be considered a highly advanced, endonormatively oriented variety in phase 4 of Schneider’s (2007) model, which is on the brink of moving on to phase 5 (cf. 4.1.1.5). Kenyan English, in contrast, was characterised as being in phase 3, which is due to lower speaker numbers than for Singaporean English, a less stable, exonormative orientation, and its confinement to largely formal domains. Afterwards, a language-typological survey of the respective local languages showed that the lingueme pool of Singaporean English is (and has always been) dominated by head-final NPs, while Kenyan English has primarily been under head-initial influence from African language genera. – Chapter 5 was dedicated to the development of a syntactic model of the NP, which is based on Bache’s (2000) functionalist model but, in order to allow for the full range of structural patterns found in corpora, adds a number of categories and accommodates results of recent studies. Thereafter, the main factors were identified and discussed which govern the structure of NPs, ranging from language-internal ones such as syntactic position and semantic class of the head noun to external influences such as modality, register and variety. Then, a model of NP complexity was developed. Following a discussion of the general notion of linguistic complexity, the complexity of NPs was broken down into the factors of length and hierarchical structure, both of which have been shown to be independent, psycholinguistically relevant proxies of actual cognitive complexity. Length can be operationalised simply on the basis of word counts; for the purpose of measuring hierarchical complexity, however, specific metrics had to be developed, which rank modifiers according to the presence or absence of syntactic structures of varying levels of cognitive complexity (cf. Berlage 2014). On this basis, both pre- and postmodifiers can be classified into three categories (1Prem )–(3Prem ) and (1Pom )–(3Pom ). Likewise, the complexity of whole NPs can be gauged according to a similar metric reaching from the simplest type (1NP ) to the most complex type (8NP ) (cf. 5.3.4). In Part B, a corpus-based study of NP structures in Singaporean English and Kenyan English was conducted against the theoretical backdrop developed in the preceding chapters. – Chapter 6 discussed the choice of corpus sections and the nature of the ICE corpora, and provided technical details about the extraction of data and their annotation as well as the statistical methods used in the analysis of corpus data.

8.2 The implications of the results for the study of New Englishes

297

– Chapter 7 was dedicated to a detailed corpus-based study of language contact and simplification in NPs from Singaporean English and Kenyan English. With regard to language contact, a wide variety of the statistical distributions of NP modification patterns were found, which confirm that NP usage is influenced by the NP structures of local indigenous languages (cf. 7.2.4). Hypotheses predicting the simplification of NPs in the two new varieties, too, could be largely confirmed. In addition, the degree of simplification was found to be consonant with the evolutionary stages of the varieties, with Singaporean English, broadly speaking, covering the middle ground between British English and Kenyan English, and Kenyan English turning out to be least complex (cf. 7.3.4). In addition, Singaporean English has been shown to exceed British complexity levels occasionally. However, the pattern of variation turned out to be highly register-specific, which allowed for a range of further conclusions. The key insight was that effects of language contact appear to be entrenched to a similar degree in the two varieties, as it has highly similar effects across both registers. Simplification, in contrast, which figures across registers in Kenyan English but is restricted to informal speech in Singaporean English, seems a more transient phenomenon, which speakers in the phase-3 variety cannot control at will, whereas speakers of a phase-4 variety have active control of this strategy.

8.2 The implications of the results for the study of New Englishes The study has a range of general implications with regard to linguistic methodology, and more specific ones in terms of the phenomenon of New Englishes.

8.2.1 Methodological implications The study, to start with, testifies to the potential of corpus-based approaches to grammatical variation. Being non-categorical, statistical phenomena, the tendencies found cannot be analysed on the basis of anecdotal evidence or through the analysis of individual linguistic items. Furthermore, it shows the potential of custom-made extraction scripts, which, in conjunction with part-of-speech tagging, have allowed for safe and time-effective extraction of complex and highly variable structures such as NPs. This challenge cannot be tackled easily by means of traditional concordance programmes, as NPs lack consistent formal properties.

298

8 Conclusion

Furthermore, the study underscores the potential of a comparative approach to New Englishes, which “combine[s] careful, intralingual-philological, variationist analysis with the broad, abstractive bird’s eye perspective that is the hallmark of language typology” (Kortmann and Szmrecsanyi 2009a: 31). It was only through the comparison of Singaporean English and Kenyan English, along with the knowledge about the typological differences between their substrates, that the syntactic effects found in S1A and W1A could be ascribed with due certainty to either variety-specific effects of language contact or universal factors of simplification. These results, which have been found across New Englishes from geographically distant regions and are consonant with the contrasting structure of the relevant substrates and the evolutionary stages assumed for the varieties in question, render the results reliable and minimise the danger of interpreting other causes of variation in terms of language contact. The results also testify to the importance of register and modality in the analysis of New Englishes. Both the effects of language contact and the impact of simplification were highly register-dependent, proving strongest in informal speech and considerably weaker in formal academic language (cf. 7.4). For this reason, Sand’s (2004: 294) emphasis on the importance of register and modality and her warning that it is “questionable when generalizations about the nature […] of any […] variety are made on the basis of data of only one text type” (Sand 2004: 295) can be endorsed. The study confirms the advantages of assuming a cognitive linguistic grounding in a study of New Englishes. It was only through the use of the cognitive notion of lingueme that language contact involving the transfer of abstract syntactic structures (rather than concrete substantive items) could be explained plausibly (cf. 1.3.2.2). In addition, the study has also capitalised on an evolutionary construal of language change. Rather than relying on an abstract notion of “structural nativisation” or viewing language contact as a process affecting linguistic systems at large, this allowed for a clear and explicit statement of the processes and motivations of individual speakers which have led to the patterns of NP usage found in present-day Singaporean English and Kenyan English, in line with the notion of methodological individualism in linguistics (cf. Keller 1984), i.e. the insight that “[l]anguages don’t change; people change language through their actions” (Croft 2000: 4) (cf. 3.2.3.1).

8.2 The implications of the results for the study of New Englishes

299

8.2.2 Implications with regard to the phenomenon of New Englishes Through its typological results, the present study confirms that there are “real and firm” (Schneider 2007: 87) dimensions of structural nativisation in New Englishes at a highly basic level of syntax, which, so far, have received hardly any attention in the study of New Englishes (cf. 1.3.1). Speakers of Singaporean English and Kenyan English are influenced in an abstract constructional choice by the structures of local languages, which affect a wide range of structural properties of their English NPs, ranging from the overall frequency of pre- and postmodified NPs to highly detailed structural options such as the syntactic elaboration of pre- and postmodifiers. This study demonstrates empirically that New Englishes do not only have their own phoneme systems and intonation patterns, their own sets of lexemes, their own tendencies in inflection and the use of verbal complementation patterns and all the other obvious and tangible innovations revealed in previous research; these varieties also develop complex “regionally patterned differences in preference” (Gupta 2010: 69) at an abstract syntactic level, which, while not visible on the basis of anecdotal evidence, can be analysed using comparable corpora and statistical methods (cf. 1). This is the level of variation which Schneider (2007: 89) ascribes to speakers “consistently select[ing] […] forms and patterns that conform to an overarching language type, with many innovations being typologically similar in nature and strengthening one specific parameter”, which, in the long run, may lead to a typological split between the native input variety and localised New Englishes. With regard to simplification, this study demonstrates effects with regard to a hierarchical metric of complexity at a very high level of granularity and specificity in a way which has hitherto played hardly any role in the field of New Englishes, despite the ubiquity of the notion of simplification in the discussion surrounding this type of variety (cf. 2.2.2.2). On this basis, the study also confirms Schneider’s (2007) Dynamic Model in its sociohistorical and linguistic analysis of Singaporean English as more advanced than Kenyan English, as the effects found are closely correlated with the evolutionary stages of the varieties under study. Kenyan English appears as a variety which has all the makings of an L1, while Singaporean English clearly tends towards the native variety in terms of complexity or, occasionally, even surpasses British English in formal style (cf. 7.3.2.4).

Bibliography Aarts, F. G. A. M. 1971. “On the distribution of noun-phrase types in English clause-structure”. Lingua 26.3, 281–93. Abdulaziz, Mohamed H. and Ken Osinde. 1997. “Sheng and Engsh: development of mixed codes among the urban youth in Kenya”. International Journal of the Sociology of Language 125.1, 1–21. Abshire, Jean E. 2011. The History of Singapore. Santa Barbara, Denver, and Oxford: Greenwood. Adelaar, Alexander. 2005. “Structural diversity in the Malayic subgroup”. In: The Austronesian Languages of Asia and Madagascar. Ed. by Alexander Adelaar and Nikolaus P. Himmelmann. London: Routledge, 202–26. Alsagoff, Lubna. 2010. “English in Singapore: culture, capital and identity in linguistic variation.” World Englishes 29.3, 336–48. Alsagoff, Lubna and Chee Lick Ho. 1998. “The grammar of Singapore English”. In: English in New Cultural Contexts: Reflections from Singapore. Ed. by Joseph Foley et al. Singapore: Oxford University Press, 127–51. Altenberg, Bengt. 1982. The Genitive versus the Of-Construction: A Study of Syntactic Variation in 17th Century English. Lund: Gleerup. Ansaldo, Umberto. 2009. Contact Languages: Ecology and Evolution in Asia. Cambridge: Cambridge University Press. Appleby, Lee L. 1961. A First Luyia Grammar: with Exercises. Rev. ed. Dar es Salaam, Nairobi, and Kampala: East African Literature Bureau. Aye, Khin Khin. 2005. “Bazaar Malay: History, grammar and contact”. PhD thesis. Singapore: National University of Singapore. http://scholarbank.nus.edu.sg/handle/10635/15028 (7 January 2015). Aye, Khin Khin. 2010. “Identity and variations in the syntax of Singapore Bazaar Malay noun phrase”. Paper presented at the International Conference on Minority and Majority: Language, Culture and Identity. Kuching, Sarawak, Malaysia. 23-24 November 2010. http://www.mymla.org/index.php/14-sample-data-articles/125-icmm2010 (1 April 2014). Aye, Khin Khin. 2013. “Singapore Bazaar Malay”. In: The Survey of Pidgin and Creole Languages. Ed. by Susanne Maria Michaelis et al. Vol. III. Contact Languages Based on Languages from Africa, Asia, Australia and the Americas. Oxford: Oxford University Press, 86–93. Baayen, Rolf Harald. 2008. Analyzing Linguistic Data. Cambridge: Cambridge University Press. Bache, Carl. 1978. The Order of Premodifying Adjectives in Present-Day English. Odense: Odense University Press. Bache, Carl. 2000. Essentials of Mastering English: A Concise Grammar. Berlin: Mouton de Gruyter. Bao, Zhiming. 2005. “The aspectual system of Singapore English and the systemic substratist explanation”. Journal of Linguistics 41.2, 237–67. Bao, Zhiming and Khin Khin Aye. 2010. “Bazaar Malay topics”. Journal of Pidgin and Creole Languages 25.1, 155–71. Bao, Zhiming and Huaqing Hong. 2006. “Diglossia and register variation in Singapore English”. World Englishes 25.1, 105–14. Barton, Juxon. 1921. “Turkana grammatical notes and vocabulary”. Bulletin of the School of Oriental and African Studies 2.1, 43–73.

302

Bibliography

Basciano, Bianca, Nancy Kula, and Chiara Melloni. 2011. “Modes of compounding in Bantu, Romance and Chinese”. Rivista di Linguistica 23.2, 203–49. Bauer, Laurie. 1998. “When is a sequence of two nouns a compound in English?” English Language and Linguistics 2.1, 65–86. Bauer, Robert S. and Stephen Matthews. 2003. “Cantonese”. In: The Sino-Tibetan Languages. Ed. by Graham Thurgood and Randy J. LaPolla. London and New York: Routledge, 146–55. Behaghel, Otto. 1930. “Von deutscher Wortstellung”. Zeitschrift für Deutschkunde 44, 81–9. Bennett, George and Alison Smith. 1976. “Kenya: From white man’s country to Kenyatta’s state 1945–1963”. In: History of East Africa. Ed. by Donald Anthony Low and Alison Smith. Vol. III. Oxford: Clarendon, 109–56. Berlage, Eva. 2014. Noun Phrase Complexity in English. Cambridge: Cambridge University Press. Biber, Douglas. 1988. Variation across Speech and Writing. Cambridge: Cambridge University Press. Biber, Douglas. 1993. “Representativeness in corpus design”. Literary and Linguistic Computing 8.4, 243–57. Biber, Douglas and Bethany Gray. 2011. “Grammatical change in the noun phrase: the influence of written language use influence of written language use”. English Language and Linguistics 15.2, 223–50. Biber, Douglas, Stig Johansson, et al. 1999. Longman Grammar of Spoken and Written English. Harlow, Essex: Longman. Bible Translation & Literacy. 2014. http://www.btlkenya.org/ (1 April 2014). Bodman, Nicholas Cleaveland. 1955. Spoken Amoy Hokkien. Vol. 1. Kuala Lumpur: Spoken Language Services Inc. Bokhorst-Heng, Wendy. 1998. “Language planning and management in Singapore”. In: English in New Cultural Contexts: Reflections from Singapore. Ed. by Joseph Foley et al. Singapore: Oxford University Press, 287–309. Bokhorst-Heng, Wendy. 2005. “Debating Singlish”. Multilingua 24.3, 185–209. Bolinger, Dwight. 1967. “Adjectives in English: attribution and predication”. Lingua 18.1, 1–34. Bongartz, Christiane M. and Sarah Buschfeld. 2011. “English in Cyprus: second language variety or learner English?” In: Exploring Second-Language Varieties of English and Learner Englishes: Bridging a Paradigm Gap. Ed. by Joybrato Mukherjee and Marianne Hundt. Amsterdam: Benjamins, 35–54. Breban, Tine. 2010a. English Adjectives of Comparison: Lexical and Grammaticalized Uses. Berlin and New York: Mouton de Gruyter. Breban, Tine. 2010b. “Is there a postdeterminer in the English noun phrase?” Transactions of the Philological Society 108.3, 213–401. Brems, Lieselotte and Kristin Davidse. 2010. “The grammaticalisation of nominal type noun constructions with kind/sort of : chronology and paths of change”. English Studies 91.2, 180–202. Brenzinger, Matthias, ed. 1992. Language Death: Factual and Theoretical Explanations with Special Reference to East Africa. Contributions to the Sociology of Language. Berlin and New York: Mouton de Gruyter. Brunner, Thomas. 2014. “Structural nativization, typology and complexity: Noun phrase structures in British, Kenyan and Singaporean English”. English Language and Linguistics 18.1, 23–48. Bruthiaux, Paul. 2003. “Squaring the circles: issues in modeling English worldwide”. International Journal of Applied Linguistics 13.2, 159–78.

Bibliography

303

Bruthiaux, Paul. 2010. “The Speak Good English Movement: A web-user’s perspective”. In: English in Singapore. Modernity and Management. Ed. by Lisa Lim, Anne Pakir, and Lionel Wee. Hong Kong: Hong Kong University Press, 91–108. Brutzer, Ernst. 2012. [1906]. Handbuch der Kambasprache. München: Lincom Europa. Buregeya, Alfred. 2006. “Grammatical features of Kenyan English and the extent of their acceptability”. English World-Wide 27.2, 199–216. Buschfeld, Sarah. 2011. “The English language in Cyprus: an empirical investigation of variety status”. PhD thesis. Cologne: University of Cologne. Buschfeld, Sarah. 2013. English in Cyprus or Cyprus English: An Empirical Investigation of Variety Status. Amsterdam: Benjamins. Bybee, Joan L. 2006. “From usage to grammar: the mind’s response to repetition”. Language 82.4, 711–33. Caldwell-Harris, Catherine, Jonathan Berant, and Shimon Edelman. 2012. “Measuring mental entrenchment of phrases with perceptual identification, familiarity ratings, and corpus frequency statistics”. In: Frequency Effects in Language Representation. Ed. by Dagmar Divjak and Stefan Thomas Gries. Berlin and Boston: Mouton de Gruyter, 165–94. Carroll, Mary et al. 2000. “The relevance of information organization to Second Language Acquisition studies”. Studies in Second Language Acquisition 22.3, 441–66. Center for Immigration Studies. 1996. “Are immigration preferences for English speakers racist?” http://cis.org/ImmigrationPreferences-EnglishSpeakers (25 March 2014). Chambers, Jack K. 2004. “Dynamic typology and vernacular universals”. In: Dialectology Meets Typology: Dialect Grammar from a Cross-Linguistic Perspective. Ed. by Bernd Kortmann. Berlin: Mouton de Gruyter, 127–45. Chan, Alice Y. W. 2004. “Syntactic transfer: evidence from the interlanguage of Hong Kong Chinese ESL learners”. The Modern Language Journal 88.1, 56–74. Chong, Euvin Loong Jin and Mark F. Seilhamer. 2014. “Young people, Malay and English in multilingual Singapore”. World Englishes 33.3, 363–77. CIA. 2016. Kenya. The World Factbook. https://www.cia.gov/library/publications/the-world-factbook/geos/ke.html (22 September 2016). Colleman, Timothy. 2011. “Ditransitive verbs and the ditransitive construction: a diachronic perspective”. Zeitschrift für Anglistik und Amerikanistik 59.4, 387–410. Collins, Laura. 2002. “The roles of L1 influence and lexical aspect in the acquisition of temporal morphology”. Language Learning 52.1, 43–94. Cooper, Thomas C. 1976. “Measuring written syntactic patterns of second language learners of German”. Journal of Educational Research 69.5, 176–83. Cos, eriu, Eugenio. 1974. [1958]. Synchronie, Diachronie und Geschichte: das Problem des Sprachwandels. München: Fink. Croft, William. 2000. Explaining Language Change. Harlow and Munich: Longman. Croft, William. 2006. “The relevance of an evolutionary model to historical linguistics”. In: Competing Models of Linguistic Change: Evolution and Beyond. Ed. by Ole Nedergaard Thomsen. Amsterdam and Philadelphia: Benjamins, 91–132. Croft, William and D. Alan Cruse. 2007. Cognitive Linguistics. Cambridge: Cambridge University Press. Dagut, Menachem and Batia Laufer. 1985. “Avoidance of phrasal verbs — a case for contrastive analysis”. Studies in Second Language Acquisition 7.1, 73–9.

304

Bibliography

Davidse, Kristin. 2004. “The interaction of quantification and identification in English determiners”. In: Language, Culture and Mind. Ed. by Michel Achard. Stanford: CSLI Publications, 507–33. Davidse, Kristin. 2009. “Complete and sort of : from identifying to intensifying?” Transactions of the Philological Society 107.3, 262–92. Davidse, Kristin, Lieselotte Brems, and Liesbeth de Smedt. 2008. “Type noun uses in the English NP: A case of right to left layering”. International Journal of Corpus Linguistics 13.2, 139–68. De Klerk, Vivian. 2003. “Xhosa English as an institutionalised variety of English: in search of evidence”. English World-Wide 24.2, 221–43. De Haan, Pieter. 1989. Postmodifying Clauses in the English Noun Phrase: A Corpus-Based Study. Rodopi: Amsterdam. De Haan, Pieter. 1993. “Noun phrase structure as an indication of text variety”. In: The Noun Phrase in English. Its Structure and Variability. Ed. by Andreas H. Jucker. Heidelberg: Winter, 85–106. Denison, David. 2006. “Category change and gradience in the determiner system”. In: The Handbook of the History of English. Ed. by Ans van Kemenade and Bettelou Los. Oxford: Cambridge University Press, 279–304. Dennett, Daniel Clement. 1995. Darwin’s Dangerous Idea: Evolution and the Meanings of Life. New York: Simon & Schuster. Department of Statistics Singapore. 2010. Census of Population 2010. http://www.singstat.gov.sg/pubn/popn/c2010acr.pdf (25 March 2014). Deutscher, Guy. 2009. “‘Overall complexity’: a wild goose chase?” In: Language Complexity as an Evolving Variable. Ed. by Geoffrey Sampson, David Gil, and Peter Trudgill. Oxford and New York: Oxford University Press, 243–51. Dik, Simon C. 1978. Functional Grammar. Amsterdam, New York, Oxford: North-Holland Publishing Company. Dixon, Robert M. W. 1997. The Rise and Fall of Languages. Cambridge: Cambridge University Press. Dryer, Matthew S. 2003. “Word order in Sino-Tibetan languages from a typological and geographical perspective”. In: The Sino-Tibetan Language. Ed. by Graham Thurgood and Randy J. LaPolla. London and New York: Routledge, 143–55. Dryer, Matthew S. 2011. “Noun-modifier order in Africa”. In: Geographical Typology and Linguistic Areas: With Special Reference to Africa. Ed. by Osamu Hieda, Christa König, and Hirosi Nakagawa. Amsterdam: Benjamins, 287–311. Dryer, Matthew S. and Martin Haspelmath, eds. 2013. WALS Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/ (1 January 2015). Ellis, Rod. 2008. The Study of Second Language Acquisition. 2nd ed. Oxford and New York: Oxford University Press. Færch, Claus and Gabriele Kasper. 1986. “Cognitive dimensions of language transfer”. In: Crosslinguistic Influence in Second Language Acquisition. Ed. by Eric Kellerman and Michael Sharwood Smith. New York: Pergamon, 49–65. Fathman, Ann. 1977. “Similarities and simplification in the interlanguage of second language learners”. In: The Notions of Simplification, Interlanguages, and Pidgins and Their Relation to Second Language Pedagogy. Ed. by S. Pit Corder and Eddy Roulet. Geneva: Droz, 30–8. Feist, Jim. 2009. “Premodifier order in English nominal phrases: A semantic account”. Cognitive Linguistics 20.2, 301–40.

Bibliography

305

Feist, Jim. 2012. Premodifiers in English: Their Structure and Significance. Cambridge et al.: Cambridge University Press. Ferreira, Fernanda. 1991. “Effects of length and syntactic complexity on initiation times for prepared utterances”. Journal of Memory and Language 30.2, 210–33. Field, Andy P., Jeremy Miles, and Zoë Field. 2012. Discovering Statistics Using R. London et al.: Sage. Fillmore, Charles J., Paul Kay, and Mary Catherine O’Connor. 1988. “Regularity and idiomaticity in grammatical constructions: The case of let alone”. Language 64.3, 501–38. Fink, Teresa Kathleen. 2005. “Attitudes toward languages in Nairobi”. MA thesis. Pittsburgh: Graduate Faculty of Arts and Sciences, University of Pittsburgh. http://d-scholarship.pitt.edu/7407/ (7 January 2015). Fischer, Roswitha. 1998. Lexical Change in Present Day English: A Corpus-Based Study of the Motivation, Institutionalization, and Productivity of Creative Neologisms. Tübingen: Narr. Gacheche, Kagure. 2010. “Challenges in implementing a mother tongue-based languagein-education policy: Policy and practice in Kenya”. POLIS Journal 4, 1–45. Garside, Roger. 1996. Readme: CLAWS part-of-speech tagger. UCREL Computing Department. Garside, Roger and Nicholas Smith. 1997. “A hybrid grammatical tagger: CLAWS 4”. In: Corpus Annotation: Linguistic Information from Computer Text Corpora. Ed. by Roger Garside, Geoffrey Leech, and Tony McEnery. London: Longman, 102–21. Gass, Susan M. and Larry Selinker. 2008. Second Language Acquisition. An Introductory Course. 3rd ed. New York and London: Routledge. Ghesquière, Lobke. 2009. “From determining to emphasizing meanings: the adjectives of specificity”. Folia Linguistica 43.2, 311–43. Giegerich, Heinz J. 2004. “Compound or phrase? English noun-plus-noun constructions and the stress criterion”. English Language and Linguistics 8.1, 1–24. Gil, David. 1998. “Patterns of macrofunctionality in Singlish noun phrases: a questionnaire survey”. In: Papers from the Fifth Annual Meeting of the Southeast Asian Linguistics Society, 1995. Ed. by Shobhana L. Chelliah and Willem J. de Reuse. Tempe: Program for Southeast Asian Studies, Arizona State University, 147–82. Githiora, Chege. 2002. “Sheng: peer language, Swahili dialect or emerging creole?” Journal of African Cultural Studies 15.2, 159–81. Githiora, Chege. 2008. “Kenya: Language and the search for a coherent national identity”. In: Language & National Identity in Africa. Ed. by Andrew Simpson. Oxford and New York: Oxford University Press, 235–52. Goldberg, Adele E. 2006. Constructions at Work: The Nature of Generalization in Language. Oxford: Oxford University Press. Görlach, Manfred. 1990. Studies in the History of the English Language. Heidelberg: Winter. Gorman, Thomas Patrick. 1974. “The development of language policy in Kenya with particular reference to the educational system.” In: Language in Kenya. Ed. by Wilfred H. Whiteley. Nairobi: Oxford University Press, 397–453. Greenbaum, Sidney. 1988. “A proposal for an international computerized corpus of English”. World Englishes 7.3, 315–15. Greenbaum, Sidney. 1996. “Introducing ICE”. In: Comparing English Worldwide: the International Corpus of English. Ed. by Sidney Greenbaum. Oxford: Clarendon Press, 3–12. Greenberg, Joseph Harold. 1963. The Languages of Africa. Bloomington: Indiana University Press.

306

Bibliography

Gries, Stefan Thomas. 2004. HCFA 3.2 — A Program for Hierarchical Configural Frequency Analysis. http://www.linguistics.ucsb.edu/faculty/stgries/research/overview-research.html (11 July 2014). Gries, Stefan Thomas. 2008. Statistik für Sprachwissenschaftler. Göttingen: Vandenhoeck & Ruprecht. Gries, Stefan Thomas. 2009. Statistics for Linguistics with R: A Practical Introduction. Berlin: Mouton de Gruyter. Gries, Stefan Thomas. 2012. “Corpus linguistics, theoretical linguistics, and cognitive/psycholinguistics: towards more and more fruitful exchanges”. In: Corpus Linguistics and Variation in English: Theory and Description. Ed. by Joybrato Mukherjee and Magnus Huber. Amsterdam: Rodopi, 41–63. Gupta, Anthea Fraser. 1989. “Singapore Colloquial English and Standard English”. Singapore Journal of Education 10.2, 33–9. Gupta, Anthea Fraser. 1994. The Step-tongue. Children’s English in Singapore. Clevedon, Philadelphia, and Adelaide: Multilingual Matters. Gupta, Anthea Fraser. 1997. “Colonisation, migration, and functions of English”. In: Englishes Around the World. Ed. by Edgar W. Schneider. Vol. I. General Studies, British Isles, North America. Studies in Honour of Manfred Görlach. Amsterdam and Philadelphia: Benjamins, 47–58. Gupta, Anthea Fraser. 1998. “The situation of English in Singapore”. In: English in New Cultural Contexts: Reflections from Singapore. Ed. by Joseph A. Foley et al. Singapore: Oxford University Press, 106–26. Gupta, Anthea Fraser. 2010. “Singapore Standard English revisited”. In: English in Singapore. Modernity and Management. Ed. by Lisa Lim, Anne Pakir, and Lionel Wee. Hong Kong: Hong Kong University Press, 57–90. Gut, Ulrike. 2007. “First language influence and final consonant clusters in the new Englishes of Singapore and Nigeria”. World Englishes 26.3, 346–59. Gut, Ulrike. 2011. “Studying structural innovations in New English varieties”. In: Exploring Second-Language Varieties of English and Learner Englishes. Bridging a Paradigm Gap. Ed. by Joybrato Mukherjee and Marianne Hundt. Amsterdam and New York: John Benjamins, 101–24. Gut, Ulrike and Lilian Coronel. 2012. “Relatives worldwide”. In: Mapping Unity and Diversity World-Wide: Corpus-Based Studies of New Englishes. Ed. by Marianne Hundt and Ulrike Gut. Amsterdam: Benjamins, 215–41. Halliday, Michael Alexander Kirkwood. 1967. Grammar, Society and the Noun. London: H. K. Lewis & Co. Halliday, Michael Alexander Kirkwood. 1987. “Spoken and written modes of meaning”. In: Comprehending Oral and Written Language. Ed. by Rosalind Horowitz and S. Jay Samuels. San Diego: Academic, 55–82. Halliday, Michael Alexander Kirkwood. 1989. An Introduction to Functional Grammar. Repr. London: Arnold. Haspelmath, Martin. 1999. “Optimality and diachronic adaptation”. Zeitschrift für Sprachwissenschaft 18.2, 180–205. Hawkins, John Alan. 1994. A Performance Theory of Order and Constituency. Cambridge and New York: Cambridge University Press.

Bibliography

307

Heine, Bernd. 1976. A Typology of African Languages: Based on the Order of Meaningful Elements. Berlin: Reimer. Heine, Bernd and Oswin Köhler. 1981. Linguistics — East Africa (Kenya, Uganda, Tanzania). Afrika-Kartenwerk. Monograph to Sheet 10. Berlin and Stuttgart: Gebrüder Bornträger. Heine, Bernd and Wilhelm J. G. Möhlig, eds. 1980. Geographical and Historical Introduction, Language and Society, Selected Bibliography. Vol. 1. Language and Dialect Atlas of Kenya. Berlin: Reimer. Heine, Bernd and Derek Nurse. 2000. “Introduction”. In: African Languages: an Introduction. Ed. by Bernd Heine and Derek Nurse. Cambridge: Cambridge University Press, 1–14. Hemphill, Marie de Kiewiet. 1963. “The British sphere, 1884–94”. In: History of East Africa. Ed. by Roland Oliver and Gervase Mathew. Vol. I. Oxford: Clarendon, 391–432. Hilpert, Martin. 2014. Construction Grammar and its Application to English. Edinburgh: Edinburgh University Press. Hirsig, René. 2000. Statistische Methoden in den Sozialwissenschaften: eine Einführung im Hinblick auf computergestützte Datenanalysen mit SPSS für Windows. 2nd ed. Vol. 2. Zürich: Seismo-Verlag. Hoffmann, Thomas. 2011. Preposition Placement in English. A Usage-based Approach. Cambridge: Cambridge University Press. Huber, Magnus. 2012. “Syntactic and variational complexity in British and Ghanaian English. Relative clause formation in the written parts of the International Corpus of English”. In: Language Complexity: Typology, Contact, Change. Ed. by Bernd Kortmann and Benedikt Szmrecsanyi. Berlin, Boston: Mouton de Gruyter, 218–42. Huddleston, Rodney and Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hudson-Ettle, Diana. 1998. “Grammatical subordination strategies in differing text types in the English spoken and written in Kenya”. PhD thesis. Chemnitz: Philosophische Fakultät der Technischen Universität Chemnitz. https://www.tu-chemnitz.de/phil/english/sections/ linguist/real/independent/eafrica/Diss_Diana/index.htm (8 January 2015). Hudson-Ettle, Diana and Tore Nilsson. 2002. “Orality and noun phrase structure in registers of British and Kenyan English”. ICAME Journal 26.2, 33–61. Hull, David L. 1988. Science as a Process: An Evolutionary Account of the Social and Conceptual Development of Science. Chicago: The University of Chicago Press. Hulstijn, Jan H. and Elaine Marchena. 1989. “Avoidance: Grammatical or semantic causes?” Studies in Second Language Acquisition 11.3, 241–55. Hunt, Kellogg W. 1970. “Recent measures in syntactic development”. In: Readings in Applied Transformational Grammar. Ed. by Mark Lester. New York et al.: Holt, Rinehart and Winston, 187–200. Ibrahim, Muhammad H. 1978. “Patterns in spelling errors”. ELT Journal 32.3, 207–12. Jiemin, Bu. 2012. “A study of relationships between L1 pragmatic transfer and L2 proficiency”. English Language Teaching 5.1, 32–43. Jucker, Andreas H. 1992. Social Stylistics. Syntactic Variation in British Newspapers. Berlin: Mouton de Gruyter. Juola, Patrick. 2008. “Assessing linguistic complexity”. In: Language Complexity: Typology, Contact, Change. Ed. by Fred Karlsson, Matti Miestamo, and Kaius Sinnemäki. Amsterdam: Benjamins, 89–108.

308

Bibliography

Kachru, Braj B. 1982. “Models for non-native Englishes”. In: The Other Tongue: English across Cultures. Ed. by Braj B. Kachru. Urbana, Chicago, and London: University of Illinois Press, 31–57. Kachru, Braj B. 1983. The Indianization of English. Delhi et al.: Oxford University Press. Kachru, Braj B. 1985. “Standards, codification and sociolinguistic realism: the English language in the outer circle”. In: English in the World. Teaching and Learning the Language and Literatures. Ed. by Randolph Quirk and Henry. G. Widdowson. Cambridge et al.: Cambridge University Press, 11–34. Kadenge, Maxwell. 2009. “African Englishes: the indigenization of English vowels by Zimbabwean native Shona speakers”. The Journal of Pan African Studies 3.1, 156–73. Keizer, Evelien. 2007. The English Noun Phrase. The Nature of Linguistic Categorization. Cambridge et al.: Cambridge University Press. Keller, Rudi. 1984. “Bemerkungen zur Theorie des sprachlichen Wandels (Eine Replik auf Dieter Cherubims und Peter Eyers Diskussionsbeiträge zu meinem Aufsatz ”Zur Theorie sprachlichen Wandels” in ZGL 10, 1982, 1–28)/Remarks on the theory of linguistic change. An answer to D. Cherubim and P. Eyer”. Zeitschrift für Germanistische Linguistik 12.1, 63. Keller, Rudi. 1994. On Language Change: The Invisible Hand in Language. London and New York: Routledge. Kellerman, Eric. 1977. “Towards a characterisation of the strategy of transfer in Second Language Learning”. The Interlanguage Studies Bulletin 2.1, 58–145. Kellerman, Eric. 1979. “Transfer and non-transfer: where we are now”. Studies in Second Language Acquisition 2.1, 37–57. Kellerman, Eric. 1983. “Now you see it, now you don’t”. In: Language Transfer in Language Learning. Ed. by Susan M. Gass and Larry Selinker. Rowley: Newbury House, 112–34. Kembo Sure, Edward. 1991. “Language functions and language attitudes in Kenya”. English World-Wide 12.2, 245–60. Kembo Sure, Edward. 2003. “The democratization of language policy: a cultural-linguistic analysis of the status of English in Kenya.” In: The Politics of English as a World Language: New Horizons in Postcolonial Cultural Studies. Ed. by Christian Mair. Amsterdam: Rodopi, 247–65. Kenya National Bureau of Statistics. 2006. 2006 Kenya National Adults Literacy Survey. http://statistics.knbs.or.ke/nada/index.php/catalog/58 (20 March 2014). Kenya National Bureau of Statistics. 2009a. “Ethnic Affiliation”. Kenya Census 2009. http://www.knbs.or.ke/index.php?option=com_content&view=article&id=151:ethnicaffiliation&catid=112&Itemid=638 (1 April 2014). Kenya National Bureau of Statistics. 2009b. Kenya Census 2009. http: //www.knbs.or.ke/index.php?option=com_content&view=article&id=149&Itemid=635 (25 March 2014). Kiarie, Wa’Njogu J. 2004. “Language and multiparty democracy in a multiethnic Kenya”. Africa Today 50.3, 55–73. Kioko, Angelina Nduku and Margaret Jepkirui Muthwii. 2003. “English variety for the public domain in Kenya: speakers’ attitudes and views.” Language, Culture, and Curriculum 16.2, 130–45. Kirby, Simon. 1999. Function, Selection, and Innateness: The Emergence of Language Universals. Oxford: Oxford University Press.

Bibliography

309

Kirk, John William Carnegie. 1905. A Grammar of the Somali Language with Examples in Prose and Verse and an Account of the Yibir and Midgan Dialects. Cambridge: Cambridge University Press. Klein, Wolfgang. 1986. Second Language Acquisition. Cambridge, New York, and Melbourne: Cambridge University Press. Klein, Wolfgang and Clive Perdue, eds. 1992. Utterance Structure: Developing Grammars Again. Amsterdam: Benjamins. Klein, Wolfgang and Clive Perdue. 1997. “The Basic Variety (or: Couldn’t natural languages be much simpler?)” Second Language Research 13.4, 301–47. Kleinmann, Howard H. 1977. “Avoidance behavior in adult Second Language Acquisition”. Language Learning 27.1, 93–107. Kolmogorov, Andrey. 1965. “Three approaches to the quantitative definition of information”. Problems of Information Transmission 1.1, 1–11. Kortmann, Bernd, Kate Burridge, et al., eds. 2004. A Handbook of Varieties of English. Berlin and New York: Mouton de Gruyter. Kortmann, Bernd and Benedikt Szmrecsanyi. 2004. “Global synopsis: morphological and syntactic variation in English”. In: A Handbook of Varieties of English. Ed. by Bernd Kortmann et al. Vol. 2: Morphology and Syntax. Berlin and New York: Mouton de Gruyter, 1142–202. Kortmann, Bernd and Benedikt Szmrecsanyi. 2009a. “Typological parameters of intralingual variability: Grammatical analyticity versus syntheticity in varieties of English”. Language Variation and Change 21.3, 319–53. Kortmann, Bernd and Benedikt Szmrecsanyi. 2009b. “World Englishes between simplification and complexification”. In: World Englishes: Problems, Properties and Prospects. Ed. by Thomas Hoffmann and Lucia Siebers. Amsterdam: Benjamins, 265–85. Kortmann, Bernd and Benedikt Szmrecsanyi. 2012. “Introduction: Linguistic complexity — Second Language Acquisition, indigenization, contact”. In: Linguistic Complexity: Second Language Acquisition, Indigenization, Contact. Ed. by Bernd Kortmann and Benedikt Szmrecsanyi. Berlin: Mouton de Gruyter, 6–34. Krashen, Stephen D. 1983. “Newmark’s ‘ignorance hypothesis’ and current Second Language Acquisition theory”. In: Language Transfer in Language Learning. Ed. by Susan M. Gass and Larry Selinker. Amsterdam: Newbury House, 135–53. Krauth, Joachim and Gustav A. Lienert. 1973. Die Konfigurationsfrequenzanalyse (KFA) und ihre Anwendung in Psychologie und Medizin: ein multivariates nichtparametrisches Verfahren zur Aufdeckung von Typen und Syndromen. Freiburg and Munich: Alber. Kusters, Wouter. 2003. Linguistic Complexity. The Influence of Social Change on Verbal Inflection. Utrecht: Landelijke Onderzoekschool Taalwetenschap. Kusters, Wouter. 2008. “Complexity in linguistic theory, language learning and language change”. In: Language Complexity: Typology, Contact, Change. Ed. by Matti Miestamo, Kaius Sinnemäki, and Fred Karlsson. Amsterdam: Benjamins, 3–22. Lado, Robert. 1957. Linguistics across Cultures. Applied Linguistics for Language Teachers. Ann Arbor: University of Michigan Press. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1. Stanford: Stanford University Press. Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Vol. 2. Stanford: Stanford University Press.

310

Bibliography

Langacker, Ronald W. 1999. “Assessing the cognitive linguistic enterprise”. In: Cognitive Linguistics: Foundations, Scope, and Methodology. Ed. by Theo Janssen and Gisela Redeker. Berlin and New York: Mouton de Gruyter, 13–60. Larsen, Iver A. 2014. Biography of Iver A. Larsen. http://www.sil.org/biography/iver-larsen-ma (1 April 2014). Larsen-Freeman, Diane. 1983. “Assessing global second language proficiency”. In: Classroom Oriented Research in Second Language Acquisition. Ed. by Herbert W. Seliger and Michael H. Long. Rowley: Newbury House, 287–304. Lass, Roger. 1990. “How to do things with junk: exaptation in language evolution”. Journal of Linguistics 26.1, 79–102. Lass, Roger. 1997. Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Laufer, Batia and Stig Eliasson. 1993. “What causes avoidance in L2 learning — L1-L2 Difference, L1-L2 Similarity, or L2 Complexity?” Studies in Second Language Acquisition 15.1, 35–48. Leech, Geoffrey N. et al., eds. 2009. Change in Contemporary English: A Grammatical Study. Cambridge et al.: Cambridge University Press. Leimgruber, Jakob R. E. 2012. “Singapore English: An indexical approach”. World Englishes 31.1, 1–14. Leimgruber, Jakob R. E. 2013. Singapore English: Structure, Variation, and Usage. Cambridge: Cambridge University Press. Lewis, M. Paul, Gary F. Simons, and Charles D. Fennig, eds. 2013. Ethnologue: Languages of the World. 17th ed. Dallas: SIL International. http://www.ethnologue.com (25 March 2014). Leyew, Zelealem and Bernd Heine. 2008. “Is Africa a linguistic area?” In: A Linguistic Geography of Africa. Ed. by Bernd Heine and Derek Nurse. Cambridge et al.: Cambridge University Press, 15–35. Li, Charles N. and Sandra A. Thompson. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. Lim, Lisa. 2001. “Ethnic group varieties of Singapore English: melody or harmony?” In: Evolving Identities. The English Language in Singapore and Malaysia. Ed. by Vincent B. Y. Ooi. Singapore: Times Academic Press, 53–68. Lim, Lisa. 2007. “Mergers and acquisitions: On the ages and origins of Singapore English particles”. World Englishes 26.4, 446–73. Lim, Lisa. 2010. “Migrants and ‘mother tongues’: Extralinguistic forces in the ecology of English in Singapore”. In: English in Singapore. Modernity and Management. Ed. by Lisa Lim, Anne Pakir, and Lionel Wee. Hong Kong: Hong Kong University Press, 19–54. Lim, Lisa, Anne Pakir, and Lionel Wee. 2010. “English in Singapore: Policies and Prospects”. In: English in Singapore. Modernity and Management. Ed. by Lisa Lim, Anne Pakir, and Lionel Wee. Hong Kong: Hong Kong University Press, 3–18. Lim, Sonny. 1988. “Baba Malay: The language of the Straits-born Chinese”. Papers in Western Austronesian Linguistics 3, 1–61. Lipps, Jonathan. 2011. “A lexical-functional analysis of Swahili relative clauses”. MA thesis. Oxford: Oxford University. http://ora.ox.ac.uk/objects/uuid:1df03c97-7889-46e2-b316-546bed53bc44 (7 January 2015). Lu, Xiaofei. 2010. “Automatic analysis of syntactic complexity in second language writing”. International Journal of Corpus Linguistics 15.4, 474–96. Lyons, John. 1977. Semantics. Vol. 2. Cambridge: Cambridge University Press.

Bibliography

311

Mair, Christian. 2003. “Kreolismen und verbales Identitätsmanagement im geschriebenen jamaikanischen Englisch”. In: Zwischen Ausgrenzung und Hybridisierung. Ed. by Elisabeth Vogel, Antonia Napp, and Wolfram Lutterer. Würzburg: Ergon, 79–96. Mair, Christian. 2007. “British English/American English grammar: Convergence in writing — divergence in speech.” Anglia 125.1, 84–100. Marr, David G. 1971. Vietnamese Anticolonialism 1885–1925. Berkeley: University of California Press. Marsh, Zöe and George W. Kingsnorth. 1972. A History of East Africa: An Introductory Survey. 4th ed. Cambridge: Cambridge University Press. Matthews, Stephen and Virginia Yip. 2003. “Relative clauses in early bilingual development: transfer and universals”. In: Typology and Second Language Acquisition. Ed. by Anna Giacalone Ramat. Berlin: Mouton de Gruyter, 39–81. Matthews, Stephen and Virginia Yip. 2011. Cantonese. A Comprehensive Grammar. 2nd ed. London: Routledge. Mazaud, Caroline. 2004. “Complex premodifiers in present-day English. A corpus-based study”. PhD thesis. Heidelberg: University of Heidelberg. Mazrui, Alamin M. and Ali A. Mazrui. 1996. “A tale of two Englishes: The imperial language in post-colonial Kenya and Uganda”. In: Post-Imperial English. Status Change in Former British and American Colonies, 1940–1990. Ed. by Joshua A. Fishman, Andrew W. Conrad, and Alma Rubal-Lopez. Berlin and New York: Mouton de Gruyter, 271–302. McArthur, Tom. 1987. “The English languages?” English Today 3.3, 9–13. McEnery, Tony, Richard Xiao, and Yukio Tono. 2006. Corpus-based Language Studies. London: Routledge. McWhorter, John H. 2008. “Why does a language undress? Strange cases in Indonesia”. In: Language Complexity: Typology, Contact, Change. Ed. by Matti Miestamo, Kaius Sinnemäki, and Fred Karlsson. Amsterdam: Benjamins, 167–90. Meierkord, Christiane. 2009. “It’s kuloo tu: recent developments in Kenya’s Englishes”. English Today 25.1, 3–11. Melchers, Gunnel and Philip Shaw. 2003. World Englishes. An Introduction. London: Arnold. Mesthrie, Rajend. 1992. English in Language Shift. Cambridge: Cambridge University Press. Mesthrie, Rajend. 2004. “Synopsis: morphological and syntactic variation in Africa and Southeast Asia”. In: A Handbook of Varieties of English. Ed. by Bernd Kortmann et al. Vol. 2: Morphology and Syntax. Berlin and New York: Mouton de Gruyter, 1132–141. Mesthrie, Rajend. 2006. “Anti-deletions in an L2 grammar: A study of Black South African English mesolect”. English World-Wide 27.2, 111–45. Mesthrie, Rajend. 2012. “Deletions, antideletions and complexity theory, with special reference to Black South African and Singaporean Englishes”. In: Linguistic Complexity. Second Language Acquisition, Indigenization, Contact. Ed. by Bernd Kortmann and Benedikt Szmrecsanyi. Berlin: Mouton de Gruyter, 90–100. Mesthrie, Rajend and Rakesh Mohan Bhatt. 2008. World Englishes. The Study of New Linguistic Varieties. Cambridge et al.: Cambridge University Press. Michieka, Martha Moraa. 2005. “English in Kenya: a sociolinguistic profile”. World Englishes 24.2, 173–86. Miestamo, Matti. 2006. “On the complexity of standard negation”. In: A Man of Measure: Festschrift in Honour of Fred Karlsson on His 60th Birthday [Special Supplement to the SKY Journal of Linguistics 19]. Ed. by Mickael Suominen et al. Turku: The Linguistic Association

312

Bibliography

of Finland, 345–56. Updated online version. http://www.linguistics.fi/julkaisut/sky2006special.shtml (15 February 2014). Miestamo, Matti. 2008. “Grammatical complexity in a cross-linguistic perspective”. In: Language Complexity: Typology, Contact, Change. Ed. by Bernd Kortmann and Benedikt Szmrecsanyi. Amsterdam: Benjamins, 23–41. Miestamo, Matti, Kaius Sinnemäki, and Fred Karlsson. 2008. Language Complexity. Typology, Contact, Change. Amsterdam and Philadelphia: Benjamins. Miller, Jim and Regina Weinert. 1998. Spontaneous Spoken Language: Syntax and Discourse. Oxford: Clarendon Press. Milroy, Lesley and Matthew Gordon. 2003. Sociolinguistics: Method and Interpretation. Malden: Blackwell Publishing. Moag, Rodney F. 1992. “The life cycle of non-native Englishes: a case study”. In: The Other Tongue. English Across Cultures. Ed. by Braj B. Kachru. Urbana and Chicago: University of Illinois Press, 233–52. Mufwene, Salikoko S. 2001. The Ecology of Language Evolution. Cambridge: Cambridge University Press. Mufwene, Salikoko S. 2008. Language Evolution: Contact, Competition and Change. London: Continuum. Mugane John, M. 1997. Paradigmatic Grammar of Gı̃kũyũ. Stanford: CSLI Publications. Mukherjee, Joybrato and Stefan Thomas Gries. 2009. “Collostructional nativisation in New Englishes. Verb-construction associations in the International Corpus of English”. English World-Wide 30.1, 27–51. Mukherjee, Joybrato and Stefan Thomas Gries. 2010. “Lexical gravity across varieties of English: An ICE-based study of n-grams in Asian Englishes”. International Journal of Corpus Linguistics 15.4, 520–48. Mukherjee, Joybrato and Sebastian Hoffmann. 2006. “Describing verb-complementational profiles of New Englishes: A pilot study of Indian English”. English World-Wide 27.2, 147–73. Muthwii, Margaret Jepkirui. 2004. “Language planning and literacy in Kenya: living with unresolved paradoxes.” Current Issues in Language Planning 5.1, 34–50. Mwangi, Serah. 2003. Prepositions in Kenyan English: A Corpus-Based Study in Lexico-Grammatical Variation. Aachen: Shaker. Mwanzi, Henry A. 1985. “African initiatives and resistance in East Africa, 1880–1914”. In: General History of Africa. Ed. by A. Adu Boahen. Vol. VII. Berkeley: Heinemann Educational Books, 149–68. Nabea, Wendo. 2009. “Language policy in Kenya: negotiation with hegemony”. The Journal of Pan African Studies 3.1, 121–38. National Council for Law Reporting. 2010. The Constitution of Kenya. http://www.kenyalaw.org/kl/index.php?id=398 (13 March 2014). Nattinger, James R. and Jeanette S. DeCarrico. 1992. Lexical Phrases and Language Teaching. Oxford: Oxford University Press. Nelson, Gerald. 1996. “The design of the corpus”. In: Comparing English Worldwide: the International Corpus of English. Ed. by Sidney Greenbaum. Oxford: Clarendon Press, 27–35. Nelson, Gerald. 2002. The Singapore Corpus. User Manual. International Corpus of English. Hong Kong.

Bibliography

313

Nelson, Gerald, Sean Wallis, and Bas Aarts. 2002. Exploring Natural Language: Working with the British Component of the International Corpus of English. Amsterdam and Philadelphia: Benjamins. Odlin, Terence. 1989. Language Transfer. Cross-Linguistic Influence in Language Learning. Cambridge, New York, and Melbourne: Cambridge University Press. Odlin, Terence. 1990. “Word-order transfer, metalinguistic awareness, and constraints on foreign language learning”. In: Second Language Acquisition/Foreign Language Learning. Ed. by Bill VanPatten and James F. Lee. Clevedon, Avon: Multilingual Matters, 95–117. Odlin, Terence. 2003. “Cross-linguistic influence”. In: The Handbook of Second Language Acquisition. Ed. by Catherine J. Doughty and Michael H. Long. Malden: Blackwell, 436–86. Okoth Okombo, Duncan. 1997. A Functional Grammar of Dholuo. Köln: Köppe. Olavarrı̀a de Ersson, Eugenia and Philip Shaw. 2003. “Verb complementation patterns in Indian Standard English”. English World-Wide 24.2, 137–61. Ortega, Lourdes. 2003. “Syntactic complexity measures and their relationship to L2 proficiency: a research synthesis of college-level L2 writing”. Applied Linguistics 24.4, 492–518. Ortega, Lourdes. 2012. “Interlanguage complexity. A construct in search of theoretical renewal”. In: Language Complexity: Typology, Contact, Change. Ed. by Bernd Kortmann and Benedikt Szmrecsanyi. Amsterdam: Mouton de Gruyter, 127–55. Orwenjo, Daniel Ochieng. 2012. “Multilingual education in Kenya: Debunking the myths”. International Journal of Multilingualism 9.3, 294–317. Owino, Francis R. 1999. Basic Data on African Languages in Kenya. Cape Town: Centre for Advanced Studies of African Society. Oyori Ogechi, Nathan. 2002. “Trilingual codeswitching in Kenya — evidence from Ekegusii, Kiswahili, English and Sheng”. PhD thesis. Hamburg: Hamburg University. http://ediss.sub.uni-hamburg.de/volltexte/2005/2749 (3 April 2014). Pakir, Anne. 1986. “A Linguistic Investigation of Baba Malay”. PhD thesis. Manoa: University of Hawaii. Pakir, Anne. 1991. “The range and depth of English-knowing bilinguals in Singapore”. World Englishes 10.2, 167–79. Paolillo, John C. 2006. “Evaluating Language Statistics: The Ethnologue and Beyond. A report prepared for the UNESCO Institute for Statistics”. https://www.academia.edu/168253/ Evaluating_Language_Statistics_The_Ethnologue_and_Beyond (28 March 2014). Pastor Gómez, Iria. 2011. The Status and Development of N+N Sequences in Contemporary English Noun Phrases. Frankfurt am Main et al.: Peter Lang. Paul, Hermann. 1886. Principien der Sprachgeschichte. Halle: Niemeyer. Perdue, Clive. 1993. Adult Language Acquisition. Vol. 2: The Results. Cambridge: Cambridge University Press. Perrott, Daisy Valerie. 1976. Swahili. 15th ed. London: Hodder and Stoughton. Plag, Ingo. 2006. “The variability of compound stress in English: structural, semantic, and analogical factors”. English Language and Linguistics 10.1, 143–72. Plag, Ingo et al. 2008. “The role of semantics, argument structure, and lexicalization in compound stress assignment in English”. Language 84.4, 760–94. Platt, John T. 1975. “The Singapore English speech continuum and its basilect ’Singlish’ as a ’creoloid’”. Anthropological Linguistics 17.7, 363–74. Platt, John Talbot, Heidi Weber, and Mian Lian Ho. 1984. The New Englishes. London et al.: Routledge & Kegan Paul.

314

Bibliography

Polomé, Edgar C. 1967. Swahili Language Handbook. Washington D. C.: Center for Applied Linguistics. Post, Mark W. 2007. “Grammaticalization and compounding in Thai and Chinese: a text-frequency approach”. Studies in Language 31.1, 117–75. Pride, John, ed. 1982. New Englishes. Rowley et al.: Newbury House Publishers. Primus, Beatrice. 2001. “Word order typology”. In: Language Typology and Language Universals. An International Handbook. Ed. by Martin Haspelmath et al. Vol. 20. 1. Berlin: Mouton de Gruyter, 855–73. Quirk, Randolph et al. 1985. A Comprehensive Grammar of the English Language. London and New York: Longman. R Development Core Team. 2013. R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing http://www.R-project.org (1 January 2015). Rasch, Björn et al. 2006. Quantitative Methoden. Einführung in die Statistik. 2nd ed. Heidelberg: Springer. Ravid, Dorit and Ruth A. Berman. 2010. “Developing noun phrase complexity at school age: A text-embedded cross-linguistic analysis”. First Language 30.1, 3–26. Ringbom, Håkan. 1978. “On learning related and unrelated languages”. Moderna Språk 72.1, 21–5. Ringbom, Håkan. 1992. “On L1 transfer in L2 comprehension and L2 production”. Language Learning 42.1, 85–112. Rosenbach, Anette. 2008. “Language change as cultural evolution: Evolutionary approaches to language change”. In: Variation, Selection, Development. Probing the Evolutionary Model of Language Change. Ed. by Regine Eckardt. Berlin: Mouton de Gruyter, 23–72. Rosenbach, Anette. 2014. “English genitive variation — the state of the art”. English Language and Linguistics 18.2, 215–62. Rutherford, William E. 1983. “Language typology and language transfer”. In: Language Transfer in Language Learning. Ed. by Susan M. Gass and Larry Selinker. Rowley: Newbury House, 358–70. Saeed, John Ibrahim. 1993. Somali Reference Grammar. Kensington: Dunwoody Press. Sampson, Geoffrey. 1994. Schools of Linguistics. Stanford: Stanford University Press. Sand, Andrea. 2004. “Shared morpho-syntactic features of contact varieties: article use”. World Englishes 23.2, 281–98. Schachter, Jacquelyn. 1974. “An error in error analysis”. Language Learning 24.2, 205–14. Schachter, Jacquelyn and William Rutherford. 1979. “Discourse function and language transfer”. Working Papers on Bilingualism 19, 3–12. Schadeberg, Thilo C. 1992. A Sketch of Swahili Morphology. 3rd rev. ed. Köln: Köppe. Schäpers, Uta Katharina Elisabeth. 2009. Nominal versus Clausal Complexity in Spoken and Written English: Theory and Description. Frankfurt am Main: Lang. Schilk, Marco. 2006. “Collocations in Indian English: a corpus-based sample analysis”. Anglia 124.2, 276–316. Schilk, Marco and Steffen Schaub. 2016. “Noun phrase complexity across varieties of English. Focus on syntactic function and text type”. English World-Wide 37.1, 58–85. Schmied, Josef. 1991. English in Africa. London: Longman. Schmied, Josef. 2004a. “East African English (Kenya, Uganda, Tanzania): Morphology and syntax”. In: A Handbook of Varieties of English. Ed. by Kate Burridge et al. Vol. 2: Morphology and Syntax. Berlin and New York: Mouton de Gruyter, 929–47.

Bibliography

315

Schmied, Josef. 2004b. “East African English (Kenya, Uganda, Tanzania): Phonology”. In: A Handbook of Varieties of English. Ed. by Kate Burridge et al. Vol. 1: Phonology. Berlin and New York: Mouton de Gruyter, 918–30. Schmied, Josef and Diana Hudson-Ettle. 1999. “Manual to accompany the East African component of the International Corpus of English”. Chemnitz University of Technology. Schneider, Edgar W. 2003. “The dynamics of New Englishes: from identity construction to dialect birth”. Language 79.2, 233–81. Schneider, Edgar W. 2007. Postcolonial English: Varieties around the World. Cambridge: Cambridge University Press. Schneider, Edgar W. 2013. “English as a contact language: the ‘New Englishes’”. In: English as a Contact Language. Ed. by Marianne Hundt and Daniel Schreier. Cambridge et al.: Cambridge University Press, 131–48. Schneider, Edgar W. 2014. “‘Transnational attraction’: New reflections on the evolutionary dynamics of World Englishes”. World Englishes 33.1, 11–32. Schwartz, Bonnie D. 1999. “‘Transfer’ and L2 acquisition of syntax: where are we now? (‘Transfer’: maligned, realigned, reconsidered, redefined)”. Newcastle and Durham Working Papers in Linguistics 5, 211–34. Seiler, Guido. 2006. “The role of functional factors in language change: an evolutionary approach”. In: Competing Models of Linguistic Change: Evolution and Beyond. Ed. by Ole Nedergaard Thomsen. Amsterdam: Benjamins, 163–82. Seliger, Herbert W. 1989. “Semantic transfer constraints on the production of English passive by Hebrew-English bilinguals”. In: Transfer in Language Production. Ed. by Hans W. Dechert and Manfred Raupach. Norwood: Ablex, 21–33. Sharma, Devyani. 2005. “Dialect stabilization and speaker awareness in non-native varieties of English”. Journal of Sociolinguistics 9.2, 194–224. Sharma, Devyani. 2009. “Typological diversity in New Englishes”. English World-Wide 30.2, 170–95. Siegel, Jeff. 2008. The Emergence of Pidgin and Creole Languages. Oxford and New York: Oxford University Press. Simo Bobda, Augustin. 2007. “Patterns of segment sequence simplification in some African Englishes”. World Englishes 26.4, 411–23. Simpson, John and Edmund Weiner, eds. 1991. The Compact Oxford English Dictionary. 2nd ed. Oxford: Clarendon Press. Sinclair, John. 1990. Collins Cobuild English Grammar. London: Collins. Skandera, Paul. 2003. Drawing a Map of Africa. Tübingen: Narr. Sørensen, Knud. 1980. “From postmodification to premodification”. In: Papers from the Scandinavian Symposium on Syntactic Variation, Stockholm, May 18-19, 1979. Ed. by Sven Jacobson. Stockholm: Almqvist & Wiksell, 77–84. Sridhar, Kamal K. and S. N. Sridhar. 1986. “Bridging the paradigm gap: second language acquisition theory and indigenized varieties of English”. World Englishes 5.1, 3–14. Stafford, Roy Lawrence. 1967. An Elementary Luo Grammar; with Vocabularies. London: Oxford University Press. Steger, Maria and Edgar W. Schneider. 2012. “Complexity as a function of iconicity. The case of complement clause constructions in New Englishes”. In: Linguistic Complexity. Second Language Acquisition, Indigenization, Contact. Ed. by Bernd Kortmann and Benedikt Szmrecsanyi. Berlin and Boston: Mouton de Gruyter, 156–91. Strang, Barbara M. H. 1970. A History of English. London: Methuen.

316

Bibliography

Szmrecsanyi, Benedikt. 2004. “On operationalizing syntactic complexity”. In: Le poids des mots. Proceedings of the 7th International Conference on Textual Data Statistical Analysis. Ed. by Gérard Purnelle, Cédrick Fairon, and Anne Dister. Vol. 2. Louvain-la-Neuve: Presses universitaires de Louvain, 1032–39. Tadmor, Uri. 2009. “Malay-Indonesian”. In: The World’s Major Languages. Ed. by Bernard Comrie. Vol. 2. London and New York: Routledge, 791–818. Tan, Peter K. W. and Daniel K. H. Tan. 2008. “Attitudes towards non-standard English in Singapore.” World Englishes 27.3–4, 465–79. Taylor, Barry P. 1975. “The use of overgeneralization and transfer learning strategies by elementary and intermediate students of ESL”. Language Learning 25.1, 73–107. The ICE Project. 2014. The International Corpus of English. http://ice-corpora.net/ice/ (1 July 2014). Thomason, Sarah G. 2005. Language Contact. An Introduction. Repr. Edinburgh: Edinburgh University Press. Thomason, Sarah G. 2008. “Why universals versus contact-induced change?” In: Vernacular Universals and Language Contacts. Evidence from Varieties of English and Beyond. Ed. by Markku Filppula, Juhani Klemola, and Heli Paulasto. New York: Routledge, 349–63. Thurgood, Elzbieta. 1998. “A description of nineteenth century Baba Malay: a Malay variety influenced by language shift”. PhD thesis. Manoa: University of Hawaii. Thurgood, Elzbieta. 2001. “The development of articles in Baba Malay”. Anthropological Linguistics 43.4, 471–90. Turnbull, Constance Mary. 1980. A Short History of Malaysia, Singapore and Brunei. Stanmore: Cassell. Van Rooy, Bertus. 2008. “A multidimensional analysis of student writing in Black South African English”. English World-Wide 29.3, 268–305. Van Rooy, Bertus. 2010. “Social and linguistic perspectives on variability in World Englishes”. World Englishes 29.1, 3–20. Van Rooy, Bertus. 2011. “A principled distinction between error and conventionalized innovation in African Englishes”. In: Exploring Second-Language Varieties of English and Learner Englishes: Bridging a Paradigm Gap. Ed. by Joybrato Mukherjee and Marianne Hundt. Amsterdam: Benjamins, 189–207. Van Rooy, Bertus and Lande Schäfer. 2003. “An evaluation of three POS taggers for the tagging of the Tswana Learner English Corpus”. In: Proceedings of the Corpus Linguistics 2003 conference, 28-31 March 2003. Ed. by Dawn Archer et al. Vol. 16. Lancaster University Centre for Computer Corpus Research on Language Technical Papers. Lancaster: Lancaster University, 835–44. Van Rooy, Bertus, Lize Terblanche, et al. 2010. “Register differentiation in East African English. A multidimensional study”. English World-Wide 31.3, 311–49. Von Humboldt, Wilhelm. 1836. Über die Verschiedenheit des menschlichen Sprachbaues und ihren Einfluſs auf die geistige Entwickelung des Menschengeschlechts. Berlin: Dümmler. Wasow, Thomas. 1997. “Remarks on grammatical weight”. Language Variation and Change 9.1, 81–105. Wee, Lionel. 2004. “Reduplication and discourse particles”. In: Singapore English: A Grammatical Description. Ed. by Lisa Lim and Joseph A. Foley. Amsterdam: Benjamins, 105–26.

Bibliography

317

Wee, Lionel. 2008a. “Singapore English: Morphology and syntax”. In: A Handbook of Varieties of English. Ed. by Rajend Mesthrie, Bernd Kortmann, and Edgar W. Schneider. Berlin and New York: Mouton de Gruyter, 593–609. Wee, Lionel. 2008b. “Singapore English: Phonology”. In: A Handbook of Varieties of English. Ed. by Rajend Mesthrie, Bernd Kortmann, and Edgar W. Schneider. Berlin and New York: Mouton de Gruyter, 259–77. Wee, Lionel and Umberto Ansaldo. 2004. “Nouns and noun phrases”. In: Singapore English: A Grammatical Description. Ed. by Lisa Lim and Joseph A. Foley. Amsterdam: Benjamins, 57–74. Welch, B. L. 1951. “On the comparison of several mean values: An alternative approach”. Biometrika 38.3–4, 330–6. Whiteley, Wilfred H. 1974a. “Some patterns of language use in the rural areas of Kenya.” In: Language in Kenya. Ed. by Wilfred H. Whiteley. Nairobi: Oxford University Press, 319–50. Whiteley, Wilfred H. 1974b. “The classification and distribution of Kenya’s African languages”. In: Language in Kenya. Ed. by Wilfred H. Whiteley. Nairobi: Oxford University Press, 13–64. Wilcox, Rand R. 2003. Applying Contemporary Statistical Techniques. Amsterdam and Boston: Academic Press. Wilcox, Rand R. and Felix D. Schönbrodt. 2014. The WRS package for robust statistics in R. Version 0.24. https://github.com/nicebread/WRS (9 January 2015). Williams, Jessica. 1987. “Non-native varieties of English: a special case of language acquisition”. English World-Wide 8.2, 161–99. Wiltshire, Caroline R. 2005. “The ‘Indian English’ of Tibeto-Burman language speakers”. English World-Wide 26.3, 275–300. Winford, Donald. 2003. An Introduction to Contact Linguistics. Malden et al.: Blackwell. Winford, Donald. 2009. “The interplay of ‘universals’ and contact-induced change”. In: Vernacular Universals and Language Contacts. Ed. by Markku Filppula, Juhani Klemola, and Heli Paulasto. New York and London: Routledge, 206–30. Winter-Froemel, Esme. 2008. “Towards a comprehensive view of language change: three recent evolutionary approaches”. In: The Paradox of Grammatical Change: Perspectives from Romance. Ed. by Ulrich Detges and Richard Waltereit. Amsterdam: Benjamins, 215–50. Winter-Froemel, Esme. 2011. Entlehnung in der Kommunikation und im Sprachwandel: Theorie und Analysen zum Französischen. Berlin and New York: Mouton de Gruyter. Wolk, Christoph et al. 2013. “Dative and genitive variability in Late Modern English: Exploring cross-constructional variation and change”. Diachronica 30.3, 382–419. Wong, Irene F. H. 1983. “Simplification features in the structure of colloquial Malaysian English”. In: Varieties of English in Southeast Asia. Ed. by Richard B. Noss. Singapore: Singapore University Press, 125–49. Wulff, Stefanie. 2003. “A multifactorial corpus analysis of adjective order in English”. International Journal of Corpus Linguistics 8.2, 245–82. Xiao, Richard. 2009. “Multidimensional analysis and the study of World Englishes”. World Englishes 28.4, 421–50. Yao, Xinyue and Peter Collins. 2012. “The present perfect in World Englishes”. World Englishes 31.3, 386–403. Yeo, Pamela Yu Hui. 2011. “A sketch grammar of Singapore Teochew”. Final year project report. Singapore: Nanyang Technological University. http://dr.ntu.edu.sg/handle/10220/7801 (7 January 2015).

318

Bibliography

Zipp, Lena and Tobias Bernaisch. 2012. “Particle verbs across first and second language varieties of English”. In: Mapping Unity and Diversity Worldwide: Corpus-Based Studies of New Englishes. Ed. by Marianne Hundt and Ulrike Gut. Amsterdam: Benjamins, 167–96. Zobl, Helmut. 1986. “Word order typology, lexical government, and the prediction of multiple, graded effects in L2 word order”. Language Learning 36.2, 159–83. Zuengler, Jane E. 1982. “Kenyan English”. In: The Other Tongue: English across Cultures. Ed. by Braj B. Kachru. Champaign: University of Illinois Press, 112–24.

Appendices

A HCFA results In what follows, the HCFA results for all analyses of categorical data in 7 produced by Gries’s (2004) HCFA script will be provided. It should be noted that only the interactions will be shown which are indicated in the respective graphs; all insignificant cases have been excluded (cf. 6.1.3 for details on how to interpret the output of Gries’s (2004) HCFA script).

ea ea sing sing

Variety 180 207 260 128

Freq 220.916,5 166.083,5 221.487,3 166.512,7

Exp 7.578,2 10.080,2 6.696,7 8.907,6

Cont.chisq < > >
>
>

Obs-exp

0.000,166,333,9 0.027,373,213

P.adj.Holm

*** *

Dec

0.044 0.036

Q

Table A.3. HCFA results shown in figure 7.3 (p. 177) — The co-occurrence patterns of subtypes of pre- and postmodifiers in S1A

Variety

Subtype

Table A.2. HCFA results shown in figure 7.2 (p. 173) — Modifier subtype and the choice between pre- and postmodification in S1A

premodified postmodified premodified postmodified

Type

Table A.1. HCFA results shown in figure 7.1 (p. 171) — The overall frequencies of pre- and postmodified NPs in S1A

322 A HCFA results

ea

AP

44

Freq 25.445,8

Exp 13.529,1

Cont.chisq >

Obs-exp 0.011,956,328,4

P.adj.Holm

sing

(2Prem )

36

Freq 23.763,1

Exp 6.301,4

Cont.chisq >

Obs-exp

0.058,542,877,4

P.adj.Holm

*

Dec

ms

Dec

0.039

Q

0.026

Q

Variety sing sing ea ea

Position

post-head pre-head post-head pre-head

126 159 176 77

Freq 164.198,9 120.801,1 145.762,5 107.237,5

Exp 8.886,5 12.079 6.272,6 8.526

Cont.chisq < > >

< < < > >
< > >
>

Obs-exp

0.003,136,317,5 0.008,275,622,5

P.adj.Holm

1

Levels ea

Variety 253

Freq 306.0968

Exp

9.2104

Cont.chisq


< < >

Obs-exp 5.291,415,650,051,58 0.009,772,291,8 0.000,330,334,2 0.004,000,026,8

P.adj.Holm 10-⁰⁰⁷

0 0 1 1 1 8 2

Complexity level

ea sing gb ea sing ea ea

Variety

s s nonsubj s nonsubj s s

Syntactic position 409 463 250 201 235 31 56

Freq 461.378,3 408.526,9 317.484,8 132.813,4 294.043,6 72.489,4 84.453,6

Exp 5.946,3 7.263,5 14.344,6 35.007,1 11.855,9 23.746,5 9.586,4

Cont.chisq < > < > < <


Obs-exp 9.12062683239482E-007 0.0015353303 0.0456052971 0.00001046 8.85312398151059E-005

P.adj.Holm

Variety

gb ea sing sing sing ea sing sing ea ea sing

Semantic class

3 – abstract 1 – concrete 3 – abstract 3 – abstract 1 – concrete 2 – activity/process 1 – concrete 2 – activity/process 1 – concrete 1 – concrete 2 – activity/process

1 1 8 1 1 5 5 1 8 5 7

Complexity level 164 197 198 177 134 115 42 50 26 61 3

Freq 256.661 106.0684 129.7421 237.7106 93.9181 76.2739 79.2723 80.0143 57.892 89.5278 15.385

Exp 33.4529 77.9549 35.9108 15.5053 17.106 19.6622 17.5247 11.2587 17.5689 9.0903 9.97

Cont.chisq < > > < > > < < < <