English in The Indian Diaspora

English in the Indian Diaspora
Varieties of English Around the World (VEAW)

A companion monograph series devoted to sociolinguistic research, surveys
and annotated text collections. The VEAW series is divided into two parts: a
text series contains carefully selected specimens of Englishes documenting
the coexistence of regional, social, stylistic and diachronic varieties in a
particular region; and a general series which contains outstanding studies
in the field, collections of papers devoted to one region or written by one
scholar, bibliographies and other reference works.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/veaw
Editor
Stephanie Hackert
University of Munich (LMU)
Editorial Board
Manfred Görlach
Cologne
Rajend Mesthrie
University of Cape Town
Peter L. Patrick
University of Essex
Edgar W. Schneider
University of Regensburg
Peter Trudgill
University of Fribourg
Walt Wolfram
North Carolina State University
Volume G50
Edited by Marianne Hundt and Devyani Sharma
Edited by
Marianne Hundt
University of Zurich
Devyani Sharma
Queen Mary University of London
John Benjamins Publishing Company

Amsterdam / Philadelphia
TM
The paper used in this publication meets the minimum requirements of
8
the American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.
CIP data is available from the Library of Congress
Varieties of English Around the World, issn 0172-7362 ; v. G50

isbn 978 90 272 4910 4 (Hb ; alk. paper)
isbn 978 90 272 6951 5 (Eb)
© 2014 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
Table of contents
List of tables vii

List of figures ix
Chapter 1
Introduction 1
Marianne Hundt and Devyani Sharma
Chapter 2
Indo-Trinidadian speech: An investigation into a popular stereotype
surrounding pitch 9
Glenda Alicia Leung and Dagmar Deuber
Chapter 3
Identity, ethnicity and fine phonetic detail: An acoustic phonetic
analysis of syllable-initial /t/ in Glaswegian girls of Pakistani heritage 29
Farhana Alam and Jane Stuart-Smith
Chapter 4
East African Indian twice migrants in Britain: Phonological variation
across generations 55
Claudia Rathore
Chapter 5
Sociophonetics and the Indian diaspora: The NURSE vowel and other
selected features in South African Indian English 85
Rajend Mesthrie and Alida Chevalier
Chapter 6
Imperfectives in Singapore’s Indian community 105
Jakob R. E. Leimgruber and Lavanya Sankaran
vi English in the Indian Diaspora
Chapter 7
Zero articles in Indian Englishes: A comparison of primary
and secondary diaspora situations 131
Marianne Hundt
Chapter 8
A lesser globalisation: A sociolexical study of Indian Englishes
in diaspora, with a primary focus on South Africa 171
Rajend Mesthrie
Chapter 9
Indo-Fijian English: Linguistic diaspora or endonormative stabilization? 187
Lena Zipp
Chapter 10
Transnational flows, language variation, and ideology 215
Devyani Sharma
Index 243
List of tables
Table 2.1 Mean F0 and F0 range for men and women

Table 2.2 Ethnicity identification of female voices as a percentage with chi-square results,
mean F0, F0 range, and phonation type
Table 2.3 Ethnicity identification of male voices as a percentage with chi-square results,
mean F0, and phonation type
Table 3.1 Numbers of Asian students who participated in the ethnography at Riverburn
High, Glasgow, according to Community of Practice
Table 3.2 Average spectral moments for all speakers according to following segment
Table 3.3 Average spectral moments for each speaker
Table 4.1 Production of postvocalic /r/ by generation (R = rhotic realisation, zero = null
realisation)
Table 4.2 Percentage of rhotic tokens by following phonological environment and
generation (pc = preconsonantal, pp = prepausal, pv = prevocalic)
Table 4.3 Phonetic realisation of rhotic tokens by generation (including linking /r/ tokens)
Table 4.4 Percentage of rhotic tokens by speaker (R = rhotic token; N = 50 tokens per
speaker)
Table 4.5 First-generation speakers’ production of postvocalic /r/ by gender (R = rhotic
realisation, zero = null realisation)
Table 4.6 Results of the logistic mixed-effects regression analysis estimating the effects
of following phonological environment, gender and age on the production of
postvocalic /r/ by first-generation participants
Table 5.1 Linguistic influences on English amongst Indians in five selected diasporic
communities
Table 5.2 Means and standard deviations for F1 and F2 or NURSE per speaker
Table 5.3 p-values in a linear regression mixed effects model for F1 and F2 by class and
gender
Table 6.1 Indian population in Singapore. Data 1821–1980 from Sandhu (1993, 775), 1990
from Shantakumar (1993, 867), and 2000–2010 from Wong (2011a)
Table 6.2 Self-reported dominant home language within the ethnically Indian population.
Data from Leow (2001) and Wong (2011a)
Table 6.3 Singapore’s main substrate languages expressing imperfective and perfective
categories
Table 7.1 Birthplace of Indian residents in New Zealand (1981, 2001)
Table 7.2a First-generation speakers in the NZ secondary diaspora
Table 7.2b Second-generation speakers in the NZ secondary diaspora
Table 7.3 Fiji Indians from Fiji (spontaneous conversations from ICE-FJ)
Table 7.4 Variable article use in the primary diaspora (Fiji)
viii English in the Indian Diaspora
Table 7.5a Variable article use in the secondary diaspora (New Zealand) – first generation
(adult migration)
Table 7.5b Variable article use in the secondary diaspora (New Zealand) – first generation
(pre-adolescence migration)
Table 7.6 Variable article use in the secondary diaspora (New Zealand) – second generation
Table 7.7a Variable article usage in first-generation Fiji Indians in New Zealand – testing for
fluctuation within the sample (Standard Deviation/Mean) – adult migrants only
Table 7.7b Variable article usage in first-generation Fiji Indians in New Zealand – testing for
fluctuation within the sample (Standard Deviation/Mean) – all first-generation
informants
Table 7.8 Null articles according to specificity of the NP
Table 7.9 Type of NP – first generation Fiji Indians in New Zealand
Table 9.1 Reported usage of Fiji English: positive versus negative
Table 9.2 Modes of usage: spoken versus written; register: informal
Table 9.3 Domains of use
Table 9.4 Functions of Fiji English
Table 10.1 Transnational activity among older Gen 2 men
Table 10.2 Transnational activity among older Gen 2 women
Table 10.3 Transnational activity among younger Gen 2 men
Table 10.4 Transnational activity among younger Gen 2 women
Table 10.5 Transnational activity and language use (older Gen 2 men)
Table 10.6 Transnational activity and language use (older Gen 2 women)
Table 10.7 Transnational activity and language use (younger Gen 2 men)
Table 10.8 Transnational activity and language use (younger Gen 2 women)
List of figures
Figure 2.1 Listener responses (N = 88) to unmodifed and modified clips pooled by gender
and ethnicity
Figure 3.1 Map of Pakistan showing main areas of provenance of the Scottish Asian
community
Figure 3.2 Schematic representation of relevant Communities of Practice for Glasgow
Pakistani high school girls
Figure 3.3 Box plots of Mean according to Community of Practice (N = 436)
Figure 3.4 Box plots of Skew according to Community of Practice (N = 436)
Figure 3.5 Box plots of Kurtosis according to Community of Practice (N = 436)
Figure 3.6 Schematic representation of a flattish spectrum cut off by a high-pass filter
Figure 3.7 Box plots of Mean in 5 Glasgow Asian speakers and 2 Glasgow non-Asian
speakers (N = 88); from Stuart-Smith (2009)
Figure 3.8 Spectrograms of the same Glasgow Asian girl producing retroflex /ʈ/ in Punjabi
motiyan ‘fat’ (top), and retracted /t/ in don’ teach (bottom); frequency location of
the bursts is marked with arrows
Figure 4.1 Map of North-western India (1890–1947), regions of origins of East African
Indians
Figure 4.2 Production of postvocalic /r/ by generation (R = rhotic realisation, zero = null
realisation)
Figure 4.3 Production of postvocalic /r/ by following phonological environment and
Figure 4.4 Phonetic realisation of rhotic tokens by generation (including linking /r/ tokens;
tapTrill = taps and trills, app = approximants)
Figure 4.5 Production of postvocalic /r/ by speaker (R = rhotic realisation, zero = null
realisation; N = 50 tokens per speaker)
Figure 4.6 First-generation speakers’ production of postvocalic /r/ by gender (R = rhotic
Figure 4.7 First-generation speakers’ production of postvocalic /r/ by age (R = rhotic
Figure 5.1 /ɜː/ and its variants in RP (from Gimson 1989, 123)
Figure 5.2 The cardinal vowels represented in the BARK scale (based on Thomas 2001, 146)
Figure 5.3 The range of variation for NURSE by speaker means in the SAIE sample (N = 21)
Figure 5.4 Boxplot showing means, median and distribution of F1 for NURSE tokens per
class and gender group
Figure 5.5 Boxplot showing means, median and distribution of F2 for NURSE tokens per
class and gender group
Figure 5.6 Cluster dendogram based on Euclidean distances between speakers
Figure 6.1 Percentage of the use of the progressive -ing in Indian English
Figure 6.2 Percentage of sentences scored as being correct by the three ethnic groups
Chapter 1
Introduction
Marianne Hundt and Devyani Sharma

University of Zurich / Queen Mary University of London
According to India’s Ministry of External Affairs, more than 20 million Indians

live outside India, either as people of Indian origin (PIOs) or non-resident Indi-
ans (NRIs). This is a relatively conservative estimate, however, as it is based on a
narrow definition of what constitutes the Indian Diaspora:
The Indian Diaspora is a generic term to describe the people who migrated from
territories that are currently within the borders of the Republic of India.
(http://indiandiaspora.nic.in [last accessed on 1/31/2014])
The Encyclopedia of the World’s Minorities (Skutsch 2005) uses a broader defini-
tion, including people who migrated from the whole of the Indian subcontinent
(including Pakistan, Sri Lanka, Bangladesh, Nepal and Bhutan). It is this wider
definition that we take as a starting point for the papers collected in this volume.
Almost all contributions in the present collection deal with migration from In-
dia, but our use of the term ‘Indian Diaspora’ refers more generally to the Indian
subcontinent.
As far as scholarly treatments of the Indian Diaspora are concerned, literary
works have received wide recognition (see the work of authors such as Salman
Rushdie, V. S. Naipaul or Jhumpa Lahiri, to name but a few). Historians, sociol-
ogists, psychologists, anthropologists and economists have focussed on aspects
of the Indian Diaspora.1 Surprisingly, the linguistics of the Indian Diaspora had,
until very recently, been given relatively little attention. Previous research on
language in the Indian Diaspora has tended to focus on the status and mainte-
nance of Indian languages (e.g. Barz & Siegel 1988; Sharma & Annamalai 2003).
One major exception has been research on the Indian Diaspora in South Africa:
Mesthrie’s work is exceptional because it focuses not only on Bhojpuri Hindi
1. Lal, Reeves and Rai (2006) provides a good overview of the historical development of the
Indian Diaspora. Jayaram (2004) and Bhatia (2007) address sociological and psychological as-
pects of migration, whereas Oonk (2007) provides an anthropological perspective. An interdis-
ciplinary approach (including economic aspects) is taken in Parekh et al. (2003).
2 Marianne Hundt and Devyani Sharma
(1991) and sociolinguistic aspects of language contact (1992a) but also describes
the local contact variety of Indian English (1992b). More recently, attitudes to-
wards South African Indian English have also been studied (Wiebesiek 2011).
With the exception of South African Indian English, varieties of English in
the Indian Diaspora are thus a fairly recent area of study. Research has been re-
stricted to individual cases rather than comparisons of language contact across
different diasporic settings.2 In bringing together research by scholars who have
been working on the use of English in the Indian Diaspora, this volume provides
the first comparative perspective on the topic. The contributions originated in a
workshop at the second ISLE (International Society for the Linguistics of English)
conference in Boston in 2011. Together, these works explore the following theo-
retical and methodological questions:
1. To what extent have the different diasporic situations resulted in different

outcomes?
2. How similar are the contact features and processes that we observe across
diasporic situations?
3. Do types of responses to dialect contact in migration relate to the relative
degree of endonormative stabilization of Asian Englishes?
4. To what extent do differences inform our understanding of community struc-
tures, attitudinal orientations and identity development?
5. Is there a relation between the rate of change/adaptation/focusing and the
degree of transnational network maintenance or other factors specific to
migration?
These questions do not simply further our understandings of the Indian Dias-
pora. They can ultimately feed into wider theoretical questions in the study of
language contact, such as the degree of universality in the process of focusing in
migration situations (Kerswill & Trudgill 2005), selection of traits from contact
feature pools (Mufwene 2001), and interactions between social access, identity,
and language change (Le Page & Tabouret-Keller 1985).
The regional spread covered by the present set of studies ranges from Singa-
pore via Africa and the UK to the Caribbean and the South Pacific. Moreover,
the studies in this volume are not restricted to primary diaspora situations but
include research into secondary diaspora contexts in the UK and New Zealand.
The contributions also span a wide range of linguistic features (phonetics, syntax
and lexis) and processes of change.
2. Hallett (2010) compares code switching in newspaper language in English-medium publi-

cations in the Indian and Jewish diaspora.
Chapter 1.â•‡ Introduction 3
Accent variation and change has always been a core focus in variationist lin-
guistics. It is therefore not surprising that four of the nine studies in this volume
examine phonetic variation.
Leung and Deuber (Chapter 2) provide a case study on pitch (a less-studied
variable in diaspora research) in Indo-Trinidadian speech and its significance as a
marker of Indian ethnicity. Using acoustically modified speech samples by Afro-
and Indo-Trinidadians, the authors asked listeners in their experiment to identify
speaker ethnicity. The results of their study show that a high-pitched voice is in-
deed associated with Indo-Trinidadians.
In Chapter 3, Alam and Stuart-Smith provide a detailed acoustic phonetic
analysis of adolescent girls in Glasgow with a Pakistani background. Combining
ethnographic fieldwork and naturalistic recordings with close instrumental analy-
sis, they find that fine phonetic distinctions in the articulation of /t/ correspond to
Communities of Practice that the girls participate in. Like Mesthrie and Chevalier’s
later chapter, this analysis points to the importance of ongoing social change and
emergent local identities in the diasporic location as central in restructuring mate-
rial originally contributed by Indian languages.
Rathore also looks at phonological variation in a UK-based Indian diaspo-
ra in Chapter 4, but with a focus on variation across different generations in a
secondary diaspora situation, i.e. migrants from East Africa in Leicester. The
patterns of variation she finds in her case study on rhoticity clearly show that
second-generation members of the community adhere to local, East Midland pat-
terns of usage whereas first-generation migrants retain rhoticity, a feature shared
with mainstream usage in subcontinental India. As far as generational differences
are concerned, the results of her research tie in with other studies on Asian mi-
norities in the UK with similar socio-economic profiles (i.e. university education
and middle-class occupations), showing, as Mesthrie and Chevalier do in the
chapter that follows, that class and other broad social factors must be considered
alongside ethnic or heritage factors.
Mesthrie and Chevalier’s study (Chapter 5) looks at ongoing sound change in
South African Indian English (SAIE). Part of the chapter offers a broader over-
view of types of change observable across phonetic features in SAIE, taking a
broad view of sources and influences, including Indian languages as well as other
local English dialects and languages. This broad discussion also includes a useful
preliminary contrast of linguistic and social factors across different Indian di-
aspora contexts worldwide, that might serve as a basis for further research into
similarities and contrasts across Indian diaspora outcomes. In the later part of the
study, Mesthrie and Chevalier focus more narrowly on variation in the use of the
nurse vowel in South Africa and specifically in SAIE. Comparing evidence from
older SAIE speakers to new data from younger groups, their results indicate, like
Alam & Stuart-Smith’s earlier chapter, that phonetic variation in the use of this
form points to emergent identities in the post-apartheid era, in particular an in-
teraction of gender with class.
Grammatical, rather than phonetic, variation is the focus of the chapters by
Leimgruber and Sankaran (Chapter 6) as well as Hundt (Chapter 7). In a study
of imperfectives in Singapore, Leimgruber and Sankaran explore the possibility
of ethnic sub-community distinctions in Singapore English. They discover dis-
tinctions among Tamil, Chinese, and Malay sub-groups and, assessing the rela-
tive contribution of substrate systems, they find that fine distinctions among the
substrate language grammars may indeed underlie these differences. The study
supports earlier work which suggests that a careful consideration of linguistic sys-
tems in contact might account for subtle differences in emergent grammars across
Indian diaspora contexts.
Hundt compares variable article use in a primary and secondary diaspora
context of Indians in Fiji and Fiji Indians in New Zealand. The expectation was
that variable article use in the secondary diaspora would be closer to the patterns
found in metropolitan varieties of English, particularly for second-generation in-
formants, but also for participants who had migrated before the onset of puber-
ty. These patterns were expected to be subject to difference in the maintenance
of transnational ties. While the data largely confirm these hypotheses, they also
reveal a considerable amount of fluctuation within the sample, which is taken
to reflect individual speakers’ language proficiency in English rather than differ-
ences in identity construction in the first generation. For the second generation,
maintenance of transnational ties appears to be a decisive factor.
Chapter 8 by Mesthrie is concerned with the lexicon of SAIE, the only va-
riety of Indian diasporic English to have received comprehensive lexicographic
treatment so far. Having described themes for a comparative study of diaspora
lexis, Mesthrie moves on to investigate cultural aspects of lexical retention and
variation. He discusses examples of semantic shift, adaptation and borrowing,
and argues that a systematic study of lexis in the Indian Diaspora adds to the
dominant language history in giving “a sense of social history from below”. As
with other studies at the level of phonetic and grammatical variation, his study of
lexical variation indicates clearly that there is a balance between heritage sources
and local contact processes in semantic change.
One of the recurrent questions in the study of world Englishes is that of shift-
ing norm orientation. According to Schneider’s (2007) influential dynamic model
of the evolution of New Englishes, a decisive step in the process is a shift from an
exonormative, colonial model to an endonormative (local) model. In Chapter 9,
Zipp compares data from corpus-based research with evidence from an attitude
survey to verify whether Fiji English has progressed to stage four in Schneider’s
model, that of endonormative stabilization, but instead of comparing language

use by Fiji Indians with usage patterns found in British English, she takes possible
links with a diasporic Indian model into account. While there is evidence of the
development of a local variety of English both in terms of usage and reported use,
the data did not yield evidence of a close connection between the English used by
Indo-Fijians and the English of India, neither in terms of actual structural simi-
larities found in corpus data nor with respect to norm orientation in the attitude
survey data.3
In the final chapter, Sharma continues the focus on the more social question
of whether an orientation to Asian norms is maintained over time. The study
assesses material (network) and ideological (attitudinal) factors in maintenance
of Indian English phonetic forms among second generation Punjabi Londoners.
A measure of transnational network activity points to declining transnational
ties and declining importance of this factor. For instance, some younger Brit-
ish Asians maintain robust use of Punjabi-derived postalveolar /t/, but this no
longer correlates with personal transnational ties, as it did for older individuals.
This mirrors other chapters (Alam and Stuart-Smith, Mesthrie and Chevalier) in
highlighting the development of local meanings for heritage features. In the ide-
ological dimension, British Asians associate educated Indian English with high
status, suggesting that this variety may show increasing pan-diasporic prestige
and influence, perhaps especially when wealthy and powerful transnational South
Asians are a salient presence.
The primary goal of this collection has been to bring together parallel, state-
of-the-art research on a single diaspora but from diverse settings. This permits a
consideration of how prevalent shared or divergent developments are in the face
of distinct linguistic and social environments. In closing, we consider here how a
few of these comparative questions – both linguistic and social – have been facil-
itated by the present collection.
Several papers address the question of whether some features of English in
the diaspora are more durable or persistent than others, i.e. whether certain ac-
cent features prevail, for instance, or some syntactic patterns are observable even
in second-generation speakers and across diasporic situations. In many studies,
certain variables appear to be relatively more durable, although there is little evi-
dence of strictly identical linguistic development across diasporic contexts. These
observations have implications for ongoing debates over Angloversals and shared
traits in New Englishes, in that they can identify some of the limits of apparent
3. Zipp’s analysis in this chapter focuses on the question of reported language use. Cognitive,
affective and conative attitudes to varieties of English in Fiji (both by Fijians and Fiji Indians)
are investigated in Hundt et al. (submitted).
similarities across contexts, highlight the importance of substrate effects, but also
indicate linguistic domains that may be more susceptible to parallel developments.
Another theme that arises across several contributions is whether most speak-
ers adapt to the new environment over time, and which social factors foster reten-
tion in general, while also selecting certain adaptations over others. A number of
studies show some retention by the second, even the third, generations, but often
with structural as well as social restructuring and reallocation. These processes
suggest that, although the development of the Indian diaspora communities is in
many ways accompanied by cultural and linguistic situations quite distinct from
Western urban and rural dialectological studies, certain processes of focusing and
reallocation arise regardless of the cultural setting and thus are likely to also per-
tain to other diaspora contexts (e.g. Trudgill 2004).
In addition, the papers address the related, socially oriented themes of ide-
ology, identity construction, and types of social embedding: To what extent are
diasporic varieties exo- or endonormative? How do community boundaries
and emerging identities affect attitudes, which in turn may influence linguistic
form? How internally heterogeneous are diasporic groups (and what are the con-
sequences of these differences)? And finally, what is the broader positioning of
English in the given diasporic setting, e.g. vis à vis the majority variety, second
language varieties, as part of a multilingual repertoire, or in a shift situation with
an English-based creole?
Among these questions, one particularly striking discovery is repeated across
several chapters in the volume. This is the finding that later generations, even if
they retain forms that derived originally from heritage language forms, almost
always show evidence of new local social meanings and clear local interactions of
social factors, e.g. strengthening interactions of gender and class. This repeated
pattern dovetails with robust findings of this kind in the variationist sociolinguis-
tic tradition (Labov 2001). This parallel between the diasporic studies and older
Western urban studies reminds us that dynamics of dialect change in the Indian
diaspora may well conform at times, though not always, to well-established prin-
ciples in urban Western sociolinguistic research.
This volume is just a first step in developing a robust comparative sociolin-
guistics of diaspora Englishes, and studies explicitly designed to compare a single
diaspora across distinct settings remain very rare. We hope that the volume offers
a starting point for innovative, comparative research into the question of language
variation and change – as well as identity formation and social change – through
the specific lens of transnational migration.
Acknowledgements
The papers in this volume were reviewed by experts in the field. We would like
to thank the following colleagues (in alphabetical order) for providing useful
comments on previous drafts of the chapters in this volume: Rakesh Bhatt, Susan
Fitzmaurice, Paul Kerswill, Sam Kirkham, Lisa Lim, Christian Mair, Andrea Sand,
Jan Tent and Kevin Watson. Stephanie Hackert provided useful feedback on the
entire manuscript. Thanks also go to Jenny Mary Keller and Georgina Wood, the
former and current editorial assistants at the English Department in Zurich, for
help with preparing the manuscript for publication. Brook Bolander (Zurich) did
a sterling job at collating the collections at the proofing stage. Thanks!
References
Barz, Richard K. & Siegel, Jeff. 1988. Language Transplanted: The Development of Overseas Hin-
di. Wiesbaden: Otto Harrassowitz.
Bhatia, Sunil. 2007. American Karma: Race, Culture, and Identity in the Indian Diaspora. New
York NY: New York University Press.
Hallett, Jill. 2010. Code switching in diasporic Indian and Jewish English-language media. In
From International to Local English-and Back Again, Roberta Facchinetti, David Crystal &
Barbara Seidlhofer (eds), 27–49. Bern: Peter Lang.
Hundt, Marianne, Zipp, Lena & Huber, André. Submitted. Attitudes towards varieties of Eng-
lish in Fiji: A shift to endonormativity?
Jayaram, Narayana (ed.). 2004. The Indian Diaspora: Dynamics of Migration. New Delhi: Sage.
Kerswill, Paul & Trudgill, Peter. 2005. The birth of new dialects. In Dialect Change: Convergence
and Divergence in European Languages, Peter Auer, Frans Hinskens & Paul Kerswill (eds),
196–220. Cambridge: CUP.
Labov, William. 2001. Principles of Linguistic Change, Vol 2: Social Factors. Malden MA:
Blackwell.
Lal, Brij. V., Reeves, Peter & Rai, Rajesh (eds). 2006. The Encyclopedia of the Indian Diaspora.
Singapore: Singapore University Press.
Le Page, Robert & Tabouret-Keller, Andrée. 1985. Acts of Identity: Creole-Based Approaches to
Language and Ethnicity. Cambridge: CUP.
Mesthrie, Rajend. 1991. Language in Indenture. A Sociolinguistic History of Bhojpuri-Hindi in
South Africa. Johannesburg: Witwatersrand University Press.
Mesthrie, Rajend. 1992a. Language in Indenture: A Sociolinguistic History of Bhojpuri-Hindi in
South Africa. London: Routledge.
Mesthrie, Rajend. 1992b. English in Language Shift. The History, Structure, and Sociolinguistics
of South African Indian English. Cambridge: CUP.
Mufwene, Salikoko. 2001. The Ecology of Language Evolution. Cambridge: CUP.
Oonk, Gijsbert. 2007. Global Indian Diasporas: Exploring Trajectories of Migration and Theory.
Amsterdam: AUP. DOI: 10.5117/9789053560358
Parekh, Bhikhu, Singh, Gurharpal & Vertovec, Steven (eds). 2003. Culture and Economy in the
Indian Diaspora. London: Routledge.
Sharma, Rekha & Annamalai, E. (eds). 2003. Indian Diaspora: In Search of Identity. Mysore:
Central Institute of Indian Languages.
Schneider, Edgar. 2007. Postcolonial English: Varieties around the World. Cambridge: CUP.
Skutsch, Carl. 2005. Encyclopedia of the World’s Minorities. London: Routledge.
Trudgill, Peter. 2004. New Dialect Formation: The Inevitability of Colonial Englishes. Edinburgh:
EUP.
Wiebesiek, Lisa. 2011. South African Indian English: A qualitative study of attitudes. World
Englishes 30(2): 251–268. DOI: 10.1111/j.1467-971X.2011.01709.x
Chapter 2
Indo-Trinidadian speech
An investigation into a popular
stereotype surrounding pitch
Glenda Alicia Leung and Dagmar Deuber

University of Freiburg / University of Münster
This paper examines the extent to which fundamental frequency (F0) contriÂ�
butes to the stereotype that Indo-Trinidadians have a distinctive way of speak-
ing, namely that they have high-pitched voices in contrast to Afro-Trinidadians,
who are perceived as having voices with a low tone. We report the results of an
experiment in which the F0 of voice samples of both Afro- and Indo-Trinidad-
ians was acoustically modified to investigate this stereotype. Listeners were pre-
sented with unmodified as well as modified samples and were asked to identify
the speaker’s ethnicity. Our results reveal that F0 is indeed one of the salient
cues which Trinidadians rely on to distinguish ethnicity. In addition, phonation
emerges as a potential ethnicity cue.
Keywords: Indo- versus Afro-Trinidadians, ethnicity, perception, F0

manipulation
1. Introduction
At the southernmost point of the Caribbean islands, Trinidad forms a two-island

state together with the much smaller island of Tobago. While Tobago has not been
significantly affected by Indian immigration, Indo-Trinidadians are now one of
the two major ethnic groups in Trinidad, the other one being Afro-Trinidadians.
In the 2000 census of Trinidad and Tobago, 40 percent of the country’s popula-
tion was recorded to be of Indian descent, 37.5 percent of African descent, 20.5
percent mixed and the remainder of unknown or other ethnicity (Central Intel-
ligence Agency 2013). The main languages of Trinidad are English, the official
language, and a mesolectal English-lexifier creole. Indo-Trinidadians have shifted
to English/Creole over time. According to a popular stereotype, their speech is
10 Glenda Alicia Leung and Dagmar Deuber
characterized by a distinctive type of pitch, but there is little empirical evidence

for this so far.
After an outline of the historical, social, and linguistic background to the
Indo-Trinidadian population (Section 2) and a brief sketch of related previous
research (Section 3), this paper will present in its main part (Sections 4–7) an
experiment designed to test the extent to which fundamental frequency (F0) – or
pitch – is an ethnicity marker in Indo-Trinidadian speech. Conclusions and an
outlook for further research will be given in the final Section 8.
2. Historical, social, and linguistic background of the Indo-Trinidadian

population
After the abolition of slavery in the British West Indies in 1834–1838, a new
source of labour was needed for the sugar plantations. Between 1845 and 1917,
approximately 144,000 Indians came to Trinidad under an indentureship scheme
(see Shepherd 2006, 306 for detailed figures of Indian migration to the West In-
dies). Only about 30,000 of these immigrants returned to India between 1850 and
1917 and some re-emigrated as a result of disappointment with what they found
there (Haraksingh 2006, 280).
The majority of Indian immigrants to Trinidad came from the Bhojpuri-Â�
speaking areas in the present-day states of Uttar Pradesh and Bihar (Mohan &
Zador 1986, 294; Mohan 1990, 22). Although there were also speakers of other
Indian languages among the immigrants, Bhojpuri established itself as the lan-
guage of the Indian community in Trinidad and came to represent the new Indo-Â�
Trinidadian identity (Mohan & Zador 1986, 294; Mohan 1990, 23).1 It was locally
known as “plantation Hindi” rather than Bhojpuri, however (Mohan & Zador
1986, 294). Having been transported from a diglossic situation in which Hindi
was the superordinate language (Mohan 1990, 25), it was generally considered a
“broken” form of that language in Trinidad, which, it has been argued, contribut-
ed to a reluctance to maintain it (Mohan 1990, 21, 25).
The first Indian indentured workers generally lived in barracks on the plan-
tations, while after 1870 village settlements became more common (Haraksingh
2006, 280). After the 1920s the life of Indians in Trinidad changed significant-
ly (Haraksingh 2006, 283). Increasing numbers moved away from agricultural
environments, a development which heightened after the 1940s (Haraksingh
2006, 280). Opportunities for education, at first provided mainly by Canadian
1. According to Mohan & Zador (1986, 294), the only Indian language that survived in Trin-
idad beyond the immigrant generation was Tamil.
Chapter 2.â•‡ Indo-Trinidadian speech 11
Presbyterian missionaries (Samaroo 1975), were increasingly made use of and

Indians began to enter professional occupations (Haraksingh 2006, 283). Since
the 1940s, successive generations of Indians have achieved increasing econom-
ic and educational success and they have become progressively integrated into
the larger Trinidad society (Haraksingh 2006, 284–286).2 In 1995–2001 the of-
fice of prime minister was held for the first time by a person of Indian ethnicity,
Basdeo Panday. The country’s current Prime Minister, Kamla Persad-Bissessar,
in office since 2010, is also of Indian ethnicity. Haraksingh (2006, 286) observes
that many of today’s concerns of Indians in Trinidad are shared by Trinidadians
of other ethnicities and that specific concerns are not so large as to create a major
division in the society. “This”, he writes “is an indication of the depth of integra-
tion of Indians into Trinidad society, but more telling is the overriding tenor of
life in Trinidad, which is more embracing and celebrates the country’s diversity”
(Haraksingh 2006, 286).
Bhojpuri has given way to English/Creole as the Indian population has be-
come assimilated. Mohan and Zador (1986) diagnosed language death in a study
based on 1979 data from rural Caroni, where Bhojpuri had survived longer than
anywhere else (Mohan & Zador 1986, 295; see also Bhatia 1988). The oldest native
speaker of the language in their sample had been born in 1930. Among young-
er members of the community they were able to find only a few speakers, who,
moreover, had only a limited, non-native knowledge of the language. The SIL
Ethnologue gives the number 15,600 remaining speakers of Trinidad Bhojpuri in
1996, with the addition that those who use the language are older adults and that
90 percent or more of them reportedly have English/Creole as their first language
(Lewis 2009).3 In another study based on the 1979 Caroni data, Mohan (1990, 30)
concludes: “The linguistic future, then, would appear to be essentially identical to
the socioeconomic future: an Indian community now completely transplanted in
Trinidad, and essentially integrated into its new home”.
3. Previous studies of Indo-Trinidadian speech
Although Indians are linguistically well-integrated into society, it is not implausi-

ble that an Indo-Trinidadian ethnolect may exist. Clyne (2000, 86) broadly defines
ethnolects as “varieties of a language that mark speakers as members of ethnic
2. Transnational ties to India are not generally maintained by Indo-Trinidadians today.

3. Despite the dwindling number of speakers, it is noteworthy that there has been renewed
interest in research on Trinidadian Bhojpuri, most notably in the research agendas of Jennifer
De Silva (2012) and Kofi Yakpo (2012).
groups who originally used another language or distinctive variety”. A natural

concern which arises in such language scenarios is the degree to which elements
of the language originally spoken by the ethnic group are reflected in the new lan-
guage the group has acquired (be it in lexicon, grammar, phonology, or prosody).
As usage of the original language decreases over time, “in the second generation
and beyond, its symbolic significance as an identity marker is transferred to a
variety of the majority language, the variety which is employed by the minority
group either generally or in in-group situations” (Clyne 2000, 86).
Very few studies have focused exclusively on ethnolects in Trinidad despite
the lay belief that Indo- and Afro-Trinidadians “talk differently”. In a macro-so-
ciolinguistic study carried out in 1992, a majority of respondents thought that
there was a difference between Indo- and Afro-Trinidadian speech; pitch, “ac-
cent”, and pronunciation were cited most often in this connection, followed by
vocabulary (Mühleisen 2001, 56). Topics which have received some attention as
regards Indo-Trinidadians and their English/Creole language usage include In-
dic loan words (Winer 2007), phonology (Winford 1978; Leung 2013), and pitch
accent (Gooden, Drayton & Beckman 2009). Though indirectly related, work has
also been done on the use of Indic lexicon by bi-racial Trinidadians of Indian and
African descent (Regis 2012), locally known as douglas.4 To date, no study has
investigated whether the grammar of Bhojpuri has had any effect on Trinidadian
English/Creole.
As regards lexicon, Winer (2007) has identified 1,844 words of Indian (main-
ly Bhojpuri) origin that can be considered part of the lexicon of Trinidad English/
Creole. Most of these words are from various specific domains such as religion,
cooking/food, kinship, and agriculture. Some are also known and used by non-In-
do-Trinidadians, but most are associated with the language use of Indo-Trinidad-
ians, who, however, have varying knowledge of them.
In terms of phonology, Winford’s major sociolinguistic survey conducted in
1970 indicated that old Indian speakers for whom Creole was a second language
had a distinctive vowel system but that younger, first language speakers con-
verged with Afro-Trinidadians (Winford 1978). In a similar vein, Leung’s (2013)
recent work on vowel variation in contemporary Trinidadian English indicates
that at a segmental level there is little difference between the vowel realizations
of Indo- and Afro-Trinidadians. That is not to say that differences do not exist at
a suprasegmental level. In fact Gooden et al.’s (2009) study of Trinidadian Creole
speech reveals that Indo-Trinidadians tend to have a low tone on stressed syl-
lables (L*) while Afro-Trinidadians use a long rise to peak on stressed syllables
4. The term dougla is derived from Hindi dogalā which means “hybrid: mixture; mongrel
animal; person of mixed descent” (Winer 2008, 311).
(L+H*). Furthermore, Drayton (2006) notes that “other Afro-Trinidadians who

interact with Indo-Trinidadians also have L* accents instead of L+H* accents on
stressed syllables” (qtd. in Gooden et al. 2009).
4. An investigation into a popular stereotype surrounding pitch
As mentioned before, there is a commonly held folk belief that Indo-Trinidadians

have a distinctive speech pattern. Indo-Trinidadian voices are often said to be
light, soft, fine, and high in pitch, in contrast to Afro-Trinidadian voices, which
are commonly described as strong, heavy, deep and low in tone.5 This stereotype
can be well observed in a short cartoon called “Vendor Rivalry” (Sattar 2009)
which won the Animae Caribe 2009 award for Best Animation. The video high-
lights ethnic pitch stereotyping by showing two vendors in full confrontation with
each other: the Indo-Trinidadian vendor is represented as having a high pitched
voice whereas the Afro-Trinidadian has a markedly deep voice.
What lay people describe as pitch is in fact a perceptual phenomenon. Al-
though frequency and pitch are related, acoustically speaking the two are not
the same. It is important to bear the distinction between production versus per-
ception in mind. In speech, humans produce frequencies which are measured
in Hertz (Hz). Frequency in correlation to pitch is known as the fundamental
frequency or F0. In contrast, when humans hear a voice, they perceive pitch or the
“auditory realization [or] impression…of that property” (Ball & Müller 2005, 8).
That is to say, in terms of acoustic and psycho-perceptual realities, humans pro-
duce frequencies, but hear pitch. For the sake of convenience, we use the word
pitch to refer to F0 although we are fully aware of the technical difference between
the two terms.
As regards perceptual research on speaker ethnicity, an extensive amount of
work has been done on European-American and African-American voices over
the last 60 years (see Thomas & Reaser 2004 for a comprehensive overview; also
Rahman 2008), ranging from studies on vowel quality to prosodic features, such
as pitch, intonation, and phonation. Recent works outside of an African-Ameri-
can context include Mendoza-Denton’s (2011) research on Chicano gang perso-
nas, Szakay’s (2012) work on voice quality as an ethnicity marker in New Zealand,
and Newmann and Wu’s (2011) work on Chinese- and Korean-Americans. Pho-
nation – which may manifest as modal, creaky, falsetto, or breathy, depending on
the state of the glottis (Laver 1994; Gordon & Ladefoged 2001) – has also received
5. These descriptors were collected from our survey participants. (See Appendix, Section 2 of
the survey.)
notable attention in relation to gendered ethnicity (Stuart-Smith 1999; Alim 2004;

Yuasa 2010; Podesva 2013, and others).
The current study endeavours to contribute to this growing body of research
on speech perception by investigating the suprasegmental cues that Trinidadians
use to assess ethnicity. We start out from the popular stereotype described above
on the basis of which we have formulated the following three hypotheses to be
tested in the study:
H1: Indo-Trinidadians use a wider F0 range than Afro-Trinidadians.

H2: Listeners will identify Indo-Trinidadians as Afro-Trinidadians if the overall
F0 is decreased.
H3: Listeners will identify Afro-Trinidadians as Indo-Trinidadians if the overall
F0 is increased.
After a detailed description of the method in Section 5, the results relating to

these hypotheses will be reported in Sections 6.1–3, followed by an analysis of
interspeaker variation in Sections 6.4–5 and further discussion of the results, also
including findings on phonation in Section 7.
5. Method
In order to examine the aforementioned hypotheses, an experiment was designed

and presented in a survey format to Trinidadians. They were asked to listen to
recordings and to identify the ethnicity of the speakers. Participants were given
two options: they could either identify the speaker as Afro- or Indo-Trinidadi-
an. A follow up question “How sure are you?” probed the certainty of the par-
ticipant’s response, to which the following five response options were provided:
extremely unsure, unsure, fairly sure, sure, and extremely sure. (See Appendix).
It should be noted that we report only the results for ethnicity identification. A
sample of 16 clips of Indo- and Afro-Trinidadian speakers was used, which was
balanced for ethnicity and gender (4 clips per cell). The clips used were taken
from sociolinguistic interviews conducted in Trinidad by the first author in 2009.
Speaker names used in this paper are pseudonyms. Clips included neither Indic
loanwords nor narratives of any theme which could bias the listener’s perception
of the speaker’s ethnicity. F0 was modified in all of the original 16 clips, thus
yielding 32 clips in total. Unmodified clips were not altered in any way, meaning
that the F0 of the clips was not manipulated. In contrast, for the modified clips,
the F0 was either increased or decreased by a constant semitone value, depending
on the speaker’s gender: +/– 1.25 for females and +/– 1 for males. This was done
using Audacity, an audio editing software. F0 was manipulated by semitones since

they are closer to human perception than Hz. How pitch was manipulated in the
clips depended not only on the gender of the speaker but also on the speaker’s
ethnicity: the F0 of the Indo-Trinidadian speakers was decreased whereas the F0
of the Afro-Trinidadian speakers was increased. Below is a summary of the mod-
ifications made:
1. The Indo-Trinidadian female voices were lowered by –1.25 semitones.

2. The Indo-Trinidadian male voices were lowered by –1 semitone.
3. The Afro-Trinidadian female voices were increased by +1.25 semitones.
4. The Afro-Trinidadian male voices were increased by +1 semitone.
Although there was a total of 32 clips, each participant was only asked to judge
16 clips (8 unmodified and 8 modified). In order for all 32 clips to be evaluated,
two versions of the survey with different clips were used for data collection. It
should be noted that each listener never heard both the unmodified and modified
voice of the same speaker.
A total of 88 responses (male = 27, female = 61) was collected in March 2011.
Each version of the survey was completed by 44 participants. Chi-square tests for
independence were conducted to establish if there was a relationship between the
dependent variable “ethnicity identification” (Afro- and Indo-Trinidadian) and
the independent variable “clip status” (unmodified and modified). Chi-square
tests were run both on pooled data (grouped by both ethnicity and gender) and
individual speaker data in an R environment.
Acoustic analysis was also conducted on the clips so as to ascertain if In-
do-Trinidadians use a wider pitch range than Afro-Trinidadians and whether
Indo-Trinidadians have a higher mean F0. All clips were segmented into into-
national phrases, using the software Praat (Boersma & Weenink 2013). In Praat,
the following F0 range settings were used according to the speaker’s gender:
75–300 Hz for males and 100–500 Hz for females. Exclusions were made in the
measurement process wherever creaky voicing occurred and wherever there was
overlap or background noise that affected the pitch curve. For each speaker’s clip,
three F0 measurements (Hz) were made: the pitch floor, the pitch ceiling, and the
mean F0 per intonational phrase. The pitch floor (F0 minimum) and pitch ceil-
ing (F0 maximum) were used to calculate pitch ranges across gender and ethnic
groups. Pitch range was calculated as the F0 difference between the pitch ceil-
ing and pitch floor. Additionally, the measurement of mean F0 per intonational
phrase was used to calculate the overall mean F0. Wilcoxon tests were conduct-
ed to look for differences in F0 values between the various ethnic and gender
groupings.
This study was originally designed to test F0 as an ethnicity assessment cue.

However, during the course of examining interspeaker variation, phonation –
namely, creaky, and modal voicing – emerged as a potential factor for ethnicity
identification. Although our research was not designed to deal with phonation,
we decided to report it because of its apparent relevance to the results. A simple
binary coding scheme, referring to “creaky and modal” voice within the intona-
tional phrase, was used to classify phonation in unmodified clips.6
6. Results
6.1 Hypothesis 1: Difference in F0 range between ethnicities
As a point of comparison, we referred to Gut’s (2009) values for mean F0 (av-

erage pitch) and F0 range (pitch range) for men and women. As can be seen in
Table 2.1, our results were consistent with mean F0 and F0 pitch ranges for both
males and females. Pitch ranges for the Indo- and Afro-Trinidadian men illustrat-
ed little difference; both the pitch floor and ceiling were nearly identical. However,
among the women, differences in pitch range were noticeable. The Indo-Trinidad-
ian women had both a higher pitch floor and ceiling than the Afro-Trinidadian
women. Moreover, the Indo-Trinidadian women had a considerably wider pitch
range (351 Hz) than their Afro-Trinidadian counterparts (258 Hz).7 Further-
more, Wilcoxon tests revealed differences in mean pitch between the Afro- and
Indo-Trinidadian women to be statistically significant, W (159) = 1267, p < .001.
Table 2.1â•‡ Mean F0 and F0 range for men and women

Gender Groups Mean F0 (Hz) F0 range (Hz)
Men From Gut (2009) 100–150 â•⁄70–250
Pooled â•⁄97–135 â•⁄72–237
Indo-Trinidadian â•⁄97–128 â•⁄72–237
Afro-Trinidadian â•⁄97–135 â•⁄72–230
Women From Gut (2009) 150–250 â•⁄80–400
Pooled 147–242 100–471
Indo-Trinidadian 213–242 120–471
Afro-Trinidadian 147–193 100–358
6. Since creaky voicing is more likely to occur at phrase boundaries, we base the coding of
creaky versus modal voicing on their occurrence within the intonation phrase.
7. Pitch range was calculated here as the difference between maximum and minimum F0.
The mean pitch for the Indo-Trinidadian females was considerably higher (M =
210 Hz) than that of Afro-Trinidadian females (M = 166 Hz). However, differenc-
es in mean F0 for the men did not prove to be significant. Thus, hypothesis 1, that
Indo-Trinidadians have a wider pitch range than Afro-Trinidadians, could not
be confirmed for the men on the basis of the sample. However, there was strong
evidence to support this hypothesis among the women.
6.2 Hypothesis 2: Response to modified Indo-Trinidadian voices
Hypothesis 2 predicted how listeners would judge the manipulated clips of the
Indo-Trinidadians when pitch was decreased. Chi-square tests revealed that lis-
teners evaluated unmodified and modified clips of the Indo-Trinidadian women
differently, χ2 (1, n = 352) = 89, p < .0001. (See Figure 2.1 for a visualization of
responses to the clips.) In unmodified clips, listeners identified the speakers as
Indo-Trinidadian 70 percent of the time. As predicted, in modified clips (F0 de-
creased), listeners evaluated the speakers as Afro-Trinidadian at a high rate of
81 percent. Similarly, for the Indo-Trinidadian men, the chi-square test indicat-
ed that listeners evaluated unmodified and modified clips differently, χ2 (1, n =
352) = 20.1, p < .0001. In unmodified clips, listeners identified the speakers as
Indo-Trinidadian 64 percent of the time. In manipulated clips (F0 decreased), lis-
teners evaluated the speakers as Afro-Trinidadian 61 percent of the time. Overall,
there is strong evidence to support hypothesis 2 seeing that listeners mistakenly
identified the Indo-Trinidadians as Afro-Trinidadians when F0 was decreased.
19%
32% 35%
39% 43% 47%
70% 64%
Afro-
Indo-
81%
61% 68% 65%
57% 53%
30% 36%
Unmod.
Mod.
Unmod.
Mod.
Unmod.
Mod.
Unmod.
Mod.
Fem-Indo Male-Indo Fem- Afro Male-Afro
Figure 2.1â•‡ Listener responses (N = 88) to unmodifed and modified clips pooled by
gender and ethnicity
6.3 Hypothesis 3: Response to modified Afro-Trinidadian voices
Hypothesis 3 predicted how listeners would perceive the Afro-Trinidadian voic-

es when pitch was increased. Chi-square tests indicated that listeners generally
evaluated unmodified and modified clips differently. This was the case for both
the males and females, (χ2 (1, n = 352) = 4.71, p = .03 and χ2 (1, n = 352) = 3.92,
p = .048, respectively). (For a visual overview of listener response to the Afro-Â�
Trinidadians’ clips, see Figure 2.1). When pitch was unmodified in clips for the
Afro-Trinidadian men, listeners correctly identified ethnicity 65 percent of the
time. In modified clips (F0 decreased), listeners evaluated speakers as Indo-Trin-
idadian at a rate of 47 percent. With the female Afro-Trinidadian voices, when
pitch was not manipulated, listeners identified the speakers as Afro-Trinidadian
at a rate of 68 percent. In modified clips (F0 decreased), listeners evaluated the
speaker as Indo-Trinidadian 43 percent of the time.
From Figure 2.1, it is clear that a majority of listeners did not perceive the
modified Afro-Trinidadian voices (male and female) as Indo-Trinidadian. Over-
all, this suggests that there was a certain degree of indeterminacy in how listeners
responded to the modified clips. In other words, from the pooled data there was
not a clear consensus for ethnicity identification with manipulated clips (cf. lis-
tener responses to modified Indo-Trinidadian clips in Section 6.2). In fact, when
the pooled data were disaggregated, a much more complex picture emerged, par-
ticularly as to how listeners responded to both the modified and unmodified clips
of Afro-Trinidadian men and women. In light of this, a more detailed analysis at
the level of the individual speakers is presented in Sections 6.4 and 6.5. Overall,
the results lend moderate support to hypothesis 3.
6.4 Interspeaker variation – Trinidadian women
Thus far, it is evident that F0 is a fairly salient cue in ethnicity identification.

Nonetheless, to gain a deeper understanding of the data, additional chi-square
tests on the clips of individual speakers were conducted for interspeaker compar-
ison. Additionally, the clips were re-inspected in Praat. Through auditory analysis
we discovered that phonation types – namely, creaky voice and modal voicing –
might have functioned as cues in delimiting ethnicity.
The disaggregated data for the Afro-Trinidadian women indicated that only
the evaluation of Joan’s speech supported hypothesis 3, χ2 (1, n = 88) = 21.84,
p = < .0001. When her clip was unmodified, she was correctly identified as Af-
ro-Trinidadian 93 percent of the time. When pitch was modified (F0 increased),
she was evaluated as Indo-Trinidadian 52 percent of the time.
Table 2.2 Ethnicity identification of female voices as a percentage with chi-square results, mean F0, F0 range, and phonation type
(ethnicity of the speaker marked with an asterisk *)
Speaker Ethnicity Clip status χ21,43 p Mean F0 (Hz) F0 range (Hz) Phonation
pseudonym identification
% Unmodified (n = 44) % Modified (n = 44)
Philippa *Afro- 77 75 0.06 ns 147 100–23 creaky
Indo- 23 25
Joan *Afro- 93 48 21.84 < .0001 193 132–358 creaky
Indo- 7 52
Marge *Afro- 52 48 0.05 ns 163 101–345 creaky
Indo- 48 52
Yvette *Afro- 48 57 0.73 ns 156 102–287 creaky
Indo- 52 43
Indrani Afro- 9 64 28.29 < .0001 224 120–356 modal
*Indo- 91 36
Rani Afro- 20 93 47.43 < .0001 213 140–371 modal
*Indo- 80 7
Chandi Afro- 39 91 26.35 < .0001 242 148–419 modal
*Indo- 61 9
Sharlene Afro- 52 75 4.91 .027 218 126–471 modal
*Indo- 48 25
For all the other Afro-Trinidadian women, their chi-square tests did not
reach statistical significance, as shown in Table 2.2. In the case of Philippa, she
was moreover perceived as Afro-Trinidadian (77 percent unmodified and 75 per-
cent modified) irrespective of pitch modification. For Yvette and Marge, there
was no clear consensus on how the ethnicity of these women was perceived since
irrespective of clip modification, their voices were judged about half the time as
Indo-Trinidadian and Afro-Trinidadian. From this interspeaker comparison, it
is evident why disaggregating the pooled data is merited as individuals were not
always assessed in the same way.
The disaggregated data for the Indo-Trinidadian women demonstrated more
consistency in how clips were evaluated, as seen in Table 2.2. All chi-square tests
for individual Indo-Trinidadian women supported hypothesis 2.
As for phonation, it is noteworthy that the Afro-Trinidadian women were
found using creaky voice whereas use of modal voicing was noted among all the
female Indo-Trinidadians speakers. (For further discussion of phonation, please
refer to Section 7.)
6.5 Interspeaker variation – Trinidadian men
The disaggregated data for the Afro-Trinidadian men revealed mixed responses
to the clips. This can be seen in Table 2.3. Only the results of Jim’s and Austin’s
chi-square tests supported hypothesis 3 (χ2 (1, n = 88) = 15.01, p = < .0001 and χ2
(1, n = 88) = 8.69, p = .003, respectively). Conversely, John’s speech was consist-
ently perceived as Afro-Trinidadian (93 percent unmodified and 91 percent mod-
ified) in spite of clip modification. One peculiar case is that of Niles. Although his
chi-square test reached significance, χ2 (1, n = 88) = 8.69, p = .014, hypothesis 3
could not be supported. As seen in Table 2.3, Niles’ unmodified clip was errone-
ously evaluated as Indo-Trinidadian 77 percent of the time. Furthermore, when
pitch was modified (F0 increased), there was little difference in how his ethnicity
was perceived (Afro-Trinidadian 48 percent versus Indo-Trinidadian 52 percent).
The disaggregated data for the Indo-Trinidadian men indicated that listeners
did not respond to individual speakers in the same way. In the case of David, irre-
spective of clip modification, his speech was consistently identified as Indo-Trin-
idadian (95 percent in both unmodified and modified clips). Another peculiarity
in the data was the listeners’ response to Ricardo’s clips, χ2 (1, n = 88) = 7.57,
p = .006. His unmodified clip was erroneously evaluated as Afro-Trinidadian 64
percent of the time. However, when pitch was modified (F0 decreased), he was
perceived as Afro-Trinidadian (89 percent). Lastly, chi-square tests for Bisram’s
and Cyrus’ speech reached statistical significance, thus supporting hypothesis 2
(χ2 (1, n = 88) = 13.31, p = < .001 and χ2 (1, n = 88) = 12.02, p = < .001, respectively).
Table 2.3 Ethnicity identification of male voices as a percentage with chi-square results, mean F0, and phonation type
(ethnicity of the speaker marked with an asterisk *)
Speaker Ethnicity Clip status χ21,43 p Mean F0 (Hz) F0 Range (Hz) Phonation
pseudonym identification
% Unmodified (n = 44) % Modified (n = 44)
John *Afro- 93 91 – ns 97 72–187 creaky
Indo- 7 9
Austin *Afro- 82 52 8.69 .003 151 90–230 modal
Indo- 18 48
Jim *Afro- 64 23 15.01 < .0001 103 75–226 creaky
Indo- 36 77
Niles *Afro- 23 48 6.03 .014 125 78–219 modal
Indo- 77 52
David Afro- 5 5 – ns 129 80–237 modal
*Indo- 95 95
Bisram Afro- 25 64 13.31 < .001 118 75–235 modal
*Indo- 75 36
Cyrus Afro- 52 86 12.02 <.001 105 73–205 creaky
*Indo- 48 14
Ricardo Afro- 64 89 7.57 .006 97 72–187 creaky
*Indo- 36 11
As regards phonation, the coding did not reveal a clear tendency among the
men according to ethnicity. (For further discussion of phonation, see Section 7.)
7. Discussion
In this section we examine individual speakers in relation to what we assume to

be “prototypical” of the respective ethnicities. What we consider prototypical is
dependent upon both production and perception data. In other words, by trian-
gulating individual speaker F0 range, phonation tendency, and listeners’ response
to clips, a rough template of a prototypical speaker emerges. We do acknowledge
that locating the precise features listeners use in ethnicity identification is not a
straightforward task, since they are likely to be relying on a bricolage of linguistic
cues (Thomas & Reaser 2004; Newmann & Wu 2011). Nevertheless, given the
semi-experimental nature of this study, we have been able to gain some insight
into the prosodic features at play.
Speakers with the most prototypical Afro-Trinidadian voices included
Philippa and John, both of whose clips were perceived as Afro-Trinidadian prior to
and post pitch manipulation. Both these speakers had small pitch ranges (139 Hz
and 115 Hz, respectively) and accordingly lowest mean F0 in their groups.8 They
also used creaky voice within the intonational phrase. In contrast, speakers with
the most prototypical Indo-Trinidadian voices included Rani and David, both of
whom had high F0 means as well as fairly wide pitch ranges (231 Hz and 157 Hz)
in their respective groups. Both speakers also used modal voicing. Examining
these prototypical voices lends insight into the potential role phonation plays
in the ethnicity identification. One may hypothesize that Trinidadians associate
modal voicing with Indo-Trinidadians and creaky voice with Afro-Trinidadians.
Analysis of less prototypical voices can also motivate this hypothesis. David,
for instance, is an interesting prototypical case to juxtapose with a less proto-
typical speaker such as Ricardo. In contrast to David, Ricardo was erroneously
perceived as Afro-Trinidadian the majority of the time when his voice was un-
modified. Ricardo has the smallest pitch range among the Indo-Trinidadian men
(115 Hz). Additionally, both his pitch floor and ceiling are low and are in the same
range as the prototypically Afro-Trinidadian male, John. His mean F0 (97 Hz) is
on the low end and he often uses creaky voice. In other words, multiple prosodic
dimensions of Ricardo’s voice are similar to that of the most prototypical Afro-Â�
Trinidadian male in the study, even though he is actually Indo-Trinidadian.
8. Pitch range expressed here is the difference between maximum and minimum F0.
For a speaker such as Joan, a tentative case for the role of phonation in ethnic-
ity perception can be made. Joan is the only female Afro-Trinidadian whose voice
was consistently judged as Afro-Trinidadian when unmodified (93 percent). At
such a high rate of Afro-Trinidadian identification, it is natural to ask which cues
we can delimit as being salient to listeners. In terms of F0 values for Joan, her
pitch floor and ceiling are quite high (similar to the range of the Indo-Trinidadian
women). In this instance, listeners do not appear to be responding to the pitch
of Joan’s voice or else they would have been more likely to perceive her as Indo-Â�
Trinidadian. Thus, there must be another cue to which listeners are responding.
Similar to the other Afro-Trinidadian women, Joan uses creaky voice. Even when
her pitch is manipulated (F0 increased), listeners are still uncertain as to her eth-
nicity, with almost half of them still perceiving her as Afro-Trinidadian (48 per-
cent Afro- and 52 percent Indo-Trinidadian identification). Since F0 manipulation
does not affect creaky voice in the clip, we tentatively deduce that listeners are
potentially responding to her use of creaky voice. This evidence is hardly strong
on its own but taken together with the clear distribution of creaky versus mod-
al voicing we observed among the Afro-Trinidadian and the Indo-Trinidadian
women, respectively, the possibility ought not to be discounted.
8. Conclusion and outlook
This study indicates that there are multiple cues that listeners use to ascertain
speaker ethnicity. Differences in F0 are indeed salient in distinguishing Afro-Â�
Trinidadians from Indo-Trinidadians. The results lend some credence to the
folk perception that Indo-Trinidadians have higher pitched voices than Afro-Â�
Trinidadians. From the small sample of voices in this study, this generalization
can be applied most strongly to the Indo-Trinidadian women since there is a
marked difference in the pitch range and mean F0 values between their speech
and that of their counterparts, the Afro-Trinidadian women. However, the data
from the study do not show this to be the case with the men. Work on a much
larger sample is desirable and necessary. As for other prosodic features, phonation
emerges as a potential ethnicity cue.
Our results suggest that an Indo-Trinidadian ethnolect may well exist in
Trinidadian English at a suprasegmental level. Outlining some general acoustic
dimensions of prototypical Afro- or Indo-Trinidadian speakers thus illuminates
an important aspect of sociolinguistic variation in Trinidad. However, attributing
a type of prosodic feature to a social factor – as this study does – is of course not
the only lens through which ethnicity can be examined. In fact, Podesva (2013)
cautions against such correlations between social factors and prosodic features:
“[I]n spite of robust correlations between phonation patterns and identity catego-
ries, the social meanings of particular phonation types are culturally specific and
should not be reduced to purely iconic or unanalyzed associations to either gen-
der or race” (Podesva 2013, 428). In our case, however, the ethnic distinction is a
locally salient one; by using an ethnic stereotype as a starting point we take into
consideration what people perceive to be relevant social categories for linguis-
tic variation. Other research paradigms, such as indexicality (Eckert 2008) and
ethnolinguistic repertoire (Benor 2010; Sharma 2011), which take into consid-
eration the social meanings associated with a linguistic variable, offer constructs
that could be useful to employ in future work. An increasing number of studies
utilize such frameworks to examine the use of prosodic cues by gender and/or
ethnicity (e.g. Sicoli 2010; Becker Forthcoming). A potential study making use
of the notion of indexicality or ethnolinguistic repertoire could examine pitch
accommodation, that is, the ways in which speakers adjust aspects of their pros-
ody, depending on the ethnicity of their addressee. We have informally observed
such a phenomenon. It is probable that accommodation depends not only on the
ethnicity of the addressee, but also on speaker stances taken in the discourse.
What remains uncertain is the extent to which suprasegmental features of
Bhojpuri/Indic languages have impacted on the English/Creole of Indo-Trini-
dadians. Since Trinidad and Guyana are similar in their ethnic composition, and
share histories of indentureship and settlement, perhaps a parallel investigation of
ethnic stereotyping in Guyanese English/Creole may prove useful in determining
whether these prosodic differences possibly come from Bhojpuri/Indic languages.
Acknowledgements
We wish to thank Kathy-Ann Drayton, Jason Leach, Dionne Ligoure, Shelley

Lyons, Richard Rogers, and Stella Williams for facilitating our data collection,
Ulrike Gut and Erik R. Thomas for their advice on the survey design, and
Benedikt Szmrecsanyi, Karen Moll and Sascha Wolfer for statistical guidance.
Any errors remain ours alone.
A very special thanks to the editors, Marianne Hundt and Devyani Sharma,
for their continued support during the preparation of this manuscript.
References
Alim, Samy. 2004. You Know My Steez: An Ethnographic and Sociolinguistic Study of Styleshifting
in a Black American Speech Community. Durham NC: Duke University Press.
Ball, Martin J. & Müller, Nicole. 2005. Phonetics for Communication Disorders. Mahwah NJ:
Lawrence Erlbaum Associates.
Becker, Kara. Forthcoming. Ethnolect, dialect, and linguistic repertoire in New York City. In
New Perspectives on the Concept of Ethnolect, Malcah Yaeger-Dror & Lauren Hall-Lew
(eds).
Benor, Sarah B. 2010. Ethnolinguistic repertoire: Shifting the analytic focus in language and
Ethnicity. Journal of Sociolinguistics 14: 159–183. DOI: 10.1111/j.1467-9841.2010.00440.x
Bhatia, Tej. K. 1988. Trinidad Hindi: Its genesis and generational profile. In Language Trans-
planted: The Development of Overseas Hindi, Ricardo K. Barz & Jeff Siegel (eds), 179–196.
Wiesbaden: Otto Harrassowitz.
Boersma, Paul & Weenink, David. 2013. Praat: Doing phonetics by computer [Computer pro-
gram, version 5.3.42]. <http://www.praat.org/> (6 March 2011).
Central Intelligence Agency. 2013. Trinidad and Tobago. In The World Factbook. <https://www.
cia.gov/library/publications/the-world-factbook/> (25 April 2011).
Clyne, Michael. 2000. Lingua Franca and ethnolects in Europe and beyond. Sociolinguistica 14:
83–89.
De Silva, Jennifer. 2012. Trinidadian Bhojpuri: Then and now. Paper presented at the Society
for Caribbean Linguistics 19th Biennial Conference. Nassau, The Bahamas. August 1, 2012.
Drayton, Kathy-Ann. 2006. Word level prosody in Trinidadian English Creole: A phonetic
analysis. Seminar paper, University of the West Indies: St. Augustine, Trinidad.
Eckert, Penelope. 2008. Where do ethnolects stop? International Journal of Bilingualism 12:
25–42. DOI: 10.1177/13670069080120010301
Gooden, Shelome, Drayton, Kathy-Ann & Beckman, Mary. 2009. Tone inventories and tune-
text alignments: Prosodic variation in ‘hybrid’ prosodic systems. Studies in Language 33:
396–436. DOI: 10.1075/sl.33.2.07goo
Gordon, Matthew & Ladefoged, Peter. 2001. Phonation types: A cross-linguistic overview. Jour-
nal of Phonetics 29: 383–406. DOI: 10.1006/jpho.2001.0147
Gut, Ulrike. 2009. Introduction to English Phonetics and Phonology. Frankfurt: Peter Lang.
Haraksingh, Kusha. 2006. Trinidad and Tobago. In The Encyclopedia of the Indian Diaspora, Brij
V. Lal (ed.), 278–286. Honolulu HI: University of Hawai’i Press.
Laver, John. 1994. Principles of Phonetics. Cambridge: CUP. DOI: 10.1017/CBO9781139166621
Leung, Glenda-Alicia. 2013. A Synchronic Sociophonetic Study of Monophthongs in Trinidad-
ian English. PhD dissertation, University of Freiburg. <http://www.freidok.uni-freiburg.
de/volltexte/9015/> (6 March 2013).
Lewis, M. Paul (ed).2009. Ethnologue: Languages of the World, 16th edn. Dallas TX: SIL Inter-
national. <http://www.ethnologue.com/> (20 June 2011).
Mendoza‐Denton, Norma. 2011. The Semiotic Hitchhiker’s Guide to Creaky Voice: Circulation
and gendered hardcore in a Chicana/o gang persona. Journal of Linguistic Anthropology 21:
261–280. DOI: 10.1111/j.1548-1395.2011.01110.x
Mohan, Peggy. 1990. The rise and fall of Trinidad Bhojpuri. International Journal of the Sociol-
ogy of Language 85: 21–30.
Mohan, Peggy & Zador, Paul. 1986. Discontinuity in a life cycle: The death of Trinidad Bhojpu-
ri. Language 62: 291–319. DOI: 10.2307/414675
Mühleisen, Susanne. 2001. Is ‘bad English’ dying out? A diachronic comparative study of at-
titudes towards creole versus standard English in Trinidad. PhiN15: 43–78. <http://web.
fu-berlin.de/phin/phin15/p15t3.htm/> (6 March 2011).
Newman, Michael & Wu, Angela. 2011. ‘Do you sound Asian when you speak English?’ Racial
identification and voice in Chinese and Korean Americans’ English. American Speech 86:
152–178. DOI: 10.1215/00031283-1336992
Podesva, Robert J. 2013. Gender and the social meaning of non-modal phonation types. In Pro-
ceedings of the 37th Annual Meeting of the Berkeley Linguistics Society, Chundra Cathcart,
I-Hsuan Chen, Greg Finley, Shinae Kang, Clare S. Sandy & Elise Stickles (eds), 427–448.
<http://linguistics.berkeley.edu/bls/past_meetings/bls37/Podes-va.pdf> (14 March 2013).
Rahman, Jacquelyn. 2008. Middle-class African Americans: Reactions and attitudes toward Af-
rican American English. American Speech 83: 141–176. DOI: 10.1215/00031283-2008-009
Regis, Ferne Louanne. 2012. Constructing identity within communities of practice: The case of
two Trinidadian Douglas. SCL Occasional Papers OP 36.
Samaroo, Brinsley. 1975. The Presbyterian Canadian mission as an agent of integration in Trini-
dad during the Nineteenth and early Twentieth Centuries. Caribbean Studies 14(4): 41–55.
Sattar, Ansar. 2009. AC past winners. Animae Caribe. <http://www.animaecaribe.com/
archived.php?mid=15/> (6 March 2011).
Sharma, Devyani. 2011. Style repertoire and social change in British Asian English. Journal of
Sociolinguistics 15: 464–492. DOI: 10.1111/j.1467-9841.2011.00503.x
Shepherd, Verene A. 2006. Jamaica. In The Encyclopedia of the Indian Diaspora, Brij V. Lal (ed.),
306–311. Honolulu HI: University of Hawai’i Press.
Sicoli, Mark. 2010. Shifting voices with participant roles: Voice qualities and speech registers in
Mesoamerica. Language in Society 39: 521–553. DOI: 10.1017/S0047404510000436
Stuart-Smith, Jane. 1999. Voice quality in Glaswegian. Proceedings of the International Congress
of Phonetic Sciences 14: 2553–2556.
Szakay, Anita. 2012. Voice quality as a marker of ethnicity in New Zealand: From acoustics to
perception. Journal of Sociolinguistics 16: 382–397. DOI: 10.1111/j.1467-9841.2012.00537.x
Thomas, Erik R. & Reaser, Jeffrey. 2004. Delimiting perceptual cues used for the ethnic labeling
of African American and European American voices. Journal of Sociolinguistics 8: 54–87.
DOI: 10.1111/j.1467-9841.2004.00251.x
Winer, Lise. 2007. Indic lexicon in the English/Creole of Trinidad. In Badjohns, Bhaaji and
Banknote Blue: Essays on the Social History of Language in Trinidad and Tobago, Lise Winer
(ed.), 348–377. St. Augustine: UWI School of Continuing Studies.
Winer, Lise. 2008. Dictionary of the English/Creole of Trinidad and Tobago: On Historical Prin-
ciples. Montreal: McGill-Queen’s University Press.
Winford, Donald. 1978. Phonological hypercorrection the process of decreolization – The case of
Trinidadian English. Journal of Linguistics 14: 277–291. DOI: 10.1017/S0022226700005909
Yakpo, Kofi. 2012. Bhojpuri in contact in Suriname and Mauritius. Invited talk, Centre for Mau-
ritian Studies, Mahatma Gandhi Institute, Moka, Mauritius. May 29 2012.
Yuasa, Ikuko Patricia. 2010. Creaky voice: A new feminine voice quality for young urban-ori-
ented upwardly mobile American women? American Speech 85: 315–337.
DOI: 10.1215/00031283-2010-018
Appendix
Section 1
CLIP 1 of 16
1. Identify the ethnicity of the speaker:

☐ Indo-Trinidadian
☐ Afro-Trinidadian
2. How sure are you?

☐ extremely unsure ☐ unsure ☐ fairly sure ☐ sure ☐ extremely sure
Section 2
1. What cues/clues did you use to identify the speaker’s ethnicity? Please note that you are
asked to consider:
a. Afro-Trinidadian females vs. Indo Trinidadian females.
b. Afro-Trinidadian males vs. Indo Trinidadian males.
I used these cues/clues to distinguish I used these cues/clues to distinguish

Afro-Trinidadian voices: Indo-Trinidadian voices:
AFRO-Trini FEMALE INDO-Trini FEMALE
 
AFRO-Trini MALE INDO-Trini MALE

 
Chapter 3
Identity, ethnicity and fine phonetic detail

An acoustic phonetic analysis of syllable-initial
/t/ in Glaswegian girls of Pakistani heritage*
Farhana Alam and Jane Stuart-Smith

University of Glasgow
English in the Indian diaspora in Britain is of much interest, both in the emer-
gence of regional ethnic Englishes (Heselwood & McChrystal 2000) and in the
potential links with language change in mainstream English (Kerswill et al.
2008). Also intriguing are the processes by which features percolate through
generations of speakers (Sharma & Sankaran 2011). At the phonological level,
subtle differences in phonetic characteristics may index locally-situated social/
ethnic identities in second and further generations (Stuart-Smith et al. 2011).
Lawson et al. (2011) show that the auditory-articulatory relationship is com-
plex, with very fine-grained phonetic characteristics potentially having salience
for a community (Docherty & Foulkes 1999). We present results of an acoustic
phonetic analysis of syllable-initial /t/ in adolescent girls of Pakistani heritage
in Glasgow. Speech data were drawn from a long-term ethnography in an in-
ner-city secondary school. Several Communities of Practice were identified
(cf. Eckert 2000), spanning a continuum from British/western to traditionally
Pakistani and/or Muslim practices. Spectral analysis of the stop bursts of /t/ re-
vealed not only clear patterning according to social practices but also gradience
for individuals within Communities of Practice. This suggests the emergence of
a local ethnic accent with subtle adaptation of heritage features reflecting new
emerging identities (cf. Harris 2006).
Keywords: sociophonetics, identity, ethnicity, consonants, Scottish
* The research reported here is being supported by an ESRC PhD studentship to Farhana
Alam, and a British Academy Small Grant (SG48480: Investigating ethnic accents) to Jane
Stuart-Smith.
30 Farhana Alam and Jane Stuart-Smith
1. Introduction
In England, Wales, and Northern Ireland, 14% of the population were reported
as belonging to a minority ethnic group (other than ‘White’) in the 2011 Census,
with most resident in urban conurbations. The British Asian community – con-
sisting of a range of intersecting diasporic communities from across the Indian
subcontinent, many of whom speak major Modern Indo-Aryan languages (e.g.
Punjabi, Urdu, Bengali) – represents the largest reported minority ethnic group
at 6.8% of the total population in England and Wales. Studies on regional British
Asian accents in England (e.g. Heselwood & McChrystal 2000; Hirson & Sohail
2007; cf. Kirkham 2011) have all shown links between ethnicity and accent, bar
Evans et al. (2007), whose study of Gujarati was carried out in West London. Eng-
lish in the Indian diaspora in the UK also shows intriguing relationships with var-
iation and change in mainstream Englishes (e.g. Kerswill et al. 2008). And closer
investigation reveals some of the mechanisms by which phonological features
may percolate within and through generations of speakers (Sharma & Sankaran
2011; see also Sharma, this volume).
In Scotland, the minority ethnic population is much smaller, reported at only
just under 4% of the whole population in the 2011 Census, but the main ethnic
group, largely located in the country’s largest city, Glasgow, is also part of the Indi-
an diaspora. 58% of the Glasgow South-Asian community is of Pakistani heritage,
and Islam is the main religion. Ethnicity intersects in complex ways with identity
for members of this community because not only are local Scottish, Asian, and
Scottish-Asian identities involved, but religion is also a factor, particularly in the
context of post-9/11 British society.
Previous auditory and acoustic analyses of Glasgow-Asian accents in second
generation speakers have shown show clearer realizations of /l/ and closer allo-
phones of the GOAT and FACE vowels than Glasgow non-Asian speakers (e.g.
Lambert et al. 2007; Stuart-Smith et al. 2011). They also show that fine phonet-
ic variation relates not just to ethnicity but specifically to local ethnic identities.
Features of phonetic interference in the first generation appear with subtly dif-
ferent phonetic characteristics indexing specific identities in second and further
generations (Stuart-Smith et al. 2011). These findings add to those of recent so-
ciophonetic research, which shows that fine phonetic variation can be systemat-
ically related not only to macro-factors such as age, gender, and social class, but
also to the construction of locally-salient social identities, including those which
relate to ethnicity (e.g. Foulkes & Docherty 2006; Khan, e.g. 2009; Docherty &
Mendoza-Denton 2012). Such studies are theoretically important because they
point to the need for models of phonological knowledge that explicitly include
social-indexical information (e.g. Foulkes 2010). However the empirical base for
Chapter 3.â•‡ Identity, ethnicity and fine phonetic detail 31
their development is still rather thin, particularly with respect to speech and the
construction of locally-salient ethnic identities.
In this paper we present the results of an acoustic phonetic analysis of syl-
lable-initial /t/ in adolescent girls of Pakistani heritage in Glasgow. Speech data
were drawn from a long-term ethnography of 97 Scottish-Asian adolescents in
an inner-city secondary school. Several Communities of Practice were identified
(cf. Eckert 2000), spanning a continuum from British/western to traditionally
Pakistani and/or Muslim practices. A spectral analysis of the stop bursts of /t/
reveals not only clear patterning according to social practices but also gradience
for individuals within Communities of Practice. This provides more evidence to
suggest the emergence of a local ethnic accent with the subtle adaptation of herit-
age features reflecting local “Glasgow-Asian”, or perhaps better new hybrid, “Glas-
wasian”, identities (cf. Harris 2006).
1.1 The Glasgow Asian community
1.1.1 Historical background

The Indian diaspora to the UK began in the fifteenth century in the form of trav-
elling seamen, servants, and peddlers, but the real expansion of a South-Asian
presence took place in the 1950s and 1960s when colonial ties catalyzed migration
with the need for industrial workers after WWII. Migration was also spurred on
by the communal turmoil which continued after the independence of India and
Pakistan in 1947, where poverty and high unemployment necessitated moving
abroad for better life prospects. The primary settlements were in large industrial
areas such as London, Birmingham, Manchester, and West Yorkshire, where mi-
grants often took low-paid factory jobs.
After a scattering of settlers in the late nineteenth century and before (Maan
1992), migration of South-Asians to Scotland generally happened as a second
wave, as workers in England were laid off and travelled north to seek new em-
ployment. Migrants formed dense multiplex social networks, especially in the
major cities of the Central Belt, Glasgow, and Edinburgh. As relatively large num-
bers moved to the same area, small communities formed, located in areas such as
Govanhill, Pollokshields, and Pollokshaws in Glasgow, with their own infrastruc-
ture catering for the specific needs of the community, offering halal meat shops
and local media (e.g. Radio Awaz, Radio Ramadan) and newspapers (e.g. the
Awaz newspaper, The Friday People community newspaper), as well as mosques,
since the majority of the new community were Muslim, alongside smaller Sikh
and Hindu Indian communities in Glasgow. The primary regions in South Asia
that formed the source of the Scottish Asian community are in Pakistan, with the
UZB. TAJIKISTAN
TURKMENISTAN CHINA
Gilgit-Baltistan
Khyber
Pakhtunkhwa
Islamabad
AFGHANISTAN Tribal Areas
Azad
Kashmir
Punjab
Balochistan
IRAN INDIA
Sindh
200 km
100 km
Figure 3.1â•‡ Map of Pakistan showing main areas of provenance of the Scottish Asian
community
majority from the Punjab area, surrounding the cities of Lahore and Faisalabad,
but also some from the area around Mirpur, the largest city in Azad Kashmir
towards the north of Pakistan (which is still under dispute in terms of rule) (see
Figure 3.1).
The Scottish Pakistani community still exhibits strong transnational ties with
communities in Pakistan. Many community members make frequent visits back
to Pakistan, typically for marriage, assessing wealth and/or assets (e.g. collect-
ing rental income from land or property owned in Pakistan), family weddings
and funerals. Many first-generation parents (and grandparents) also take their
children back to their homeland during school holidays to learn more about their
heritage and develop relationships with their Pakistani relatives. At the same time,
the political turmoil in Pakistan, particularly since the military government was
overthrown in 2008, has resulted in some decrease in visits to the country. Fur-
thermore, in recent years it seems that fewer second and third generation Scottish
Asians visit Pakistan for holidays and would rather go abroad elsewhere. Popular
destinations are countries like Turkey, Morocco, and Egypt, where it is still pos-
sible to experience Islamic culture but without the cultural and familial responsi-
bilities of visiting relatives in Pakistan (cf. Shaw 1994).
1.1.2 Demography
The demographics of the communities in the Indian diaspora are not evenly dis-
tributed across the UK. As noted at the outset, there is a large disparity in general
in the size of the non-White minority ethnic populations in England and Wales
(14% of the overall population in 2011) and Scotland (just under 4% in 2011).
In England and Wales in 2011, around 6.8% of the non-White minority ethnic
population are reported as originating from South Asia (Indian 2.5%, Pakistani
2%, Other Asian 1.5%, Bangladeshi 0.8%). However, in Scotland, the South-Asian
diaspora formed only 2.7% of the total population with the predominantly Mus-
lim Pakistani community forming the largest proportion (Pakistani 0.9%, Indian
0.6%, Other Asian 0.4%, Bangladeshi 0.07%). Even though the Pakistani commu-
nity in Scotland is much smaller demographically, it is proportionately larger than
its English counterparts.
1.1.3 Sociopolitical situation

It is difficult to talk about a unified British Pakistani Muslim community in terms
of belief system, since there are sectarian and ideological divisions, e.g. Sunni,
Shia, Salafi, Ahmadiyya. The majority of the Scottish Pakistani Muslim com-
munity are Sunni, though there are many offshoots (e.g. Barelvi and Deobandi).
Despite these differences, all these communities have shared isolation from the
mainstream caused by widespread Islamophobia and xenophobia, particularly in
the media, since 9/11 and 7/7. The British Muslim community is experiencing
heightened challenges to their identity, both outwardly from the perceived hostile
environment, but also inwardly in the options they choose to project their iden-
tity, e.g. assimilation, integration or independence from mainstream British so-
ciety. There has been a range of responses to national Islamophobic views among
young British Muslim Pakistanis, from a degree of Islamic revival in younger gen-
erations to counteract gross categorization, to complete abandonment of heritage
in order to distance oneself from any stereotypes; others appear to sit on the fence
and pick and choose elements from both cultures and traditions as required.
More generally, Ballard (1994, 31) observes that identities and cultures, like
languages, are codes, which are totally context-dependent. A new generation (or
even members within a generation) can shift identities depending on domain. For
example, they may adopt certain behaviours in some religio-cultural situations
and other behaviours in Western situations – thus constantly reinterpreting and
re-evaulating their daily lives – in a way unknown to their predecessors who were
strongly linked to their single heritage background. This integration, or rather
daily negotiation of coexisting identities, is captured in Harris’ (2006) discussion
of young British Asians in London. Drawing on Hall’s (1992) notion of “cultures
of hybridity”, Harris (2006, 1–2) proposes: “Brasian … [suggesting] a continu-
ous flow of everyday life and cultural practices in which, at any given moment,
both British and particular South Asian derived elements are always co-present”.
Alam (2007) had made a similar, independent, observation for the linguistic and
social practices of her Glasgow-Asian high school girls – more of whose data are
presented here. Alam also preferred a blended term, “Glaswasian” as opposed to
“Glaswegian Asian” to express the co-present identities of both being of the city of
Glasgow and at the same time enjoying their Pakistani Muslim heritage.
1.1.4 Linguistic context

The Scottish-Pakistani community brought with it a complex and already diglos-
sic language situation from Pakistan, with Urdu as the High variety and Pun-
jabi, and some other local languages and/or dialects (such as Mirpuri) as Low
varieties (see Masica 1993); classical Arabic functions as another, highly restrict-
ed, High variety, used for Islamic religious practices, such as reciting the Holy
Qur’an or performing congregational prayers. The micro-economy created by the
Scottish-Pakistani community has acted as a language maintenance mechanism
by creating an environment where it is more natural to use the community lan-
guages. Despite this, language loss is apparent in 2nd, 3rd, and later generations,
especially outside the home domain (Fishman, e.g. 1991). No formal statistics
exist to document the progress of language loss, but an illustration is provided by
reports on heritage language use by members of this sample. Nearly all of the 97
Asian high school children reported that they spoke English almost exclusively
outside the home, using Punjabi or Urdu only for emphasis, mockery or mimicry
outside the home domain. Their reports are confirmed by their usage during the
interviews, with only very occasional instances of code-switching. Code-switch-
ing itself is used as a way of expressing humour in an “English” conversation,
with some stereotyping of accents and mocking of older generations. In addition,
Punjabi/Urdu words, phrases, or idioms may be used when an English equiva-
lent does not exist (e.g. Urdu mamu for ‘maternal uncle’) or to refer to typical,
culturally specific domestic activities (e.g. roti ‘unleavened Asian bread’ or salan
‘curry’). During the interviews there was very limited use of code-switching, and
only largely when the interviewer herself code-switched, which then allowed the
students to speak the heritage language without embarrassment. In general they
were reluctant to be deemed as “TeePees” (short for the phrase ‘Typical Pakistanis’
which has negative connotations associated with people born in Pakistan). Many
thought they were too “cool” to speak the heritage language which might work
against an urban/modern Scottish identity.
In general, Urdu and Punjabi are only used to speak to older relatives who
perhaps cannot speak English and also to older family members as a sign of re-
spect. Between peers, English is the dominant language. There is some attempt
to retain the heritage language by younger generations through teaching their
children, but this tends to be very limited and is not considered very important, in
comparison with the aspirations to achieve well in the Scottish education system
and other extra-curricular pursuits. As a result these 2nd+ generation members of
the community exhibit transitional bilingualism and reduced proficiency in Pun-
jabi and Urdu, accompanied by borrowing of English loanwords and reduction
of the Punjabi and Urdu morphology system. Any attempts at formal language
maintenance (e.g. Saturday morning classes, GCSEs and A-Levels) are focused on
the High variety, Urdu, rather than Punjabi with erosion well underway.
Alongside usual community language use, both British Asians and the major-
ity population recognize particular enregistered varieties (Agha 2003) such as the
form of stylized Indian English represented in the TV comedy, Goodness Gracious
Me, or the more recent TV family sitcom, Citizen Khan (cf. Citizen Kane), set
in Birmingham and drawing on very stereotypical portrayals of the first genera-
tion. Such a style is usually relegated to first generation speakers and is a source
of amusement for younger generations. In Scotland, enregistered Scottish-Asian,
and even Glasgow-Asian, styles also exist, with typically more syllable-timed
rhythm and exaggerated use of a retroflex articulatory setting. The shopkeeper
“Navid”, in the BBC Scotland comedy Still Game, played by a local Asian celebrity,
Sanjeev Kohli, is a good example of a well-known local Asian stereotype. The per-
ceptual salience of colloquial regional British-Asian accents is far less marked but
is still present. Both Heselwood & McChrystal (2000) for Bradford, and Lambert
et al. (2007) for Glasgow, found accurate discrimination of Asian versus non-
Asian accents, even when the contributing phonetic features were rather subtle.
1.2 The realization of /t/ in Glasgow Asian
Using retroflex stops for the voiceless, and voiced, coronal plosives, /t/ and /d/, is
stereotypical of stylized Indian-English accents and is also found in many first-
generation British-Asian speakers who were born outside the UK and who are
also proficient speakers of a modern Indo-Aryan language such as Punjabi or
Urdu, which have retroflex consonants (Bhatia 1993; Bhardwaj 1995; see also
Sharma & Sankaran 2011). Second-generation British-Asian speakers may also
show auditorily retracted realisations of /t/ (Heselwood & McChrystal 2000),
though these usually sound more postalveolar than retroflex (e.g. Kirkham 2011).
In Glasgow, auditory analysis of /t/ in 2nd+ generation Asian speakers
showed retracted, postalveolar stops (including some ejectives, Lambert et al.
2007). Glasgow-Â�English typically shows fronted, denti-alveolar, or even dental
allophones for this stop (Stuart-Smith 1999). Auditorily identified postalveolar
stops also contributed 54% of the overall variation for /t/ in the speech of the
Glasgow Asian girls examined by Alam (2007) with significant differences ac-
cording to the social, cultural, and religious practices in which they were engaged.
In this paper we return to Glasgow Asian, and in particular to the extended
ethnographic study of Pakistani Muslim heritage high school girls carried out by
the first author. Here we increase our resolution of the phonetic detail by using an
acoustic phonetic analysis to consider the realization of syllable-initial /t/.
Our main research questions are:
1. What are the acoustic characteristics of /t/ for these speakers?

2. Does fine phonetic variation pattern according to locally-salient ethnic
identities?
3. How does such variation intersect with that related to phonetic constraints?
But before we try to answer these questions, we consider aspects of methodology.

First we outline the data collection and speaker sample. Second we reflect briefly
on the phonetic analysis itself.
2. Methodology
2.1 Data collection
The wider study from which these results are drawn is based in the sociolinguistic
Communities of Practice (CoP) framework (e.g. Eckert 2000). This approach uses
long-term ethnographic methods to access members of communities, observe
their shared developing social practices, and gradually to discern the Communi-
ties of Practice to which members belong. It offers an excellent way of investigat-
ing the kind of subtle social differentiation which needs to be identified if we are
to consider the development of specific patterns of speech variation in conjunc-
tion with local ethnic identities.
The first part of the study from which these data are drawn involved a long-
term, ethnographic fieldwork in an inner-city high school in Glasgow. Riverburn
High (pseudonym) has the highest proportion of ethnic minority pupils in the
city (45%), with Asian Pakistani children making up 30% of this (“Ethnic Figures
for Secondary Schools 2005–2006”, Glasgow City Council Pupil Census). The first
author, who is herself a female member of the Glasgow Pakistani Muslim com-
munity, carried out participant observation over three years, by hanging out with
students during lunchtimes, breaks and free periods, and by attending school
social events. She soon felt herself to be part of the school environment as she
was often mistaken for a pupil and reprimanded by teachers in corridors, giving
credence to the fact she was not a figure of institutional authority and someone
the pupils could trust. Various pupil hangouts emerged, often highlighting key
identity differences within the ethnic community, from Indian takeaways to deli
sandwich shops. Extensive field-notes on the interactions were made allowing for
general observations and the identification of a series of Communities of Practice
existing for the Asian students, especially the girls, who are the focus of this study.
2.2 Communities of Practice in Riverburn High, Glasgow
Even before 9/11, young female British Asians encountered particular challenges
of plural ethnic identities, neatly summarized in a line from the hit film, Bend
it like Beckham: “Who wants to cook aloo gobi when you can bend a ball like
Beckham?” In the aftermath of 9/11, the tensions between adhering to traditional
cultural values, or moving to more westernized practices, gained the now height-
ened factor of religion.
The qualitative sociolinguistic analysis of the school-based ethnography re-
vealed sharp (informal) social segregation between Asian and non-Asian stu-
dents. A total number of 117 students/informants aged between 14 and 18 years
old participated in the fieldwork, 97 Asians and 20 non-Asians. Digital recordings
of speakers were made during informal interviews between the researcher and
pairs or triads of speakers, using AT831b microphones and a flashcard Micro-
track recorder with a sampling rate of 44,100 Hz/16 bit. Approximately 60 hours
of recordings were made over the three years of ethnography.
Table 3.1 shows the distribution of the Asian students according to Commu-
nity of Practice. Within the Asian girls the Communities of Practice that were
salient are shown in Figure 3.2, which gives a schematic representation. The ovals
represent the CoPs, with size reflecting the proportion of female participants in
each. The overlapping areas represent some fluidity and similarity in social prac-
tices, and the vertical arrow shows the social space through which the Shifters
moved. The grey-filled ovals indicate the three CoPs represented here.
Table 3.1â•‡ Numbers of Asian students who participated in the ethnography at Riverburn
High, Glasgow, according to Community of Practice
Communities of Practice for Asian students Girls Boys
Wannabes â•⁄7 â•⁄3
Messabouts 14 10
Moderns 10 â•⁄3
Religionistas â•⁄6 â•⁄0
Shifters 25 11
Conservatives â•⁄8 â•⁄0
Totals 70 27
Wannabes
Messabouts
Moderns
Religionista Shifters
Conservatives
Figure 3.2â•‡ Schematic representation of relevant Communities of Practice for Glasgow

Pakistani high school girls. The grey-filled ovals indicate the three CoPs represented in
this study
Membership of the different Communities of Practice related to girls’ affilia-

tion to western ideologies as opposed to eastern ideologies, spanning a multidi-
mensional continuum from very traditionally Pakistani and/or Muslim, to very
British/western, values and practices.
In this paper we present results from six 17–18 year-old girls, selected on the
basis of their participation as core members of three Communities of Practice
(pseudonyms are given for the girls):
1. Conservatives (Inaya, Aneela): These girls strongly adhere to traditional Pa-

kistani cultural values and practices, favouring earlier marriage without pre-
vious relationships and staying at home to raise children, though still seek
education. They wear plain clothes and the hijab (headscarf), with little or no
make-up or jewellery.
2. Moderns (Zahida, Huma): The Modern girls wear trendy, fashionable cloth-
ing, with visible make-up and generally no headscarf. They express aspira-
tions for continuing in further education and holding careers. They want to
date and talk about boys – they are relatively daring but do not act outside of
the bounds of accepted community norms.
3. Messabouts (Asma, Naazi): Visibly similar in many ways to the Moderns,
the Messabouts are different in that some of their practices (e.g. drinking,
smoking, drugs, swearing) are not acceptable to the community and could
lead to serious consequences. Some even wear the headscarf but often this is
imposed and not through personal choice, which may explain their rebellious
practices in the school domain – often split personalities. Nevertheless their
socializing is still restricted to within the Glasgow Asian community, mak-
ing them different again to the Wannabes (who effectively align with White
students).
Auditory impressions of the realization of /t/ in these speakers were of fine-

grained, complex articulatory differences along a kind of continuum which can
be rather coarsely captured in two parameters, front-/backness of the articulation,
ranging from dental to postalveolar, and variation in the active articulator, or con-
figuration of the tongue tip or blade. Conservatives sounded more apico-postal-
veolar (auditorily retracted but also more tip), whilst Messabouts’ variants seemed
more laminal-dental (auditorily fronted but also more blade). Genuine retroflex
stops were only identified in the handful of Punjabi code-switches.
2.3 Phonetic analysis of sociophonetic variation
Trying to pin down precise articulations using even the finest grained traditional
auditory phonetic transcription was very difficult for these highly subtle differ-
ences (very similar problems occurred when trying to deal with sociophonetic
variation in /s/ in Glasgow, Stuart-Smith 2007). This was, on the one hand, a mo-
tivation for using an acoustic phonetic analysis. But it also led us to question more
broadly how we might best describe and identify the patterns underlying this
kind of sociophonetic variation, and more fundamentally what different kinds of
phonetic analysis – auditory, acoustic and articulatory – are able to show.
Auditory phonetic analysis is typically used for sociolinguistic analysis of
phonetic/phonological variation. This involves the representation of phonetic
variation using the symbols of the International Phonetic Alphabet (IPA), with
more (narrow) or less (broad) detail represented as the analyst decides (Ogden
2009). Auditory transcription requires the analyst to categorize the auditory con-
tinuum of variation in “articulatory terms” (i.e. the analyst’s kinaesthetic interpre-
tation of the kinds of articulatory strategy that the speaker might be thought to
be using; Catford 2001). The result of this kind of analysis, even at the fine level, is
fairly gross, discrete categories which make strong assumptions about the articu-
latory gestures underlying auditory objects, for example voiceless retroflex plosive
[ʈ], retracted voiceless alveolar plosive [t‒], and so on. Despite the IPA symbols and
their labels, the result is auditory, not articulatory, objects. Furthermore, the fine-
grained variation which is pertinent to the community may not be easily audible
and/or discriminable, even to trained phoneticians (Docherty & Foulkes 1999).
At the same time, numerous sociolinguistic studies have repeatedly shown statis-
tically robust sociolinguistic patterning of auditorily-identified units with social
factors, including ethnicity (e.g. Labov 2001; Mendoza-Denton 2008).
A stock response to the apparent shortcomings of auditory transcription is
to invoke the scientific superiority of acoustic phonetic analysis. Acoustic rep-
resentations of speech show chunks of speech in terms of amplitude and fre-
quency over time, e.g. waveforms, spectrograms, and spectra. Acoustic analysis
is persuasive because it allows the analyst to view speech in particular ways, and
it facilitates the objective quantification of aspects of the speech signal, in terms
of e.g. duration and/or frequency. The result is continuous acoustic measures for
specific aspects of the signal specified by the analyst. The advantages are that such
data permit robust statistical analysis, and the data are fine-grained and contin-
uous – the output of the observed measures have not been placed into analyti-
cal categories. At the same time, acoustic analysis yields acoustic objects, whose
relationship both to articulation and to perception must be inferred and/or ex-
trapolated (e.g. Johnson 2011). Such relationships, particularly between acoustic
and articulatory representations, are less straightforward than they might appear.
Models such as source-filter theory make necessarily simplistic assumptions
about cavities and their resonances, whereas actual articulations entail complex
variable filters. Acoustic measures are also partial, taking only a few dimensions.
So a form of categorization is imposed by the analyst’s choice of which measure
to use to represent a phenomenon, even when other characteristics may not be
captured. For example, the classic representation of vowels using the first two
formants cannot at the same time include information about lip-rounding, which
is captured by the third formant. Acoustic analysis is also rather more subjective
than it might appear, since it is strongly determined by the analyst’s decisions
about how and where to take measures (Ogden 2009).
Perhaps an answer is provided by articulatory phonetic analysis. Different
techniques provide visual, dynamic representations of some aspect of speech
articulation, e.g. tongue-palate contact (electropalatography) or tongue surface
configuration (Ultrasound Tongue Imaging). Despite anxieties regarding socio-

linguistic naturalness, in fact such visible equipment has not been found to affect
speech style (e.g. for UTI, see Lawson et al. 2008). Quantification of aspects of
articulatory representations is possible for some techniques and under develop-
ment for others. The result is dynamic and static representations of articulatory
configurations that otherwise might be inferred (from auditory and/or acoustic
analyses). At the same time, articulatory phonetic analysis does require some ex-
pertise to implement data capture and then to interpret the data; quantification
for some techniques is not straightforward. But more importantly, whilst articu-
latory representations may appear “direct”, they are in fact removed from actual
articulation, being both partial and processed (e.g. most ultrasound films are very
slow at 30 Hz, may not show all of the tongue, and/or tongue tip, and only present
a “sketch” of tongue movement).
These three different kinds of representation of speech variation share the
characteristics of each offering only partial sketches of the data and also of being
connected to each other in complex ways. Ideally any sociophonetic description
would be based on data drawn from all three points of the “speech data triangle”.
Take, for example, a recent analysis of derhoticisation of postvocalic /r/ in Scot-
tish English, in e.g. car, card (Stuart-Smith et al. 2014). Auditory analysis is ex-
tremely difficult, yielding different categorization according to different analysts
and fuzzy/inbetween categories. Acoustic analysis shows ambiguous patterns of
weakened energy exactly where formant transitions are expected. And articulato-
ry analysis shows how tongue gestures are present, but delayed so that they coin-
cide with the end of voicing. It is clear that each kind of phonetic analysis yields
complementary objects, each valid and yet distinct.
These reflections provide a backdrop to our choice of methodology, and es-
pecially the results and their interpretation. Here we chose to attempt to capture
fine-grained phonetic variation in the realization of /t/ which might be carrying
social meaning for these speakers by using acoustic (spectral) analysis of the stop
burst because this: (a) yields continuous measures; (b) provides several acoustic
indices to capture (a few) aspects of the spectrum (a slice taken at the stop burst);
and (c) avoids having to assign variants to auditory (i.e. quasi-articulatory) cate-
gories when the auditory differences between “fronted” and “retracted” variants
of /t/ are likely to relate as much to the overall shape of the constriction, captured
grossly in terms of “apical” – tip or “laminal” – blade, as to any differences in
“front-back” dimension of tongue tip/blade. At the same time, we note that using
this form of analysis can only provide a partial sketch of this aspect of these girls’
speech.
2.4 Acoustic phonetic analysis
The speech recordings were uploaded, with full orthographic transcripts, to the
database/search engine ONZE Miner (now LaBB-CAT; Fromont & Hay 2008).
All possible tokens of syllable-initial /t/ were extracted into Praat, and problem-
atic tokens were removed. The sound files were high-pass filtered at 450 Hz and
low-pass filtered at 12000 Hz to remove extraneous low-frequency energy and to
prevent aliasing, or the appearance of false peaks in the spectrum. FFT spectra
were taken using a 10 ms Hamming window manually centred on the stop burst.
We considered different aspects of the spectral energy of the stop bursts
using a spectral moment or “centre of gravity” analysis (Forrest et al. 1988; cf.
Harrington 2010), which models the spectrum as a “single normal distribution
which may reflect the dominant front cavity formant” (Wrench 1995, 460); in
other words, this is like thinking of a spectrum as having a main peak (area) of
acoustic energy, with particular dimensions (a bit like a mountain rising out of
a mountain range). Characteristics of the overall distribution of spectral energy
are captured using four measures: the mean is the midpoint frequency at which
the energy under the curve on either side of the point is equal; the spread is the
“bandwidth” of the energy either side of the mean; the skew refers to the asym-
metry of the energy surrounding the mean; and kurtosis to the peakiness of the
distribution. The measures were taken using Praat algorithms to calculate: “centre
of gravity”, “standard deviation”, “skew” and “kurtosis”.
It is possible to make simple predictions about frequency of the mean and
size of the front cavity immediately in front of the articulatory constriction, such
that a longer front cavity will show a lower mean, and a shorter front cavity a
higher one. But it is important to note that the measures are also determined by
both the overall shape of the articulatory front cavity and also by the shape of
the constriction itself. So, for example, what determines the patterning of spec-
tral energy for different variants of /t/ will not only be the distance between the
end of the constriction created by the tip or blade of the tongue and the lips (the
fronting/dental or backing/alveolar of place of articulation), but also by the kind
of front cavity that is created, and by the different dimensions and shape of the
constriction formed by the tip or the blade of the tongue. Thus, whilst we might
expect a shift from a postalveolar to an alveolar articulation to raise the mean as
the distance between the tip/blade and teeth is reduced, further fronting and/or
dentalisation of a /t/ does not simply shorten the front cavity and increase the
mean even more. Rather it can result in the near-absence of the front cavity and
substantial difference in the shape of the entire constriction. This leads to an over-
all shift in the shape of the acoustic spectrum – dental articulations typically show
spectra which are rather flat and lack any kind of overall peak.
More generally, mapping these measures onto notions of particular articula-

tions without instrumental phonetic articulatory data, such as electropalatogra-
phy, is not straightforward (cf. Sundara 2005, 1035). At the same time, moment
analysis has been shown to provide a good acoustic characterisation of obstruent
articulations (Forrest et al. 1988; Ladefoged 2003) and even to be able to distin-
guish between relatively fine differences in place of articulation, e.g. between al-
veolar and dental stops in Canadian English/French bilingual speakers (Sundara
2005). It is also the case that different datasets analysed using different settings
(e.g. frequency range for taking spectral moments) also introduces some variabil-
ity into measures taken for the “same” sounds across different studies. This in turn
leads to a more complex relationship between measures, auditory impressions,
and predicted articulations.
Statistical analysis of the measures was carried out using 3-factor ANOVAs
followed by Bonferroni post hoc tests, to test for effects of Preceding/Following
Segment and Community of Practice, as well as their interactions. Given that
each CoP is represented here by only two girls, we also ran the ANOVAs with
Speaker in place of CoP. Only results significant with p < .05 are reported.
3. Results
3.1 Adjacent phonetic segment
Results for each adjacent segment were inspected and grouped into larger catego-
ries on the basis of similar results (which aligned with phonetic similarity, e.g. mid/
low front vowels, back rounded vowels). Preceding segment showed no effect on
any of the spectral moments, nor were there any interactions. Following segment
showed significant effects for all four spectral moments (mean: [F(4, 436) = 3.788,
p = .005], spread [F(4, 436) = 9.259, p < .000], skew [F(4, 436) = 4.334, p = .001]
kurtosis [F(4, 436) = 3.267, p = .012]) though without any interactions with Com-
munity of Practice or Speaker; hence they are presented separately (Table 3.2).
Table 3.2â•‡ Average spectral moments for all speakers according to following segment
Following segment Mean Spread Skew Kurtosis n
I 2338 1856 1.55 5.09 104
ɛ, a, ai 2329 1901 1.66 5.04 â•⁄81
o, ɔ 2389 1658 1.37 4.67 â•⁄72
ʉ 2761 1809 1.00 3.17 151
w, r 1802 1222 1.89 9.73 â•⁄28
Post hoc tests showed the highest mean before /ʉ/, which also showed the
lowest skew. The mean was lower before the rounded consonants /w/ and /r/,
which also showed the lowest spread and highest kurtosis.
3.2 Community of Practice
A significant effect of Community of Practice was found for the Mean [F(2, 436) =
5.050, p = .007], Skew [F(2,436) = 7.657, p = .001], and Kurtosis [F(2, 436) =
4.145, p = .044]. Post hoc tests show that the Conservative girls have a higher
Mean than both the Modern and Messabout girls, who are not different from
each other (Figure 3.3). Skew and Kurtosis are different for each Community of
Practice: the values for both Skew (Figure 3.4) and Kurtosis (Figure 3.5) rise from
Conservative through to Messabouts. These results suggest an increasing positive
tilt to the spectrum, and an increasingly peaked distribution, which at first sight
seemed counterintuitive since auditorily it appears that the Messabouts show the
most laminal dental stops, for which we would expect a rather flat spectrum. We
explore a possible explanation in the discussion in Section 4.
5000
4000
Mean_Hz
3000
2000
1000
Conservative Modern Messabout

CoP
Figure 3.3â•‡ Box plots of Mean according to Community of Practice (n = 436)

6.000
4.000
Skew
2.000
0.000
–2.000

CoP
Figure 3.4â•‡ Box plots of Skew according to Community of Practice (n = 436)
40.000
30.000
Kurtosis
20.000
10.000
0.000

CoP
Figure 3.5â•‡ Box plots of Kurtosis according to Community of Practice (n = 436)

Table 3.3â•‡ Average spectral moments for each speaker

CoP Speaker Mean Spread Skew Kurtosis n
Messabouts Asma 1901 1402 2.04 9.04 â•⁄30
Naazi 2024 2037 2.31 8.29 â•⁄55
Moderns Zahida 2588 2057 1.26 2.51 105
Huma 1778 1466 2.14 9.23 â•⁄71
Conservatives Aneela 2782 1870 0.75 1.48 â•⁄94
Inaya 3004 1528 0.68 2.96 â•⁄81
3.3 Individual speakers
We also wanted to investigate the extent to which the individual speaker vari-
ation aligned with Community of Practice membership, and so we ran all the
ANOVAs again, substituting Speaker for CoP. This improved the explanation of
variance as indicated by the R2 by about 10% to almost 50% for all measures. An
effect of Speaker was found for all four measures: Mean [F(5, 436) = 3.410, p =
.005], Spread [F(5, 436) = 9.158, p < .000], Skew [F(5, 436) = 4.334, p = .001], and
Kurtosis [F(5, 436) = 5.074, p < .000]. The results are summarised in Table 3.3.
Post hoc tests for Mean, Skew, and Kurtosis measures confirmed the grouping
of Aneela and Inaya as “Conservative” and Asma and Naazi as “Messabouts”. But
for the same measures, the “Modern” girls were split: Zahida consistently pat-
terned with the Conservatives and Huma with the Messabouts. Spread showed a
completely different pattern, cutting across CoP, with Aneela, Zahida and Naazi
showing higher values and Inaya, Huma and Asma showing lower ones.
4. Discussion
The spectral moment analysis of the stop burst of /t/ showed a consistent effect of
Community of Practice for three measures (Mean, Skew, Kurtosis), which were
supported and informed by subsequent analysis of the individual speaker vari-
ation. Specifically, these measures show systematic variation with membership
of the Conservative and the Messabout CoPs, which capture shared engagement
in more traditional Pakistani culture and Islamic religious practices on the one
hand and bold ventures into forbidden western social practices on the other. The
Modern group is split: acoustic variation groups one girl with the Conservatives
and the other with the Messabouts. Interestingly, the same two girls show the op-
posite split patterning for the realization of the lateral (Stuart-Smith et al. 2011).
This suggests that the social-indexical value of phonetic variation for the Modern
450 Hz (high pass filter)
Figure 3.6â•‡ Schematic representation of a flattish spectrum cut off by a high-pass filter.
The arrow points to the apparent ‘peak’, and positive tilt/skew that emerges as the
moment analysis attempts to model this shape as a normal distribution. A low
concentration of the Mean would also be predicted
Community of Practice may lie in variation across clusters of features, as opposed

to any single feature (cf. Eckert 2008).
As noted in 2.3, it is difficult to relate the spectral measures exactly to spe-
cific (auditorily identified) articulatory configurations, but we can make some
suggestions. The lower Mean in the Messabout girls is consistent with the flat-
ter spectrum which is typical of dentals, where the longer constriction can also
damp the spectrum (see Sundara’s 2005 discussion of Canadian French data). If
the spectrum is flat, there is no appreciable overall peak and the mean will be low,
especially given the overall drop in acoustic energy at higher frequencies, leaving
the main concentration of energy at the low end of the spectrum. This would align
with these girls sounding as if they were producing more laminal dental plosives.
The boxplots of the Mean, and Skew, and Kurtosis show a clear inverse as-
sociation, and these measures are also significantly correlated with each other
(p < .000). This also presumes a rather flat energy distribution, with a lower Mean
the flatter the distribution. We suspect that the positive tilt and increased peakiÂ�
ness probably reflect an artifact of modeling a flatter distribution cut off at low
frequencies with high-pass filter (cf. Flipsen et al. 1999; see Figure 3.6). It seems
likely that future analysis may well benefit from using psychoacoustic transfor-
mation of the linear Hertz spectra into non-linear auditory spectra, which ap-
proximates better what listeners hear, and which may thus also be important for
uncovering the social meanings carried by this aspect of speech production.
We also found that, as might be expected, variation in spectral energy was
consistently affected by the following segment, (cf. Harrington 2010, 302ff.).
Before /w/ and /r/, both of which show lip-rounding and hence an appreciable
lengthening of the front cavity, /t/ showed the lowest Mean. The highest Mean
occurred before the BOOT vowel (corresponding to English GOOSE and FOOT),
which in Scottish English is /ʉ/, typically a central or front vowel. The raised Mean
may indicate a degree of unrounding of this vowel, though further acoustic analy-
sis of the Glasgow-Asian vowel system is needed in order to understand better the
coarticulatory effects of this vowel on /t/ release.
It is difficult to compare absolute values of moment analyses across different
studies because they are sensitive to recording condition and analytical proce-
dure, and this is particularly so for Skew and Kurtosis (Sundara 2005). Never-
theless we do notice a correspondence between the patterning of the Mean with
ethnic and social identity in these Glasgow-Pakistani girls, and with ethnicity in
Glasgow male speakers. The Mean values of the Messabouts (and Huma), heard
as laminal dental stops, are lower than those for the Conservatives (and Zahida),
thought to be apico-postalveolars. This is very similar to the pattern found in
a small-scale study of Glaswegian males, where a lower Mean in Glasgow non-
Asian males (laminal dental) was found compared to a higher Mean in Glasgow
Asian males (postalveolar), see Figure 3.7 (Stuart-Smith 2009).1 This suggests that
4000
3000
Mean_Hz
2000
1000
Glasgow Asian Glasgow non-Asian

Ethnicity
Figure 3.7â•‡ Box plots of Mean in 5 Glasgow Asian speakers and 2 Glasgow non-Asian
speakers (n = 88); from Stuart-Smith (2009)
1. The absolute values of the female speakers are about 400 Hz higher than those of the male
speakers, cf. Sundara (2005).
the Messabout girls may be accommodating to Glasgow non-Asian norms in the

realization of this feature. It will be extremely interesting to see how the Wannabe
CoP patterns for /t/ in later analysis.
The absolute Mean values for the auditory postalveolars heard in the Glasgow
Asians (Conservative girls and Glasgow Asian males) are somewhat lower than
those found recently in Sheffield Asian speakers (Kirkham 2011). However we
note that the range of variation in the two studies is rather different. Kirkham
(2011) has a range of 4000–6000 Hz for the analysis of the fine-grained differences
in (auditory) alveolar/postalveolar variants of Sheffield Asian /t/. Here the range
is between around 700–5000 Hz, reflecting a larger range of auditory variants,
including dentals which we predict to show a much lower mean given the flat
spectrum. The difference may also relate to differences in data collection and in
the realization of a different linguistic phonetic contrast, given that the Sheffield
non-Asian realization of syllable-initial /t/ is typically alveolar. A comprehensive,
co-ordinated sociophonetic study of regional British Asian accents would be ex-
tremely enlightening.
Most of our discussion has focused on the Messabouts and their den-
tal-sounding variants. What are we to make of the Conservative girls and their
more retracted, postalveolar realizations? Our auditory impressions do not sup-
port direct phonetic interference from Punjabi/Urdu retroflexion as an explana-
tion for postalveolar /t/ in these 2nd+ generation Glasgow-Pakistani girls since
their stops never sound retroflex. Models of language contact do not necessarily
predict identical phonetic qualities in cases of language transfer. However, the
speculation here is also supported by both Kirkham (2011) and Sharma and
Sankaran (2011). Sharma and Sankaran did a detailed measure of bilingual
language use and found it to be only weakly significant. Kirkham also suggests
based on ethnographic observation, that bilingual use is not likely to be a factor.
Direct acoustic comparison in this study is tricky given the scarcity of relevant
code-switched items in the interviews. Figure 3.8 shows spectrograms of the only
phonetically comparable pair of tokens, Punjabi retroflex /ʈ/ and Glasgow Asian
auditorily retracted /t/, both preceding /i/ and following /o/ (with greater or lesser
degree of nasalization).
The spectral characteristics of the two stops show some differences. The burst
is more intense and compact for the retroflex than the postalveolar, but more
notable is the release phase. The retroflex stop shows very little aspiration and
swift resumption of modal voicing for the vowel, whereas the postalveolar stop
is followed by a substantial period of noisy aspiration. This is only a small indi-
cation, but it supports our auditory impressions. The “retracted” variants which
are surfacing in young Glasgow-Asian 2nd+ generation speakers, indexing local-
ly specific ethnic identities, including those with strong affiliations to Pakistani
104
Frequency (Hz)
0
0 0.3638
Time (s)
104
Frequency (Hz)
0
0 0.3303
Time (s)
Figure 3.8â•‡ Spectrograms of the same Glasgow Asian girl producing retroflex /ʈ/ in
Punjabi motiyan ‘fat’ (top), and retracted /t/ in don’ teach (bottom); frequency location
of the bursts is marked with arrows
heritage and culture, are not instances of phonetic transfer from the heritage lan-
guage. Rather, they appear to be phonetically hybrid entities (cf. Harris 2006),
reflecting and encoding Glaswegian and (traces of) Punjabi/Urdu phonetics to
greater or lesser degrees. Here postalveolar /t/ is more like Glaswegian in showing
a longer burst and some noisy aspiration (especially before /i/). But at the same
time, the main frequency location of the burst energy is rather similar across the
two stops. Such hybridity at the fine phonetic level is also clearly evident in the
realization of clearer dark laterals by Glasgow Asians as opposed to Glasgow non-
Asians and again patterning with Community of Practice in some of the same
girls (Stuart-Smith et al. 2011). New social-indexical variation is emerging to ex-
press new – locally salient – hybrid identities.
Finally, the close alignment of ethnic and social identity with fine-grained
phonetic patterning, alongside phonetically-governed variation, suggests both
that – at some level, probably subconscious – speakers may exercise a high degree
of control over speech production, and that listeners may be sensitive to very sub-
tle differences. This in turn provides further motivation for the development of
models of phonological knowledge which are phonetically and socially rich (e.g.
Johnson 1997).
5. Conclusions
This small-scale sociophonetic study of /t/ in Pakistani Muslim girls at a Glasgow

High School suggests that ethnicity is integrally linked with locally-salient iden-
tity, and hence that fine phonetic variation which indexes ethnicity is in fact in-
dexical of local ethnic identity. These are, however, only preliminary results from
a larger study underway by the first author. The next steps involve analyzing /t/
for more speakers for each Community of Practice and for more Communities of
Practice. Other variables will also be considered, including FACE and GOAT, which
also appear to index local ethnic identities for this community. We also intend to
consider the data using auditorily transformed spectra.
More generally, our results raise a number of questions for further research:
To what extent are these differences in speech production available to listeners
within and outside the community? What kinds of articulatory configurations
underlie this kind of patterning? And to what extent are realizations like postal-
veolar /t/ in Glasgow Asian similar to those in regional Asian accents elsewhere
in the UK and even further afield? Just how locally embedded has the former ret-
roflex /t/ become as it has travelled – and continues to travel – through Englishes
in the Indian diaspora?
References
Agha, Asif. 2003. The social life of cultural value. Language and Communication 23: 231–273.
DOI: 10.1016/S0271-5309(03)00012-0
Alam, Farhana. 2007. Language and identity in ‘Glaswasian’ adolescents. MLitt dissertation,
Glasgow University.
Ballard, Roger. 1994. Desh Pardesh: The South Asian Presence in Britain. London: Hurst.
Bhardwaj, Mangat Rai. 1995. Colloquial Punjabi. London: Routledge.
Bhatia, Tej. 1993. Punjabi. London: Routledge.
Catford, John Cunnison. 2001. A Practical Introduction to Phonetics. Oxford: OUP.
Docherty, Gerry & Foulkes, Paul. 1999. Derby and Newcastle: Instrumental phonetics and var-
iationist studies. In Urban Voices: Variation and Change in British Accents, Paul Foulkes &
Gerry Docherty (eds), 47–71. London: Arnold.
Docherty, Gerry & Mendoza-Denton, Norma. 2012. Speaker-related variation – Sociophonetic

factors. In The Oxford Handbook of Laboratory Phonology, ed. by Abigail C. Cohn, Cécile
Fougeron & Marie K. Huffman (eds), 44–60. Oxford: OUP.
Eckert, Penelope. 2000. Linguistic Variation as Social Practice. Oxford: Blackwell.
Eckert, Penelope. 2008. Variation and the indexical field. Journal of Sociolinguistics 12: 453–476.
DOI: 10.1111/j.1467-9841.2008.00374.x
Evans, Bronwen, Mistry, Ajay & Moreiras, Caroline. 2007. An acoustic study of first- and sec-
ond-generation Gujerati immigrants in Wembley: Evidence for accent convergence? Proc
16th ICPhS, Saarbrücken, 1741–1744.
Fishman, Joshua. 1991. Reversing Language Shift: Theory and Practice of Assistance to Threat-
ened Languages. Clevedon: Multilingual Matters.
Flipsen, Peter, Shriberg, Lawrence, Weismer, Gary, Karlsson, Heather & McSweeny, Jane. 1999.
Acoustic characteristics of /s/ in adolescents. Journal of Speech, Language, and Hearing
Research 42: 663–677.
Forrest, Karen, Weismer, Gary, Milenkovic, Paul & Dougall, Ronald. 1988. Statistical analysis of
word-initial voiceless obstruents: Preliminary data. JASA 84: 115–123.
DOI: 10.1121/1.396977
Foulkes, Paul. 2010. Exploring social-indexical knowledge: A long past but a short history. Lab-
oratory Phonology 1: 5–39. DOI: 10.1515/labphon.2010.003
Foulkes, Paul & Docherty, Gerry. 2006. The social life of phonetics and phonology. Journal of
Phonetics 34: 409–438. DOI: 10.1016/j.wocn.2005.08.002
Fromont, Robert & Hay, Jen. 2008. ONZE Miner: The development of a browser-based research
tool. Corpora 3: 173–193. DOI: 10.3366/E1749503208000142
Hall, Stuart. 1992. The question of cultural identity. In Modernity and its Futures, Stuart Hall,
David Held & Tony McGrew (eds), 273–316. Cambridge: Polity.
Harrington, Jonathan. 2010. Phonetic Analysis of Speech Corpora. Oxford: Wiley-Blackwell.
Harris, Roxy. 2006. New Ethnicities in Language Use. London: Palgrave.
DOI: 10.1057/9780230626461
Heselwood, Barry & McChrystal, Louise. 2000. Gender, accent features and voicing in Panjabi-Â�
English bilingual children. Leeds Working Papers in Linguistics 8: 45–69.
Hirson, Allen & Sohail, Nabiah. 2007. Variability of rhotics in Punjabi-English bilinguals. Proc
16th ICPhS, Saarbrücken, 1501–1504.
Johnson, Keith. 1997. Speech perception without speaker normalization: An exemplar model.
In Talker Variability in Speech Processing, Keith Johnson & John Mullennix (eds), 145–165.
San Diego CA: Academic Press.
Johnson, Keith. 2011. Acoustic and Auditory Phonetics, 3rd edn. Oxford: Wiley-Blackwell.
Kerswill, Paul, Torgersen, Eivind & Fox, Sue. 2008. Reversing ‘drift’: Innovation and diffusion in
the London diphthong system. Language Variation and Change 20: 451–491.
DOI: 10.1017/S0954394508000148
Khan, Arfaan. 2009. Ethnicity and language. In English Language: Description, Variation and
Context, Jonathan Culpepper, Francis Katamba, Paul Kerswill, Ruth Wodak & Tony
McEnery (eds), 243–256. Houndmills: Palgrave.
Kirkham, Sam. 2011. The acoustics of coronal stops in British Asian English. Proc 17th ICPhS,
Hong Kong, 1102–1105.
Labov, William. 2001. Principles of Linguistic Change: Social Factors. Oxford: Blackwell.
Ladefoged, Peter. 2003. Phonetic Data Analysis. Oxford: Blackwell.
Lambert, Kirsten, Alam, Farhana & Stuart-Smith, Jane. 2007. Investigating British Asian ac-
cents: Studies from Glasgow. Proc 16th ICPhS, Saarbrücken, 1509–1511.
Lawson, Eleanor, Scobbie, Jim & Stuart-Smith, Jane. 2008. Articulatory insights into language
variation and change: Preliminary findings from an ultrasound study of derhoticization
in Scottish English. University of Pennsylvania Working Papers in Linguistics 14 (2): Papers
from NWAV 36: 102–110.
Lawson, Eleanor, Scobbie, Jim & Stuart-Smith Jane. 2011. The social stratification of tongue
shape for postvocalic /r/ in Scottish English. Journal of Sociolinguistics 15: 256–268.
DOI: 10.1111/j.1467-9841.2011.00464.x
Maan, Bashir. 1992. The New Scots: The Story of Asians in Scotland. Edinburgh: John Donald.
Masica, Colin P. 1993. Indo-Aryan Languages. Cambridge: CUP.
Mendoza-Denton, Norma. 2008. Homegirls: Language and Cultural Practice among Latina
Youth Gangs. Oxford: Blackwell.
Ogden, Richard A. 2009. An Introduction to English Phonetics. Edinburgh: EUP.
Sharma, Devyani & Sankaran, Lavanya. 2011. Cognitive and social forces in dialect shift: Grad-
ual change in London Asian speech. Language Variation and Change 23: 399–428.
DOI: 10.1017/S0954394511000159
Shaw, Alison. 1994. A Pakistani community in Oxford. In Desh Pardesh: The South Asian Pres-
ence in Britain, Roger Ballard (ed.). London: Hurst & Company.
Stuart-Smith, Jane. 1999. Glasgow: Accent and voice quality. In Urban voices: Accent studies in
the British Isles, Paul Foulkes & Gerry Docherty (eds), 201–222. London: Edward Arnold.
Stuart-Smith, Jane. 2007. Empirical evidence for gendered speech production: /s/ in Glaswe-
gian. In Change in Phonology [Laboratory Phonology 9], Jennifer Cole & Jose Hualde
(eds), 65–86. Berlin: Mouton de Gruyter.
Stuart-Smith, Jane. 2009. Investigating Ethnic Accents. Final Report to British Academy
(SG-48480).
Stuart-Smith, Jane, Timmins, Claire & Alam, Farhana. 2011. Hybridity and ethnic accents: A
sociophonetic analysis of ‘Glaswasian’. In Language Variation – European Perspectives III
[Studies in Language Variation 7], Frans Gregersen, Jeffrey K. Parrott & Pia Quist (eds),
43–57. Amsterdam: John Benjamins. DOI: 10.1075/silv.7.04stu
Stuart-Smith, Jane, Lawson, Eleanor & Scobbie, James M. 2014. Derhoticisation in Scottish
English: A sociophonetic journey. In Advances in Sociophonetics [Studies in Language Var-
iation 15], Chiara Celata & Silvia Calamai (eds), 59–96. Amsterdam: John Benjamins.
Sundara, Megha. 2005. Acoustic-phonetics of coronal stops: A cross-language study of Canadi-
an English & Canadian French. JASA 118: 1026–1037. DOI: 10.1121/1.1953270
Wrench, Alan. 1995. Analysis of fricatives using multiple centres of gravity. Proc 13th ICPhS,
Stockholm, 460–463.
Chapter 4
East African Indian twice migrants in Britain

Phonological variation across generations*
Claudia Rathore
Recent years have seen a rapid increase of sociolinguistic interest in the use of
English in the British Asian diaspora. The focus of this work has usually been
on locally-born speakers (e.g. Heselwood & McChrystal 2000; Hirson & Sohail
2007; Cheshire et al. 2011; Sharma 2011; Stuart-Smith et al. 2011) with some
studies also looking into cross-generational variation (Evans et al. 2007;
McCarthy et al. 2011; Sharma & Sankaran 2011). The present paper contributes
to this growing body of research by providing insight into patterns of dialect
variation and change among East African Indians in Leicester, a community
of South Asian twice migrants who settled in Britain via East Africa in the late
1960s and early 1970s. Like Fiji Indians in New Zealand (Hundt, this volume),
East African Indians in Britain present an interesting case study for the linguis-
tic outcomes of dialect contact in secondary diaspora situations, an under-re-
searched type of contact setting. The aim of this paper is to determine how the
complex migration pattern of East African Indians in Leicester has influenced
their variety of English and, furthermore, whether and how linguistic patterns
change across generations. To this end, I examine variation in the use of post-
vocalic /r/ in a group of first- and second-generation migrants. The results in-
dicate that, despite a strong sense of affiliation with East Africa, first-generation
speakers have predominantly maintained Indian English patterns in their use of
this variable whereas second-generation subjects show accommodation to the
local variety of British English. Evidence from the community’s social history
accounts for the findings.
Keywords: twice migrants, East African Indians, cross-generational variation,

dialect contact, British Asian Englishes, postvocalic /r/, sociophonetics
* This paper is based on parts of the author’s unpublished PhD dissertation (Rathore 2013).
I am grateful to Marianne Hundt, Devyani Sharma and an anonymous reviewer for helpful
comments on the draft version. All remaining shortcomings are my own.
56 Claudia Rathore
1. Introduction
In the years after the Second World War, Great Britain became one of the ma-
jor destinations of migration from the Indian subcontinent. Today, the British
Asian diaspora is the largest non-white ethnic minority group in the UK. In the
2011 Census, 2,984,670 residents of England and Wales (5.3 percent of the to-
tal population) identified as Indian, Pakistani, or Bangladeshi (ONS 2012, table
KS201EW).1 Since patterns of settlement were determined by the labour shortage
created by the war, the majority of South Asians are concentrated in the country’s
main conurbations, with Indians accounting, for instance, for 28.3 percent of the
local population in Leicester, Pakistanis for 20.4 percent of the local population in
Bradford, and Bangladeshis for 32 percent of the local population in Tower Ham-
lets, London (ONS 2012, table KS201EW). Furthermore, British Asians show var-
iation not only along ethno-national lines but also in terms of religious affiliation
and regional-linguistic origins (see e.g. Brown 2006).
Due to the enormous impact of South Asian immigration on the country’s
ethnic profile, sociolinguistic interest in the British Asian diaspora has consid-
erably increased in recent years. Much of this work has been concerned with the
maintenance and transmission of features derived from the heritage languages or,
more generally, the emergence of contact-induced innovations. The few studies
that included first-generation South Asian migrants invariably found them to ex-
hibit L1-influenced accent traits (Evans et al. 2007; McCarthy et al. 2011; Sharma
& Sankaran 2011). With the exception of Evans et al. (2007) and McCarthy et
al. (2011), research also demonstrated that L1-derived features such as retract-
ed variants of /t/, clear variants of coda /l/, and monophthongal realisations of
face and goat tend to be transmitted to second-generation speakers.2 Among
locally-born migrants, L1-derived accent traits were further shown to undergo
“reallocation”, a process of new dialect formation “where two or more variants
in the dialect mix survive the levelling process but are refunctionalised, evolving
new social or linguistic functions in the new dialect” (Britain & Trudgill 1999,
245). Moreover, some of the features generated within South Asian and other
ethnic minority communities were found to spread beyond these groups and to
be adopted by white British English (BrE) speakers (e.g. Khan 2006; Fox 2007;
1. The three terms refer to the modern nation states of India, Pakistan and Bangladesh here.
Elsewhere in this study, however, the word “Indian” is used in a more general sense to denote
persons who originate from the Indian subcontinent.
2. See e.g. Rampton (1995), Heselwood & McChrystal (2000), Harris (2006), Hirson & Sohail
(2007), Stuart-Smith et al. (2011), Kirkham (2011), Sharma (2011) and Sharma & Sankaran
(2011).
Chapter 4.â•‡ East African Indian twice migrants in Britain 57
Kerswill et al. 2008; Cheshire et al. 2011). Crucially, these findings suggest that
the varieties of English spoken by British Asians and other ethnic minority groups
play an important role as a source of innovation and change in present-day BrE.
The present paper focuses on patterns of dialect variation and change in the
East African Indian community in Leicester, the largest city in the East Midlands.
East African Indians are distinctive among British Asians in that they represent
a secondary or double diaspora situation. Their migration history involved two
stages rather than one: a journey from South Asia to East Africa followed by a
journey from East Africa to the UK. Except for Hundt’s (this volume) research on
Fiji Indians in New Zealand, twice-migrant communities have not, to the best of
my knowledge, been an explicit concern in sociolinguistic work on English in the
Indian diaspora. Nonetheless, such complex migration patterns raise a number of
questions that are of great interest to the study of the linguistic consequences of
migration and contact. For example, how are the varieties of English spoken by
immigrant groups influenced by contact with the dialects, languages, and cultures
of three societies, and does this lead to linguistic outcomes that are different from
those found in direct migrant communities? Do twice migrants retain traits de-
rived from the languages of their original homeland, do they adopt features of the
variety of English spoken in their first host community, or do they show accom-
modation to the dialect of their second (current) host community? Moreover,
how do patterns of linguistic variation change across generations?
In this study, I explore these questions by examining the extent to which the
patterns of linguistic variation displayed by members of the East African Indian
community in Leicester parallel those attested in Indian English (IndE), East Af-
rican English (EAfE), and East Midlands English (EMidE). More specifically, I in-
vestigate variation in the use of postvocalic /r/, one of the most widely researched
variables in sociolinguistics since Labov’s (1966; 1972) ground-breaking work on
New York City, across two generations of migrants. My aim is to answer the fol-
lowing two research questions:
1. Does the use of postvocalic /r/ by the first generation show any parallels to
IndE, EAfE, or EMidE?
2. Does the production of postvocalic /r/ change from the first generation to the
second generation and if so, how?
First-generation migrants are defined as speakers who grew up in East Africa

and undertook migration to Leicester in adulthood whereas second-genera-
tion migrants are speakers who were raised in Leicester. The motivation behind
this approach is that IndE, EAfE, and EMidE are the dialects of English most
closely related to the community’s migration history. “Indian English” is used in
the broader of the two senses of the word found in the literature to denote the
58 Claudia Rathore
second-language varieties of English spoken on the Indian subcontinent (Wells

1982, 624), the original homeland of East African Indians in Leicester. “East Afri-
can English” refers to English spoken as a second language in Kenya, Uganda, and
Tanzania (Schmied 2008, 150), their first host community, and “East Midlands
English” is the native variety of English spoken in Leicester (Hughes et al. 2005,
68–71), their current host community.
In the following sections, I first provide an overview of the migration history
of the East African Indian community in Leicester followed by a discussion of
postvocalic /r/ in IndE, EAfE, and EMidE as well as a description of the method of
the present study. I then present and discuss the findings. I first analyse variation
at group level, looking at the overall frequency of occurrence of postvocalic /r/ in
prepausal and preconsonantal position (rhoticity) and at the phonetic realisation
of rhotic tokens. This is complemented by considerations on linking /r/ and intru-
sive /r/. Furthermore, I explore some of the factors that constrain variation in the
use of postvocalic /r/ within the first generation and conclude by examining the
findings in the light of the community’s social history.
2. Socio-historical background
2.1 Indians in East Africa
Merchants from the Indian subcontinent had been trading along the East African
coast for at least 2,000 years (Gregory 1993). However, large-scale settlement of
the region only began in the 1890s, when Germany and Britain gained possession
of the territories corresponding to present-day Kenya, Uganda, and Tanzania and
the European expansion into the interior created new employment opportunities
for migrants from South Asia. In the early years of colonial rule, some 32,000 in-
dentured labourers, mostly Sikhs from the Punjab, were imported by the British
for the construction of the Uganda Railway, around one fifth of whom settled
permanently after the end of their contracts (Gregory 1993, 160–162). The over-
whelming majority of arrivals were, however, free migrants who were attracted to
East Africa by the potential for economic improvement offered by the retail trade,
other areas of commerce, and the artisan sector, as well as the possibilities of em-
ployment as civil servants, policemen, and soldiers in the colonial administration,
or as clerks in private companies. While some of the free migrants, especially civil
servants, arrived from Goa, most came from the north-western part of the Indian
subcontinent, particularly Kutch, Kathiawar, inland Gujarat, Sindh and, to a lesser
extent, Maharashtra (Gregory 1993, 14, 37; Oonk 2006, 255; see Figure 4.1). The
East African context thus differed from other major destinations of South Asian
Amritsar
Lahore
P U N J A B
New Delhi
B A L U C H I S TA N
R A J A S T H A N
Hyderabad
N Karachi
A R A B I A N S I N D
KUTCH Ahmedabad 23o
S E A Jemnagar Rajkot Baroda
GUJARAT Bhavnagar
Porbandar Broach
Diu Surat
0 100 200 300 MI. Bombay

I N D I A N Poona
0 200 400 KM. MAHARASHTRA
O C E A N
Pangim
GOA
Figure 4.1â•‡ Map of North-western India (1890–1947), regions of origins of East African
Indians (Gregory 1992; reprinted in Gregory 1993, 11)
movement during the colonial period, e.g. Natal, Fiji, and Trinidad, where inden-
tured migration from North and South India often prevailed (see the respective
contributions in this volume).
By the time the East African territories gained Independence in the early
1960s, Indians numbered about 175,000 in Kenya, 77,500 in Uganda, 92,000 in
Tanganyika, and 20,000 in Zanzibar (Oonk 2006, 256). They always remained a
small community, never exceeding 2 percent of the total population. Their lin-
guistic heritage included several Indo-Aryan languages (mainly Gujarati and
Kutchi, followed by Punjabi, Sindhi, Marathi, and Konkani), but they often also
acquired Kiswahili and English as second languages (Gregory 1993; Oonk 2006).
With regard to religion, East African Indians belonged to five major faith groups,
divided into a variety of sects: Hindus, Muslims and, in smaller numbers, Jains,
Sikhs, and Christians. Crucially, they were an extremely successful community
in economic terms, coming to predominate at the lower and middle levels of the
colonial administration and in the many business areas to which they had turned
(see the extensive discussion in Gregory 1993).
60 Claudia Rathore
The financial achievements of East African Indians stood in stark contrast to

the lack of opportunities from which the African population suffered. After Inde-
pendence, the new governments adopted a policy of Africanisation of the econ-
omy in an attempt to improve the situation, which included a series of measures
restricting work permits and trading licenses for non-citizens and, in Tanzania,
the nationalisation of the main industries, banks, and insurances (Gregory 1993).
These measures prompted a mass exodus of Indians from East Africa. In the late
1960s and early 1970s, about half of the Kenyan Indian population and 40 percent
of Tanzanian Indians emigrated (Oonk 2006, 256). In Uganda, General Idi Amin
took the even more drastic decision of expelling non-citizen Indians in August
1972, giving them 90 days to leave the country (see Marett 1989 for a detailed ac-
count of the expulsion). After this period was over, the overwhelming majority of
Ugandan Indians had gone. They had been allowed to bring only £50 per head of
family and their personal belongings with them so that the assets they left behind
amounted to an estimated £100–150 million.
During the exodus, East African Indians scattered around the world. While
relatively few returned to the Indian subcontinent, many migrated to Canada or
the US, and some went to places like Malawi, Australia, and New Zealand or to
cities in continental Europe (Brown 2006; Oonk 2006). Due to the former colo-
nial connections, however, Britain was by far the preferred destination of East
African Indians, with most resettling in London or Leicester.
2.2 East African Indians in Leicester
According to an estimate of the Leicester Council for Community Relations,

some 19,500 East African Indians were living in Leicester by 1978, accounting for
about half of the city’s South Asian population; 10,500 had come from Uganda,
7,000 from Kenya, and 2,000 from Tanzania (Phillips 1981, 104).
It is commonly acknowledged that East African Indians differed markedly
from the direct South Asian migrants who had preceded them (see e.g. Brown
2006). The latter were often unskilled male workers from rural areas who had
primarily migrated for economic reasons. Many of them had reached the UK via
chain migration and were only later joined by their families. In contrast, East Af-
rican Indians were political refugees. They tended to come from urban areas, had
pursued a variety of occupations in East Africa and migrated to Britain as family
units. Moreover, it was common for them to have formal qualifications and good,
or fairly good, levels of spoken English (Robinson 1993, 234–235).
For many East African Indians, resettlement in the UK initially involved a
drop in social status (Marett 1989, 8). They frequently had to move into high-
ly overcrowded, privately rented accommodations and take up factory jobs or
other types of low-status work. As shown in Marett (1993) and Robinson (1993),
however, the history of the community is characterised by remarkable upward
social mobility. Thus, within a few years the socioeconomic situation of many
East African Indians improved considerably as they quickly moved into the own-
er-occupier sector, re-qualified, took up white-collar occupations, and/or became
self-employed.
2.2.1 Transnational ties, identity and language

The evidence from the present study indicates that, for East African Indians, re-
settlement in Britain generally involved a rather clean break with their lives in the
first host community, with many retaining only weak (if any) “real” ties with East
Africa (see also Robinson 1993). By contrast, many first-generation East African
Indians have maintained strong links with the Indian subcontinent to this day.
Most interviewees reported that they returned to India regularly to visit relatives
or for travelling purposes. Moreover, Indian cultural and religious traditions still
form an important part of their heritage. It is therefore particularly striking that
first-generation informants tend to express strong affiliation with East Africa and
a clear sense of distinctiveness vis-à-vis direct migrants from South Asia. This
sense of separateness has, however, largely vanished among second-generation
participants, who mostly describe themselves in terms of a more general British
Asian identity.
With regard to language use, the complex multilingual situation that charac-
terised the Indian diaspora in East Africa is still reflected among first-generation
East African Indians in Leicester today, even though the general use of English
has increased at the same time. The research conducted for this study suggests
that, while there are older members of the community who still have no or only
very limited proficiency in English, the linguistic repertoire of many first-genera-
tion migrants includes two or more Indian languages, English, and often (some)
Kiswahili. Within the first generation, the use of Indian languages continues to
be very widespread, especially in personal contexts, and intriguingly, Kiswahili is
still occasionally used as well, mostly as a means of reinforcing the sense of bond-
ing among community members. In the case of Gujarati at least, the influence of
Kiswahili is also visible in the use of lexical borrowings, especially for food and
household items, e.g. fagio ‘broom’ and kijiko ‘spoon’. Great importance is gener-
ally attached to Indian heritage languages as carriers of the community’s cultural
identity, which is also reflected in the common practice of sending children to
mother tongue classes organised by temples and community centres. Nonetheless,
informant reports indicate that among British-born speakers, proficiency levels
and the amount of daily use of Indian languages vary greatly. In the present study,
for instance, some second-generation informants claimed almost categorical use
62 Claudia Rathore
of Gujarati with family members, including siblings, whereas others stated that
they had only very limited knowledge of the language and mostly used English at
home. The extent to which Indian heritage languages will be maintained in subse-
quent generations is therefore difficult to predict.
3. Postvocalic /r/
3.1 Rhoticity, linking /r/ and intrusive /r/
The present study investigates the variable presence of /r/ in syllable coda posi-
tion, an environment commonly referred to as “postvocalic /r/” or “non-prevocal-
ic /r/”. Whereas in “rhotic” varieties of English /r/ is pronounced in all positions,
“non-rhotic” dialects permit an overt phonetic realisation of /r/ only when it
precedes a vowel, as in brat or carry, but not when it occurs before a consonant
or pause, as in bark or star ǁ (Wells 1982, 75–76, 218). Most non-rhotic native
varieties of English are also characterised by a phenomenon known as “linking
/r/” in which historical /r/ is articulated in word-final position when followed by
a word or morpheme beginning with a vowel as in “fear anything” and “fearing”
(Wells 1982, 219). In many non-rhotic accents, this pattern has been extended
to lexical items without etymological word-final /r/, leading to the appearance
of an “intrusive /r/” in phrases like “the idea isn’t” /ðiː aɪdɪəɹ ɪznt/ and “drawing”
/dɹɔːɹɪŋ/ (Wells 1982, 223–227).
3.2 Postvocalic /r/ in IndE, EAfE, EMidE
Research on IndE reveals a continuum from non-rhoticity to rhoticity. Accord-

ing to Bansal (1990) and Pingali (2009), there is a highly prestigious non-rhotic
form of IndE, labelled Standard Indian English Pronunciation (SIEP) by the lat-
ter, which is spoken by a small minority of the population only. In contrast, the
majority of IndE speakers, including educated ones, are reported to have (varia-
bly) rhotic accents (e.g. Bansal 1990; Nihalani et al. 2004; Gargesh 2008). Varia-
ble rhoticity is also attested among first-generation Indian immigrants in the San
Francisco Bay area (Sharma 2005), whereas South African Indian English and
Indo-Fijian English are non-rhotic (Mesthrie 2008; Tent & Mugler 2008). Little
is known about the occurrence of linking /r/ and intrusive /r/ in IndE. Pingali
(2009, 20) claims that SIEP displays the former but not the latter, but it is not clear
whether this also holds for other forms of IndE. It may be noted, however, that ac-
cording to Schneider (2004, 1125), intrusive /r/ “occurs hardly at all in Africa and
Asia”. Phonetically, IndE /r/ is most commonly described as a tap or trill, although
approximants are also attested (CIEFL 1972; Wiltshire & Harnsberger 2006;
Gargesh 2008). Variation between taps, trills, and approximants is also found in
diasporic varieties (e.g. Sharma 2005; Mesthrie 2008). Furthermore, quantitative
studies of IndE on the subcontinent and in the diaspora demonstrated that, as in
other English dialects, both the occurrence of coda /r/ and the phonetic quality of
/r/ are constrained by a number of factors (see Section 6).
As Simo Bobda (2001, 269) points out, EAfE is still an under-researched va-
riety, particularly with regard to phonology. However, the existing evidence indi-
cates that, like most African Englishes, the varieties of English spoken in Kenya,
Uganda, and Tanzania are non-rhotic (Simo Bobda 2001; Schmied 2008). No in-
formation appears to be available on the occurrence of linking /r/ and intrusive
/r/ in EAfE. With regard to the phonetic quality of /r/ in prevocalic position, EAfE
shows variation between tapped, trilled, and approximant realisations (Simo
Bobda 2001; Schmied 2008).
Like most other dialects of England, EMidE has lost postvocalic /r/ (Hughes
et al. 2005). While there are no descriptions of linking /r/ and intrusive /r/ in
this variety, it is noteworthy that the presence of both phenomena is well attested
in non-rhotic BrE dialects, including those of the North (e.g. Broadbent 1991;
Foulkes 1997; Barras 2010). It is thus likely that the production of both etymo-
logical and unetymological word-final /r/ before vowels is permitted in EMidE as
well. Before vowels, northern English /r/ is predominantly realised as a post-al-
veolar approximant, but alveolar taps also occur, particularly in intervocalic posi-
tion (Wells 1982; Hughes et al. 2005).
4. Method
4.1 Data collection and participants
The present study is based on data from 25 first- and second-generation speakers
who were interviewed in 2007 and 2009. The first-generation group included six
female and five male participants aged between 42 and 80, who had been born
and raised in Kenya or Uganda and had migrated to Leicester in adulthood (at
the age of 18 or later). Eight spoke Gujarati as L1, two Punjabi, and one Kutchi,
although, like many first-generation East African Indians, most were in fact mul-
tilingual. All had acquired English as a second language before arriving in the UK,
mainly through English-medium education in East Africa. The second generation
comprised seven female and seven male subjects, aged from 17 to 41. All had
been born and bred in Leicester, apart from two participants who had arrived
64 Claudia Rathore
in England at the age of 2 and 4, respectively.3 In contrast to the first genera-

tion, second-generation speakers exclusively had Gujarati backgrounds. Most had
completed university-level education or were (in two cases) doing their A-levels
with the aim of attending university. Informants from both generations mainly
had middle-class occupations, with some holding positions at managerial level.
Interviews lasted from 20 minutes to two hours, with an average length of an
hour. They consisted of a relatively open-ended conversation in which I elicited
information about the speakers’ biographical background, language proficiency,
use and attitudes, as well as cultural and ethnic affiliations, while also taking up
whichever topics they enjoyed talking about.
4.2 Auditory analysis
50 tokens of postvocalic /r/ in prepausal and preconsonantal environments were

extracted for every speaker, generally starting after the first 15 minutes of the
interview. Following Wells (1982, 219–220), the prepausal category comprised
instances of word-final coda /r/ occurring at “the end of an utterance, [before] a
pause, or a major syntactic boundary” (when I was younger ǁ), whereas the precon-
sonantal category included tokens of coda /r/ followed by a consonant across “an
optional morpheme or word boundary” (force, scared, my father came). Instances
of linking /r/ across word boundaries (a poor area) were coded but not included in
the main rhoticity analysis. The reason is that both rhotic and non-rhotic speakers
may pronounce postvocalic /r/ in these contexts, so that they are indistinguisha-
ble on the surface. Tokens of linking /r/ within words (fearing) were excluded, as
they were very infrequent and no variation was noticed in their realisation. Lastly,
overt phonetic realisations of intrusive /r/ (idea(r) of my mother) were noted but
not analysed quantitatively, since in spontaneous conversations the environments
in which this phenomenon may surface are very rare.
Token extraction was limited to content words and restricted to three tokens
per lexical item. To facilitate comparison with existing quantitative work on post-
vocalic /r/, a binary classification of the dependent variable was used for the main
analysis. Tokens with any kind of audible consonantal constriction, ranging from
r-coloured vowels to vowels followed by approximant, trilled, or tapped realisa-
tions of /r/, were categorised as “rhotic” while tokens with no audible consonantal
constriction (including tokens in which a vowel was followed by a glottal stop;
3. While in the families of most second-generation speakers both parents came from East
Africa, in three cases the speaker’s mother came from India. Though not ideal for comparative
purposes, this is representative of the community in that it reflects a common tendency among
East African Asian men, who often married women from the Indian subcontinent.
Foulkes 1997) were classified as “non-rhotic”. In order to gain further insight into
possible parallels with IndE, EAfE, and EMidE, rhotic tokens were further sub-
divided into “approximants” vs. “taps/trills”. Although each variety shows some
variation in the phonetic quality of /r/ (see Section 3.2), the two categories can be
generally regarded as representative of English English and non-English English
variants, respectively.
4.3 Statistical analysis
A generalised linear (logistic) mixed-effects regression analysis was carried out to

shed further light on variation in the overall frequency of postvocalic /r/ among
first-generation speakers.4 This type of statistical technique is an extension of tra-
ditional logistic regression, as performed, for instance, by Goldvarb (e.g. Sankoff
et al. 2012). Both linear and logistic mixed-effects models are a relatively recent
development in statistics (Baayen 2008; Johnson 2009). Unlike ordinary regres-
sion analysis, they do not require independence of observations. Their assump-
tions therefore provide a better match to datasets consisting of natural language
samples in which tokens tend to be grouped by speaker and lexical item (see
Johnson 2009 for a full discussion). The distinction between two types of pre-
dictors, fixed effects and random effects, enables mixed models to evaluate the
influence of independent variables such as social class or syllable stress (the fixed
effects) while simultaneously taking into account variation across individual
speakers and lexical items (the random effects).
The logistic mixed-effects regression analysis estimated the influence of three
fixed effects and their interactions: following phonological environment (cod-
ed as a factor with the levels prepausal and preconsonantal), gender (coded as a
factor with the levels female and male), and age (coded as a continuous predic-
tor).5 These independent variables were chosen because of their attested effect on
postvocalic /r/ in IndE and/or other English dialects (see Section 6.1). Speaker
and lexical item were included in the model as random effects in order to control
for imbalances in the dataset as well as inter-speaker and inter-word variability.
4. The statistical analysis was restricted to the first generation because, as shown in Sec-
tion 5.1, second-generation speakers exhibited almost categorical non-rhoticity.
5. The analysis was run in R (R Development Core Team 2011), using the glmer function
from the lme4 package (Bates et al. 2011). Note that there is an important difference between
the terminology used by Goldvarb and that of lme4 (as well as most other statistical packages):
Goldvarb’s “factor groups” are usually referred to as “factors” in other programs, and Goldvarb’s
“factors” as “factor levels” (Johnson 2009, 361). In the present study, I use the terms “factor” and
“factor levels”.
66 Claudia Rathore
Following Johnson (2009), the overall significance of fixed effects and their inter-
actions was evaluated via likelihood ratio chi-squared tests.
5. Results
5.1 Variation across generations
In total, 1,250 tokens of postvocalic /r/ in prepausal and preconsonantal position

were coded across the 25 speakers.6 Table 4.1 and Figure 4.2 show the proportions
of rhotic and non-rhotic tokens for each generation. First-generation speakers
pronounced coda /r/ 19.3 percent of the time. With zero realisations appearing in
roughly four tokens out of five, the group as a whole clearly exhibits a tendency
towards non-rhoticity. At the same time, though, the frequency of rhotic variants
is high enough as to deserve consideration, particularly when compared to the
pattern displayed by second-generation informants. Among the latter, only five
tokens of postvocalic /r/ out of 700 (0.7 percent) are rhotic. The difference be-
tween the generations is remarkable not only because it is rather large, amounting
Table 4.1â•‡ Production of postvocalic /r/ by generation (R = rhotic realisation, zero = null
realisation)
Generation Postvocalic /r/
R% zero % N
First 19.3 80.7 550
Second â•⁄0.7 99.3 700
Table 4.2â•‡ Percentage of rhotic tokens by following phonological environment and gener-
ation (pc = preconsonantal, pp = prepausal, pv = prevocalic)
Generation Following phonological environment
pc pp pv
% N % N % N
First 14.9 444 37.7 106 70.3 37
Second â•⁄0.3 576 â•⁄2.4 124 77.6 85
6. As mentioned in Section 4.2, both linking /r/ and intrusive /r/ were excluded from the
main analysis and are discussed separately here. Unless stated otherwise, they are therefore not
included in the overall figures for postvocalic /r/ presented in this study.
1.0
0.8
Production of postvocalic /r/
Zero
0.6
0.4
0.2
R
0.0
First Second
Generation
Figure 4.2â•‡ Production of postvocalic /r/ by generation (R = rhotic realisation, zero =

null realisation)
First generation Second generation

1.0
1.0

0.8
0.8
0.6
0.6
Zero
Zero
0.4
0.4
0.2
0.2
R
R
0.0
0.0
pc pp pv pc pp pv
Following phonological environment Following phonological environment
Figure 4.3â•‡ Production of postvocalic /r/ by following phonological environment and

to almost 20 percent, but also because it is likely to be qualitative rather than

quantitative, with the first generation displaying variable rhoticity and the sec-
ond generation near-categorical non-rhoticity. The contrast is even more evident
when considering variation across individual speakers within each group.
68 Claudia Rathore
In addition to prepausal and preconsonantal coda /r/, 122 tokens of word-fi-

nal prevocalic (linking) /r/ were extracted during the analysis. Both first- and
second-generation speakers exhibited a strong tendency to pronounce coda /r/
in this context. As shown in Table 4.2 and Figure 4.3, the frequency of rhotic to-
kens in this position surpasses 70 percent for both groups. At first sight, first- and
second-generation participants also display the same constraint hierarchy for
following phonological environment, with coda /r/ being articulated most com-
monly in word-final prevocalic contexts and least commonly in preconsonantal
ones. However, since for second-generation informants the number of rhotic to-
kens in prepausal and preconsonantal position is very low (3 and 2, respectively),
the apparent pattern of prepausal environments favouring rhoticity over precon-
sonantal ones may be a chance effect for this group. A more plausible interpreta-
tion may therefore be that the two groups are characterised by two fundamentally
different patterns: while the first generation exhibits variable presence of postvo-
calic /r/ in all three environments, the second generation shows near-categorical
absence of coda /r/ in preconsonantal and prepausal contexts and variable but
frequent occurrence of rhotic variants in linking /r/ contexts. Rhotic variants are
more common in linking /r/ contexts for the second generation (77.6 percent)
than for the first (70.3 percent), but the difference is rather small. Furthermore,
12 instances of unetymological word-final prevocalic /r/ occurred in the sections
of the interviews that were analysed. Even if no more than a few tentative spec-
ulations can be made about intrusive /r/ here, it is noteworthy that they were all
produced by second-generation speakers.
Due to the low number of overt realisations of prepausal and preconsonantal
coda /r/ produced by the second generation, word-final prevocalic contexts were
included in the analysis of the phonetic quality of rhotic tokens. Once again, Ta-
ble 4.3 and Figure 4.4 reveal a sharp contrast between the generations: whereas
the first generation displays an overwhelming tendency towards the use of taps
and trills (93.9 percent), the second generation exhibits a diametrically opposed,
almost exclusive, preference for approximants (93 percent).
Finally, an inspection of the rhoticity levels of individual subjects shows that
the two generations also clearly differ in terms of the amount of variation found
Table 4.3â•‡ Phonetic realisation of rhotic tokens by generation (including linking /r/
tokens)
Generation Rhotic tokens
taps and trills % approximants % N
First 93.9 â•⁄6.1 132
Second â•⁄7.0 93.0 â•⁄71
1.0
App
Phonetic realisation of rhotic tokens
0.8
0.6
Tap trill
0.4
0.2
0.0
First Second
Generation
Figure 4.4â•‡ Phonetic realisation of rhotic tokens by generation (including linking /r/
tokens; tapTrill = taps and trills, app = approximants)
within each group (see Table 4.4 and Figure 4.5). Considering that the second
generation displays an average frequency of rhotic tokens of 0.7 percent, it is not
Table 4.4â•‡ Percentage of rhotic tokens by speaker (R = rhotic token; N = 50 tokens per
speaker)
First generation Second generation
speaker R% speaker R%
S â•⁄0 C 0
G â•⁄2 H 0
T â•⁄6 L 0
W â•⁄8 M 0
K 12 Q 0
J 14 R 0
X 14 U 0
I 26 V 0
N 26 Y 0
E 52 A 2
F 52 B 2
D 2
O 2
P 2
70 Claudia Rathore

Zero First generation
0.8
0.4
R
0.0
E F G I J K N S T W X
Speaker
Second generation
Zero
0.8
0.4
R
0.0
A B C D H L M O P Q R U V Y
Speaker
Figure 4.5â•‡ Production of postvocalic /r/ by speaker (R = rhotic realisation, zero = null
realisation; N = 50 tokens per speaker)
surprising that inter-speaker variability is close to zero for this group, with scores
of individual participants ranging from 0 percent to 2 percent. In contrast, the
rhoticity levels of first-generation informants vary from 0% to 52%. Such a high
degree of inter-speaker variability raises the question of which sources of influ-
ence constrain the use of postvocalic /r/ within this group. This issue is addressed
in Section 5.2.
5.2 Variation within the first generation
Section 5.1 demonstrated that, for the first generation, rhoticity is clearly re-
lated to the phonological environment, with coda /r/ being articulated consid-
erably more often in prepausal contexts (37.7 percent) than in preconsonantal
ones (14.9 percent). Table 4.5 and Figure 4.6 suggest that gender, the second of
the independent variables examined here, only shows a weak correlation with
the frequency of postvocalic /r/. Female participants exhibit a somewhat higher
rhoticity level than male speakers (21 percent and 17.2 percent, respectively), but
the difference is small (3.8 percent).
In contrast, the first generation’s production of postvocalic /r/ displays a clear
correlation with age, the third predictor included in the analysis. As Figure 4.7
Table 4.5â•‡ First-generation speakers’ production of postvocalic /r/ by gender (R = rhotic

Gender Postvocalic /r/
R% zero % N
female 21 79 300
male 17.2 82.8 250
1.0
0.8
Zero
0.6
0.4
0.2
R
0.0
Female Male
Gender
Figure 4.6â•‡ First-generation speakers’ production of postvocalic /r/ by gender (R = rhotic

shows, older participants tend to use rhotic variants much more frequently than
their younger counterparts.
Table 4.6 shows the results of the logistic mixed-effects regression analysis
which assessed the effect of following phonological environment, gender, and age
and their interactions on the occurrence of postvocalic /r/ among first-generation
speakers. Since none of the interactions turned out to be significant, they were
removed from the model. The output of the glmer function differs from that of
Goldvarb in several ways (see Johnson 2009). Two of these differences are briefly
explained here in order to facilitate the interpretation of the table. The first one is
that, while Goldvarb expresses coefficient units as factor weights or probabilities,
glmer returns coefficient estimates in log-odds, “the natural (base e) logarithm
72 Claudia Rathore
1.0
0.8
0.6
Zero
0.4
0.2
R
0.0
40 45 50 55 60 65 75 80
Age
Figure 4.7â•‡ First-generation speakers’ production of postvocalic /r/ by age (R = rhotic

of the odds, where the odds are the probability of an event occurring, divided by
the probability of it not occurring” (Johnson 2009, 361). Log-odds are used here
because the model includes a continuous predictor, age, for which coefficient es-
timates cannot be converted into factor weights. The second difference is that the
effects of factors are represented in terms of treatment contrasts. In this method,
one level of each factor (here: preconsonantal, female) is selected as the default
level. Default levels are combined in the model’s intercept and assigned a coeffi-
cient value of 0, while the other levels (prepausal, male) are assigned coefficient
values which represent the difference of each level from the default level (Baayen
2008, 102–103).
As shown in the second part of Table 4.6, the three independent variables
are clearly ranked in terms of statistical significance. Following phonological
environment and age emerge as very significant predictors of rhotic variants
(p < 0.001 and p = 0.001, respectively). In contrast, gender has no significant effect
on the production of rhotic tokens (p = 0.964). The log-odds coefficient estimates
reported in the first part of the table reveal that the effect size (relative strength)
of following phonological environment (1.776) is considerably larger than that
of gender (0.029), which is close to zero. As Johnson (2009, 362) observes, for
continuous predictors glmer estimates the probability of rule application for a
Table 4.6â•‡ Results of the logistic mixed-effects regression analysis estimating the effects
of following phonological environment, gender and age on the production of postvocalic
/r/ by first-generation participants
Generalised linear mixed model fit by the Laplace approximation
Model fit statistics
deviance â•⁄414.517
log likelihood –207.258
df 6
Total N 550
Random effects Variance Std. dev. N
lexical item 1.961 1.400 215
speaker 0.738 0.859 â•⁄11
Fixed effects Estimate Std. error z value p
intercept –10.145 1.924 –5.274 < 0.001
follow. phon. environment: prepausal â•⁄â•⁄1.776 0.399 â•⁄4.450 < 0.001
age: +1 â•⁄â•⁄0.119 0.028 â•⁄4.174 < 0.001
gender: male â•⁄–0.029 0.631 –0.046 â•⁄â•›â•›0.964
Overall significance of fixed effects* Chisq Chi df p
follow. phon. environment 18.413 1 < 0.001
age 10.334 1 â•⁄â•›â•›0.001
gender â•⁄0.002 1 â•⁄â•›â•›0.964
* Assessed via likelihood ratio chi-squared tests.
one-unit increase of the independent variable (here: one year). If one considers
that a yearly increase of 0.119 log-odds corresponds to a difference of 1.19 log-
odds between, say, 50- and 60-year-old speakers, the effect size of age appears to
be rather large too.
The coefficient estimates also provide insight into the internal constraint hi-
erarchy or direction of correlation of the three predictors. For following phono-
logical environment, the positive coefficient value for prepausal contexts (1.776)
indicates that coda /r/ is more likely to be pronounced in this position than in
preconsonantal environments. For gender, the negative value for males (–0.029)
suggests that male speakers are slightly less rhotic than female speakers (even
though, as seen above, the difference is not statistically significant). Lastly, the
positive coefficient estimate for age (0.119) points to a positive correlation with
rhoticity: the older the participants are, the greater the probability of their pro-
nouncing postvocalic /r/. The mixed-effects regression analysis thus confirms the
trends which emerged from the distributional analysis of the data.
74 Claudia Rathore
6. Discussion
6.1 First generation
The first research question of this paper was whether the production of postvo-
calic /r/ by the first generation resembles IndE, EAfE, or EMidE. The analysis in
Section 5 showed that, with an average rhoticity level of 19.3 percent and consid-
erable variation across individual speakers, the first generation can be regarded as
variably rhotic. Though comparably low, the use of rhotic variants clearly sets the
variety of English spoken by this group apart from EAfE and EMidE, which both
lack postvocalic /r/. Parallels therefore have to be sought in IndE, which is also
variably rhotic (except for South African Indian English and Indo-Fijian English).
There is great variation in the rhoticity levels reported in previous quantitative
work on IndE, with some groups, e.g. the Tibeto-Burman L1 subjects of Wiltshire
(2005) and the New Delhi informants of Chand (2010), displaying a high usage
of rhotic variants. Although frequencies clearly have to be compared with caution
across studies, it is striking that the overall percentage of rhotic tokens produced
by the first generation is very similar to the casual speech score of 20.5 percent
reported by Sahgal and Agnihotri (1988) for Delhi IndE speakers who attended
prestigious English-medium schools. Moreover, it closely resembles the average
frequency of 17 percent displayed by the Gujarati English speakers of Wiltshire
& Harnsberger (2006). This is particularly intriguing, considering the preponder-
ance of informants from a Gujarati background in the present study.
It is hard to draw any inferences about parallels with IndE, EAfE, and EMidE
from the first generation’s strong tendency to articulate /r/ in linking /r/ contexts
(70.3 percent), since little is known about the occurrence of this phenomenon in
the three varieties and overt realisations of /r/ in this position are found in both
rhotic and non-rhotic accents. Further insights may be gained, however, from the
phonetic quality of rhotic tokens. The overwhelming preference for tapped and
trilled realisations (93.9 percent) displayed by the first generation clearly diverges
from EMidE, which predominantly shows approximant realisations of /r/. While
alternation between taps, trills, and approximants is attested in both IndE (includ-
ing diasporic varieties) and EAfE, taps and trills are the variants most commonly
reported for the former. Moreover, there is once again a striking parallel with
the results of Wiltshire and Harnsberger (2006), whose Gujarati English speakers
used taps most often, followed by trilled and approximant realisations.
As regards inter-speaker variability within the first generation, the tendency
of this group to pronounce postvocalic /r/ more frequently in prepausal than pre-
consonantal position confirms a pattern well attested in sociolinguistic research
on coda /r/ (e.g. Labov 1966; 1972; Barras 2010; Nagy & Irwin 2010). Chand
(2010) reported that phonetic environment (coded as a combination of preceding

vowel and following segment) significantly constrained /r/-deletion among her
Delhi IndE speakers. Although differences in categorisation make a comparison
of the ranking of factor levels difficult, it is noteworthy that in her study precon-
sonantal coda contexts, either with a full vowel (fourth) or a schwa nucleus (bird),
favoured zero realisations of coda /r/ the most, thus pointing to a similar pattern.
In contrast, the absence of a clear gender difference in rhoticity levels seems
to diverge from the patterns reported for postvocalic /r/ in IndE. Both Sahgal and
Agnihotri (1988) and Chand (2010) found gender to be an important predictor of
the occurrence of coda /r/ among their informants, with female speakers showing
a greater preference for zero realisations, the prestige variants, than male speak-
ers. It may very well be, however, that in the present study the apparent lack of
gender-related variation is simply due to the different age distributions of female
and male speakers: whereas four women out of six are older than 60, only one
man out of five is. Since rhoticity strongly correlates with age, the predominance
of older female speakers is therefore likely to have contributed to a higher overall
rhoticity level for women. Due to the small number of speakers, it is not possible
to distinguish a “true” gender effect from individual variation here, but it is worth
mentioning that if one compares female and male speakers of similar age, the
constraint ranking for gender is actually reversed, with women using fewer rhotic
variants than men (see Rathore 2013).
Lastly, age-related variation in the use of postvocalic /r/ is attested in research
on IndE as well. In Sahgal and Agnihotri’s (1988) investigation, younger speakers
pronounced coda /r/ less frequently than older speakers, whereas the informants
of Chand (2010) displayed a decrease in rhoticity from the oldest to the mid-
dle-aged generation, followed by an increase in rhotic variants in the youngest
generation. In both studies, the differences across age groups were used as ap-
parent-time evidence for changes in progress. It seems, however, likely that for
first-generation East African Indians in Leicester the correlation between age and
rhoticity is in fact a reflection of the effect of other constraints. In their investi-
gation of the Southall Punjabi community in London, for instance, Sharma and
Sankaran (2011) found that time spent in the UK was the most significant social
factor predicting the use of retroflex /t/ among first-generation informants:
India-born participants who have lived in the U.K. for 3–12 years show a sharp
decline in use of /ʈ/. These individuals are often negotiating work situations with
British English speakers and struggling to find a place in their new environment.
(…) Intriguingly, /ʈ/ use increases again among the long-stay (> 12 years) group.
These individuals have often settled with families in very Asian networks and
may have less need to accommodate to British English or may have regained
confidence in their original variety. (Sharma & Sankaran 2011, 417–418)
76 Claudia Rathore
It is difficult to assess the influence of length of residence on the use of postvocalic

/r/ among the first-generation participants of the present study because, with the
exception of two speakers who had lived in Leicester for 11 and 24 years, respec-
tively, they had all lived in the city for 30 years or more at the time of the interview
and were well settled in their community. The similar dates of migration of East
African Indians to the UK make it (potentially) possible, though, to investigate
the influence of another factor related to age, namely age of arrival. Since in this
study the majority of informants moved to Leicester in their early twenties, a larg-
er speaker sample would be needed to evaluate the effect of this predictor system-
atically. Nevertheless, it is striking that the four participants who arrived at a later
age (31–62) are also the ones who display the highest rhoticity levels (speakers I,
N, E and F; see Table 4.4 and Figure 4.5).
It is also possible, however, that the correlation between age and rhoticity
results from different amounts of contact with, as well as a changing need to ac-
commodate to, the local variety of BrE in the course of individuals’ lives. Within
the first generation, speakers aged above 60 articulate postvocalic /r/ much more
frequently (34 percent) than those who are aged 60 or younger (7 percent). This
is remarkable in that this age division coincides with the distinction between in-
formants who are part of the working population and those who are retired. As
mentioned above, Sharma and Sankaran (2011) note that the Southall subjects
who made least use of retroflex /t/ often had to deal with BrE speakers in the em-
ployment domain. Auer, Barden and Grosskopf (2000) report a case study of an
eastern German migrant in western Germany who had adopted standard variants
of several phonological variables (cited in Kerswill 2006, 2277). After losing his
job, though, his social network and attitude to life in western Germany changed
and he went back to using pronunciation features characteristic of his home prov-
ince. Hence, it could be that retired first-generation informants increase their use
of rhotic variants as the employment domain becomes irrelevant to them and
they feel less need to accommodate to non-rhotic EMidE.
In sum, the variable rhoticity of first-generation participants thus points to
the maintenance of IndE patterns. Of course, a certain degree of influence from
non-rhotic EAfE and EMidE cannot be excluded on linguistic grounds alone, and
some accommodation to the local variety of BrE is indeed likely for this group.
As shown above, however, their relatively low average rhoticity level and strong
preference for taps and trills are remarkably similar to the patterns reported for
IndE speakers from comparable socioeconomic and/or regional-linguistic back-
grounds (with the exception of Chand 2010). Both the apparent absence of a gen-
der effect on the first generation’s use of postvocalic /r/ and the strong correlation
between age and rhoticity displayed by this group are intriguing and deserve fur-
ther investigation with a larger, more balanced speaker sample.
6.2 Second generation
The second research question was whether and how the production of postvocalic
/r/ changes across generations. The analysis in Section 5.1 revealed that both first-
and second-generation speakers display a strong tendency towards pronouncing
coda /r/ in linking /r/ environments (77.6 percent for the latter). Beyond this (su-
perficial) similarity, however, second-generation participants clearly diverge from
their parents’ generation in showing near-categorical non-rhoticity (0.7 percent)
and a strong preference for approximant realisations of rhotic tokens (93 percent).
Moreover, they also produced 12 rhotic tokens of unetymological word-final /r/
in prevocalic environments. These differences suggest that this group has not
adopted the patterns of postvocalic /r/ found among first-generation informants.
Instead, the second generation’s use of this variable is most plausibly accounted
for in terms of accommodation to EMidE, since non-rhoticity and approximant
realisations of /r/ are characteristic of this variety as well, and both linking /r/
and intrusive /r/ are widely found in non-rhotic native English dialects, includ-
ing those of the North of England (see Section 3.2). As regards the frequency of
linking /r/ in BrE, there seems to be considerable cross-dialectal variation. For
instance, Foulkes (1997) found that, with the exception of working-class males,
younger speakers in Derby exhibited levels of linking /r/ of around 90 percent
whereas younger speakers in Newcastle upon Tyne produced much lower scores
(30–40 percent for working-class speakers and around 55 percent for middle-class
speakers). It is thus difficult to assess whether the second generation’s use of this
feature matches the patterns found in BrE in quantitative terms.
Non-rhoticity as a result of accommodation to the local variety of BrE is
also attested among UK-born Punjabi-English bilinguals in London. Hirson and
Sohail (2007) found that, while subjects who self-identified as Asian tended to
articulate coda /r/, those who self-identified as British Asian displayed non-rhotic
accents. The similarity is noteworthy in that, as mentioned in Section 2.2.1, the
second-generation participants of the present study almost invariably described
themselves as British Asian too. Moreover, the British Asian-identified group of
Hirson and Sohail (2007) showed less variation in the phonetic realisation of /r/
than Asian-identified informants. A clear preference for approximant /r/ was also
reported for second-generation subjects from the London Bangladeshi commu-
nity, who showed similar accommodation to local BrE patterns (McCarthy et al.
2011).
78 Claudia Rathore
6.3 The role of socio-historical factors
The first generation’s retention of IndE patterns of postvocalic /r/ confirms the
maintenance of L1-derived features reported for other first-generation South
Asian migrants in Britain (Evans et al. 2007; McCarthy et al. 2011; Sharma &
Sankaran 2011) and elsewhere (Sharma 2005). As noted in Section 2.2.1, however,
first-generation participants expressed strong emotional affinity with East Africa
and a clear sense of distinctiveness from direct South Asian migrants. Why, then,
is this not reflected in their use of postvocalic /r/?
The social history of the community accounts for this pattern. Both the re-
search on the history of East African Indians (e.g. Gregory 1993) and the evidence
from the interviews indicate that Indians tended to live in ethnically segregated
areas in East Africa and to display a strong orientation towards their own com-
munities. Intermarriages were rare and the limited contact that occurred with the
African population was generally of a relatively superficial type, often in the form
of dealings with customers or interaction with African employees. As knowledge
of English was not widespread in the African population at the time (Schmied
2008, 153), communication mostly took place in Kiswahili. Furthermore, dur-
ing the colonial period there was continuous input of English from the Indian
subcontinent through incoming migrants. Indians who grew up in East Africa
generally acquired English in school contexts and, crucially, the transmission of
the language occurred mainly via teachers imported from South Asia. It was also
quite common for migrants to return to their original homeland for further edu-
cation. These factors are likely to have led to limited contact with, and influence
from, EAfE, and to have favoured the retention of IndE forms. The maintenance
of ties with the Indian subcontinent in Britain seems to have further supported
this trend.
The second generation’s adoption of EMidE patterns of postvocalic /r/ par-
allels the findings of Evans et al. (2007) and McCarthy et al. (2011), whose sec-
ond-generation subjects accommodated to the local south-eastern English dialect
for the features analysed. As discussed in Section 1, however, the retention (and
reallocation) of L1-derived accent traits by locally-born speakers is widely attest-
ed in research on the British Asian diaspora. It seems likely that social factors,
such as affiliation with a British Asian identity (see Section 6.2) or socioeconom-
ic status, play an important role in the second generation’s divergence from this
trend. With regard to socioeconomic background, the university-level education
and middle-class occupations of most second-generation participants clearly dis-
tinguish this group from the working-class communities that have often been in-
vestigated in previous research on British Asians (e.g. Heselwood & McChrystal
2000; Khan 2006; Cheshire et al. 2011). The parallels with the findings of Evans
et al. (2007) are particularly relevant in this respect, since the second-generation
speakers of the present study resemble the London Gujarati informants not only
in terms of regional-linguistic background but also in terms of socioeconomic
status. According to Evans et al. (2007, 1744), their subjects’ tendency to approx-
imate SSBE results from their social mobility. It is plausible to assume that this
factor has also led to accommodation to non-rhotic EMidE on the part of sec-
ond-generation East African Indians in Leicester.
However, a word of caution is required here. As retention of L1-derived ac-
cent traits has also been reported for locally-born South Asians from lower mid-
dle-class backgrounds (e.g. Sharma & Sankaran 2011), social mobility cannot be
the sole factor accounting for the second generation’s adoption of EMidE patterns.
Sharma’s (2011) study of the Punjabi community in Southall demonstrated that
British-born speakers may avoid L1-influenced features during a sociolinguistic
interview but use them in more personal contexts. One possibility that therefore
needs to be considered is that the second-generation participants of the present
study have adjusted their speech to the formality of the interview setting. Anoth-
er explanation concerns the linguistic variable chosen for analysis. It may well
be that second-generation East African Indians in Leicester approximate EMidE
patterns for postvocalic /r/ but diverge from this dialect along other dimensions,
doing identity work through the use of L1-derived traits such as monophthongal
face and goat, retracted realisations of /t/ or clear variants of coda /l/. All in all,
it thus seems unlikely that a single factor accounts for the second generation’s use
of postvocalic /r/.
7. Conclusion
This paper explored the question whether the dynamics of dialect variation and
change in double diaspora situations differ from those found in direct migrant
communities. To this end, I investigated variation in the use of postvocalic /r/
across two generations of East African Indians in Leicester. The results indicate
that the production of postvocalic /r/ by first-generation speakers closely paral-
lels the patterns attested for this variable in IndE. These patterns have, however,
not been acquired by second-generation participants, who have accommodated
to EMidE usage instead. The maintenance of L1-derived features is well attested
among other first-generation South Asian migrants and, although British-born
speakers were often found to diverge from the local dialect, convergence towards
the use of BrE forms is also attested for some second-generation British Asians,
particularly those from socially mobile backgrounds. For postvocalic /r/ at least,
the linguistic outcomes of the secondary diaspora situation experienced by East
80 Claudia Rathore
African Indians in Leicester hence do not seem to differ substantially from the
patterns reported for direct British Asian migrants. Intriguingly, the first genera-
tion’s retention of IndE forms suggests that, even if previous migration to another
country results in a strong sense of a separate identity among twice migrants, this
does not necessarily lead to linguistic divergence. Since the patterns of variation
exhibited by East African Indians in Leicester appear to have been determined
by socio-historical factors, it could very well be, though, that other outcomes are
found in double diaspora situations shaped by different conditions, e.g. Fiji Indi-
ans in New Zealand (Hundt, this volume) or Suriname Indians in the Netherlands.
References
Auer, Peter, Barden, Birgit & Grosskopf, Beate. 2000. Long-term linguistic accommodation
and its sociolinguistic interpretation: Evidence from the inner-German migration after
the Wende. In Dialect and Migration in a Changing Europe, Klaus Mattheier (ed.), 79–98.
Frankfurt: Peter Lang.
Baayen, R. Harald. 2008. Analyzing Linguistic Data: A Practical Introduction to Statistics Using
R. Cambridge: CUP. DOI: 10.1017/CBO9780511801686
Bansal, Ram Krishna. 1990. The pronunciation of English in India. In Studies in the Pronunci-
ation of English: A Commemorative Volume in Honour of A.C. Gimson, Susan Ramsaran
(ed.), 219–230. London: Routledge.
Barras, William Simon. 2010. The Sociophonology of Rhoticity and R-Sandhi in East Lanca-
shire English. PhD dissertation, University of Edinburgh.
Bates, Douglas, Maechler, Martin & Bolker, Ben. 2011. lme4: Linear mixed-effects models using
S4 classes. R package version 0.999375-42. <http://CRAN.R-project.org/package=lme4>
(2 January 2013).
Britain, David & Trudgill, Peter. 1999. Migration, new-dialect formation and sociolinguistic
refunctionalisation: Reallocation as an outcome of dialect contact. Transactions of the Phil-
ological Society 97(2): 245–256. DOI: 10.1111/1467-968X.00050
Broadbent, Judith. 1991. Linking and intrusive r in English. University College London Working
Papers in Linguistics 3: 281–302.
Brown, Judith M. 2006. Global South Asians: Introducing the Modern Diaspora. Cambridge:
CUP. DOI: 10.1017/CBO9780511807657
Chand, Vineeta. 2010. Postvocalic (r) in urban Indian English. English World-Wide 31(1): 1–39.
DOI: 10.1075/eww.31.1.01cha
Cheshire, Jenny, Kerswill, Paul, Fox, Sue & Torgersen, Eivind. 2011. Contact, the feature pool
and the speech community: The emergence of multicultural London English. Journal of
Sociolinguistics 15(2): 151–196. DOI: 10.1111/j.1467-9841.2011.00478.x
CIEFL. 1972. The Sound System of Indian English [Monograph 7]. Hyderabad: Central Institute
of English and Foreign Languages.
Evans, Bronwen G., Mistry, Ajay & Moreiras, Caroline. 2007. An acoustic study of first- and
second-generation Gujarati immigrants in Wembley: Evidence for accent convergence? In
Proceedings of ICPhS XVI, Saarbrücken, 6–10 August 2007, 1741–1744.
Foulkes, Paul. 1997. English [r]-sandhi. A sociolinguistic perspective. Histoire Epistémologique

Langage 19(1): 73–96. DOI: 10.3406/hel.1997.2573
Fox, Sue P. 2007. The Demise of Cockneys? Language Change among Adolescents in the ‘Tra-
ditional’ East End of London. PhD dissertation, University of Essex.
Gargesh, Ravinder. 2008. Indian English: Phonology. In Varieties of English, Vol. 4: Africa, South
and Southeast Asia, Rajend Mesthrie (ed.), 231–243. Berlin: De Gruyter.
Gregory, Robert G. 1992. The Rise and Fall of Philantropy in East Africa: The Asian Contribu-
tion. New Brunswick NJ: Transaction.
Gregory, Robert G. 1993. South Asians in East Africa: An Economic and Social History, 1890–
1980. Boulder CO: Westview Press.
Harris, Roxy. 2006. New Ethnicities and Language Use. Houndmills: Palgrave.
DOI: 10.1057/9780230626461
Heselwood, Barry & McChrystal, Louise. 2000. Gender, accent features and voicing in Panja-
bi-English bilingual children. Leeds Working Papers in Linguistics and Phonetics 8: 45–70.
Hirson, Allen & Sohail, Nabiah. 2007. Variability of rhotics in Punjabi-English bilinguals. In
Proceedings of ICPhS XVI, Saarbrücken, 6–10 August 2007, 1501–1504.
Hughes, Arthur, Trudgill, Peter & Watt, Dominic. 2005. English Accents and Dialects: An In-
troduction to Social and Regional Varieties of English in the British Isles, 4th edn. London:
Hodder Arnold.
Johnson, Daniel Ezra. 2009. Getting off the GoldVarb standard: Introducing Rbrul for mixed-ef-
fects variable rule analysis. Language and Linguistics Compass 3(1): 359–383.
DOI: 10.1111/j.1749-818X.2008.00108.x
Kerswill, Paul. 2006. Migration and language. In Sociolinguistics/Soziolinguistik: An Interna-
tional Handbook of the Science of Language and Society, Vol. 3, 2nd edn, Klaus Mattheier,
Ulrich Ammon & Peter Trudgill (eds), 2271–2285. Berlin: De Gruyter.
Kerswill, Paul, Torgersen, Eivind & Fox, Susan. 2008. Reversing ‘drift’: Innovation and diffusion
in the London diphthong system. Language Variation and Change 20: 451–491.
DOI: 10.1017/S0954394508000148
Khan, Arfaan. 2006. A Sociolinguistic Study of Birmingham English: Language Variation and
Change in a Multi-Ethnic British Community. PhD dissertation, Lancaster University.
Kirkham, Sam. 2011. The acoustics of coronal stops in British Asian English. Proceedings of
ICPhS XVII, Hong Kong, 17–21 August 2011, 1102–1105.
Labov, William. 1966. The Social Stratification of English in New York City. Washington DC:
Center for Applied Linguistics.
Labov, William. 1972. Sociolinguistic Patterns. Philadelphia PA: University of Pennsylvania
Press.
Marett, Valerie. 1989. Immigrants Settling in the City. London: Leicester University Press.
Marett, Valerie. 1993. Resettlement of Ugandan Asians in Leicester. Journal of Refugee Studies
6(3): 248–259.
McCarthy, Kathleen, Evans, Bronwen G. & Mahon, Merle. 2011. Detailing the phonetic en-
vironment: A sociophonetic study of the London Bengali community. In Proceedings of
ICPhS XVII, Hong Kong, 17–21 August 2011, 1354–1357.
Mesthrie, Rajend. 2008. Indian South African English: Phonology. In Varieties of English, Vol. 4:
Africa, South and Southeast Asia, Rajend Mesthrie (ed.), 188–99. Berlin: De Gruyter.
Nagy, Naomi & Irwin, Patricia. 2010. Boston (r): Neighbo(r)s nea(r) and fa(r). Language Vari-
ation and Change 22: 241–278. DOI: 10.1017/S0954394510000062
82 Claudia Rathore
Nihalani, Paroo, Tongue, Ray K., Hosali, Priya & Crowther, Jonathan. 2004. Indian and British
English: A Handbook of Usage and Pronunciation, 2nd edn. Oxford: OUP.
ONS (Office for National Statistics). 2012. Census 2011. <http://www.ons.gov.uk/ons/guide-
method/census/2011/index.html> ( 2 January 2013).
Oonk, Gijsbert. 2006. East Africa. In The Encyclopedia of the Indian Diaspora, Brij V. Lal, Peter
Reeves & Rajesh Rai (eds), 254–262. Honolulu HI: University of Hawaii Press.
Phillips, Deborah. 1981. The social and spatial segregation of Asians in Leicester. In Social In-
teraction and Ethnic Segregation, Peter Jackson & Susan J. Smith (eds), 101–21. London:
Academic Press.
Pingali, Sailaja. 2009. Indian English. Edinburgh: EUP.
DOI: 10.3366/edinburgh/9780748625949.001.0001
R Development Core Team. 2011. R: A language and environment for statistical computing.
R Foundation for Statistical Computing, Vienna, Austria. <http://www.R-project.org/>
(2 January 2013).
Rampton, Ben. 1995. Crossing: Language and Ethnicity among Adolescents. London: Longman.
Rathore, Claudia. 2013. Dialect Variation and Change in the Diaspora: A Sociophonetic Study
of the East African Indian Community in Leicester, UK. PhD dissertation. University of
Zurich.
Robinson, Vaughan. 1993. Marching into the middle classes? The long-term resettlement of
East African Asians in the UK. Journal of Refugee Studies 6(3): 230–247.
Sahgal, Anju & Agnihotri, Rama Kant. 1988. Indian English phonology: A sociolinguistic per-
spective. English World-Wide 9(1): 51–64. DOI: 10.1075/eww.9.1.04sah
Sankoff, David, Tagliamonte, Sali & Smith, Eric. 2012. Goldvarb LION: A variable rule applica-
tion for Macintosh. Department of Linguistics, University of Toronto. <http://individual.
utoronto.ca/tagliamonte/goldvarb.htm> (2 January 2013).
Schmied, Josef. 2008. East African English (Kenya, Uganda, Tanzania): Phonology. In Varieties
of English, Vol. 4: Africa, South and Southeast Asia, Rajend Mesthrie (ed.), 150–163. Berlin:
De Gruyter.
Schneider, Edgar W. 2004. Global synopsis: Phonetic and phonological variation in English
world-wide. In A Handbook of Varieties of English, Vol 1: Phonology, Edgar W. Schneider,
Kate Burridge, Bernd Kortmann, Rajend Mesthrie & Clive Upton (eds), 1111–1137.
Berlin: De Gruyter.
Sharma, Devyani. 2005. Dialect stabilization and speaker awareness in non-native varieties of
English. Journal of Sociolinguistics 9(2): 194–224. DOI: 10.1111/j.1360-6441.2005.00290.x
ual change in London Asian speech. Language Variation and Change 23: 399–428.
DOI: 10.1017/S0954394511000159
Simo Bobda, Augustin. 2001. East and Southern African English accents. World Englishes
20(3): 269–284. DOI: 10.1111/1467-971X.00215
Stuart-Smith, Jane, Timmins, Claire & Alam, Farhana. 2011. Hybridity and ethnic accents: A
sociophonetic analysis of ‘Glaswasian’. In Language Variation – European Perspectives, III:
Selected Papers from the 5th International Conference on Language Variation in Europe
(ICLaVE 5), Copenhagen, June 2009 [Studies in Language Variation 7], Frans Gregersen,
Jeffrey K. Parrott & Pia Quist (eds), 43–57. Amsterdam: John Benjamins.
DOI: 10.1075/silv.7.04stu
Tent, Jan & Mugler, France. 2008. Fiji English: Phonology. In Varieties of English, Vol. 3: The
Pacific and Australasia, Kate Burridge & Bernd Kortmann (eds), 234–266. Berlin: De
Gruyter.
Wells, John C. 1982. Accents of English: An Introduction. Cambridge: CUP.
DOI: 10.1017/CBO9780511611759
Wiltshire, Caroline R. 2005. The ‘Indian English’ of Tibeto-Burman language speakers. English
World-Wide 26(3): 275–300. DOI: 10.1075/eww.26.3.03wil
Wiltshire, Caroline R. & Harnsberger, James D. 2006. The influence of Gujarati and Tamil L1s
on Indian English: A preliminary study. World Englishes 25(1): 91–104.
DOI: 10.1111/j.0083-2919.2006.00448.x
Chapter 5
Sociophonetics and the Indian diaspora

The NURSE vowel and other selected features
in South African Indian English*
Rajend Mesthrie and Alida Chevalier

This chapter explores the potential of phonetic and sociophonetic research in

characterising varieties of English in the Indian diaspora, using South African
Indian English (SAIE) as a focal point. We first identify key phonetic features
that lend themselves to such comparative study. In this regard we note that ret-
roflexes are recessive in SAIE, aspiration shows patterns that are intermediate
between Indian English (IE) and varieties like White South African English
(White SAE), and interdental fricatives /ð/ and /θ/ are closer to IE, while dif-
fering in respect of aspiration. In characterising the differences between IE as
usually described in the literature and SAIE, three factors are crucial: (a) the
give-and-take between the English of speakers of North Indian and South Indi-
an origins in South Africa, (b) the influence of varieties of South African Eng-
lish, and (c) internal sociolinguistic developments within the SAIE community
relating to age, gender and class. This last aspect is illustrated in detail in the
second half of the chapter via a sociophonetic study of the nurse vowel.
Keywords: age, aspiration, gender, NURSE vowel, retroflexion, social class,

sociophonetics, South African Indian English, (th) variation
* We would like to acknowledge the National Research Foundation’s South African Research
Chairs Initiative (64805) for their financial support. Thanks also to Martin Hilpert, Tracey
Toefy and Yolandi Ribbens-Klein for assistance on matters statistical. In particular we thank
the interviewees from Chatsworth and surrounding areas of Durban for taking the time to
speak to us.
86 Rajend Mesthrie and Alida Chevalier
1. Introduction
This paper focuses on an area in which there has been little research to date,
viz. the comparative phonetics of the English of Indians in the diaspora (EID in
short). The default assumption in contact sociolinguistics is that diasporic com-
munities will pick up the dominant language of a new territory to varying degrees
in the first (migrating) generation, and that the second generation will be fluent
in that language, and approximate the sociophonetics of one of its local varieties,
whilst possibly retaining substrate features (see Rayfield 1970 for the erstwhile
USA melting pot). For EID this substrate influence could be directly from an
ancestral Indian language, or via IE (i.e. the English of India) as matrilect. The
robustness of an identifiable IE in the diaspora will depend on an array of social
and demographic factors like age of arrival of individuals, size and cohesion of
the community, degrees of social acceptance or seclusion in the wider society,
access to education, contact with the Indian homeland, socio-economic mobility
and degree of maintenance of an ancestral Indian language. In addition, language
attitudes are important as Sharma (2005) shows in her detailed study of the rela-
tion between EID and the more dominant American English in the San Francisco
area. There are further complexities like (a) the existence of other influential lan-
guages than English, and (b) the wide range of possible substrate Indian languag-
es. Table 5.1 gives a sense of the differential linguistic influences on the English of
Indians in five selected diaspora contexts.
Together with other social factors (like intense segregation in South Africa)
these influences result in clearly differentiable varieties of EID. For example, the
phonetics of English of Indians in Mauritius is noticeably influenced by French
and French Creole, and in Trinidad by general Caribbean English and Creole. In
the USA Indians of the second generation onwards are by and large fully assimi-
lated to US English in a way that they are not to the local variety of White English
in South Africa. In the UK, although a great deal of assimilation to the dominant
local forms of English is apparent, there are also relatively robust forms of survival
of South Asian English in its own right in various communities (see Fox 2007 for
London’s East End) or as a stylistic option within a repertoire that includes more
dominant forms of British English (see Sharma 2011; Rampton 1995).
For a comparative global comparison of EID, two initial simplifying assump-
tions are necessary: firstly, that IE is a sort of matrilect for the diasporic varieties,
or at least the most viable centre of comparison.1
1. Even if many migrants in the nineteenth-century diasporas (see Mesthrie, this volume)
did not speak English, the few who did would have been influential from the start as teachers,
interpreters, translators and community leaders (see Badassy 2002 for South Africa).
Chapter 5.â•‡ Sociophonetics and the Indian diaspora 87
Table 5.1 Linguistic influences on English amongst Indians in five selected diasporic communities
Territor y UK USA Trinidad Mauritius South Africa
Main substrates for Indians Panjabi Various (Gujarati, Bhojpuri Bhojpuri Tamil
Gujarati Kannada) etc. Telugu
Bengali Bhojpuri
Gujarati
Extent of Indian English as matrilect Great Great Not great (19th C) Not great (19th C) Not great (19th C)
in migrating generation (mid-20th C on) (mid-20th C on)
Dominant variety of English UK varieties US varieties English Creole and None British-oriented
British English (Some British English South African
previously) English
Other influential languages – – – French Afrikaans
Creole-French
Other languages of lesser influence – Spanish – Chinese etc. Zulu
on EID Xhosa
Tswana
Sotho, etc.
English as L1 among 2nd or + + + – +
subsequent generations [with Creole]
The second simplifying assumption is that IE is a single entity, despite differ-

ences according to Indic, Dravidian, Tibeto-Burman and Austronesian substrates
(see Gargesh 2004 on some regional differences within IE; Kachru 1983, 31) on
Dravidian characteristics; Wiltshire 2005 on Tibeto-Burman influences). We can
then assess similarities and differences from a broad template of well-known IE
features. Using these descriptions we could make initial enquiries along the fol-
lowing lines for segmental phonological similarities and differences:
a. Consonants:
– To what extent is retroflexion maintained as a prominent feature of an
EID?
– Is aspiration used differently from the local prestige variety?
– What are the reflexes of /θ/ and /ð/?
– Are /v/ and /w/ always differentiated?
– Is /h/ usually breathy voiced?
– Has postvocalic /r/ been maintained (in the Indian form of a partially
devoiced trill)?
b. Vowels:
– Are /eɪ/ and /oʊ/ diphthongal or long monophthongs?
We may ask what local features have been adopted by an EID that contribute to
its distinctiveness not only from IE but from other EIDs: e.g. glottal stops in the
UK; fronted bath and t-tapping in the USA; short front-vowel raising in South
Africa, Australia and New Zealand; and Canadian raising of diphthong onsets in
house and price.
Finally, we would need to pay attention to the internal dynamics of particular
EIDs in relation to ongoing sociolinguistic differentiation not reported to be sali-
ent in traditional IE – e.g. gender, social class and age variation. We will illustrate
this approach with reference to SAIE, relying on older descriptions in Bughwan
(1970) and Mesthrie (1992), and more recently, Chevalier (2011) for the speech
of younger people in a situation of rapid social change. Section 3 will deal with
the broad phonetic questions raised above, making special reference to the case
of South Africa, while Section 4 will offer a detailed socio-phonetic analysis of a
single variable that characterises further differentiation among young people, the
NURSE vowel. It is first necessary to give a brief overview of the development of
SAIE.
2. Background to SAIE
South Africa is historically an important territory in the Indian diaspora. For

much of the twentieth century it had the largest Indian community outside Asia
(though it has since been overtaken by South Asian immigration into the UK and
USA). Moreover, South Africa has been of special significance in Indian history
and politics since Gandhi’s 20-year sojourn in the country (1893–1913), which
fostered his philosophy of passive resistance to British colonisation. The com-
munity today labelled ‘Indian’ in South Africa dates to 1860, with the large-scale
movement of indentured workers from one British ruled territory (India) to an-
other (the colony of Natal). In the period 1860–1911, over 150,000 men, women
and children from villages and districts in what are now Bihar, Uttar Pradesh,
Tamil Nadu and Andhra Pradesh were transported to Natal to establish the cane-
fields that count amongst the most extensive in the world (in addition to smaller
plantations of tea, coffee and tobacco). The main languages brought to Natal were,
in descending order, Tamil, Bhojpuri-Hindi, Telugu and Urdu. Shortly afterwards
smaller numbers of people from north-west and central-west India migrated as
small-scale traders rather than semi-forced labour, speaking mainly Gujarati, but
also Konkani and Meman (Sindhi). The diaspora in South Africa thus was differ-
ent from many other territories outside Asia in having Tamil as the most spoken
language (in the region of 40%, Mesthrie 1991, 16), and no communities speaking
Panjabi, Marathi, Bengali, Malayalam or Kannada, i.e. languages which are well
represented in later diaspora communities in the UK and USA, cited above. The
Indian and South African governments put a halt to immigration in 1911 for
different reasons, and only a trickle of immigrants subsequently entered South
Africa via neighbouring African territories like Mozambique.2 Nevertheless, a
majority of immigrants chose to stay on in South Africa, where they were con-
tractually entitled to a small plot of land for market gardening upon completion
of their indenture.
Although Indian languages remain important for cultural purposes and are
taught in some schools at an elementary level, a slow process of language shift
has been in operation since the 1960s (Mesthrie 1992, 27–33), with a distinctive
variety of English arising in the province of Natal. The distinctiveness of this va-
riety was partly due to South Africa’s rigid policy of segregation or apartheid, and
the fact that not many Indians could afford to travel to India. In other provinces
Indians were influenced by Afrikaans as well, and overlap between ‘Coloured’ and
2. Whereas India was concerned with the maltreatment of migrants, within South Africa
there was growing Anti-Asiatic agitation once the plantation labour needs of the colony had
been satisfied.
Indian English is strong outside the province of KwaZulu-Natal (Mesthrie 2012). It

is the KwaZulu-Natal variety that has been best studied (Bughwan 1970; Mesthrie
1992; Pienaar 2007; Chevalier 2011) as the majority of Indians reside in this prov-
ince. With the weakening and eventual collapse of apartheid in the 1990s, citizens
of India speaking IE and a host of languages not previously well represented in
South Africa (especially Bengali, Panjabi and Malayalam) have contributed to the
enrichment of Indian culture and presence in the country. No study exists of the
give-and-take that is presumably occurring between IE and SAIE, an opportunity
awaiting future researchers.
3. Some salient features of SAIE phonetics in relation to IE
The ‘language shift’ version of South African Indian English spoken by a major-
ity today in KwaZulu-Natal offers a clear window for EID studies. This section,
which aims to give a broad overview of phonetic features, is based on Bughwan
(1970), and Mesthrie (1992, 2004). For convenience we refer to consonants in
capitals, especially P, T and K.
3.1 Retroflexion
The realisation of /t/ and /d/ in SAIE varies between alveolar and retroflex. Ret-
roflexes are receding in several ways:
a. Not all speakers use them;

b. Speakers who used retroflexes as children may gradually use less and less of
these in favour of alveolar (or post-alveolar) [t] and [d];
c. Retroflexion is not as strongly articulated as in IE in terms of degree of curl of
the tongue tip and force with which it articulates against the palate.
Nevertheless, many speakers who do not exhibit much retroflexion retain it un-
consciously as a stylistic device, for emphasis and reprimanding children (I’m
telling you). Retroflexion is also evident if one wants to sound more Indian in cer-
tain contexts (e.g. a funeral speech in which an Indian way of speaking English is
considered appropriate, in tandem with prayers in Sanskrit and/or another Indian
language or Arabic). Retroflex /ɳ/ can be heard in SAIE, usually homorganically
with a following [ʈ] and [ɖ] (e.g. in want, wand) or even in the neighbourhood of
a retroflex [ʈ] or [ɖ] (e.g. in ten, den). It occurs saliently in children’s articulation
of the word aunty [ɑːɳʈiː]. The same is true of /l/ and, to a much lesser extent, /s/,
which may be variably alveolar, post alveolar or retroflex in words like cold, tell
and past. Thus SAIE takes the middle ground between IE, in which retroflexes are
the norm, and White SAE, in which alveolar or even pre-alveolar pronunciations
of /t d s z n l r/ occur.
3.2 Aspiration
In IE /p t k/ are unaspirated in all positions, with Indic speakers identifying

these sounds with the unaspirated counterparts in their languages rather than
the strongly aspirated series /ph th kh/ (Kachru 1983, 29). Of the two Dravidian
languages spoken in South Africa, Tamil does not have a phonemic differentia-
tion between aspirated and unaspirated consonants, though Telugu does. As with
retroflexion there is variability amongst younger speakers with: (a) some having
the distribution of mainstream varieties of English (aspiration with initial /p t k/,
but not in other positions), (b) some occasionally having a stronger degree of
aspiration than in standard speech, and (c) others having no aspiration. Mesthrie
(2004, 962) notes an interesting intermediate system among some speakers in his
circle of acquaintances. This subset of speakers show the following patterns:
P is always unaspirated before /ɑː/, /ɔː/, /ʊ/, /eɪ/, /oʊ/ and /ɛə/. Thus park, pork,
put, pot, pay, poke, pair all have unaspirated initial P. Likewise P is always un-
aspirated before /r/ and /l/, e.g. in pray and play. This means that /r/ and /l/ are
voiced [in SAIE] in contrast to many varieties of English in which the aspiration
on initial consonants causes /r/ and /l/ to become voiceless. In all other contexts
whether P is aspirated or not, depends on the particular word. Taking P before
/e/ as an example, the following words always have aspiration – pen, pebble, pet;
whereas penny, pepper, petal, peck are always unaspirated. It has still to be re-
searched whether there is intra-speaker variability (i.e. pronouncing the same
word differently) or variation across speakers. Speakers who produce aspiration
invariantly with initial P, T, K would be judged as putting on a ‘Speech and Dra-
ma’ accent. The dialect has minimal pairs like phea and pee; phiece and piss (pro-
nounced [piːs]). It also has near-minimal pairs like phet and petal, phen and pencil.
(Mesthrie 2004, 962)
The social explanation for this pattern is that these speakers have struck a balance,
a kind of sociolinguistic fudging in the sense of Chambers and Trudgill (1998,
110–113), between the North and South Indian patterns, with no aspiration be-
fore certain back vowels, certain diphthongs and both liquids, and with aspiration
on P, T, K being word-dependent in all other contexts. Delbridge (2006) investi-
gated patterns of aspiration from Mesthrie’s (1992) data base, with a selection of
34 speakers, subject to analysis in the VARBRUL programme (Robinson et al.
2001). She found that P, T and K each patterned differently regarding aspiration.
For illustrative purposes we focus on just (initial) P. In the structural environment,

the factor group of ‘following segement’ was significant. Weightings above 0.5 oc-
curred for (in descending order): [uː], [ʊə] > [aʊ] > [ɪə] > [e] > [æ] > [aɪ] > [eɪ] >
[ɑː] > [iː]. The following had weightings below 0.5 showing low levels of aspira-
tion: [ɪ] > [ɜː], [ɛə] > [ɒ], [oʊ] > [ɔː], [ɔɪ] > [l], [r], [j] > [ʊ], [ʌ]. Gender, lectal level
and ancestral language were also identified as significant. Females aspirated more
than men (with VARBRUL weightings of 0.66 vs 0.41). Acrolectal speakers had
higher levels of aspiration than mesolectal speakers, who themselves had higher
levels than basilectal speakers (weightings of 0.68 vs 0.47 vs 0.07). For ancestral
languages, speakers of Tamil, Urdu, Telugu and Bhojpuri-Hindi (0.56 vs 0.56 vs
0.52 vs 0.49) had considerably higher weightings than Gujarati speakers (0.21).3
Not selected as significant were: preceding segment, age, education, social status,
and urban-rural residence. A follow-up study with present-day younger speakers
would be good: our impression is that younger middle-class speakers now have
higher levels of aspiration; we are less certain about working-class speakers.
3.3 Interdental fricatives
It is well known that /θ/ and /ð/ are rare in languages of the world, and that many
varieties of English replace them by some other close non-fricative realisation
(see Mesthrie & Bhatt 2008, 126–127). In Tamil phonology [ð] is an allophone of
/t̪/ in intervocalic position, except in clusters like /t̪t̪/ and /nt̪/ (which are realised
as [t̪t̪] and [nd̪] respectively (Asher 1985, 212). Gargesh (2004, 998) notes that
“the dental fricatives /θ/ and /ð/ are non-existent in Indian English. The aspirated
voiceless stop [t̪h] is realized for /θ/; the voiced stop [d] is realised for /ð/ – as in
thin = [t̪hin] and then = [d̪en]. In South India the alveolar stop /t/ is often used
instead of /θ/.” The option taken in SAIE is to use unaspirated dental stops [t̪]
and [d̪]. There appears to be no influence of the Tamil allophonic [ð] in SAIE
(or IE). Note that the SAIE [t̪] is unaspirated, in contrast to that of the IE matri-
lect. Although outsiders sometimes hear SAIE [t̪] and [d̪] as alveolars, there is no
mistaking the dental stop quality which brings these sounds very much within
the Indian (especially North Indian) phonological systems. In fact, alveolar /t/ in
some words of standard English is pronounced with a [t̪] in SAIE and can be said
to belong to the TH (i.e. /t̪/) class: tooth, teeth, tonsils, tongue, trash, taught all have
an initial voiceless dental stop.
3. Since Gujarati has aspiration of P, T, K, this result might be relatable to their greater contact
with India and IE, in which – as we have seen – P, T, K are unaspirated.
3.4 The treatment of /h/
In contrast to RP, general US English and White SAE, where /h/ is a voiceless fric-
ative, /h/ is usually a breathy-voiced fricative [ɦ] in SAIE. This is especially true
of speakers of North Indian ancestry. Working-class speakers of Tamil ancestry
tend to drop /h/, in keeping with Tamil phonology in which /h/ is absent, except
variably in some loanwords. A dropped /h/ is compensated for by a glottalised
onset, giving rise to a discernible rise in pitch, contrasting with the pitch lowering
effect of an initial [ɦ]. However, a dropped /h/ may also be replaced by the glides
[j] or [w] as in [jɪl] for ‘hill’ and [laɪvliwʊd] for ‘livelihood’. Hypercorrect forms
are also common in this subgroup with forms like [haʊtaʊs] occasionally occur-
ring for ‘outhouse’. It is remarkable that differences in the treatment of /h/ should
survive for 150 years in South Africa, and remain salient (especially in peoples’
stereotypes) as a marker of Tamil versus other ancestry. We caution, however, that
(a) not all people of Tamil ancestry ‘drop their aitches’; (b) those who do so do
it variably; (c) /h/ dropping leaves the phonetic trace of a glottalised onset, and
(d) that linguistic differences between North and South Indian ancestry are slowly
being levelled out.
3.5 The treatment of /v/ and /w/
Since these are not generally distinguished as phonemes in Indian languages, a

striking characteristic of IE is an overlap in the use of [v] and [w], either of which
may be found in words like vet, wet, vine, wine etc. Gargesh (2004, 998) gives the
example of variation in medial position of [pɑːvər] ~ [pɑːwər] ‘power’. V/w varia-
tion is reported for SAIE in the 1960s by Bughwan (1970). In our experience this
is characteristic mainly of older speakers, even in the 1970s, especially those with
some contacts with India. SAIE speakers born in South Africa generally differen-
tiate /v/ and /w/, although /v/ is a weakly articulated fricative [ʋ] rather than [v].
3.6 Postvocalic /r/
Unlike IE, where postvocalic /r/ may be realised as a partially devoiced fricative
or trill in far, car etc., SAIE is generally non-rhotic. A few exceptions occur with
some first names (Harshad, Kabir, Kirti), surnames (Varma, Parbhu), retentions
from Indian languages (karma ‘fate’, arthi ‘prayer’), the letter ‘r’ [ɑːr], and in Ire-
land [aɪrlənd].
3.7 Vowels
Trudgill and Hannah (1985, 106) note that in some varieties of IE /ɑː/ corresponds
to RP /ɑː/ and /ɔː/. SAIE does not have this characteristic, with pairs like Carl and
call, car and core always distinguishing /ɑː/ and /ɔː/. In one respect SAIE differs
from general SAE and RP: like IE and some varieties of US English it consistently
retains [ɑː] after w: hence wall, war, wash, walk, warm, wharf all have [ɑː], not
[ɔː]. Only the word water, which in vernacular styles usually has [ɑː], is corrected
to [ɔː] in formal speech. This word has the status of a social shibboleth, marking
status differences between speakers who show the corrected from and those who
do not. More recently, SAIE speakers attending non-racial high schools where,
however, the norms of White SAE prevail in the classroom, have learnt with sur-
prise that the word award has the prestige variant [ɔː], not the [aː] they were used
to hearing from their teachers in the Indian primary (or junior) schools.
Another merger noted by Trudgill and Hannah (1985, 106) concerns RP /ɒ/
and /æ/ becoming IE /a/. These two phonemes are distinct in SAIE, with pairs like
hot and hat clearly differentiable, usually occurring as [ɒ] and a slightly raised [æ̝]
respectively.
The often noted monophthongisation of [eɪ] and [oʊ] in IE was once charac-
teristic of older bilinguals in South Africa, but a clear diphthongal quality is the
norm for most SAIE speakers today. The SAIE pronunciations are in fact closer
to RP than the general White SAE forms, which have lowered onsets, hence [ɛɪ]
and [ɔʊ]. In the same vein, traditional SAIE does not show glide weakening in
the PRICE and MOUTH diphthongs, unlike what passes as the prestige norm in
White SAE.
The interplay between North and South Indian phonologies in SAIE merits
investigation, since speakers were roughly equally divided in numbers. Although
some differences persist and a trained ear can still make out whether speakers
older than about 40 are of North or South Indian background, the give-and-take
between phonologies is evident even for older speakers. One feature shows the in-
fluence of the Tamil majority within SAIE speakers: the realisation of final schwa
as an open [ɛ]. Older speakers of Bhojpuri-Hindi-Urdu use an [e] for the final
schwa of words like sugar, in contrast to [ɛ] among speakers of Dravidian back-
ground. Amongst young speakers the [ɛ] has won out, except in acrolectal varie-
ties which use a more standard [ə].
4. A case study: Variation in the NURSE vowel among younger SAIE

speakers
In this section, we examine in some detail a single vowel in SAIE with the inten-
tion of showing the complexities surrounding EID phonetics, if we factor in local
social variation. Developments among second and third generation speakers with
respect to the classic social variables of age, gender, status and class may increase
variation in an EID. We report on a study by Chevalier (2011) of NURSE among
younger SAIE speakers, using current approaches in sociophonetics.
As mid-central long vowels are not generally found in the Indian languag-
es of relevance to South Africa (Masica 1991; Asher 1985), NURSE must have
presented a potential source of difficulty to first generation learners. Gargesh
gives the following variants for IE: [ɜː] > [ʌ] > [əː] > [a]. In South Africa older
speakers recorded in the 1980s showed some variation, having [ɜː] or a front [eː].
Mesthrie (1992, 141) notes that [e:] occurs particularly among older speakers of
Bhojpuri-Â�Hindi-Urdu background in words like servant, bird etc. The difficulty of
approximating the general SAE system can be seen in older speakers of the period
showing a merger between NURSE and FAIR, with both frequently having [e:].
Other older speakers fluctuate between fair and fur (= [fɜː] ~ [feː]) and likewise
hair and her (= [hɜː] ~ [heː]). These might be seen as internal EID variants arising
from second language learners’ attempts to match their interlanguage system with
native vowel patterns.
At the same time other patterns enter the pool of variants. In the 1980s small-
scale, part-time Speech and Drama schools tried to inculcate a ‘posh’ RP-oriented
or Cultivated SAE accent to replace the more stigmatised SAIE among young
children.4 This resulted in the variants [ɛ̙ː], or possibly [aː], entering as a poten-
tial prestige pronunciation. These variants were perhaps also meant to avoid the
local White SAE raising, fronting and rounding of NURSE, which Lass (1995)
transcribes as [øː]. Gimson’s discussion of RP in the 1980s (1989, 124) shows
/ɜː/ to range considerably along a central line, as a sound ‘in the half-close region
or slightly above to the half open region or slightly below’ (as in Figure 5.1). The
SAIE variants include this range, but – as we shall see – also include realisations
that are further back (in the mid back region) and in the low front region.
Chevalier (2011) studied younger SAIE speakers in order to ascertain the ex-
tent to which traditional SAIE was accommodating to the wider norms of South
African society, given the collapse of apartheid in the 1990s, a few years after
4. White SAE is usually described in terms of a continuum comprising an RP-oriented ‘Culti-

vated’ variety, and more localised ‘General’ and ‘Broad’ varieties (Lass 1995), in ways parallel to
Australian English.
3:
††
(Mainstream)
Figure 5.1â•‡ /ɜː/ and its variants in RP (from Gimson 1989, 123)
Mesthrie’s fieldwork of 1987–1988. She also sought to ascertain whether social

class differences were increasing within SAIE as predicted by Mesthrie (1992,
221). In the period 2008 to 2009, Chevalier interviewed 22 SAIE speakers from
the city of Durban. Working-class speakers were interviewed in Chatsworth and
Merebank, while middle-class speakers came from various suburbs in Durban,
and were interviewed either in their homes or on the campus of the University
of Cape Town.5 Interviews followed a fairly traditional Labovian format, with a
word list at the end using an adaptation of Wells’ (1982) lexical sets.
This was the first close study of variation in SAIE to capitalise on advances
in sociophonetics, specifically the techniques of acoustic analysis (via PRAAT),
normalization (via the BARK METHOD) and allied statistical analysis (via R).
PRAAT measurements were undertaken manually of all clear tokens of nurse,
with segmentation at one third of the vowel duration. The number of formants
tracked by PRAAT was manually changed, and varied between 4 and 6, depend-
ing on which setting provided the clearest tracking of the formants per token.
We heeded acousticians’ advice against including tokens preceded or followed by
/l m n r/, since they affect readings of a preceding vowel. Note that since SAIE,
like other varieties of SAE, is non-rhotic, words like fur, cur, sir, herd all count as
valid tokens. (In fact they constitute the full set, since all NURSE words are spelt
with a following ‘r’.) A total of 813 tokens were available for analysis. Regarding
speaker variation the original sample of 22 speakers was reduced to 21, on account
of one outlier who did not sound as if he belonged to the SAIE community at all
5. Recording equipment was either a Marantz PMD661 MKII or an Olympus DS5000, relying
on their internal microphones.
and whose NURSE tokens gave surprising readings in the region of FLEECE.6
The sample was divided into 10 male and 11 female speakers. In terms of social
class – determined by speakers’ area of residence mainly, and confirmed by family
and schooling details supplied during the interview – the sample consisted of 11
middle-class (MC) and 10 working-class (WC) informants. All the MC speakers
had attended a formerly White school at some stage, none of the WC speakers had.
However, for two of the latter multi-racial educational contacts became stronger
at a tertiary institution.
Before speaker means could be compared they had to be subjected to a nor-
malization procedure to factor out the effects of variation in vocal tract sizes. We
tried out two methods (Watt & Fabricius and BARK) and came to the conclusion
that the latter was the more convenient choice at this stage of analysis, since it re-
lied only on NURSE tokens.7 The BARK normalization has the disadvantage that
the vowel space is warped so that it no longer corresponds with the traditional
16
i
14
y
12 u
e
Z3–Z1 (F1)
ø
10 ʉ ɤ o
ɑ
8
œ ɒ
6
œ
a
4
0 1 2 3 4 5 6 7 8 9 10
Z3–Z2 (F2)
Figure 5.2â•‡ The cardinal vowels represented in the BARK scale

(based on Thomas 2001, 146)
6. We hazard a guess that this might well be due to a minor speech defect not alluded to in the
interview.
7. By contrast, Watt-Fabricius (Watt & Fabricius 2002) relies on three other reference vowels
which were not easy to determine for SAIE. With more time and resources we hope to inves-
tigate SAIE – and other dialects of English in South Africa – via more comprehensive meth-
ods of automatic formant extraction via Forced Alignment and Vowel Extraction (FAVE; see
Rosenfelder et al. 2011).
16
14 i
12
e
Z3–Z1 (F1)
ø ɤ
10
ɑ
8
6
a
4
0 1 2 3 4 5 6 7 8 9 10
Z3–Z2 (F2)
Figure 5.3â•‡ The range of variation for NURSE by speaker means in the SAIE sample
(N = 21)
Table 5.2â•‡ Means and standard deviations for F1 and F2 or NURSE per speaker
Class Sex F1 Mean SD F2 Mean SD
MC F 9.085 0.605 3.972 0.796
MC F 6.219 0.646 3.024 1.660
MC F 6.321 1.011 2.839 1.377
MC F 7.756 1.200 3.512 1.255
MC F 8.725 0.782 2.536 0.827
MC F 6.158 0.504 2.507 0.917
MC M 9.333 0.804 5.174 1.820
MC M 8.602 0.590 3.475 0.747
MC M 9.575 0.482 4.432 0.523
MC M 8.650 0.747 3.760 0.647
MC M 6.030 0.793 2.962 1.567
WC F 7.145 1.244 3.019 1.473
WC F 8.669 0.807 4.335 0.821
WC F 7.463 1.610 3.045 1.438
WC F 6.411 1.557 3.461 1.486
WC F 8.509 1.249 2.792 1.266
WC M 9.644 0.378 4.628 0.886
WC M 9.739 0.529 4.564 1.257
WC M 9.870 0.588 4.586 1.012
WC M 8.621 0.432 4.171 0.779
WC M 8.663 0.499 3.784 0.749
vowel chart. Figure 5.2 shows how the traditional Cardinal Vowel chart would
appear via a BARK normalization (based on Thomas 2011; Mesthrie, Chevalier
& Dunne 2013).8 Figure 5.3 shows the array of variants for nurse speaker means
in our sample. The means appear to cover the vowels usually transcribed as [äː],
[ɜː], [əː], [ʌ], [ɤ̈] and [ɤ].9
Table 5.2 provides the means and standard deviations per social class and
gender groups for each of F1 (which represents vowel height) and F2 (which rep-
resents vowel backness or ‘advancement’). Degree of rounding cannot be meas-
ured acoustically with certainty, and fortunately for our study the SAIE speakers
do not appear to use lip rounding as a salient aspect of NURSE articulation, un-
like White SAE (see Section 4).
We used two methods of ascertaining the sociolinguistic distribution of the
NURSE tokens. The first measured F1 and F2 independently, using a linear re-
gression mixed-effects model (Bates 2005) in R, treating gender and class as fixed
factors. The results are reported visually in terms of a boxplot (also known as a
‘box and whiskers’ plot) per class and gender group and summarised in terms of
significant differences in Figure 5.4 below. Figure 5.4 gives the boxplot for F1 for
NURSE. F1 correlates roughly with vowel height. As is customary in boxplots,
the * denotes the mean value per social group, and the central box the mid-values
(between 25th and 75th percentile), with the dark horizontal line within it show-
ing the median. The broken line above the box represents the upper quartile, i.e
the 25% of the data that have higher values than the upper limit of the box. Sim-
ilarly, the broken line below the box indicates the lower quartile, i.e. the 25% of
the data that have lower values than the lower limit of the box. Outlier tokens are
signalled by small open circles. Figure 5.5 shows the boxplot for F2, thus giving an
indication of NURSE realisations on a front-central-back vowel axis.
Table 5.3 gives the p-values for comparisons of the gender and class groups,
with the significant values shaded in grey. It shows that there is a significant dif-
ference for gender overall for each of F1 and F2, with females having on average
lower F1 (= lower NURSE vowels) and lower F2 (= fronter NURSE vowels). This
difference in gender is stronger in the WC (for both F1 and F2) than in the MC
(where it is significant for F2 but not F1).
The second method that we adopted considers F1 and F2 together, measures
the Euclidean distance between any and all speaker means, and on this basis pro-
duces a cluster plot in R (see Figure 5.6).
8. Note the empty space in the bottom right hand corner.

9. By way of comparison the traditional NURSE vowel in Black SAE is [e], and in Coloured
varieties of SAE show [ɜː] for WC speakers and [Ø:] for MC speakers (Toefy 2014).
NURSE
10
8
F1
MC F MC M WC F WC M
Speaker group
Key: M = Male, F = Female, MC = Middle-class, WC = Working-class
Figure 5.4â•‡ Boxplot showing means, median and distribution of F1 for NURSE tokens
per class and gender group
Table 5.3â•‡ p-values in a linear regression mixed effects

model for F1 and F2 by class and gender
F1 F2
Males vs Females 0.0058 0.0005
Males by class 0.2151 0.3461
Females by class 0.8033 0.506
MC vs WC 0.3149 0.3151
MC by gender 0.2204 0.046
WC by gender 0.0008 0.0007
NURSE
10
8
F1
MC F MC M WC F WC M
Speaker group
Figure 5.5â•‡ Boxplot showing means, median and distribution of F2 for NURSE tokens
per class and gender group
11
10
Cluster 4
9
Z3–Z1 (F1)
Cluster 3
8
MCF
Cluster 2
MCM
7
WCF
Cluster 1 WCM
6
5
1 2 3 4 5 6 7 8
Z3–Z2 (F2)
Figure 5.6â•‡ Cluster plot based on Euclidean distances between speakers

The four clusters in Figure 5.6 can be roughly correlated from bottom left to
upper right with cluster 1 showing a relatively fronter and lower realisation of
NURSE; cluster 2, a relatively fronter and mid realisation; cluster 3, a (standard)
mid-central realisation; and cluster 4, a relatively backer and mid realisation. We
then ascertain the social characteristics of these groupings. Females tend to fall
into clusters 1 and 2 (9 of 11 speakers), males into clusters 3 and 4 (9 of 10 speak-
ers). This is confirmation that the results are similar whether we take F1 and F2
together or independently (see Table 5.3). Again, the results are less clear-cut for
class: no patterns are discernible for MC but WC speakers fall into clusters 2, 3
and 4 (9 of 10 speakers). WC males are the most cohesive with all 5 falling into
clusters 3 and 4; middle-class females tend to fall into cluster 1 (3 of 5 speakers;
and none in cluster 4). There is thus a dichotomy between MC females and WC
males.10 The remaining two groups are intermediate: WC females are closer to
MC females (4 of 5 speakers in clusters 1 and 2; and none in cluster 4), while MC
males are diffuse, not predominating in any cluster and spread over clusters 1
(n = 1), 3 (n = 2) and 4 (n = 2). It would appear then that younger SAIE speakers
are beginning to show all the hallmarks of a Labovian system of variation, with
the intersection of gender and class being crucial (Labov 2001; Eckert 1989).
To summarise, whereas older SAIE speakers vary between [eː] and [ɜː] real-
isations of NURSE, younger speakers in this study have a different range of vari-
ation. For them Cultivated SAE or RP-like realisations ([ɜː] or [əː]) are common,
but so too are “adjacent” variants that carry gender and class connotations, but
only within SAIE. Lowered and fronted NURSE in the region of [ɛ̙ː] or [ä] are
more associated with middle-class and female affiliation. This variant appears to
be based on attempts to inculcate an RP value in speech and drama classes, but
which misses the target somewhat. In contrast young working-class males show
a mid-vowel in the region of [ɜː] but often further back, i.e. towards [ɤ] or [ʌ].
None of the speakers in our sample approximate the general White SAE variant
[øː], though it would not be surprising if a larger and closer study shows such a
crossover effect for some speakers.
5. Conclusion
In this case study we have attempted to show the potential of a broad phonet-
ic analysis in characterising SAIE (and by implication other varieties of EID)
in terms of aspects of IE phonology (retroflexion, aspiration and other salient
10. There is only one exception: a MC female in cluster 3.

features). We give indications of levelling or fudging of certain variables like as-

piration of P T K, and of the influence of particular Indian language phonologies
as in [ɛ]-like realisations of final schwa, which is most likely to be the result of
the specific influence from Tamil speakers. We also demonstrate social class and
gender innovations that are less tied to IE developments, by a detailed analysis of
NURSE realisations among younger SAIE speakers. Here, the picture is that of a
movement away from the L2 norms of the previous generation, without neces-
sarily adopting the variants found in White SAE. Gender differences appear to
generate and sustain new variants.
References
Asher, Ronald E. 1985. Tamil. Kent: Croom Helm.

Bates, Douglas M. 2005. Fitting linear mixed models in R. R News 5: 27–30.
Badassy, Prinisha. 2002. Turbans and Top Hats: Indian Interpreters in the Colony of Natal:
1880–1910. Honours thesis, University of Natal.
Bughwan, Devi. 1970. An Investigation into the Use of English by the Indians in South Africa,
with Special Reference to Natal. PhD dissertation, University of South Africa.
Chambers, Jack & Trudgill, Peter. 1998. Dialectology, 2nd edn. Cambridge: CUP.
DOI: 10.1017/CBO9780511805103
Chevalier, Alida. 2011. Social Class Differentiation in South African Indian English: A Socio-
phonetic Study of Three Vowel Variables. MA thesis, Linguistics Section, University of
Cape Town.
Delbridge, Shelly. 2006. Aspiration in South African Indian English. MA thesis, Linguistics
Section, University of Cape Town.
Eckert, Penelope. 1989. The whole woman: Sex and gender differences in variation. Language
Variation and Change 1(3): 245–68. DOI: 10.1017/S095439450000017X
Fox, Sue. 2007. The Demise of Cockneys?: Language Change in London’s ‘Traditional’ East End.
PhD dissertation, University of Essex.
Gimson, Alfred Charles. 1989. An Introduction to the Pronunciation of English, 4th edn.
London: Edward Arnold.
Gargesh, Ravinder. 2004. Indian English: Phonology. In A Handbook of Varieties of English,
Vol. 1: Phonology, Edgar W. Schneider, Kate Burridge, Bernd Kortmann, Rajend Mesthrie
& Clive Upton (eds), 992–1002. Berlin: Mouton de Gruyter.
Kachru, Braj. 1983. The Indianization of English. New Delhi: OUP.
Labov, William. 2001. Principles of Linguistic Change, Vol 2: Social Factors. Malden MA:
Blackwell.
Lass, Roger. 1995. South African English. In Language and Social History: Studies in South Afri-
can Sociolinguistics, Rajend Mesthrie (ed.), 89–106. Cape Town: David Philip.
Masica, Colin. 1991. The Indo-Aryan Languages. Cambridge: CUP.
Mesthrie, Rajend. 1991. Language in Indenture: A Sociolinguistic History of Bhojpuri-Hindi in
Mesthrie, Rajend. 1992. English in Language Shift: the History, Structure and Sociolinguistics of
South African Indian English. Cambridge: CUP.
Mesthrie, Rajend. 2004. Indian South African English: Phonology. In A Handbook of the Vari-
eties of English, Vol. 1: Phonology, Edgar W. Schneider, Kate Burridge, Bernd Kortmann,
Rajend Mesthrie & Clive Upton (eds), 953–963. Berlin: Mouton de Gruyter.
Mesthrie, Rajend. 2012. Enthnicity, substrate and place: The dynamics of Coloured and Indian
English in five South African cities in relation to the variable (t). Language Variation and
Change 24(3): 371–395. DOI: 10.1017/S0954394512000178
Mesthrie, Rajend & Bhatt, Rakesh. 2008. World Englishes: The Study of New Linguistic Varieties.
Cambridge: CUP.
Mesthrie, Rajend, Chevalier, Alida & Dunne, Timothy. 2013. A study of variation in the BATH
vowel among white speakers of South African English in five cities. University of Pennsyl-
vania Working Papers in Linguistics 19(2): article 15. <http://repository.upenn.edu/pwpl /
vol19/iss2/15/> ( 21 February 2014).
Pienaar, Leela. 2007. Towards a Corpus of Indian South African English: An Investigation of
Lexical and Syntactic Features in a Spoken Corpus of Contemporary ISAE. MA thesis,
Rhodes University (Grahamstown).
Rampton, Ben. 1995. Crossing: Language and Ethnicity Among Adolescents. London: Longman.
Robinson, John, Lawrence, Helen & Tagliamonte, Sali. 2001. Goldvarb 2001 Users’ Manual.
Department of Language and Linguistic Science, University of York. <http://www.york.
ac.uk/depts/lang/web-stuff/goldvarb/>
Sharma, Devyani. 2005. Dialect stabilisation and speaker awareness in non-native varieties of
Thomas, Erik R. 2011. Sociophonetics: An Introduction. Houndmills: Palgrave.
Toefy, Tracey. 2014. Sociophonetics and class formation: A study of working- and middle-class
English in Cape Town’s Coloured Community. PhD Thesis: Linguistic Section, University
of Cape Town.
Trudgill, Peter & Hannah, Jean. 1985. International English: A Guide to Standard Varieties of
English. London: Edward Arnold.
Rayfield, Joan R. 1970. The Languages of a Bilingual Community. The Hague: Mouton.
Rosenfelder, Ingrid, Fruehwald, Joe, Evanini, Keelan & Yuan, Jiahong. 2011. FAVE (Forced
Alignment and Vowel Extraction) Program Suite. <http://fave.ling.upenn.edu>
Watt, Dominic & Fabricius, Anne. 2002. Evaluation of a technique for improving the mapping
of multiple speakers’ vowel spaces in the F1~F2 plane. Leeds Working Papers in Linguistics
and Phonetics 9(9): 159–173.
Wells, John C. 1982. Accents of English, 1: An Introduction. Cambridge: CUP.
DOI: 10.1017/CBO9780511611759
Wiltshire, Caroline R. 2005. The ‘Indian English’ of Tibeto-Burman language speakers. English
World Wide 26(2): 275–300. DOI: 10.1075/eww.26.3.03wil
Chapter 6
Imperfectives in Singapore’s
Indian community*
Jakob R. E. Leimgruber and Lavanya Sankaran

University of Freiburg / Kings College London
This chapter examines the sociolinguistics of the Indian community in Singa-

pore, with a particular emphasis on the use of the -ing marker among Tamils.
The Indian diaspora in Singapore is of interest because the community differs
from others in many respects: although a minority of less than 10 percent, Indi-
ans enjoy constitutional recognition, primarily through the use of Tamil as one
of the four official languages. Furthermore, they are a firmly established ethnic
group within the country, both socially and economically, having been instru-
mental in early colonial times and during the foundation phase of present-day
Singapore English. Our study, drawing on data collected from 96 informants
coming in equal parts from the Tamil, Chinese, and Malay communities, inves-
tigates the over-extension of -ing as a marker of all imperfectives, including sta-
tives and non-delimited habituals. We found the Tamils rating -ing as acceptable
significantly more frequently than the other two groups in the case of statives
and non-delimited habituals. As this parallels the Tamil aspectual system, our
findings strongly support a substratist explanation for the Indian Singapore
English aspect system.
Keywords: Singapore English, ethnic variation, sociolinguistics of Indian

Singaporeans, Indian diaspora in Singapore, Tamil
* Based on a poster presentation at ISLE-2 in Boston (Leimgruber and Sankaran 2011).

Thanks to Devyani Sharma and Marianne Hundt for many useful comments on an earlier ver-
sion, to Francesco Cavallaro and Lionel Wee for help with disseminating the questionnaire, and
to Marlene ElsÃ¤sser for data processing. Many thanks also to Marie Koh for help with Mandarin
and Cantonese, to Yurni Irwati Binte Mohamed Said for help with Malay, and to Amy Oh for
help with Hokkien.
106 Jakob R. E. Leimgruber and Lavanya Sankaran
1. Introduction
The Indian community of Singapore is the smaller of two officially recognised

minorities in the city-state: at just over 9 percent, they are outnumbered by the
indigenous Malays (13 percent) and the majority Chinese (74 percent). Despite
their small number (348,000 out of a total resident population of 3.77 million),
Indians are intimately tied to the history of Singapore and to its present-day pop-
ulation. This is due to three main reasons: firstly, ethnic Indians have been present
in the city-state since its founding in 1819; secondly, Indians were significantly
represented in the colonial civil service and the upper echelons of society (a sit-
uation that continues, somewhat, to this day, with a respectable number of Indi-
an politicians); and thirdly, and perhaps relatedly, they are the ethnic group that
boasts the highest rate of English as a home language.
Singapore’s Indian community is different from some other communities of
the Indian diaspora in several ways. For one, there is the high level of English use
at home (Indians 42 percent, Chinese 33 percent, Malays 17 percent), which is,
in part, a result of language policies in place in Singapore, which put a strong em-
phasis on English as the medium of education and generally the country’s “work-
ing language”, something also seen in Trinidad (Leung & Deuber). Another way
in which Singapore differs is in its (at least statutory) high regard for the Indian
community: the Indians are a recognised indigenous “race”, and are assigned one
of the country’s four official languages, Tamil. This choice of Tamil was motivated
by its status as the majority language in the community.
Given the high level of English use in the Indian community as compared to
the other two, larger Chinese and Malay communities, the question arises as to
how and to what extent the Englishes spoken by these three groups differ. Ethno-
linguistic variation in Singapore having, thus far, largely been confined to pho-
netics and phonology, there is a need for more research on variation in grammar.
The aspect system is of particular interest in the Singaporean case, since it has at-
tracted some interest, particularly with respect to its origin in a putative Chinese
substrate (Bao 2005). Coupled with the potential similarities with Indian English
(Sharma 2009), this raises the question as to how the aspect system of Indian Sin-
gapore English behaves relative to the other two ethnic varieties.
2. Historical background
Singapore, one of the last few remaining city-states, is an island nation of some
700 km2 located at the southern tip of the Malay Peninsula. This former British
colony, founded in 1819, currently has a population of just over 5 million. The
Chapter 6.â•‡ Imperfectives in Singapore’s Indian community 107
location of the port city on the lucrative shipping route from Hong Kong to India
and, eventually, via the Suez Canal, to Europe and Britain, made it a prime choice
for settlement by the British, as well as for subsequent immigrants in search of
work or business opportunities. While the indigenous, pre-British population was
around 1,000 (Turnbull 1996, 5), immigration from various parts of the world
soon boosted that number. The population of Singapore is traditionally divid-
ed into four “ethnic” groups: the 2010 census reports 74.1 percent of Chinese,
13.4 percent of Malays, 9.2 percent of Indians, and 3.3 percent of “Others” (Wong
2011a). These four groups are the present-day result of immigration, primarily
from southern Chinese provinces, from colonial British territories on the Indian
subcontinent, and from the Malay Archipelago.
While there was contact between India and pre-colonial Singapore and Ma-
laya (Sandhu 1969, 21–22), the first substantial group of Indians in modern Sin-
gapore arrived with the British in 1819. There were, according to Sandhu (1993,
774), around 120 Indians on Stamford Raffles’ landing party, mostly sepoys (‘sol-
diers’), lascars (‘sailors’), and servants, but also at least one trader. The sepoys were
from the Bengal Native Infantry, and the servants from a Bazaar Contingent. This
first group was “known locally as ‘Bengalis”’ and in all likelihood “came from
what is now eastern Uttar Pradesh and northwest Bihar, then the principal re-
cruiting ground for sepoys of the Bengal Native” (Lal 2006, 176). As the port of
Singapore prospered and the settlement grew, it attracted more immigrants from
the three main regions mentioned earlier. The Indians came either directly from
India or from pre-existing British settlements in Malaya (primarily Malacca and
Penang on the west coast of the Peninsula), and can be grouped, for convenience,
into three categories. The first comprises convicts and soldiers and is a testimony
to Singapore’s early days as a penal colony. Convicts were brought in from all over
British Southeast Asia as well as directly from India. Many were trained, by their
guards, in useful trades and contributed extensively to the building and main-
tenance of the infrastructure of early Singapore: in the words of Sandhu (1993,
775), “for years the history of these convicts was the history of the Public Works
Department”. Upon release, they either returned to India, moved to neighbouring
Malaya, or stayed on in Singapore. The soldiers were part of the British Indian
Army, and largely stationed in their barracks with little contact with the local
population. Their presence was also transitory in that they could be redeployed
to other locations in case of military need. However, “while few […] remained
in Singapore at the end of their tour, the persistence of a dhobi community from
early times suggests that some members of the Bazaar Contingent left the garri-
son to assume the role of civilian immigrants in the new settlement” (Lal 2006,
176). Their linguistic make-up is hard to define with certainty; convicts “repre-
sented a cross-section of Indian society” (Lal 2006, 177) and came from a range of
ethnic and linguistic backgrounds. As for the soldiers, Turnbull (1996, 126) also
mentions the Indian Army 5th Light Infantry, which consisted entirely of Panjabi
Muslims.
The second group of early Indians includes traders and labourers. Indentured
labourers were in high demand, initially for rubber plantations and tin mines,
later for building and construction work. “The vast majority came from Tamil
Nadu” (Lal 2006, 178). The Indians also had a near-monopoly in the laundry
business. Traders and businessmen arrived both directly from India and from
the Straits Settlements, attracted by the business opportunity offered by the tar-
iff-free port of Singapore. These merchants were instrumental in building up the
local economy, and in creating and maintaining economic ties with both the other
Straits Settlements and the Subcontinent. Tamils were in a majority in this group,
hailing from both Ceylon and the mainland, though a number of them would
have come from further inland.
The third category consists of English-educated migrants, who were drawn
in to help administer the colony as civil servants, teachers, interpreters, clerks,
and lawyers. Many came from Sri Lanka and South India, with Malayalis and Sri
Lankan Tamils in a majority (Lal 2006, 178). The police force (recruited from the
1870s) was also largely Indian, with many Sikhs (Mani 1993, 790–791) who were
also “sought after as security personnel” (Lal 2006, 178). This group certainly had
an impact on the English language in Singapore because, though comparatively
small (see Table 6.1), they were strongly represented in the civil service and the
education system. Their language would have been English, complemented with
their ethnic languages (Tamil, Malayalam, Panjabi, Gujarati, etc.). In a classroom
setting, the colonial standard of English would have been the target variety.
The demographic evolution of the Indian population of Singapore is given in
Table 6.1, which shows data from the first census in 1821 to the most recent one
in 2010. The census category “Indian” comprises a host of populations from the
Indian sub-continent, including Bangladesh and Sri Lanka. Self-reported mem-
bership of a particular sub-group in the “Indian” ethnic group, as given in the
2010 census, gives some indication of where in India the initial Indian migration
to Singapore originated. The largest groups were the “Tamil” (54.2 percent), the
“Malayalee” (7.6 percent), “Hindi” (3.8 percent), and “Sikh” (3.7 percent), with
6 other named groups and many more at less than 2 percent of the ethnically
Indian population. These numbers include changes since the 2000 census: Tamil
respondents have dropped by 4.1 percentage points, while Hindis increased by
2.2 points and “Others” by 5.2 points. Table 6.1 shows the percentage of Indians
that are likely to have a Dravidian language as their (current or ancestral) moth-
er tongue – “likely” because of the problematic equation between ethnic (sub-)
group and language.
Table 6.1â•‡ Indian population in Singapore. Data 1821–1980 from Sandhu (1993, 775),
1990 from Shantakumar (1993, 867), and 2000–2010 from Wong (2011a)
Year Number of Indians % of total population Dravidians as % of the
Indian ethnic group1
1821 â•⁄â•⁄â•⁄â•‹132 â•⁄2.8
1871 â•⁄11,501 11.8 86.6 (1881)
1891 â•⁄16,035 â•⁄8.7 78.0
1911 â•⁄28,454 â•⁄9.2 84.3 (1921)
1931 â•⁄50,860 â•⁄9.4 79.3
1947 â•⁄68,978 â•⁄7.7 74.1
1957 124,084 â•⁄9.0 79.9
1970 145,169 â•⁄7.0 80.6
1980 154,632 â•⁄6.4 71.9
1990 190,900 â•⁄7.0 72.5
2000 257,866 â•⁄7.8 66.7
2010 348,119 â•⁄9.2 61.7
These sub-groups are frequently termed “dialect groups” in Singapore, a char-

acterisation which, as pointed out elsewhere (Leimgruber 2013), is a misnomer,
because language and ethnicity are blended beyond recognition. Assignment to a
particular sub-group is based on self-reporting during census exercises. As a re-
sult, a largely English monolingual respondent can still identify as ethnically Ta-
mil. Furthermore, the term dialect glosses over the massive linguistic differences
among Indian languages, which belong to two entirely separate language families
(Dravidian and Indo-Aryan).
Singapore’s English-speaking population has grown rapidly in recent decades
(see below). In colonial times, English was restricted to British rulers and Eu-
ropean merchants in the city as well as to a small minority of educated subjects
(predominantly Indians, but also others) working in the civil service and in the
education sector. It was the Eurasian and the Peranakan (a wealthy mixed Malay-Â�
Chinese community, the “Straits-born Chinese”) communities that were first to
switch to English as a language of education and as a home language. Wealthy
families in the Indian and Chinese community soon followed suit. The shift to-
wards English is ongoing, with stratification along socio-economic, educational,
1. The naming of sub-groups within the ‘Indian’ category has changed several times over the
various censuses. Thus, for instance, between 1931 and 1970 there was a group ‘Ceylon Tamil’
in addition to ‘Indian Tamil’. ‘Dravidians’ in this table refers to those identifying as ‘(Indian/
Ceylon) Tamil’, ‘Malayali’, and ‘Telugu’, but does not include the ‘Other Indian’ category, which
may well contain some speakers of Dravidian languages.
and generational dimensions. As a rule, the young are English-dominant, English

having been the sole medium of education in all state schools since 1987.
3. Singapore’s Indian community
A first broad way of defining the group ‘Indians in Singapore’ may be in terms of
nationality and residency status. Indians can be divided into two groups: (1) Sin-
gapore citizens, known as ‘Singaporean Indians’, or ‘local Indians’, who are the
locally born second, third, fourth or even fifth generation descendants of immi-
grants from South Asia and (2) more recent Indian immigrants or ‘Indian Indi-
ans’ who can be sub divided into (a) low-income unskilled workers and (b) high
income professionals or entrepreneurs. Group (a) immigrants are only allowed
work permits which allow them to reside in Singapore for a limited period of
time. By contrast, group (b) immigrants are given employment passes which are
renewable and which give them eligibility to apply for permanent residency or
even Singapore citizenship. All Foreign Nationals or ‘Indian Indians’ have strong
transnational ties with India, unlike local Singaporean Indians.
There is no real ‘pan-Indian’ identity in Singapore because of the diverse
economic, language and cultural backgrounds. This heterogeneity is evident
at several levels: The ethnic diversity in the population was mentioned above.
Religious diversity also exists, the population consisting of 59 percent Hindus,
22 percent Muslims, and 13 percent Christians. In terms of linguistic diversity, it
is unfortunate that the census only records Tamil and “other Indian languages”,
such that a closer analysis of the latter is less straightforward. One number that
the census does report, however, is that of the “language most frequently spo-
ken at home”. The 2010 census (Wong 2011a) reports home language use to have
changed somewhat since the previous (2000) census: the self-reported dominant
home language has seen an increase of English and a decrease of both Tamil and
Malay (see Table 6.2). The presence of Malay as an important minority language
among the Indians will be discussed later; suffice it to say that it is associated with
Table 6.2â•‡ Self-reported dominant home language within the ethnically Indian
population. Data from Leow (2001) and Wong (2011a)
Dominant home language 2000 2010
English 35.6 41.6
Tamil 42.9 36.7
Malay 11.6 â•⁄7.9
‘Other Indian languages’ â•⁄9.3 13.2
Muslim Indians who share their religion with the Malays, the other major minor-
ity of Singapore. The shift towards English is a phenomenon observed across the
whole population, the education policy of having English as the main medium
of instruction having probably played an important role in this shift. It remains,
however, that at 41.6 percent, the Indians are the ethnic group with the highest
percentage of English as a home language: this number is at 32.6 percent for the
Chinese and 17.0 percent for the Malays.
There are few other tangible ways in which the Indians are distinctly different
from the rest of the Singaporean population. For instance, their socioeconomic
distribution, as measured by gross monthly income, patterns almost exactly like
that of the Chinese population (unlike the Malays, who have higher percentag-
es in the lower income groups). Similarly, traditional settlement patterns, which
used to follow a strongly segregational system under colonial rule, have broken
down substantially, especially since the introduction of ethnic quotas in public
housing new towns, where 87.7 percent of the population lives (Wong 2011b).
What remains is the traditional area of Little India along Serangoon Road, which
is still reputed as a prized tourist spot for its concentration of Indian restaurants
and businesses. Notwithstanding Little India, it is fair to say that in the city-state,
Indians are a minority in any neighbourhood.
It is perhaps in the domain of language policy that the Indian population is
treated somewhat differently from the other two indigenous ethnic groups. A first
difference resides in the governmental efforts aimed at engineering the popula-
tions’ linguistic skills. There have been two major campaigns with two targets: the
Speak Mandarin Campaign, launched in 1979 and aimed largely at the Chinese
population, and the Speak Good English Movement, launched in 2000, which
targeted the whole (English-speaking) population. The annual Speak Mandarin
Campaign (SMC) has had different emphases in the past, but it is broadly con-
cerned with promoting the use of Mandarin and demoting the use of other va-
rieties of Chinese. The SMC has been the subject of much research (see, inter
alia, Bokhorst-Heng 1999; Wee 2006) and has, if census data are to be trusted,
been highly effective: whereas in 1980, 10 percent of Chinese spoke Mandarin as
a home language and 60 percent used the “dialects”, these numbers had changed,
by 2010, to 48 percent and 19 percent, respectively (Pecotich & Shultz 2006; Wong
2011a). The Speak Good English Movement (SGEM), also subjected to scholarly
research (see e.g. Rappa & Wee 2006), has been targeting English users and ini-
tially aimed to eradicate the use of “Singlish” (Colloquial Singapore English) in
favour of Standard English. Its success is much more difficult to measure, since it
is unclear what kind of English (Singlish or Standard) respondents to the census
question on main home language actually use. In sum, Mandarin and Standard
English benefit from enormous governmental attention, have their status and
airtime on public broadcasting promoted, and are being commented upon fa-
vourably year-round by ministers and officials, often with a rhetoric involving
cultural heritage (Mandarin) and economic capital (English, but increasingly also
Mandarin). In stark contrast, Malay and Tamil, the other two official languages,
while usually acknowledged as being important “mother tongues”2 (more on this
term later) of the two minority ethnic groups, do not benefit from wide-ranging
campaigns such as the SMC or the SGEM. Particularly when the discussion is
framed in economic terms, Tamil is rarely mentioned, as business with India is
overwhelmingly carried out in English. Given the status of Malay in Singapore’s
neighbouring countries (especially Malaysia), Malay enjoys a slightly more priv-
ileged position, cemented, in part, in its constitutional status as the national lan-
guage (in addition to its official status, shared with English, Mandarin, and Tamil).
A second difference lies in the so-called “mother tongue” policy, which as-
signs a “mother tongue” to each of the three ethnic groups: Mandarin for the Chi-
nese, Malay for the Malays, and Tamil for the Indians. The ethnic mother tongues
are official languages, joined in this status by the ethnically-neutral English. The
mother tongue policy has two main aspects: one of cultural policy and one of
educational policy. The cultural policy is closely linked to the use of English as
the main working language of the republic: being the dominant language in edu-
cation, in politics, in public administration, and in the (white-collar) workplace,
its spread was perceived to be a threat to the cultural heritage of the indigenous
population, particularly because of its perceived association with potentially un-
desirable Western values. To counter this, English is presented, in public rhetoric,
as being “dissociated from Western culture”, and by being “referred to as a global
rather than a Western language” (Alsagoff 2010, 342). While the utilitarian value
of English is highlighted – being, as it is, the international lingua franca so impor-
tant for the economic development of the country – it is denied, officially, native
speaker status (Wee 2003). This is where the mother tongues come in, which are
“presented as repositories and mediums of ethnic culture and identity” (Alsagoff
2010, 342). The mother tongues are therefore deemed important, worthy of study,
and as being in need of protection in the face of the potential shift towards Eng-
lish, given the elevated position of English in the country. Mother tongues are,
as a result, taught at a high level in school, under a policy of bilingual education
(Pakir 1991; Pakir 2001; Dixon 2005).
The mother tongue of the Indian community, in this system, is designated
as Tamil. However, the Indians are in the unusual situation that, since the 1990s,
members of their ethnic group have the option of choosing from one of six mother
2. See following paragraph for a definition of this term in the Singaporean context.
tongues: Tamil, Bengali, Gujarati, Hindi, Panjabi, or Urdu. Thus, arguably, the In-
dians have more choice than the other ethnic groups in what language they learn at
school – not a small concern, since the mother tongue results make up 25 percent
of the year’s mark (in primary school) and since one might well be at an advan-
tage if the “second language” (after English, the medium of instruction) learned at
school is also a language actually spoken at home. It should be noted, however, that
the mother tongue–ethnicity pairing is not mandatory, and that pupils (i.e. their
parents) can choose to enrol for any mother tongue (e.g. an Indian taking Chinese,
or a Chinese taking Malay). That said, it would appear, based on the census data
for literacy, that the predicted ethnic mother tongues are most usually chosen by
pupils/parents – except in the case of the Indians: 79.9 percent of Chinese respond-
ents claimed literacy in Chinese, 88.9 percent of Malays claimed literacy in Malay,
whereas only 49.6 percent of Indians claimed literacy in Tamil.3 Among the Indi-
ans, too, literacy in non-official languages is highest at 19.6 percent (2.3 percent
for the Chinese and 5.9 percent for the Malays), which is unsurprising, seeing as
Tamil is the only official “Indian” language, whereas there were and are many other
languages spoken in the community, both traditionally, and as a result of more
recent immigration. Additionally, Mandarin and Malay are languages that play a
certain role in the Indian community: Indian Muslims tend to be conversant to
some extent in Malay for religious purposes (Malay ethnicity and Muslim religion
are intimately connected in Singapore and Malaysia), and there is anecdotal evi-
dence that Mandarin is being used informally as a tool for communication with
the majority group, albeit in informal (e.g. school yard) settings.
Given the historical and contemporary linguistic background of Indians in
Singapore, it is reasonable to assume that there are some differences to be expect-
ed between the English spoken by them and the other two major groups. Malay,
and particularly Chinese languages are, after all, typologically rather different
from Dravidian languages (e.g. Chinese/Malay SVO vs. Dravidian SOV, Chinese/
Malay uninflected for tense vs. Dravidian tense inflection, isolating Chinese vs.
agglutinative Dravidian/Malay). For it is indeed Dravidian languages that must
have had an effect on a putative Indian Singapore English ethnic variety, particu-
larly if the proportions given in Table 6.1 are indicative of a possible “founder
effect” (Mufwene 1995).
3. These data from Wong (2011a) take into account the resident population over 15 years
of age. They are presented, in the census release, in terms of combinations of languages, e.g.
‘Chinese only’, ‘English and Tamil only’, ‘English, Malay, and Tamil only’, etc. Combinations not
taken into account here are ‘Non-official language only’, ‘Other two languages only’, ‘English
and non-official language only’, and ‘Other three or more languages’.
4. Ethnic variation in Singapore English
Research on linguistic variation between ethnic groups in Singapore has been

largely concerned with pronunciation differences at all phonetic levels (see e.g.
Lim 1996; Tan 1999; Deterding & Poedjosoedarmo 2000; Lim 2000; Suzanna &
Brown 2000; Huang 2003; Gut 2005; Deterding 2007; Tan 2012). Some (Gut 2005;
Deterding 2007) are concerned with measurements of fine phonetic detail, while
others (e.g. Lim 2000; Deterding & Poedjosoedarmo 2000; Tan 2012) are more
interested in identification tasks: thus Lim (2000) reports that correct identifica-
tion of the speaker’s ethnicity correlates strongly with the level of formality in this
speaker’s language use (the more informal, the higher the correct identification).
Further, Tan (2012) reports that informants in the 19–29 age range were less likely
to correctly identify the ethnicity of recorded voices (what she calls “accent deaf-
ness”) as well as a tendency for ethnic accents to be misidentified as Chinese. She
explains this by the dominance of the Chinese population in Singapore, making
their variety the perceived “default Singaporean accent” (Tan 2012, 14). A certain
amount of convergence towards Chinese Singapore English would certainly be
unsurprising, given the majority status of its speakers.
As far as grammatical differences go, out of the wealth of research done on
Singapore English grammar, there has been very limited research involving eth-
nic variation. Many substratist accounts are based on Chinese varieties. While
some focus on Hokkien and Cantonese (the majority varieties in the initial con-
tact period), others (e.g. Bao 2005; Bao & Lye 2005; Bao 2009) prefer to use Man-
darin, arguing for considerable similarities between dialects of Chinese (a point
of contention, see Siegel 2012). Previous research by Leimgruber (2009, 194) in-
volving all three ethnic groups in comparable sizes has shown some statistically
significant differences in the use of discourse particles: Indians used significantly
fewer particles than the Chinese and the Malays in informal settings but over-
took the Chinese in informal settings. Other grammatical features investigated in
that study (aspect markers and existential constructions) were used similarly by
all three groups. These results notwithstanding, grammatical differences between
the varieties of the three main ethnic groups are rarely discussed; this paper is an
attempt to contribute towards filling this gap.
5. Aspect in English, Sinitic, and Dravidian
Comrie (1976) posits that there are two main aspectual perspectives in all lan-
guages: the perfective and the imperfective. The perfective aspect focuses on a sit-
uation from the outside, as a single unanalysable whole, where there is no attempt
to view the individual phases of the situation. The perfective is thus often said
to denote a “complete” situation with the beginning, middle, and end presented
in their entirety. In English, the simple past tense form (in regular and irregular
verbs) convey perfective aspect as well as the past tense meaning. Imperfectivity
presents a situation as one that is ongoing and focuses on its inside without speci-
fying its initial or final endpoints. While many languages have a single category to
express imperfectivity, there are others that use grammatical means to only par-
tially express the imperfective meaning, and some where the imperfective notion
is subdivided into a number of distinct categories. Imperfective aspect includes
both the habitual and the durative (or continuous) notions, where durativity en-
compasses both non-progressiveness and progressiveness (Comrie 1976, 24–25).
Progressive aspect has connotations of dynamism and continuous successive
changing phases. In English it is marked with an auxiliary and the inflection -ing.
It follows, therefore, that the progressive marker does not generally combine well
with stative verbs, which may be continuous but are non-dynamic. In English, the
sentence She was know-ing him well is regarded as being ungrammatical, because
the stative verb know is incompatible with the progressive -ing marker. However,
Comrie (1976, 25) additionally puts forth the suggestion that the English -ing
marker could be “a kind of imperfective” because it can occur in a wide range of
imperfective constructions, even sometimes in stative contexts such as when a
state is temporary, e.g. I’m loving this flower motif.
In this paper we investigate whether the progressive marker -ing is over-ex-
tended in Singapore English. We will be replicating Sharma’s (2009) study, which
examined imperfectivity in Indian English and in Singapore English. Sharma
(2009) specified four different imperfective categories which may or may not
licence the use of -ing, namely progressives, statives, delimited habituals, and
non-delimited habituals. An example of each sentence type is provided in (1),
with sentences marked for ungrammaticality in Standard English. We used these
categories in our study as well and, like Sharma, we also included some perfective
sentences, which were combined with -ing. Note that the imperfective sentences
specified below were used in the questionnaire as part of our main study.
(1) a. I’m writing a letter. [Progressive]
b. â•›*This bottle is containing one litre of juice. [Stative]
c. I’m eating a lot these days. [Delimited habitual]
d. â•›*Ahmed is brushing every morning and evening.
[Non-delimited habitual]
e. â•›â•›*She was falling down suddenly. [Perfective]
In standard varieties of English, the progressive -ing marker is naturally ac-

ceptable with progressive sentences such as (1a), but it is also acceptable with
delimited habituals such as the sentence in (1c). A delimited habitual is a sub-type

of habitual,4 which consists of an adverbial that temporally binds the situation
specified, giving the sentence a temporary and episodic reading. Thus, it cannot
be seen as stative and is compatible with the progressive -ing marker. The same
cannot be seen in the other sub-category of habituals, i.e. non-delimited habituals
such as (1d). A non-delimited habitual is not time bound by a temporal adverbial
and, accordingly, the habitual situation specified occurs repeatedly without there
being a specified end point. It therefore has a stative interpretation. It has a sim-
ilar reading to sentence (1b), which is unambiguously a stative sentence, where
its stative predicate is not dynamic and thus incompatible with the -ing marker.
Therefore, it cannot be combined felicitously with the progressive -ing marker in
Standard English. Sentence (1e) consists of a perfective predicate, which is bound
by the punctual time adverbial suddenly. This gives the sentence a non-continu-
ous interpretation and therefore is incompatible with the progressive -ing. Thus,
to summarise, of the four imperfective categories and the one perfective category,
only the progressive and the delimited habitual environments licence the use of
the progressive -ing marker in standard, metropolitan varieties of English.5
Table 6.3 shows how the above mentioned imperfective and perfective sen-
tences are expressed in Singapore’s main substrate languages, i.e. Mandarin, Can-
tonese, Hokkien, Malay and Tamil.
Given the success of the Speak Mandarin Campaign (SMC) and other forms
of institutional promotion of the language since the 1980s, Mandarin has, in
the last 30 years, been replacing many of the other Chinese languages such as
Hokkien, Cantonese, Teochew, etc. This makes Mandarin the most relevant Chi-
nese substrate language influencing Singapore English presently, which is why
we have chosen it as the representative Chinese substrate in our current study.
It is important, however, to distinguish between the Mandarin spoken in Main-
land China and the Mandarin spoken in Singapore, as “the Singaporean variety of
Mandarin, like Malay, has been affected by contact with other Chinese languages
in Singapore” (Sharma 2009, 175–176).
4. Interestingly, while English does have a separate ‘habitual’ aspect, it marks it only in the
past tense, i.e. ‘John used to work here’ (Comrie 1976, 25).
5. There is some indication that these textbook rules on the use of the progressive may be
crumbling. Hundt and Vogel (2011), for instance, observe that progressives can mark perfect
meanings in some ENL varieties. Pfaff et al. (2013) further note the ‘new’ function of the past
progressive to mark recentness (as in I was just reading this article).
Table 6.3 Singapore’s main substrate languages expressing imperfective and perfective categories
Imperfective Perfective
Progressive Stative Delimited habitual Non-del. Habitual
English I’m writing (prog) a This bottle *[is I’m eating (prog) a lot He *[is brushing] (prog)/ She *[was falling]
letter. containing] (prog)/ these days. brushes his teeth every (prog)/fell down
contains juice. morning. suddenly.
Mandarin Wǒ zài (prog) xiě yīfēng Zhè píng (*zài) (prog) Wǒ zùijìn (*zài) (prog) Tā (*zài) (prog) měitiān Tā (*zài) (prog) túrán
̍
xìn. hányǒu guǒzhī. chī le hěn duō. zǎoshang shuā yā. jiān diédǎo.
Cantonese Ngo5 se2-gan2 (prog) Ni1 zeon1seoi2 zong1zyu6 Ngo5 ni1 paai4 sik6 Koei5 mui5 ziu1z dou1 Koei5 mou4dyun1dyun1
soen3. (*gan2) gwo2zap1. (*gan2) hou2do1 je5. caat3 (*gan2) ngaa4. dit3dou2 (*gan2).
̍
Hokkien Góa ló (prog) siá sìn. Chit-ê (*ló) chun ū chiap. Góa chit lēng-kang (*ló) I tak chá-khí (*ló) lù I tông-thut-kan (*ló)
chiah chiâⁿ-chōe. chhùi-khí. poah-tó.
Malay Saya sedang (prog) Botol ini (*sedang) Saya *[sedang (prog) Dia *[sedang (prog) Dia *[sedang (prog)
menulis surat. (prog) mengandungi jus. memakan]/makan banyak memberus]/berus giginya jatuh]/terjatuh tiba-tiba.
beberapa hari ini. setiap pagi.
Tamil Naan kadithathai ezruthi- Intha kuppi pazrarasam- Naan ippo thellam niraiya Avan thinamum kaalaiyil Aval thidirendru
kondiru-(impf)-kkiren. kondiru-(impf)-kkirathu. saappittu-kondiru- pal theythu-kondiru- vizrunthu-*kondiru
(impf)-kkiren. (impf)-kkiraan. (impf)-nthaal.
Gloss: Imperfective: impf; Progressive: prog.

̍
̍
5.1 A brief description of the various imperfective systems of Singapore’s

main substrate/adstrate languages
Chinese: Progressive aspect in Mandarin is expressed by the marker zài, which

can only be used in dynamic, continuous contexts. Mandarin also possesses an
imperfective marker -zhe, which has a relatively restricted use within certain
imperfective contexts. It is primarily used to mark temporary result states (Sun
2006) and is actually not used in many imperfective constructions such as habitu-
als, simultaneity, and persistence. There is also extensive optionality in the use of
-zhe as it is conditioned by prosodic, semantic, and word order factors. In many
cases, it is omitted altogether. Convergence across the various Chinese substrates
seems to reinforce the lack of pressure to mark imperfectivity using -zhe (Sharma
2009, 187; Ansaldo 2004).
In Cantonese the progressive aspect is marked with a verbal suffix -gan2;
there is also a continuous marker -zyu3, similar to Mandarin -zhe. Like Mandarin
zài, -gan2 appears not to be licenced for non-progressive imperfectives. The same
can be said for Hokkien, where the preverbal marker ló expresses the progressive
aspect, but no other imperfectives (Chappell 1992).
Malay: The progressive aspectual perspective is marked with the form sedang
in standard Malay. The form sedang is used in fairly restricted contexts and cannot
be used to express future reference or simultaneity. This Malay progressive mark-
er operates very differently from the English progressive -ing in that it cannot
occur in a wide range of imperfective contexts like -ing does. Another form that
can also be analysed as a progressive marker is the prefix meN- (Soh & Nomoto
2009, 148). This marker, like sedang, is incompatible with stative verbs, indicating
that in Malay there is no special motivation to mark imperfectivity in all contexts
(Svalberg & Chuchu 1998).
Tamil: Tamil belongs to the southern branch of the Dravidian language fam-
ily. Its morphology is agglutinating. Tense is obligatorily marked using overt suf-
fixes whereas aspect is optionally marked using auxiliaries. In (2) we see the use
of the finite form of a verb predicate that consists of an aspectual auxiliary and a
tense suffix.
(2) Verb stem + Aspect + Tense + Person Number Gender (PNG)
The imperfective perspective in Tamil is expressed using the almost fully gram-
maticalised6 -kondiru marker. Note that -kondiru is an imperfective marker,
6. Aspect markers in Tamil are each in various stages of grammaticalisation. The ones that
are fully grammaticalised are primarily aspectual, whereas those that are in the early stages of
grammaticalisation mainly express attitude. Since aspect is a category that is still in the process
rather than a progressive marker. It therefore combines with all verb type catego-
ries, either expressing continuity with statives or progressive aspect with dynamic
verb types (Annamalai 1997, 57).
5.2 Motivations for our present study
In this paper we investigate whether the progressive marker -ing is over-extended

in Singapore English, distinguishing between the three main Singapore English
varieties, namely the Chinese, Malay, and Tamil varieties. We will be replicating
Sharma’s (2009) study, which examined imperfectivity in Indian English and in
Singapore English.
Kortmann and Szmrecsanyi (2004) observe that over-extension of the pro-
gressive is found in bilingual postcolonial varieties. Sharma (2009) finds that its
occurrence in Indian English is extremely frequent. In all the four imperfective
categories, i.e. progressives, delimited habituals, statives, and non-delimited ha-
bituals, for instance, she found a robust over-extension of -ing to non-delimited
habituals and statives. Figure 6.1 illustrates this clearly.
Sharma (2009) also investigated the use of the progressive -ing in Singapore
English and concluded that while there is some evidence of the over-extension of
-ing, it is nowhere near the extent to which it occurs in Indian English. Ho and
Platt (1993, 189), in fact, claim that “in general, the state-process distinction holds
for Singaporean Chinese learners of English”. Sharma (2009) uses a substratist
100
90
80
70
60
50
40
30
20
10
0
Progressive Delimited Statives Non-delimited Perfectives
habituals habituals
Figure 6.1â•‡ Percentage of the use of the progressive -ing in Indian English
of grammaticalisation, there is considerable variability in its usage across dialects and idiolects
(Schiffman 1999, 81, 104).
explanation to justify this difference between the use of -ing in Indian English
and in Singapore English. She argues that Chinese systems account for the few
instances of the over-extension of the progressive in Singapore English whereas
Hindi, the primary language spoken by her Indian subjects, obligatorily marks
the imperfective. Indian English speakers, therefore, seem to be “recasting -ing as
a general imperfective marker” (Sharma 2009, 183).
In this chapter, we investigate the use of the progressive -ing marker in Singa-
pore English in more depth. The socio-historical summary given in Sections 1–3
raises several possibilities. Firstly, it is likely that the Indian ethnic group fea-
tures a variety of English different from the other groups: Indian teachers were
instrumental in the early education system, and presently exhibit the highest level
of English use as a home language; a faster shift to English in the community
(accompanied with more education-induced standardization) is to be expected.
Secondly, it is, conversely, possible that differences between ethnic groups are
less than obvious, given the numerical predominance of the Chinese population
and the absence of segregationist measures (in fact, the presence of integrationist
policies plays a significant role); convergence is thus not to be excluded in what
remains, after all, a rather small national speech community.
Nonetheless, we stipulate there being at least three varieties of Singapore
English, distinguishing between the varieties spoken by the Mandarin speaking
Chinese, the Malays, and the Tamils. We explore whether there are differences
in the progressive -ing marker used by the Chinese, Malays, and Tamils in Sin-
gapore English and verify whether a substratist explanation can account for any
differences if they exist. Specifically, we investigate (a) whether the progressive
is over-extended in all varieties of Singapore English to the same extent and
(b) whether the substrate languages, i.e. Singapore Mandarin, Malay, and Tamil
can account for any of the differences, if they do exist.
6. Methodology
The present study draws on data collected from the three main ethnic groups via
administration of an online questionnaire. The subjects ranged from ages 16 years
to 30 years.7 This age group was chosen because of the Speak Mandarin Campaign
(SMC), launched in 1979, which would have unanimously affected the Chinese
participants in this study. We thus involved only a younger generation of Singa-
poreans to be assured that all the Chinese subjects’ variety of Singapore English
7. Only one subject was 40 years old and she was Tamil.
would be affected mainly/solely by Singapore Mandarin, given that we are testing

whether there is a substrate influence on the participants’ use of the progressive
-ing. We were thus selective in terms of our age group because it was important
that all the ethnic groups in our study had the same mother tongue language ex-
perience at school.
Only Mandarin speaking Chinese have been chosen to represent the Singa-
pore ethnic Chinese for this study, not only because of the effects of the SMC
(which led to Mandarin becoming the dialect spoken by the newer generation of
Chinese Singaporeans), but also because it is one of Singapore’s official languages
and features prominently in the education system. The group chosen for the study
would thus be representative of the younger Chinese Singaporeans who study/
studied Mandarin at school and who speak it at home with their families. The Ma-
lay subjects participating in this study all study/studied Malay at school and also
speak it at home. The Tamil Singaporeans represent the Indian subjects in this
study. The socio-cultural history outlined in the previous sections makes a strong
argument for Dravidian languages having had the most profound influence on
a putative Indian Singapore English ethnic variety. For ease of comparison, we
have chosen the Tamil ethnic group to represent the Indians because their mother
tongue Tamil, besides being one of the official languages in Singapore, is also one
of the oldest members of the Dravidian language family in Singapore. Tamil can
therefore be reasonably expected to represent the other Dravidian languages that
may have had a strong influence on the Indian Singapore English variety. The
Tamil speaking subjects in this study all learn/learned Tamil at school and also
sometimes speak it at home. This group is ideally positioned to compare against
the Hindi speaking Indian subjects in Sharma’s (2009) study in order to investi-
gate the substratist explanation in more detail.
There were 32 subjects from each ethnic group. An almost equal number of
males and females from each group participated, reducing the possibility of gen-
der bias. All subjects were asked what mother tongue they learnt at school and
what language they spoke most of the time at home. Based on their answers, they
were placed in one of the three main ethnic groups, i.e. Chinese, Malay, or Tamil.
Note that all subjects were either students or working adults who spoke English
as an L1.
The questionnaire comprised of four different types of imperfective sentenc-
es, i.e. progressives, statives, delimited habituals, and non-delimited habituals as
well perfective sentences, which were all combined with the -ing marker (see (1)
in Section 5 for more detail).
The subjects were given five sentences from each imperfective category, three
perfective sentences as well as a few filler sentences. The online survey system
automatically mixed them into a random order. The subjects’ task was to mark
a given sentence as being either “correct” or “incorrect”. If they marked a sen-

tence as being “incorrect”, they were asked to provide what they felt was a correct
version of the sentence. They were also asked whether the given sentence in the
questionnaire was one that they would use in their everyday speech. The subjects
were told that there were no right or wrong answers and that the questionnaire
was employed to elicit people’s opinions, not to test their English language ability.
There was evidence that the subjects corrected sentences based on their everyday
language use and not on the perceived norms of correctness. One example of this
is when given the sentence, ‘My knee is paining’, one of the informants wrote ‘My
knee is pain’ as the alternative. Note that if a subject marked a given sentence as
being incorrect, but then provided an alternative sentence of the same aspectual
construction as the given sentence, it was coded as being correct. Also, if the sub-
ject marked the given sentence as being incorrect but then claimed that he would
still use it in his speech, it was coded as being correct.
7. Results
The results, summarised in Figure 6.2, indicate that all three ethnic groups find
progressive predicates combined with the -ing marker acceptable almost 100 per-
cent of the time. More than 50 percent of the time, they also mark delimited habit-
uals combined with -ing as being correct, with no significant difference amongst
the three ethnic groups (χ2-test, 4 d.f., p > 0.25 for all combinations). With re-
gards to non-delimited habituals, however, the Chinese and Malays find them
acceptable with -ing only 15 percent and 18 percent of the time, respectively. The
Tamils, on the other hand, find them acceptable 26 percent of the time. Here the
difference is statistically significant (p < 0.05) between the Tamils on the one hand
and the Chinese and Malays on the other (the difference between the latter two
groups being non-significant at p > 0.125). In the case of statives, the percentage
of acceptability is even lower. The Chinese and Malays find statives with -ing ac-
ceptable at a rate of only 7 percent and 8 percent, respectively, while the Tamils
find them acceptable 15 percent of the time. This difference in the acceptance
rates of -ing with statives, between the Chinese and Malays versus the Tamils is,
again, statistically significant (p < 0.025), whereas it is not significant between the
Chinese and the Malays (p > 0.124).
The percentage of perfective predicates found acceptable with -ing is low
amongst all three ethnic groups, which is to be expected. This is because the per-
fective sentences in the questionnaire consisted of punctual verbs (e.g. fall) mod-
ified by punctual adverbs (e.g. suddenly), which created an environment that did
100%
90%
80%
70%
60% Tamils
50% Chinese
40% Malays
30% *
20% *
10%
0%
Progressive Delimited Statives Non-delimited Perfectives
habituals habituals
Figure 6.2â•‡ Percentage of sentences scored as being correct by the three ethnic groups
not licence the use of the progressive -ing marker in either Standard English or
basilectal ethnic varieties of SgE. Thus the verbs in the given perfective sentences
are only compatible with a perfective aspectual marker and not a progressive one.
Interestingly, the results show that even with regard to these perfective sentences,
among the three ethnic groups, the Tamils show the highest rate of acceptance of
the -ing progressive marker. The Tamils mark perfectives combined with -ing as
being correct 13 percent of the time as opposed to the Chinese and Malays who
mark them as being correct only 9 percent and 8 percent of the time respectively.
However, these differences are not significant (p > 0.09 in all cases).
To summarise, with regard to the predicates that are not acceptable with -ing
in standard metropolitan varieties of English, namely statives, non-delimited ha-
bituals, and perfectives, Tamils show a significantly higher acceptability rate when
compared to the Chinese and Malays. Their acceptance of -ing with progressive
predicates is also one of the highest in comparison with the other ethnic groups.
Interestingly, the rate of the Tamils’ acceptability of -ing with delimited habit-
uals is low in comparison with the Chinese and almost on a par with the Malays.
This is unusual given that the Tamil group consistently outperforms the other
two ethnic groups with regard to the acceptability rate of -ing with all the oth-
er sentence type categories. We would, in fact, expect the Tamil group to rate
the acceptability of -ing with delimited habituals very high, higher than even the
Chinese group, especially since in Standard English, delimited habituals are ac-
ceptable with the progressive -ing marker. This discrepancy in the results will be
addressed in the next section.
8. Discussion
This study is closely modelled on Sharma’s (2009) corpus-based study of the

over-extension of the progressive -ing marker in imperfective and perfective con-
texts. Sharma investigated this phenomenon in Indian English and in Singapore
English, looking at substrate influences from Hindi and Singapore Mandarin to
account for her results. She found that in the Singapore component of the Inter-
national Corpus of English (ICE-Sing), there is a very slight over-extension of
-ing, which is nowhere near the extent to which -ing is over-extended in Indian
English (based on ICE-India).
In our study we first propose that there is no single variety of Singapore Eng-
lish and that there are differences in the varieties spoken by the Chinese, Malays,
and Tamils, at least in their use of the progressive -ing marker. We look at Singa-
pore Mandarin, Malay, and Tamil as being the three main substrate languages that
could influence the use of -ing in Singapore English and suggest that the gram-
matical differences in these languages account for the differences in our results.
Like Sharma’s (2009) study, our results show that there is indeed evidence of
over-extension of the progressive -ing marker in all varieties of Singapore English.
Our results, however, highlight that the Tamil group shows a significantly higher
rate of -ing over-extension when compared to the Chinese and Malay groups.
This is not surprising if we compare the Singapore Mandarin, Malay, and Tamil
aspectual systems, which we claim influence the different varieties of Singapore
English, as spoken by the three ethnic groups.
To recall, in Singapore Mandarin the progressive marker zài is only used in
strictly progressive contexts, while the imperfective marker -zhe has a highly vari-
able usage and is in many cases omitted in stative and in many other imperfective
contexts. Malay’s progressive markers, sedang and meN-, like zài in Mandarin,
can only be used in strictly progressive contexts. Note that Malay does not even
possess an imperfective marker that can be used in stative contexts. An examina-
tion of the Chinese and Malay results suggests, therefore, that both the Chinese
and Malay imperfective systems reduce the pressure to mark imperfectivity overt-
ly in the Singapore English spoken by the Chinese and Malays, respectively, in
non-progressive contexts. There is a very slight over-extension of the -ing marker
evidenced in the Chinese and Malay groups, but it is not as pronounced as the
over-extension of the -ing marker evidenced in the Tamil group.
The substratist explanation proposed by Sharma (2009) for differences be-
tween Indic- and Sinitic-influenced English varieties seems, on the surface, to also
account for the Tamil’s higher rate of acceptance of -ing across all the imperfective
categories. The Tamil results suggest that the -ing marker is possibly a relexified
form of the Tamil imperfective marker kondiru. This is particularly plausible if we
bear in mind that Tamil uses the same marker, kondiru, to denote imperfectivity
and it is acceptable with all verb types including progressives and statives.
But while we can trace effects from the Tamil substrate in our study, we find
that the overall robustness of the pattern of the over-extension of the -ing marker
is much higher in Sharma’s (2009) study of Indian English (where Hindi is the
substrate language in question) when compared to our study of Singapore Eng-
lish, in which Tamil is the substrate language in question. Sharma (2009, 185)
states that “Hindi is a strict imperfectivity-marking system, such that all finite
clauses must be marked as either perfective or imperfective”. Hindi has an imper-
fective form (non progressive and/or habitual) -ta, which is never optional in ha-
bitual and stative contexts as well as a progressive form rahna, which performs a
much stricter function than the broad scope of the English -ing. Sharma suggests
that “due to its extended range, the form -ing appears to equally map to rahna and
-ta and Indian English speakers interpret it as a global imperfectivity marker”
(Sharma 2009, 185).
One of the reasons why the substrate effects in our Tamil Singapore English
sample are much slighter than what is observed in Sharma’s Indian English sam-
ple may be due to the differences between the aspectual systems of Tamil and
Hindi. Unlike Hindi, where aspect is marked obligatorily, in Tamil it is often not
explicitly marked due to pragmatic considerations that are related to politeness,
shared perceptions, or the nature of truth propositions etc. (Schiffman 1999, 103).
Thus, -kondiru does not necessarily occur in imperfective contexts in Tamil even
though it is licensed to do so. In many instances where the imperfective is not
overtly marked in Tamil, the verb is neutral where aspect is concerned. In some
cases, the tense marker or a temporal adverbial in the sentence may convey an
aspectual perspective rather than an overt aspectual marker itself.
To return to the case of delimited habituals, which were presented with overt
time-bound adverbials in the questionnaire, the results show that Tamils are sur-
prisingly less likely than the Chinese to accept the putative imperfective marker
-ing. This is the only context where the Tamils do not over-extend the -ing marker
more than the Chinese and Malays. This result, however, could be explained by
the fact that the time-bound adverbials used in the delimited habitual sentences
dissuaded the Tamils from using an overt aspectual marker. Recall that in Tamil,
aspect marking is optional and tense marking is obligatory. Our substrate account
would therefore posit that the temporal adverbial and tense marker already pres-
ent in the delimited habitual constructions conveyed the imperfective aspectual
perspective, without the need for the Tamils to include the -ing marker. The high-
er (though statistically insignificant, p > 0.25 in all cases) acceptance rate of -ing
with delimited habituals by the Chinese, observed in Figure 6.1, could, likewise,
be a result of the function of the Chinese progressive marker zài in marking de-
limited habituals in that language (Yang & Bateman 2002).
To summarise, our study highlights the fact that even within Singapore Eng-
lish varieties, there can be significant differences in how -ing is over-extended.
The results of our short survey show that there is a significant difference in the
grammaticality judgments of statives and non-delimited habituals marked with
-ing between Tamil respondents on the one hand, and Chinese and Malay re-
spondents on the other. For other imperfectives, however, the difference between
the three groups did not prove to be significant. The explanation we offer to ac-
count for the differences in ethnic varieties with respect to the -ing marking of
statives and non-delimited habituals draws on the grammatical means used in the
substrate languages involved: the Tamil marker kondiru is used to mark all im-
perfectives, whereas the Chinese marker 在 zài and the Malay markers meN- and
sedang are progressive markers that are restricted to “the typical ‘action-in-pro-
gress’ meaning” (Li & Shirai 2000, 24). The progressive markers in Chinese and
Malay, in other words, cannot be used with stative (Soh & Nomoto 2009) or punc-
tual/telic predicates which accounts for the low acceptance rate of -ing with sta-
tives, non-delimited habituals, and even perfectives by the Chinese and Malay
subjects. The progressive markers in Chinese and Malay, however, can be com-
bined with delimited habituals (Li & Shirai 2000, 211), which is one of the ways to
account for the observation that -ing marking has high acceptance rates with de-
limited habituals in all three groups. However, since delimited habituals are also
possible in ENL varieties, it is not necessary to fall back on an explanation involv-
ing substrate influence in this particular instance. Our study nonetheless shows
that there is a significant difference between Tamils and non-Tamils, i.e. Chinese
and Malays, where the latter much more clearly reject the inflection on statives
and non-delimited habituals, following the model of standard varieties of native
English. Thus, the progress/movement/tendency towards a unified, pan-ethnic
Singapore English, while clearly underway, is by no means complete: even though
the findings by Tan (2012), for instance, suggest that accent distinctions are erod-
ing, we show that subtle and perhaps fairly stable (and certainly statistically sig-
nificant) differences among groups can be identified with close analysis.
We have also drawn comparisons with Sharma’s (2009) results and have
highlighted the fact that, despite the parallels, our results show that the Tamils’
over-extension of -ing in Singapore English is not at all at the same scale as Hindi
speakers’ over-extension of -ing in Indian English. The results therefore indicate
an important difference between Tamil speakers’ Singapore English and Hindi
speakers’ Indian English; in Indian English -ing is extended to all imperfective
aspectual constructions, whereas this is not the case in Tamil speakers’ Singapore
English. We have accounted for this by illustrating the aspectual differences in the
respective substrate languages.8
In conclusion, it would appear that the aspectual systems of the various sub-
strate languages, at least with respect to the categories explored here, show some
evidence of ethnic differences in the grammar of Singapore English, differences
which, thus far, have not been noted in the literature (with the possible exception
of different rates of copula-deletion, see Sharma & Rickford 2009). We explain
this difference by proposing that the grammatical systems of the substrate lan-
guages of the informant groups have a direct influence on the resulting varieties
of Singapore English. Therefore, accounts that draw on a single substrate language
(such as Bao 2005) need to be rethought as providing evidence only for a sub-
set of speakers of Singapore English, namely that associated with said substrate
language.
References
Alsagoff, Lubna. 2010. English in Singapore: Culture, capital and identity in linguistic variation.
World Englishes 29: 336–348. DOI: 10.1111/j.1467-971X.2010.01658.x
Annamalai, E. 1997. Adjectival Clauses in Tamil. Institute for the Study of Languages and Cul-
tures of Asia and Africa (ILCAA). Tokyo: Tokyo University of Foreign Studies.
Ansaldo, Umberto. 2004. The evolution of Singapore English: Finding the matrix. In Singapore
English: A Grammatical Description [Varieties of English Around the World G33], Lisa
Lim (ed.), 127–149. Amsterdam: John Benjamins.
Bao, Zhiming. 2005. The aspectual system of Singapore English and the systemic substratist
explanation. Journal of Linguistics 41(2): 237–267. DOI: 10.1017/S0022226705003269
Bao, Zhiming. 2009. One in Singapore English. Studies in Language 33(2): 338–365.
DOI: 10.1075/sl.33.2.05bao
Bao, Zhiming & Hui Min, Lye. 2005. Systemic transfer, topic prominence, and the bare condi-
tional in Singapore English. Journal of Pidgin and Creole Languages 20 (2): 269–291.
DOI: 10.1075/jpcl.20.2.03zhi
Bokhorst-Heng, Wendy. 1999. Singapore’s speak Mandarin campaign: Language ideological
debates in the imagining of the nation. In Language Ideological Debates, Jan Blommaert
(ed.), 235–265. Berlin: de Gruyter.
Chappell, Hilary. 1992. Towards a typology of aspect in Sinitic languages. Zhongguo Jingnei
Yuyan ji Yuyanxue: Hanyu Fangyan (Chinese Languages and Linguistics: Chinese Dia-
lects) 1(1): 67–106.
Comrie, Bernard. 1976. Aspect. Cambridge: CUP.
8. Note that the two studies in question made use of different methodologies. Sharma’s (2009)
study drew examples from corpus data whereas the present study made use of acceptability
ratings collected through a survey.
Deterding, David. 2007. The vowels of the different ethnic groups in Singapore. In English in
Southeast Asia: Varieties, Literacies and Literatures, David Prescott (ed.), 2–29. Newcastle
upon Tyne: Cambridge Scholars.
Deterding, David & Poedjosoedarmo, Gloria R. 2000. To what extent can the ethnic group of
young Singaporeans be identified from their speech? In The English Language in Singapore:
Research on Pronunciation, Adam Brown, David Deterding & Ee Ling Low (eds), 1–9.
Singapore: Singapore Association for Applied Linguistics.
Dixon, L. Quentin. 2005. The bilingual education policy in Singapore: Implications for second
language acquisition. In 4th International Symposium on Bilingualism, James Cohen, Kara
T. McAlister, Kellie Rolstad & Jeff MacSwan (eds), 625–635. Somerville MA: Cascadilla
Press.
Gut, Ulrike. 2005. The realisation of final plosives in Singapore English: Phonological rules and
ethnic differences. In English in Singapore: Phonetic Research on a Corpus, David Deterding,
Adam Brown & Ee Ling Low (eds), 14–25. Singapore: McGraw-Hill.
Ho, Mian Lian & Platt, John. 1993. Dynamics of a Contact Continuum. Oxford: OUP.
Huang, Vivien Shu Jun. 2003. Distinguishing between the Chinese, Malay and Indian Ethnic
Varieties of Singapore English: A Study of Voice Onset Time. Honours thesis. Department
of English Language and Literature, National University of Singapore.
Hundt, Marianne & Vogel, Katrin. 2011. Overuse of the progressive in ESL and learner Eng-
lishes – Fact or fiction? In Exploring Second-Language Varieties of English and Learner Eng-
lishes: Bridging a Paradigm Gap [Studies in Corpus Linguistics 44], Joybrato Mukherjee &
Marianne Hundt (eds), 145–166. Amsterdam: John Benjamins. DOI: 10.1075/scl.44.08vog
Kortmann, Bernd & Szmrecsanyi, Benedikt. 2004. Global synopsis. Morphological and syn-
tactic variation in English. In A Handbook of Varieties of English, Vol. 2: Morphology and
Syntax, Bernd Kortmann, Kate Burridge, Rajend Mesthrie, Edgar W. Schneider & Clive
Upton (eds), 1122–1182. Berlin: De Gruyter.
Lal, Brij V. (ed). 2006. The Encyclopedia of the Indian Diaspora. Singapore: Editions Didier
Millet.
Leimgruber, Jakob R.E. 2009. Modelling Variation in Singapore English. DPhil thesis, Univer-
sity of Oxford.
Leimgruber, Jakob R.E. 2013. The management of multilingualism in a city-state: Language
policy in Singapore. In Multilingualism and Language Diversity in Urban Areas: Acqui-
sition, identities, space, education [Hamburg Studies on Linguistic Diversity 1], Peter
Siemund, Ingrid Gogolin, Monika Schulz & Julia Davydova (eds), 227–255. Amsterdam:
John Benjamins.
Leow, Bee Geok (ed.). 2001. Census of Population 2000: Statistical Release 2: Education, Lan-
guage and Religion. Singapore: Department of Statistics.
Leimgruber, Jakob R.E. & Sankaran, Lavanya. 2011. Imperfectives in Singapore English: New
evidence for ethnic varieties? Poster presented at the 2nd triennial conference of the Interna-
tional Society for the Linguistics of English, Boston, MA, USA, 17–21 June 2011.
Li, Ping & Shirai, Yasuhiro. 2000. The Acquisition of Lexical and Grammatical Aspect. Berlin:
De Gruyter.
Lim, Lisa. 1996. Prosodic Patterns Characterising Chinese, Indian and Malay Singapore Eng-
lish. PhD dissertation, University of Reading.
Lim, Lisa. 2000. Ethnic group differences aligned? Intonation patterns of Chinese, Indian and
Malay Singaporean English. In The English Language in Singapore: Research on Pronuncia-
tion, Adam Brown, David Deterding & Ee Ling Low, 10–21. Singapore: Singapore Associ-
ation for Applied Linguistics.
Mani, A. 1993. Indians in Singapore society. In Indian Communities in Southeast Asia, Kernial
Singh Sandhu & A. Mani (eds), 787–809. Singapore: Times Academic Press.
Mufwene, Salikoko S. 1995. The founder principle in Creole genesis. Diachronica 13: 83–134.
DOI: 10.1075/dia.13.1.05muf
Pakir, Anne. 1991. The range and depth of English-knowing bilinguals in Singapore. World
Englishes 10: 167–179. DOI: 10.1111/j.1467-971X.1991.tb00149.x
Pakir, Anne. 2001. Bilingual education with English as an official language: Sociocultural
implications. In Georgetown University Round Table on Language and Linguistics, 1999:
Language in our Time, James E. Alatis & Tan Ai-Hui (eds), 11–30. Washington DC:
Georgetown University Press.
Pecotich, Anthony & Shultz, Clifford J. (eds). 2006. Handbook of Markets and Economies: East
Asia, Southeast Asia, Australia, New Zealand. New York NY: Sharpe.
Pfaff, Meike, Bergs, Alexander & Hoffman, Thomas. 2013. I was just reading this article – On the
expression of recentness and the English past progressive. In The Verb Phrase in English:
Investigating Recent Language Change with Corpora, Bas Aarts, Joanne Close, Geoffrey
Leech & Sean Wallis (eds), 217–239. Cambridge: CUP.
Rappa, Antonio L. & Wee, Lionel. 2006. Language Policy and Modernity in Southeast Asia: Ma-
laysia, the Philippines, Singapore, and Thailand. New York NY: Springer.
Sandhu, Kernial Singh. 1969. Indians in Malaya: Immigration and Settlement. Cambridge:
Cambridge University Press.
Sandhu, Kernial Singh. 1993. Indian immigration and settlement in Singapore. In Indian Com-
munities in Southeast Asia, Kernial Singh Sandhu & A. Mani (eds), 775–788. Singapore:
Times Academic Press.
Schiffman, F. Harold. 1999. A Reference Grammar of Spoken Tamil. Cambridge: CUP.
Shantakumar, G. 1993. “The Indian Population of Singapore: Some Implications for Develop-
ment. In Indian Communities in Southeast Asia, Kernial Singh Sandhu & A. Mani (eds),
866–909. Singapore: Times Academic Press.
Sharma, Devyani. 2009. Typological diversity in new Englishes. English World-Wide 30(2):
170–195. DOI: 10.1075/eww.30.2.04sha
Sharma, Devyani & Rickford, John. 2009. AAVE/Creole copula absence: A critique of the im-
perfect learning hypothesis. Journal of Pidgin and Creole Languages 24(1): 53–90.
DOI: 10.1075/jpcl.24.1.03sha
Siegel, Jeff. 2012. Constraints on substrate transfer revisited. Journal of Linguistics 48(2): 473–
478. DOI: 10.1017/S0022226712000060
Soh, Hooi Ling & Nomoto, Hiroki. 2009. Progressive aspect, the verbal prefix meN-, and stative
sentences in Malay. Oceanic Linguistics 48(1): 148–171. DOI: 10.1353/ol.0.0026
Sun, Chaofen. 2006. Chinese: A Linguistic Introduction. Cambridge: CUP.
DOI: 10.1017/CBO9780511755019
Suzanna, Bte Hashim & Brown, Adam. 2000. The [e] and [æ] vowels in Singapore English.
In The English Language in Singapore: Research on Pronunciation, Adam Brown, David
Deterding & Ee Ling Low (eds), 84–92. Singapore: Singapore Association for Applied
Linguistics.
Svalberg, Agneta M.-L. & Hjh Fatimah Bte Hj Awg Chuchu. 1998. Are English and Malay
Worlds apart? Typological distance and the learning of tense and aspect concepts. Inter-
national Journal of Applied Linguistics 8: 27–60. DOI: 10.1111/j.1473-4192.1998.tb00120.x
Tan, Ying Ying. 1999. Intonation Patterns of the Chinese, Malay and Indian Varieties of Singa-
pore English: A Substratist Approach. Honours thesis, National University of Singapore.
Tan, Ying Ying. 2012. Age as a factor in ethnic accent identification in Singapore. Journal of
Multilingual and Multicultural Development 33(6): 569–587.
DOI: 10.1080/01434632.2012.682730
Turnbull, Constance Mary. 1996. A History of Singapore: 1819–1988, 3rd edn. Singapore: OUP.
Wee, Lionel. 2003. Linguistic instrumentalism in Singapore. Journal of Multilingual and Multi-
cultural Development 24: 211–224. DOI: 10.1080/01434630308666499
Wee, Lionel. 2006. The semiotics of language ideologies in Singapore. Journal of Sociolinguistics
10: 344–361. DOI: 10.1111/j.1360-6441.2006.00331.x
Wong, Wee Kim (ed.). 2011a. Census of Population 2010. Statistical Release 1: Demographic
Characteristics, Education, Language and Religion. Singapore: Department of Statistics.
Wong, Wee Kim (ed.). 2011b. Census of Population 2010. Statistical Release 2: Households and
Housing. Singapore: Department of Statistics.
Yang, Guowen & John A. Bateman. 2002. The Chinese Aspect System and its Semantic Inter-
pretation. COLING ‘02 Proceedings of the 19th International Conference on Computational
Linguistics, Vol. 1, 1–7. DOI: 10.3115/1072228.1072259
Chapter 7
Zero articles in Indian Englishes

A comparison of primary and secondary
diaspora situations*
Marianne Hundt
The omission of articles where British or American English require either a

definite or indefinite article is a typical feature of Indian English (IndE). Sharma
(2005b) found that substrate influence played a role for the use of indefinite
one in IndE, whereas pragmatic functions (givenness and modification) played
a role in the use of zero articles (see also Sedlatschek 2009, 227). Definite the
and indefinte a are also omitted by Fiji Indians in Fiji (Mugler & Tent 2008)
and those who have moved to a secondary diaspora in places like New Zealand
or Australia. Sharma (2005a) investigated the use of articles by first-generation
immigrants from India in the US. Her study shows that zero articles are a fea-
ture that is retained in the diaspora context, even by speakers who are otherwise
close to using standard, native-like English. The data for this paper come from
fieldwork in Fiji (spontaneous conversations) and sociolinguistic interviews
conducted in the Fiji Indian Diaspora in Wellington, New Zealand. Record-
ings from the secondary diaspora include both first- and second-generation
migrants, and the study thus investigates whether zero articles are retained in
the speech even of people who acquired their English in a predominantly Eng-
lish-speaking environment. The study also aims to link the use of zero articles to
informants’ construction of identity in the secondary diaspora.
Keywords: variable article use, Fiji Indians, primary and secondary diaspora
* Funding for the fieldwork in Fiji in 2013 was provided by the University of Zurich’s Stiftung
Wissenschaftliche Forschung, project F-62420-01-01. Hanna Parviainen, Adina Staicov and
Lena Zipp as well as Katrin Forrer, Anne Gardner, Katja Meyer, Anja Neukomm-Herrmann
and Simone Pfenninger all helped with the transcription of audio files from Fiji and New Zea-
land, respectively. For comments on earlier versions of the paper, I would like to thank Carolin
Biewer, Andrea Sand, Gerold Schneider, Devyani Sharma and Jan Tent.
132 Marianne Hundt
1. Introduction
In many contact varieties of English, definite and indefinite articles are omitted
in contexts where they are required in first-language varieties of English (ENL).
This holds both for institutionalized varieties of English as a second language
(ESL) and for learner Englishes (EFL).1 It is therefore not surprising that we find
“zero” or “null” articles2 in different varieties of Indian English (IndE), both on
the sub-continent (see e.g. Agnihotri et al. 1994 or Sedlatschek 2009) and in dias-
pora contexts (see e.g. Mesthrie 1992; Mugler & Tent 2008).
Sharma (2005a; 2005b) investigated the use of articles by first-generation im-
migrants from India in the US. Her study shows that zero articles are a feature
that is retained in the diaspora context, even by speakers who are otherwise close
to using standard, native-like English. No previous study has compared variable
article usage in primary and secondary diaspora situations. Such a comparative
study could show whether first-generation migrants are likely to have adapted to
the majority ENL speakers in the secondary diaspora. A study that also samples
second-generation migrants would further enable us to investigate whether zero
articles are retained in the speech even of people who acquired their English in a
predominantly English-speaking environment. Taking such a comparative angle
is the main aim of this paper. The focus will be on Indo-Fijians in Fiji and New
Zealand.
The Indo-Fijian diaspora is of particular interest for several reasons. In the
primary diaspora in Fiji, Indians at one point (the 1960s) outnumbered the in-
digenous population and thus constituted an ethnic majority rather than a mi-
nority. Large-scale migration of Indo-Fijians to Canada, the US, Australia and
New Zealand started in the late 1980s (see Section 2.1). Of the possible secondary
diaspora contexts, New Zealand is particularly interesting because of its relatively
small and homogeneous host community and the resulting greater “visibility” of
the Indian diaspora:
1. See e.g. Platt et al. (1984), Williams (1987), Sand (2004) and Filppula et al. (2009). A cri-
tique of these and related studies can be found in Sharma (2012b). Williams (1987, 167) dis-
cusses variable article usage in the context of similarities across institutionalized second lan-
guage varieties and second languages acquired in a native speaker context – as an example of
“vulnerability” of English, i.e. constructions that are difficult to acquire for both types of learner
(NSs mark definiteness, ESL speakers mark specificity). In addition to article omission, both
ESL and EFL speakers insert articles where they are not typically used by ENL speakers.
2. Sharma (2005b, 545) argues that, since the term “zero article” has been used in reference to
bare NPs in standard English, studies of article omission in nonstandard Englishes (including
contact varieties such as Indian English) should use the term “null article”. In this paper, “zero”
is also used to refer to instances of article omission in Fiji Indian English.
Chapter 7.â•‡ Zero articles in Indian Englishes 133
Although the Indian diaspora in New Zealand is small on a global scene, the
issues of diversity and multiple identities are amply illustrated […]. Further, with
New Zealand’s relatively small population of just over 4 million, and a proactive
immigration policy, the proportion of its population that identifies as Indian has
increased rapidly over the last two decades. (Friesen & Kearns 2008, 212)3
The main reason for choosing zero articles as a case study is that they seem to be
such a persistent feature of English in Indian diaspora contexts. Sharma (2005a,
194) found that they were a relatively “[…] stable, incipient non-standard system
shared to some extent by all speakers”. The fact that speakers self-correct occa-
sionally and that hypercorrect use of articles is attested indicates that zero articles
are above the level of speakers’ awareness. At the same time, my informants do
not mention them as a feature typical of Indian Englishes (unlike accent and as-
pects of para-verbal communication), so they are unlikely to be a candidate for a
sociolinguistic marker or even stereotype. If they are found in the speech of sec-
ond-generation migrants this is likely to be the result of either substrate influence
or dialect contact with speakers from the first generation.
From a structural point of view, zero articles are interesting because they are
not simply learner errors but arise from typological differences between languag-
es that have articles and those that do not, as Platt, Weber and Ho (1984, 52–59)
point out: languages like English mark the difference between definite and indefi-
nite NPs whereas those like Chinese and Hindi mark the difference between spe-
cific and non-specific. Thus, article usage in speakers of Indian Englishes cannot
simply be attributed to different degrees of proficiency. This finds corroborative
evidence in Sharma’s (2005a, 205) comparative study of article usage and other
features:
Even if Indian English article use derives from late stage SLA, at least two aspects
of the new usage suggest that it may be a relatively stable system. First, the more
proficient speakers in the continuum show little evidence of being at other inter-
mediate learning stages, and so the contrast between their article absence rates
and lack of other SLA features is quite stark; a couple of these individuals have
grown up using English and consider it to be on an equal footing with their other
native languages. Second, speakers share a strikingly similar system of principles
for article use, deriving from an interaction between language transfer and dis-
course universals […].
3. Friesen and Kearns (2008, 212) continue to point out that “the impacts of the Indian dias-
pora on the New Zealand ‘mainstream’ have been considerable, in cultural, economic, social
and political terms.” An example of this would be the fact that the national museum (Te Papa)
organized various exhibitions that featured the Indian diaspora, among them one on Indian
weddings (see http://www.tepapa.govt.nz/LEARNING/AAINAA/default.htm).
134 Marianne Hundt
Furthermore, taking typological differences into account will explain not only the
absence of indefinite articles in non-specific contexts but also the use of one and
the demonstrative determiners this/these and that/those in specific contexts (for
more examples, see Section 4.1.2).4
(1) <$B><#>mom got it from one nurse (ICE-FJ, S1A-014)
(2) but I come back to my- in this country- Ø new country where I belong to now.
(NZ-Fji-1, female, MS)
Example (2) is interesting because it shows variation between a demonstrative

and a zero article, but the zero article is used in an NP where the head noun is
premodified by an adjective.
Previous studies (Agnihotri et al. 1994, 184f. and Sharma 2005b) have shown
that articles are more likely to be used in NPs without premodification, while pre-
modification increases the likelihood of article omission. According to Sharma’s
(2005b, 558) study, “bare nouns require overt articles, modified nouns are more
likely to be associated with omission of the article, and quantified nouns actually
favor null articles”. Sharma (2005b, 551) has also shown that indefinite articles are
more likely to be left out with non-specific NPs, something that she attributes to
transfer and the typological difference between English and Hindi. The usage of
definite and zero definite articles did not produce such an obvious pattern in the
English of her informants: “Where a gap occurs in the L1 (i.e., no definite article),
we do not find a matching absence of use in the L2 English grammar; instead,
there appears to be a nearly even rate of overt and null use of the definite article”
(Sharma 2005b, 552). Thus, while article usage in Sharma’s study reflected degree
of proficiency to some extent, it also showed signs of stabilizing new dialect for-
mation as far as the distribution of indefinite null articles was concerned.
4. Examples from the Fiji component of the International Corpus of English (ICE-FJ) are iden-
tified by their file names. Examples from the New Zealand diaspora project are referred to by
generation (NZ-Fiji-1 vs. NZ-Fiji-2), gender and the speaker code. Emphasis has been added
throughout. According to Sharma (2012a, 2087), language contact with different first languages
has resulted in regional stratification of this variable: “[…] this, that, and some are used more
frequently by South Indian speakers in place of English articles a and the as Dravidian lan-
guages permit the use of demonstrative and quantifier forms in these contexts.” Mugler & Tent
(2008, 551) refer to indefinite one as “perhaps the most distinctive characteristic of the Pure Fiji
English determiner system […]” pointing out that this is something typical of creoloid and L2
Englishes.
The hypotheses to be tested in this paper are that:5
1. Variable article usage will exhibit a growing closeness to ENL usage in the
following way:
a. speakers in the secondary diaspora will omit fewer articles than speakers
in the primary diaspora;
b. within the secondary diaspora, second-generation Indo-Fijians will omit
fewer articles than first-generation Indo-Fijians;
c. within the first generation, those speakers that migrated to New Zealand
before the onset of puberty will omit fewer articles than speakers who
migrated as adults.
2. With respect to different types of articles and NPs it can be expected that,
overall,
a. indefinite articles will be more commonly omitted in non-specific NPs
whereas the absence of a definite article in the L1 is less likely to show a
direct influence;
b. article omission is more likely to occur in NPs that show premodification
whereas article retention is more likely to occur in otherwise bare NPs.
Sharma (2005a, 196) points out that it is important to link results on language
use to evidence on attitudes towards in-group and out-group language use in the
diaspora. I will therefore discuss the quantitative results against speaker attitudes
in the secondary diaspora, taking into account the maintenance of transnational
ties (see Sharma, this volume).
In Section 2 of my paper, I will give some background on the Indian diaspora
in Fiji and the secondary Fiji Indian Diaspora in New Zealand. For the secondary
diaspora, qualitative data from the interviews will be used to evaluate the differ-
ent degrees to which transnational ties are being maintained. In Section 3, I will
describe the subset of recordings on which this paper is based, define the variable
context investigated, discuss problems of categorization, and detail data retrieval
and analysis. The results will be presented and discussed in Section 4.
5. Note that these hypotheses abstract away from possible differences amongst informants
caused by their educational background. Where available, this factor will be considered in the
evaluation of the results.
136 Marianne Hundt
2. The Indian diaspora in Fiji and the secondary Fiji Indian Diaspora
2.1 Indians in Fiji
Indians who migrated to Fiji are largely part of the second6 or colonial diaspora,
when large numbers of Indians where recruited as indentured labourers to work
on the tea, cotton, and sugar cane plantations of the British empire. The “recruit-
ment” did not always happen on a voluntary basis and working conditions on
the plantations were very close to slavery. The contract or “agreement” that the
labourers signed came to be referred to as girmit and the indentured labourers
referred to themselves as girmitiyas. Migration of indentured labourers to the
Fiji Islands started relatively late, namely in 1879.7 In the thirty years of girmit
migration, almost 61,000 people arrived from India to work on the sugar cane
plantations of the newly established British colony; the majority (45,000) came
from modern-day Uttar Pradesh and Bihar in the north and the remainder from
the South of India; they were from different casts and represented a cross-section,
socially, of Indian agricultural society (Lal 2006, 46ff. and Srebnik 2008). From
1900 onwards, free Indian settlers arrived in Fiji who were mainly of Punjabi and
Gujerati origin (the former mostly agriculturalists, the latter merchants).
Ethnically, Fijians and Indians stayed compartmentalised, as Srebrnik (2008,
91) points out:
There has been little in the way of cultural borrowing or adaptation and virtu-
ally no inter-marriage between the groups. As a result, and in contrast to more
integrated countries such as Mauritius or Trinidad, no “creolised” culture has
developed.8
Consequently, a complex political situation and ethnic tensions in Fiji have led to
a series of military coups (see e.g. Srebrnik 2008).9 The first one in 1987 directly
opened the sluices of Indian emigration from Fiji. The Indian population, which
at one point constituted a little more than 50 per cent, has dropped to under
6. Historians (e.g. Lal 2006) distinguish three phases, the first diaspora, which refers to the
trading movements of Indian merchants up to the beginning of the nineteenth century; the
second diaspora set in after the abolition of slavery in the first half of the nineteenth century;
the third diaspora is the post-war migration of professionals that is still in full swing.
7. The first colony to receive labourers from India was Mauritius in 1834.
8. For Indian English in Trinidad and Tobago, see the paper by Leung & Deuber (this volume).
9. Since the constitutional crisis in 2009, Fiji has been administered by an interim government
and, at the time of writing, is still awaiting democratic elections.
38 per cent.10 In the Encyclopedia of the Indian Diaspora, the current situation is
summed up as follows (Lal 2006, 382):
After more than a century, Indo-Fijians still struggle for political equality in the
land of their birth. The deeply felt but often unacknowledged need of the human
soul to belong, to have a place of one’s own, to be rooted, is denied them. How
long, they ask, should a people live in a place before they are allowed to call it
home? ‘From Immigration to Emigration’: that may in time come to be the epi-
taph of Fiji’s Indo-Fijian community.
2.2 Indo-Fijians in New Zealand
2.2.1 Migration and demographics

The 1987 coup coincided with a significant change in migration policy in New
Zealand, making migration to the South Pacific neighbour a lot easier than it had
previously been (see Friesen & Kearns 2008, 213). Since that coup, more than
120,000 Indians have left Fiji for Australia, New Zealand, the US, and Canada
(Lal 2006, 282; Srebrnik 2008, 90). Obtaining accurate statistics on the size of the
Fiji Indian Diaspora in New Zealand is difficult, however, as the census data tend
to list people as “Fiji-born” and do not give additional information on the ethnic
background of those who emigrated from Fiji; in addition, the second-generation
Fiji Indians do not appear in the statistics at all.11 In the self-identification part
of the census in New Zealand, “Fiji Indian” or “Indo-Fijian” are not provided as
labels.12 The 2006 census gives 37,74613 as the figure for people of Fiji origin, but
this figure must comprise ethnic Fijians and Indians, some people of mixed eth-
nicity and even, possibly, people of other ethnic backgrounds from Fiji. A much
more conservative figure is therefore based on the information of the birthplace
of Indian residents in New Zealand from the New Zealand Census of Population
and Dwellings.14
10. See http://www.statsfiji.gov.fj [last accessed 05.07.2013].

11. See, for instance, the New Zealand census homepage at http://www.stats.govt.nz/census/
2006-census-data/quickstats-about-culture-identity/quickstats-about-culture-and-identity
[last visited 09.22.2008].
12. Similar problems of obtaining accurate statistics on the Fiji Indian Diaspora also apply to
Australia (Jan Ten, p.c.).
13. http://www.stats.govt.nz/census/2006-census-data [last accessed 09.22.2008].
14. http://www.stats.govt.nz/datasets/population/census-of-population-and-dwellings.htm
[last accessed 09.22.2008].
138 Marianne Hundt
Table 7.1â•‡ Birthplace of Indian residents in New Zealand (1981, 2001)

NZ Fiji India Other Total
1981 â•⁄ 5,160 (44.5%) â•⁄ 1,617 (14.0%) â•⁄ 3,615 (31.2%) 1,185 (10.2%) 11,577
2001 17,946 (28.65%) 19,593 (31.3%) 19,053 (30.4%) 6,054 (9.7%) 62,646
2.2.2 Transnationalism
In this section, I will briefly outline the notion of the term “transnationalism” and
illustrate the degree to which it plays a role in the Fiji Indian Diaspora in New Zea-
land. Friesen & Kearns (2008, 225) define transnationalism as “a set of ongoing
linkages”. These can either be seen as consisting of actual social networks span-
ning across national borders (social morphology), as a “diaspora consciousness”
that involves dual or even multiple identities, or as social reproduction (Friesen
& Kearns 2008, 220, based on Vertovec 1999). By “diaspora consciousness” they
mean that, depending on the circumstances, “a person may be a Bengali, an In-
dian and/or a New Zealander of Indian, Fijian or other origin” (Friesen & Kearns
2008, 225). On an everyday basis of cultural practice and thus social reproduc-
tion, transnationalism may involve “hybrid forms of fashion, music and other art
forms […]” (Friesen & Kearns 2008, 220) “[…] through festivals like Diwali […]
as well as through websites, media attention, the increasing profile of Bollywood
and even languages regularly heard on the street” (Friesen & Kearns 2008, 226).
In my fieldwork I discussed all three aspects of transnationalism with inform-
ants. They maintained social networks with their family and friends in Fiji to
varying degrees, drew on the various options available to them for their identity
construction in different contexts, and commented on cultural and religious prac-
tice that linked them to their Fiji Indian origin, but also showed aspects of hybrid-
isation. It is clearly beyond the scope of this paper to give detailed background
information on these three dimensions of transnationalism for every informant
in the sample. In this section, I will briefly discuss the question whether, with
large-scale migration from Fiji since the late 1980s and a recent increase in migra-
tion from India, a pan-Indian identity is emerging in the New Zealand context.
According to Friesen & Kearns (2008, 222), diaspora consciousness as an aspect
of transnationalism is evidenced in the Indian diaspora in New Zealand:
The duality of being Indian and New Zealander is one aspect of this conscious-
ness, but the multiple identities of many Indian New Zealanders is also apparent.
Many ethnic associations have been formed to facilitate the maintenance of a
specific Indian regional culture. […] At the same time, there is a strong impera-
tive to establishing a pan-Indian identity and political presence in New Zealand.
The subjects that Friesen & Kearns interviewed did not all agree on the last aspect,
i.e. whether a pan-Indian identity existed in New Zealand (2008, 222f.). Inter-
estingly, they found a difference of opinion between the Indians from India and
those from Fiji:
Even these [Indo-Fijian community, M.H.] have a sense of ‘Indianness’, stem-
ming from the acquisition by most Indo-Fijians of Fiji Hindi, from a sense of
solidarity arising from the binary polarisation of the Fijian political system and
from their love of Bollywood. The last of these was referred to by others as a
pan-Indian phenomenon, since even many of those whose first language is not
Hindi are passionate about Bollywood. (Friesen & Kearns 2008, 223)
My informants commented on the experience of being “othered” by Indians from

India at times, and some of this seems to be based on the history of the Fiji Indian
Diaspora:
yeah/ the- the experiences that I’ve had with people from India is they don’t
consider you/ people from Fiji as one of them as- as being full-- fully Indian//
um// and that’s because of the whole indenture process of what happened
there um// they think that// it’s hard to- they don’t think that you’re one of
them/ so I find that they have sort of negative attitudes towards people from-
uh Indians from Fiji and especially in- in New Zealand like there’s- there’s
quite a lot from both- both countries and when you like they don’t really
interming – there’s not really much intermingling like the- you know the
sort of Fiji Indians all have their own little community and their own little
thing going on their own little associations and things whereas um/ the India
Indians will have the same/ there’s not a lot of intermingling going on because
of that sort of negative connotations// (34:50) (NZ-Fji-2, female, DV)
The awareness of boundaries may also be reinforced linguistically, i.e. by the fact
that Fiji Indians speak a markedly different variety of Hindi (see Section 2.3):
yeah yeah um/ yeah/ but um/ it’s <laughs>/ cause you know our Hindi is
broken from India so when you talk to India Indians they um/ think that you
can’t talk it properly so/ it’s quite interesting having conversations with them
about that <laughs>/
[…]
Indians from Fiji obviously we speak the same type of Hindi/ the broken
Hindi <laughs>/ so um it does make you feel a bit conscious though of how
you’re sounding and how you’e saying/ … (NZ-Fji-2, female, DV)
Informants are conscious of a trend towards using English amongst Fiji Indians:
140 Marianne Hundt
[…] I guess the reason I talk so much in English is because/ like I- I just
have such a diverse group of friends and it’s just easier and the same with my
brother/ like the people that we’ve interacted with are mo- mostly uh raised
here as well even if they were born in Fiji like family friends and what not they
were all raised here the ones we have known so my parents have known them
for quite a while so/ everyone converses in English like it’s just easier just you
know/ mm/ (NZ-Fji-2, female, DV)
In the long run, this may actually foster the development of a pan-Indian identity
in New Zealand, something that appears to have become possible for the sec-
ond-generation immigrants (despite a strong sense of boundaries between differ-
ent groups):
[…] with my parents’ generation/ they’re all pretty much <laughs> sticking
with their own// I’d- I can’t really imagine them all being in one/ like just in
one Indian group// uhm/ even though at uni/ like you’re friends with different
types of Indians/ but I don’t really know whether/ uhm/ whether it’ll all one
day eventually become one// like I don’t really see it happening/ just I don’t
know I guess ’cause we just see like our parents/ like they just socialise within
their own// like my my parents are uhm friends with a lot of like different
types of Indians/ but uhm it is always like a division and I think that probably
does have quite a bit of an influence on my generation as well/ even though
you do socialise with different Indians but/ I don’t know if that would one day
eventually become a whole/ as such because like the different Indian cultures
always have uhm like different social like social events// uhm like there’d be
like/ like a Punjabi/ they have like a social every now and then/ same with
Fijian Indians/ same with Gujeratis/ but they have their own different social
events// but then there’s also times uhm/ have you ever heard of Asia 2000//
ah// uhm it started quite a while ago/ but uhm every year around Diwali
time/ in like the civic centre and uhm in the town hall there’d be just like
a weekend where it’s just celebrating uhm Ø Indian New Year/ and so they
have like dance competitions and they have just like different performances/
different stalls/ they’d like have a bit of outdoor and indoor/ just like every-
one/ like even non-Indians/ like go to it as well ’cause they’re just interested
in that and everything// and that brings everyone together ’cause like all
the different Indian cultures go as well as non-Indians/ but that’s uhm yeah
kind of showing like a bit of a development/ that it’s not so like segregated so
much/ like people are uhm coming together with different Indian cultures
[…] (NZ-Fji-2, female, SiN)
Most of my informants maintained close contact with relatives and friends both
in the Fiji Islands and secondary diaspora contexts through regular visits.15 Some
(but obviously not all) had also visited India. A factor that fosters the mainte-
nance of transnational ties for the Fiji Indian secondary diaspora is the continued
migration from Fiji. SN, a second-generation female informant, mentioned the
recent arrival of relatives who brought with them their baby son and thus pro-
vided her with an opportunity to use Hindi rather than English with this relative.
Continued migration from Fiji might thus, to some extent, counteract the on-go-
ing language shift from Hindi to English in this community (see Section 2.3).
2.3 Sociolinguistics of language contact in the primary and secondary

diaspora
As in other contexts of the colonial diaspora, a special, local variety of Hindi de-
veloped in Fiji through language and dialect contact but it also has characteristics
of a former pidgin variety. Fiji Hindi, according to Siegel (1987, 187) is “a na-
tivized immigrant koiné of second generation Indians, a stable linguistic variety
which results from contact between varieties which are subsystems of the same
linguistic system”. Fiji Hindi is the mother tongue of the majority of Fiji Indians
in Fiji today, but it remains primarily a spoken variety;16 in school (if it is taught),
standard Hindi is used alongside English. In addition, English is favoured over
Hindi as a school subject and medium of instruction by the Indian community:
The Indians are conscious of the privileged position of the Fijians especially with
regard to land and see in the acquisition of English, a means of advancement, at
least a secure job and possibly employment overseas. (Geraghty 1984, 70)
As a result, a lot of Fiji Indians, while growing up with a variety of Hindi as their
mother tongue, are literate in English but not in standard Hindi, where the Deva-
nagari script may pose an additional obstacle to obtaining literacy skills, espe-
cially in the absence of formal schooling. English is the first language of a small
minority of people in Fiji, but mostly those of a mixed Fijian and European back-
ground (Tent & Mugler 2008, 235). English is the language that dominates the
media and it was used for Parliamentary debates until the constitutional crisis in
15. At the time that I was revising this article, for instance, one of my informants told me that
she and her older daughter were about to attend the wedding of one of her nieces in Canada
while her husband was due to fly out to Hamburg to attend the wedding of one of his nephews.
16. Prasad (in preparation) describes the standardizing grammar of Fiji Hindi, which is written
using the Latin alphabet, e.g. in text messages.
142 Marianne Hundt
2009. It is the lingua franca of Fiji and the second language of most native Fijians
and Fiji Indians. It is spoken with varying degrees of proficiency; but as detailed in
Section 3.1, the data used for this study come from acrolectal speakers of Fiji Eng-
lish (FE) with Hindi as their first language.17 There is both overlap and variation
between the English spoken by Fijians in Fiji and those of Indian background.18
Shameem’s (1995) study of the secondary diaspora in Wellington, New Zea-
land, found relatively high (but declining) proficiency levels in Fiji Hindi among
Fiji-born teenagers in the community; more importantly still, she found evidence
of a declining use of Fiji Hindi (FH) in the home and predicted the imminent loss
of both Fiji and standard Hindi in the community:
At present the situation seems to be that, unless the individual or the community
as a whole makes a conscious decision to maintain their language, FH proficiency
will continue to decline and almost certainly the language will not be spoken at
all by the next NZ-born generation of Indo-Fijians in Wellington.
(Shameem 1995, 298)
Shameem collected her data in 1991 and 1993. Even though my fieldwork did
not attempt to validate her results, it was obvious that the youngest children at
Hindi school in Wellington, while probably maintaining a good level of passive
knowledge, were struggling even with simple utterances like the Hindi equivalent
of I am a boy or My name is… In other words, the Fiji Indian community in Wel-
lington appears to be one undergoing language shift from Fiji Hindi to English.
To sum up, while the Fiji Indians in Fiji are speakers of English as an institu-
tionalized second language in a country where two vernacular languages – Fijian
and Fiji Hindi – are in widespread daily use, those who have migrated to New
Zealand are surrounded by a majority of ENL speakers. Like a lot of minority
17. Even though other Indian languages (e.g. Punjabi and Gujerati) are spoken by some Fiji
Indians, the language most widely understood and spoken is Fiji Hindi. (On Indian languages
other than Hindi, see e.g. Mugler (1998).)We did not sample any speakers with another Indi-
an language for the spoken conversations of ICE-Fiji. Only four Fiji Indians in our sample of
spontaneous conversations claimed to having had English as their first language, two of these
from a mixed family background (Fijian and Indian). Native Fijians claim English as their first
language more often in the spontaneous conversations sampled for ICE-Fiji.
18. For a detailed study of similarities and differences between the two Englishes as regards the
use of prepositions, see Zipp (2014). She also provides a more detailed account of the sociolin-
guistic background of English in Fiji. For the status of Fiji Indian English along the stages of
Schneider’s (2007) developmental model of new Englishes, see Zipp (this volume). For a recent
study of Fijian-Fiji English, see Biewer (forthcoming).
language communities, they have to invest into keeping their community lan-
guage alive, e.g. in the local Hindi school.19
3. Data and data retrieval
3.1 Recordings
The data for this study come from fieldwork conducted in New Zealand (2007/08)
and Fiji (2010). The former were obtained in sociolinguistic interviews with the
fieldworker, mostly with one informant at a time, but occasionally also with
groups of informants. A subset of interviews (six first-generation informants and
three second-generation informants; see Table 7.2) were selected for this study.
All speakers in the secondary diaspora are fluent speakers of English with at least
secondary if not tertiary education (adults). To a certain extent, this reflects the
immigration policies of New Zealand, which favour skilled and highly educated
immigrants over those with minimal education (see Friesen & Kearns 2008, 217).
As indicated in the introduction, the data within the first generation is further
sub-divided into those speakers that migrated to New Zealand as adults (MS, SaS,
SN, PM) and those who migrated shortly before puberty (SS, NK).
Table 7.2aâ•‡ First-generation speakers in the NZ secondary diaspora

Speaker Gender Age Occupation Migration Background
initials (year)
MS female 50s school teacher/care-giver 1988 married to a NZer
SaS male 60s journalist 1987 spent time in GB and NZ
before migration
SN female 50s school-teacher/consultant 1988 actively involved in
Hindi School
PM male 60s diplomat/manager 1987 studied in NZ
SS female 20s social worker 1988 was younger than ten
when family came to NZ
NK female 20s single mother, went to 1987 was younger than ten
polytech when family came to NZ
19. See http://www.hindischool.wellington.net.nz/ [last accessed 20.06.2012]. The Hindi

School in Wellington is run by volunteers, who invest a substantial amount of their free time in
trying to keep the community language alive.
144 Marianne Hundt
Table 7.2bâ•‡ Second-generation speakers in the NZ secondary diaspora

Speaker Gender Age Occupation Residence in NZ Background
initials since
NL male 10s student NZ-born spent extended periods in Fiji
SiN female 20s student 1988 was less than a year old when
family migrated
DV female 20s student 1988 was four when family
migrated
Table 7.3â•‡ Fiji Indians from Fiji (spontaneous conversations from ICE-FJ)
ICE-FJ file name Speaker 1 Speaker 2
S1A-014 female, 35–39 female, 20–24
S1A-020 male, 20–24 male, 20–24
S1A-021 male, 20–24 male, 20–24
S1A-030 female, 16–19 female, 16–19
S1A-031 male, 30–34 male, 45–49
S1A-044 female, 25–29 male, 25–29
The comparative data from Fiji are taken from the spontaneous conversations
sampled for the Fiji component of the International Corpus of English (ICE-FJ).20
These were recorded on the campus of the University of the South Pacific (USP)
during a field trip to Suva in August/September 2010. To meet the requirements
of the ICE sampling frame, only acrolectal speakers of FE were sampled, and all
the Indian informants from Fiji can thus be considered fluent users of English
as a second language. Six files (each approximately 2,000 words) and a total of
twelve speakers were selected (for details, see Table 7.3). The first language of all
informants included in this study as well as the first language of their parents is
Fiji Hindi.
As far as the two diaspora contexts are concerned, the study is based on slight-
ly different kinds of data: spontaneous conversations from the primary and soci-
olinguistic interview data from the secondary diaspora. The question is whether
the interview situation – i.e. interaction with a speaker from outside the com-
munity (on several levels) – is likely to have had a significant impact on the use
of articles. In general, speakers are likely to have adapted to the relatively formal
situation and will be monitoring their speech more carefully than in interaction
20. See http://www.research-projects.uzh.ch/p14682.htm for further information.

with peers. The following exchange shows that this does not necessarily influence
the use (or non-use) of articles:
(3) Interviewer: Did you have a traditional Indian wedding?
Informant: No no just Ø registry marriage. Yeah. (NZ-Fji-1, female, MS)
In the interview situation, informants might at best be omitting slightly fewer

articles than they would be in a spontaneous conversation. Some passages in the
conversations recorded in Fiji, on the other hand, show that informants are often
very aware of the “outsider” collecting the data, even if (as was the case) the field-
worker is not present during the recording. This can be illustrated with the fol-
lowing example where speaker A explains what grog is (a highly unlikely speech
act with a local):
(4) <$A><#>grog is <}><->a</-><,><=>a</=></}> Fijian drink and we
<}><->use</-><,><=>drink</=></}> a lot of grog it’s a <,>traditional
Fijian drink <,,> and <,,> we are continuing <}><->with the</-><=>with
our</=></}> tradition <{><[><,></[> every time we <?>come into our
home</?> (ICE-FJ, S1A-021)
In other words, even in the spontaneous conversations recorded in Fiji, a lot of the
informants are probably using their most standard-like variety. Again, this does
not mean that informants are not omitting articles. The following example from
the same conversation illustrates this:
(5) <$B><#>ok right now I don’t have any <,,> but I had in Ø past <,,> <}>
<->she</-> uh <=>she</=></}> a hot girl so I had a good luck
Note that this utterance also contains a hypercorrect indefinite article, probably
triggered by the relative formality of the situation (and a zero copula, another
feature found in many contact varieties of English).
On the whole, my data are likely to be more homogeneous than Sharma’s
(2005a; 2005b) data because all informants use English on a regular basis, i.e. it
is a strong or even dominant language for them. In Sharma’s study, the degree of
bilingualism amongst her first generation migrants was an important factor. In
my study, the Indians in Fiji constitute a relatively homogenous group of acrolec-
tal ESL speakers. In the secondary diaspora, the degree of assimilation to the host
country is likely to play a more important role than degrees of bilingualism.
146 Marianne Hundt
3.2 Definition of the linguistic variable
As pointed out in the introduction, the focus in this paper is on variable article
usage in contexts where ENL varieties require either a definite or an indefinite
article. I will briefly define these contexts in this section and then move on to
discuss some problematic cases. Problems concern variability in ENL varieties,
patterns typical of (Fiji) Indian English, and the fact that the data derive from
unscripted speech.
One aim of this study is to enable comparability with Sharma’s (2005a; 2005b)
findings. I therefore tried to follow her definition of the variable as closely as pos-
sible. The following are the examples of definite (6) and indefinite (7) null articles
that she gives (Sharma 2005a, 200f.):
(6) I asked Ø bus driver which way to go.
(7) a. I’m looking for Ø job. (non-specific)
b. I met Ø friend of yours. (specific)
Contexts in which articles are optional in ENL Englishes were excluded from the
counts. This is the case with plural NPs ((The) students were late for class), prop-
er names, in particular names of countries (e.g. Gambia vs. the Gambia),21 and
therefore the following instances of zero articles and a hypercorrect article were
excluded from the analysis:
(8) a. <$B><#>well I <{><-><?>was</?></-><=>managed</=></}> to watch
some of the games <,> semifinal final and some of the <?>full</?>
matches <,,> well my team did not actually win <,> <?>it was Ø
Netherlands</?> (ICE-FJ, S1A-021)
b. <$A><#><[>uh</[><{> <,>yeah <,,> my team was also <,> Ø Netherlands
(ICE-FJ, S1A-021)
21. Since Kjellmer’s (2002) study on article use with names of countries has revealed some
surprising results, I decided to exclude this variable context altogether, despite the fact that
an article might be obligatory with country names that have a plural form (Jan Tent, p.c.). In
spoken commentary or news writing, however, the article is occasionally omitted even in ENL
varieties, as the following examples from COCA illustrates:
(i) Ambulances and fire trucks rushed to deal with what is one of the worst disasters in
Netherlands history. (COCA, 1992, SPOK)
(ii) Was identified by scouts of Dutch power Ajax (Amsterdam) in 1996 and brought to
Netherlands. (COCA, 2000, NEWS)
(iii) Not just because of Netherlands and Mexico, whose enthusiastic fans are expected to
flock to the central Florida city. (COCA, 1994, NEWS)
c. <$B><#>though it managed to beat the <{><[>Brazil <,> no?</[>

(ICE-FJ, S1A-021)
Even though plural NPs allow for optional article usage in standard English, in-
stances with quantifiers often require an article as part of the quantifier phrase.
Sharma coded for modification and quantification22 and therefore included plural
nouns in her analysis (2005b, 547, 548), giving the following as an examples: It’s
a problem for a lot of families and In fact you find Ø lot of Andhra guys here rather
than Bangalorians.
3.2.1 Optional article use in standard English

One problem for article use in different Englishes is that there are a host of excep-
tional instances of optionality. Not all can be discussed here in great detail. The
example that Sharma (2005a, 201) gives is (the) taxation of income; further exam-
ples given in Sharma (2005b, 563) are fixed constructions such as most of the N,
all of the N, such a N and adverbial uses of phrases like a little and a lot. Additional
contexts would be nouns with single role referents (e.g. queen, president, professor,
manager), which show variable article usage when they are used as subject com-
plements in ENL varieties (see Hundt fc.-a). Thus, in the following example, only
sugar mill (but not manager) was counted as an example of a null definite article:
(9) But she is still there cos her husband is manager of Ø sugar mill.
By extension, this also includes contexts like the following one where MC (master
of ceremonies) is a single-role referent on a particular occasion:
(10) Um I just was a MC to a wedding last weekend. (NZ-Fji-1, male, PM)
The definition of the variable is further confounded by the fact that there is both
ongoing change (see e.g. Rastall 1995; Ilson 1995) and regional variation. A pos-
sible candidate for ongoing change is the following example from Friesen and
Kearns (2008, 212):
22. As Sharma (2005b, 547) points out, “[t]he category of “quantified” […] was fairly broadly
interpreted and included ranking adjectives (e.g., the first N, the last N, the best N), quantifying
phrases (e.g., a lot of N, a few of N, the whole N), and numeral modifiers (e.g., the eleventh N).
148 Marianne Hundt
(11) The significance of Auckland as migrant destination is again illustrated in the

fact that 71 per cent of the Indian population added to the New Zealand total
between 1986 and 2001 was based in Auckland […].23
Article use is also variable in ENL varieties with institutional nouns, especially if
insiders refer to the senate of a university or experts discussing politics refer to the
government or the cabinet. It is therefore difficult to decide whether the following
instances of a bare NP uses of cabinet and government would have to be included
among the null articles or not (the example is part of the spoken ICE-FJ compo-
nent but from a file not included in this study):
(12) A: good evening/ welcome to have your say// it’s been a dramatic past eight
hours starting with cabinet this morning/ and the consequent rejection of a
draft memorandum of agreement with the fiji islands council of trade unions
by government// (S1B-026)
A shibboleth of regional variation between British and American English, for in-
stance, are phrases such as be in/go to (the) hospital (see e.g. Bolinger 1996; Algeo
2006; Hundt fc.-b).
On this problem, Sharma (2005b, 563) remarks in a footnote:
For the purposes of this study, standard British and American English are treated
as generally equivalent systems. This is not always the case, as in dialectal variants
such as in (the) hospital. Such cases would have been omitted from the dataset,
but none were encountered during coding.
They did occur in my data (see example (13)) but were excluded from the analysis.
(13) <$B><#>my dad my dad is good <,,> recovered really well after <,,> being in
the hospital for four days (ICE-FJ, S1A-014)
All instances where article usage is variable in L1 English were omitted, even
though they are at times difficult to group under a particular heading, as the fol-
lowing instance:
(14) I’ve got a great affection for India and um I’m always proud of India’s achieve-
ments because um there’s a part of us which will always be rooted in India.
(NZ-Fji-1, male, SaS)
23. In the Corpus of Contemporary American English (COCA), there is only one example in
a set of 17 occurrences retrieved with the search string significance of N as without an article
preceding the following noun.
The problem with optional article usage in ENL varieties is that it makes for a
certain amount of inter-annotator disagreement and thus comparisons between
different studies on zero articles have to be taken with a grain of salt.
3.2.2 Features of Fiji Indian English and (Fiji) Indian expressions

In Fiji (Indian) English, some plural nouns are not marked for number. Gram-
matically, they should be treated like plural nouns, however, and they were there-
fore excluded from the statistics like all other plural nouns. An example would be
the following use of the set phrase back in the days:
(15) <$B><#><[>yeah exactly <,> back in the day</[></{> when the mobiles
<{><[><,> first started (ICE-FJ, S1A-044)
Some phrasal expressions have indigenized variants in IndE. One example are
light verb constructions, which are also attested without indefinite articles in IndE
(see Hoffmann et al. 2011). The patterns with the indefinite article were therefore
extracted from the transcripts despite the fact that no instances with a null article
are attested in my data.
(16) And he said “Oh everybody says ‘Oh can I have a bite can I have a bite?’”
(NZ-Fji-1, female, SN)
More problematic are instances where a definite article is used instead of an indef-
inite article (or vice versa). Sharma (2005a; 2005b) does not comment on them.
Example (17) illustrates the problem:
(17) Peter Ustinov spoke with the Austrian um German-Russian accent
(NZ-Fji-1, male, SaS)24
I decided to categorize them as instances of hypercorrect in-/definite articles even

though they are, strictly speaking, not the same as a(n) in/definite article used in
the context where L1-varieties of English have a bare NP. But they could not be
counted as an instance of a(n) in-/definite article nor as an instance of a bare NP,
either.
Another regionalism is the use of specialized, technical vocabulary with dif-
ferent articles than in ENL varieties. In the ICE-FJ data, Moodle (the name of an
e-learning platform) is fairly consistently attested with a definite article, so the
question arises whether they ought to be classified as instances of hyperdefinite
24. Alternatively, this could also be an instance of a hypercorrect definite article instead of a
demonstrative that if the speaker (a journalist) was aware of the tendency for speakers of Indian
Englishes to use demonstratives instead of definite articles.
150 Marianne Hundt
articles (when compared with ENL varieties) or whether instances without the
article need to be considered as null article contexts (locally).
(18) <$A><#>oh there are questions put on the moodle (ICE-FJ, S1A-014)
In the end, taking ENL varieties as the yardstick of comparison, I decided to clas-
sify them as cases of hyperdefinite article use. By the same token, some lexical
items typical of (Fiji) IndE such as girmit or Ramayan were assumed to require an
article in ENL varieties even though they are loanwords or might even be consid-
ered proper names. Sometimes intuition on whether a noun is likely to require an
article in ENL varieties is supported by analogy, i.e. the Ramayan or the Bhaga-
vad Gita and the Bible. Instances of these nouns without an article would thus be
counted as null article contexts (even though none were attested in my spoken
data; example (20b) is from the written component of ICE-FJ):
(19) It was part of the Girmit-. (NZ-Fji-1, male, SaS)
(20) a. … and the men would read out of the Ramayan/ or the Bhagavad
b. Gita/ all the Sanskrit t-- Vedas and texts (NZ-Fji-1a, female, NK)
c. According to Panchanga the events in Ø Ramayana are said to have taken
place about one million and three hundred years ago.
(ICE-Fiji, W2F-016)
Because plural NPs do not usually allow for articles, number can be an addition-
al problem in the classification of loanwords. A case in point would be salwar
kameez. It is a count noun, but can be both singular and plural (even though sal-
war kameezes is also attested). Because of its indeterminate status I did not count
the example in (21) as an instance of a null article and would, if they had been
attested, have also excluded cases with a definite article from the statistics.
(21) with woman [NZE plural of ‘women’, MH] it’s quite easy to tell cause um in
India they’d wear like salwar kameez or a sari or something whereas Fijian
Indians they’d wear like a/ a dress or like actual pants or something
(NZ-Fiji-2, male, NL)
3.2.3 Dealing with spoken data

Spoken data also pose special problems to the analyst that, to my knowledge, have
not been discussed in previous studies on variable article use. One example would
be self-corrections. The question is whether both NPs in such contexts should
be counted separately or not. In my opinion, this depends to some extent on the
kind of self-correction. In the following instances, where the speakers self-correct
from a definite to an indefinite article, only the NP with the indefinite article was
counted as a relevant variable context:
(22) a. And I always say that Brigadier-General Major-General Rabuka deserves

to be put in the- in a permanent … hall of- of infamy as a rogue
b. Sometimes the- the teacher would say “Copy a page of the dictionary”
<laughs> and give us a penny to in the- in a can. (NZ-Fji-1, female, MS)
c. But I think there was the- there was a big hole somewhere.
(NZ-Fji-1, female, SN)
By the same rationale, the following was not included in the statistics because the
speaker self-corrects to a bare NP, which would also be possible in ENL varieties:
(23) The um non-Muslims were given the right to pay tax for the protecti-- for
protection from Muslims. (NZ-Fji-1, male, SaS)
In the following instance, the speaker self-corrects from a definite article to a bare
NP; this instance was counted as one attestation of a bare NP, despite the fact that
in the immediate context the noun is repeated again with the definite article:
(24) And if you think a lot you always realize that the establishment- Ø establish-
ment is always for the rich and those who are powerful. And that the estab-
lishment [context] will always um not stand up for the downtrodden.
Other contexts are more problematic, for instance if the modifier is changed and
the indefinite article is not repeated, as in the following example:
(25) I was a Hindu so I went to a Hindu school-- Indian school.
The example is further complicated by the fact that, as pointed out in Section 3.2.1,
ENL varieties have variable article use in such constructions (i.e. be or go followed
by an institutional noun), so I decided to exclude it from the dataset altogether. In
example (26), finally, there seems to be a self-correction from a bare NP to an in-
definite article, but the head of the NP is different. In this case, I decided to count
both instances separately, the first as an instance of a zero article, the second as an
instance of an indefinite article:
(26) And they recognize our language because I mean Ø Fiji-Hindi dialect the Fiji-
Hindi language is uh similar to a dialect of Hindi spoken in uh Uttar Pradesh
in India. (NZ-Fji-1, male, SaS)
Instances that change mid-construction are similar to self-corrections. They were

excluded from the counts if the change was from article to a different kind of de-
terminer, as in example (27):
152 Marianne Hundt
(27) we are continuing <}><->with the</-><=>with our</=></}> tradition <{>

<[><,></[> every time we <?>come into our home</?> (S1A-021)
Another methodological problem in the analysis of variable article usage arises

from ellipsis. The following is likely to be an instance where the nominal head of
the NP phone was ellipted from a one GB but could be inferred from the context:
(28) <$A><#><[>so he</[></{> said that it was a one GB [ellipsis] <,,> and I think
the general one is <?>what</?> five <,,> something something <,,>
(ICE-FJ, S1A-044)
Similarly, the noun goalkeeper is not used in the following utterance but retrieva-
ble from the context of the conversation:
(29) <$B><#>oh that goalkeeper
<$A><#>yeah <,,> <{1><[1>I mean</[1> he was a good [ellipsis] <{2>
<[2><,,></[2> for his size <,> for his looks (ICE-FJ, S1A-020)
In example (30), the head noun question is also recoverable from the context:
(30) <$B><#><unclear>words</unclear> <,> oh <unclear> words </unclear> <,>

yeah I <?>did it</?> <{><[><,,></[> but uh <,,> for two or three questions
<,> I don’t know the answer to the second [ellipsis] which …
Examples (28), (29), (30) and similar cases were treated as all other instances
of ellipsis, though, i.e. excluded from the analysis despite the fact that the head
nouns were recoverable from the context.
Another issue arising from the nature of spoken data is the question how to
treat variable article usage that could be influenced by turn-taking and overlap. If
subjects continue another speaker’s turn, for instance, and that turn contains an
article, then the NP is considered to contain an article too and is not counted as
an instance of a bare NP, particularly if it is overlapping with a continuation of the
first speaker’s turn, as in the following example:
(31) <$A><#><[>it’s not</[></{> a job job I know it’s not like a grand job it’s not
a <{1><[1><,></[1> respectable reputable job but <,> it’s a <{2><[2>job</[2>
<$B><#><[1>well paying job</[1></{1> <,,> <[2>yeah</[2></{2 >
(ICE-FJ, S1A-044)
In the following example, the article is omitted turn-initially, which is also a con-
text where we find article omission in ENL varieties and which therefore was not
included among the zero articles; the second instance of the same phrase (funny
thing) occurs towards the end of the same turn and here the omission of the defi-
nite article is a lot less “natural” from the point of view of L1-varieties of English:
(32) <$A><#><[>funny thing was</[></{> <,> we <,> both different people did
our assignment and funny thing is that we almost got a similar answer
(ICE-FJ, S1A-020)
Finally, the phonetic environment sometimes makes it extremely difficult to de-

cide whether speakers have used an article or not. This may be the case for indefi-
nite articles between two liquids, for instance, as in the following example:
(33) woman [NZE plural variant for women] wear ?a? really Indian top…
(NZ-Fiji-2, male, NL)
If the word preceding a definite article ends in a dental fricative it is often difficult
or even impossible to determine whether it is followed by a definite article or not,
as in:
(34) in this world of uhm you know with ?the? whole Hindu thing of getting mar-
ried and obliging to what your husband says (NZ-Fji-1a, female, NK)
This also holds in the following example since the variant that is usually required
before a word that starts in a vowel (i.e. [ði:] instead of [ðə]) is not necessarily
used in FE:
(35) <$B><#><}><->no it’s</-> <,> <=>no I</=></}> just finished off with ?the?
<?>assignment</?> (ICE-FJ, S1A-031)
All unclear instances that arose from the phonetic environment were excluded
from the counts.
In my analysis, I did not code for the variant forms of the indefinite article, i.e.
I did not distinguish between instances where the indefinite article was adjusted
to the following word or where it was not, as in the following cases:25
(36) a. I call it Hindustani in Fiji because uh it is a amalgam of Hindi and Urdu.
b. like if you’re just calling someone like a idiot or something
(NZ-Fji-2, female, SiN).
These instances proved somewhat problematic because we are dealing with spo-
ken data and it is easy to transcribe them incorrectly as hesitation markers. This
happened, for instance, with a essay which initially was erroneously transcribed
as uh essay:
25. Occasionally, the variant an is used preceding a noun that starts in a consonant, as in “[…]
record your voice over bring it over match them together and do it like an proper news presenta-
tion” (ICE-FJ, S1A-014).
154 Marianne Hundt
(37) She wrote … uh a presentation she did at school- a essay or she- whatever she
had to do about … uh I forgot. (NZ-Fji-1, female, SN)
Thus, in the analysis of the transcripts, special attention was given to NPs preced-
ed by hesitation markers, which were routinely checked against the recordings.
3.3 Data retrieval and post-editing
The interviews and conversations were orthographically transcribed. Relevant in-

stances of variable article usage were manually retrieved by reading through the
transcripts (referring back to the recordings whenever necessary.) All variable
contexts were transferred into a FileMaker database and classified according to
the type of (null) article and structure of the NP (bare vs. modified vs. quantified).
A total of 1735 instances of variable article usage (definite/zero-definite and in-
definite/zero-indefinite) were thus retrieved from the data, including instances of
hypercorrect articles (or “article insertion” in the terminology of Sharma 2012b,
215). In a subset of my data, I also searched for article substitution separately,
i.e. instances where one or the demonstrative determiners this/these or that/those
were used in specific contexts instead of an article. If it is attested, this variable
might be a slightly more subtle indication of continuing substrate influence in the
second-generation migrants than article omission (see Section 4.1.2).
4. Results and discussion
4.1 Quantitative results
4.1.1 Article omission

The results for individual speakers are given in the tables in the Appendix. For
each sub-group in the secondary diaspora, they are ranked in descending order
of frequency of zero articles. (Hits for individual speakers in the primary Fiji di-
aspora were too small to rank the speakers.) The tables below summarize the data
for the diaspora in Fiji (Table 7.4) and the secondary diaspora in New Zealand.
The latter are divided into first-generation adult migrants (Table 7.5a) and those
who arrived in New Zealand before the onset of puberty (Table 7.5b); the findings
for the second-generation informants, including some who were very young (pre-
school age) when they migrated, are summarized in Table 7.6.
The results in Tables 7.4–7.6 confirm hypothesis 1, namely that there is a
growing closeness to ENL usage as we move further away from the Fiji Indian
Diaspora. In other words, the Indians in Fiji display the highest degree of article
Table 7.4â•‡ Variable article use in the primary diaspora (Fiji)

Bare Modified Quantified Total % Zero
definite 148 â•⁄53 13 214
zero-definite â•⁄27 â•⁄13 â•⁄5 â•⁄45 17.4
indefinite â•⁄54 â•⁄42 â•⁄6 102
zero-indefinite â•⁄12 â•⁄19 â•⁄5 â•⁄36 26.1
Total 241 127 28 396
all articles 202 â•⁄95 19 316
all zero articles â•⁄39 â•⁄32 10 â•⁄81 20.5
Table 7.5aâ•‡ Variable article use in the secondary diaspora (New Zealand) –
first generation (adult migration)
definite 268 126 30 424
zero-definite â•⁄23 â•⁄31 â•⁄4 â•⁄58 12.0
indefinite 203 147 46 396
zero-indefinite â•⁄13 â•⁄23 10 â•⁄46 10.4
Total 507 327 90 924
all articles 471 273 76 820
all zero articles â•⁄36 â•⁄54 14 104 11.3
Table 7.5bâ•‡ Variable article use in the secondary diaspora (New Zealand) –
first generation (pre-adolescence migration)
definite 33 11 10 â•⁄54
zero-definite â•⁄1 â•⁄2 â•⁄1 â•⁄â•⁄4 â•⁄6.9
indefinite 35 31 11 â•⁄77
zero-indefinite â•⁄2 â•⁄8 â•⁄0 â•⁄10 11.5
Total 71 52 21 144
all zero articles â•⁄3 10 â•⁄1 â•⁄14 â•⁄9.7
omission in variable contexts where ENL varieties require them (20.5 per cent
overall), whereas the second-generation speakers in the New Zealand diaspora
omit only very few articles (4.1 per cent overall), with first-generation migrants
in the secondary diaspora in between (11.3 per cent and 9.7 per cent overall for
the adult and pre-puberty migrants, respectively). As a mirror image of null ar-
ticles, we might also briefly consider the use of hypercorrect articles in the three
156 Marianne Hundt
Table 7.6â•‡ Variable article use in the secondary diaspora (New Zealand) –
second generation
definite â•⁄29 â•⁄11 â•⁄5 â•⁄45
zero-definite â•⁄â•⁄1 â•⁄â•⁄2 â•⁄0 â•⁄â•⁄3 6.3
indefinite â•⁄79 â•⁄89 44 212
zero-indefinite â•⁄â•⁄1 â•⁄â•⁄5 â•⁄2 â•⁄â•⁄8 3.6
Total 110 107 51 268
all zero articles â•⁄â•⁄2 â•⁄â•⁄7 â•⁄2 â•⁄11 4.1
contexts: Fiji Indians in the primary diaspora score highest at 5.5 per cent (23/419),
the first-generation speakers in New Zealand reach 3.3 per cent (36/1104) and the
second-generation speakers have the lowest score at 1.5 per cent (4/272).26
Returning to the use of null articles, a chi-square test reveals that the differ-
ences between the three main groups prove significant (at p < 0.000 and df = 2),
whereas the difference between adult and pre-puberty migrants in the first-gener-
ation group does not (Yates chi-square = 0.18, p < 0.67 and df = 1). This may part-
ly have to be attributed to the small sample size of the pre-puberty migrants (two
speakers in one recording, but only part of the interview was analyzed, yielding
a total of only 144 variable contexts). Adding to this sample would therefore give
more robust results. But variation among first-generation speakers in the second-
ary diaspora is also quite marked. One speaker (MS) has rates of article omission
(19 per cent and 18.6 per cent for definite and indefinite null articles, respectively)
that are close to the rates found in the sample taken from ICE Fiji (16.4 per cent
and 26.1 per cent) than to the average rate of the first-generation migrants. At
the other end, another first-generation migrant (SaS) has rates (3.4 per cent and
3.1 per cent) that are even lower than the average rates found in the second-gen-
eration speakers in New Zealand (6.3 per cent and 3.6 per cent). This is confirmed
by relating the mean scores of the sample (first generation Fiji Indians) to the
standard deviation (see Table 7.7a).27
If the measure “standard deviation/mean” approaches or exceeds 100 per cent,
then the fluctuation within the sample is greater for this value than any tendencies
of the sample as a whole. It will be interesting to see whether these quantitative
26. The percentages are calculated against the total of relevant contexts, including definite and
indefinite, zero-definite and zero-indefinite as well as hypercorrect definite and indefinite arti-
cles. Unclear cases were excluded from the statistics.
27. The mean values and standard deviations are given in a table in the Appendix.
Table 7.7aâ•‡ Variable article usage in first-generation Fiji Indians in New Zealand – testing
for fluctuation within the sample (Standard Deviation/Mean) – adult migrants only
Bare Modified Quantified
definite 28% 60% â•⁄59%
zero-definite 63% 37% 115%
indefinite 49% 56% â•⁄59%
zero-indefinite 53% 63% 200%
Table 7.7bâ•‡ Variable article usage in first-generation Fiji Indians in New Zealand –
testing for fluctuation within the sample (Standard Deviation/Mean) – all first-
generation informants
definite 10% 52% 11%
zero-definite 16% 13% 21%
indefinite 12% 12% 12%
zero-indefinite 12% 12% 41%
findings are in line with speakers’ comments on self-identification and attitudes

towards their host community as well as the degree to which transnational ties are
maintained within the older first-generation informants. There is less fluctuation
if the younger first-generation informants are included in the sample, as Table 7.7b
indicates.
This means that the significant difference between the first- and second-gen-
eration informants found in the chi-square test more accurately describes the
difference between all of the first-generation informants against the second-gen-
eration informants.
A closer look at the second-generation informants shows that the degree to
which they omit articles does not match up with their biographies: DV, who was
4 years old when her parents moved to New Zealand, has the lowest proportion
of zero articles whereas SiN, who was only a year old when she came to New Zea-
land, has the highest proportion of zero articles, with NL, a New Zealand-born
male informant, ranging between those two. It will therefore be interesting to see
whether we find any support in the qualitative analysis of the interview data when
we look at these speakers’ attitudes towards their own community and the host
community as well as differences between speakers in the maintenance of trans-
national ties (see Section 4.2).
With respect to the second hypothesis, my study confirms Sharma’s (2005a;
2005b) results of systematic use of null articles that reflects substrate influence
158 Marianne Hundt
Table 7.8â•‡ Null articles according to specificity of the NP

Indefinite Indefinite Definite
specific28 NPs non-specific NPs NPs
N % N % N %
primary diaspora â•⁄5/81 6.3 29/55 52.7 â•⁄45/259 17.4
secondary diaspora
1st generation 10/364 2.7 45/163 27.6 â•⁄58/482 12.6
secondary diaspora
2nd generation â•⁄0/158 0 â•⁄8/62 12.9 â•⁄â•⁄3/48 â•⁄6.3
Total 15/602 2.5 81/280 18.9 106/789 13.4
Table 7.9â•‡ Type of NP – first generation Fiji Indians in New Zealand

Bare Modified Yates chi-square (1df) p-value
definite 301 177 12.49 0.0004
zero-definite â•⁄24 â•⁄38
indefinite 238 235 10.19 0.0014
zero-indefinite â•⁄15 â•⁄41
from Hindi (see Table 7.8): a chi-square test confirms that the likelihood of an in-
definite article being omitted from a non-specific NP is significantly higher than
it is to be omitted from a specific NP; this holds across the dataset as a whole as
well as for the two diaspora settings separately. My findings also confirm Sharma’s
result in that substrate influence is far less noticeable with definite articles: even in
the primary diaspora, zero definite articles are well below 50 per cent.
As far as differences between different types of NPs are concerned, only the
figures obtained for the first-generation migrants to New Zealand were large
enough to allow for statistical testing. Note that modification and quantification
were collated in Table 7.9 under the heading “modified”. The figures include both
older and younger migrants.
Raw frequencies for article omission in the second generation are too low to
test for statistical significance, but with only two zero articles in bare NPs and 9
in modified NPs they support the general trend exhibited by the first-generation
28. Following Sharma (2005b, 547), I included predicative NP constructions among the specif-
ic NPs.
migrants. My data thus confirm Sharma’s (2005a; 2005b) finding that articles are
more likely to be omitted from modified NPs than from bare NPs.29
4.1.2 Article substitution

Examples in (1) illustrate that Fiji Indians in Fiji and the secondary diaspora in
New Zealand occasionally use one as a substitute for the indefinite article. The ex-
amples in (38) reveal that it is attested mostly in the primary and less often in the
secondary diaspora. Interestingly, the secondary diaspora informant (example c.)
self-corrects from one to an article when the noun is repeated (and then, charac-
teristically, has a null article in the modified NP that follows):
(38) a. <$B><#><}><->he had</-><=>he got</=></}> one accident but it was
not uh <,> no serious one (ICE-FJ, S1A-020)
b. he was just looking the other side <?>when</?> one vehicle came which
was just <w>know</w>? <,> a minor one (ICE-FJ, S1A-020)
c. <$A><#>past girlfriend I had one girlfriend and she was not good with
me <{><[><,,> like</[> <,,> she used me (ICE-FJ, S1A-021)
d. in Fiji we have one whole class we teach the class- Ø whole class the same
subject (NZ-Fji-1, female, MS)
A search for this feature in the speech of second-generation migrants in New

Zealand did not provide any evidence that article substitution persists in the lan-
guage of speakers who, by and large, have assimilated to the host culture’s variety
of English. Similarly, examples of demonstrative that or this replacing a definite
article are found in the language of first-generation migrants, only:
(39) a. oh this is all these precious foods that we grew up with
(NZ-Fji-1a, female, NK)
b. and the fact that you have given us this honour of being with you and
talking to you and sharing tea with you – it’s great (NZ-Fji-1, female, SN)
c. she just- uh the tongue twisted right away that accent everything and
more so much vocabulary. (NZ-Fji-1, female, SN)
d. the English colonialism of Fiji never ever accepted that Fiji Indians could
one day run that country and the whole policy of Fiji of the English
government was to compartmentalise races. (NZ-Fji-1, male, SaS)
e. The only reason why I left was that I was a journalist and when um Major-
General Rabuka imposed uh censorship on our news I had there and then
decided to leave this country. (NZ-Fiji-1, male, SaS)
29. In the primary diaspora, 39/202 articles are omitted in bare NPs against 42/114 in modified
NPs, a difference that still proves significant at p = 0.014 (if a little less so).
160 Marianne Hundt
f. You will recall the first Indians to Fiji landed on those- the shores of those/
this<?>30 beautiful country on May fourteen<th?> eighteen seventy-nine.
(NZ-Fiji-1, male, SaS)
In examples (38d–f), reference is to Fiji by someone speaking in New Zealand,

making the substitution particularly obvious.
With respect to article substitution, my findings again substantiate the results
that Sharma (2005b, 552) obtained: “The use of one might be more noticeable in
Indian English and thus cited more often in nonquantitative studies, due to its
greater salience rather than its actual quantitative frequency”.
4.2 Qualitative results
In this section, I will come back to the question whether quantitative differences
between speakers are also reflected in their attitudes towards their own speech
community and the host community (see Sharma 2005a, 196). Data on this aspect
were only collected from the secondary diaspora in New Zealand. I will focus on
two individuals from the first generation and compare the two female informants
from the second generation. Both sets of informants can be said to provide so-
ciolinguistic puzzles of some kind because, at first glance, their article use seems
to be at odds with their identity construction and, in the case of the second-gen-
eration migrants, the two female informants, the age when they started to learn
English.
4.2.1 First-generation immigrants (PM vs. SS)

As pointed out in Section 4.1, first-generation Fiji Indians omit articles to varying
degrees, with some coming closer to the rate of omission found in the sample
from ICE-FJ, others approximating the rate of omission of the second-generation
informants. PM is a male informant who studied in New Zealand and migrated
when he was in his early forties. He worked as a diplomat for the Fijian gov-
ernment and held a well-paid job in New Zealand at the time of the interview.
Furthermore, he is a member of the Rotary Club and was the president of the Fiji
Indian Association when the fieldwork was conducted. He has come to identify
strongly as a New Zealander, though:
30. Note that vowel length is often neutralized in FE, and it is therefore not altogether clear
whether the change in this context is from plural those to singular this or plural these. The fol-
lowing head noun makes a self-correction from those to this somewhat more likely.
I think I- I would- my own preference is I would be- I’m a New Zealander

of Fiji Indian background. You know I’d be very comfortable with that. I- I
hold a New Zealand passport I’m a New Zealand citizen I have contributed
the last 20 years to this country so I’m a New Zealander in every respect of
the word but I’m of Fiji Indian background and I’m very proud of that. Like
everybody should be proud of their background because that’s what we are
we don’t have any option to get to change or not to change. So you know I’m
a Fiji- I’m an- Fiji Indian you know we are different from Indians from India
both in terms of our language and our food and our culture to a degree you
know. Although we can speak Hindi uh we also have Fiji Hindi which is dif-
ferent from Hindi from India. So yeah I’m a New Zealander now and um of
Fiji Indian background. So if somebody asked me what you are I would say
I’m a New Zealander and they want to know more I’ll say I’m from Fiji I’m a
Fiji Indian but I now am a New Zealander. (NZ-Fji-1, male, PM)
The other informant, (SaS), is of about the same age and migrated at the same
time. He is a journalist who spent time both in Britain and New Zealand before
he left Fiji. Moreover, as a journalist, he is probably very aware of the language
he uses. He is actively involved in the Fiji Indian community in Wellington as a
board member of the Fiji Indian Association. Unlike PM, however, he strongly
identifies as an Indian from Fiji:
I’m a first-generation um Fiji Indian in New Zealand and um I came here in
nineteen eighty-seven to be chief sub-editor for National Radio at Wellington/
and I came with my then wife and two children/ and uh they actually- they
took to New Zealand but my son to this day regards Fiji as home/ so does my
daughter/ […] and so do I/ […] well home is where you’re born/ the way I
look at it is home is where you’re born/ and uh while India is our ancestral
home/ we were born in Fiji and uh we can’t deny that/ so we grew up there/ we
have very pleasant and fond memories of growing up/ of schooling/ of educa-
tion/ of our parents uh thriving/ so how can we not have also good memories
of that country/ we remember the rivers/ we remember where we swam/ we
remember where we used to dive from the bridges/ we remember swinging
from the trees into the rivers/ so Fiji is home/ and anyone who says that-/ uh
any Fiji Indian who says New Zealand is their home I think is playing around
with-/ uh playing around with words a bit (NZ-Fji-1, male, SaS)
In line with these remarks is the fact that SaS claims to have maintained his FE
Indian accent. However, the different identity constructions by PM and SaS do
not match up with the informants’ use of articles: It is PM who has a high ra-
tio of article omission whereas SaS comes close to informants from the second
162 Marianne Hundt
generation. This suggests that zero articles are unlikely to be consciously em-
ployed for the purpose of linguistic identity construction.
4.2.2 Second-generation immigrants (DV vs. SiN)

DV’s and SiN’s opinions on relations amongst Indians in New Zealand were dis-
cussed in Section 2.2.2. Both came to New Zealand with their parents in 1988
when they were very young, but they omit articles in standard contexts to differ-
ent degrees. Even though DV was older when her parents moved to New Zealand,
she omits fewer articles than SiN. Examples of null articles the speakers used are
given in (40) and (41):
(40) they- they’ve always said they did it for us and- and for themselves too I guess
just better opportunities and- and um more science/ Ø more developed coun-
try than- than Fiji/ (NZ-Fji-2, female, DV)
(41) a. just like Ø pronunciation of some words (NZ-Fji-2, female, SiN)
b. every year around Diwali time/ in like the civic centre and uhm in the
town hall there’d be just like a weekend where it’s just celebrating uhm Ø
Indian New Year (NZ-Fji-2, female, SiN)
c. I think like it’s important to have Ø good balance if you’re gonna be living
in a country that’s not like a main/ like Ø Indian country
(NZ-Fji-2, female, SiN)
We might have expected DV to omit articles more often than SiN because she
was older when her parents moved to New Zealand. A likely explanation why SiN
omits articles more often is that she has been back to Fiji on a regular basis:
it does// like not-- I probably would call New Zealand more home/ but it [Fiji]
still is like a big part of me because I go there quite often ’cause my grandpar-
ents still live there/ so I go back like every two, three years/ so it’s still Ø pretty
big, important part [like home// […] usually like two three weeks each time
we go there// sometimes we’ve been there for like six weeks/ like the whole of
summer// but uhm/ yeah it’s still a really big important part// I wouldn’t say
it’s home/ but kind of as home/ but not as much as New Zealand/ but I don’t
know/ kind of hard to explain/ but yeah// I still think it’s really important/ to
me// (NZ-Fji-2, female, SiN)
SiN is also in touch with parts of her family who migrated to New Zealand more
recently, including relatives with young children who arrived speaking only Fiji
Hindi:31
31. The positive attitude towards Hindi does not mean that SiN has negative attitudes towards
the host country; on the contrary: “if you’re living in a country that’s not India and not Fiji that
I’ve got a little cousin/ he’s almost two now/ but uhm ’cause he was born in
Fiji/ he probably came to New Zealand/I mean he was probably like eight,
six months old/ seven months old/ uhm/ and he had like/ he’s got two older
brothers who’re like four and nine when they came/ and so they didn’t know
English at all because they just spoke Hindi over in Fiji/ so I still do// or not
so much now/ but uhm I used to always talk to them in Hindi// but now
because the two older boys have uhm started school/ like they’re – they’re
quite uhm good with English/ so I speak to them in English and Hindi// but
uhm yeah like my little cousin/ the baby one/ I used to always talk to him in
Hindi//<laughs>/ it’d be a bit of both/ but probably more Hindi// yeah so yeah
didn’t really feel weird// (NZ-Fji-2, female, SiN)
DV, on the other hand, even though she went back to Fiji with her parents when
she was a child, seems to be more dominant in English (see her previous com-
ment on using English with Fiji Indians). This might partly be due to the fact that
she has memories of having her Hindi commented on when she went back to visit
relatives in Fiji:32
um/ when I was little/ um but yeah/ when I was little I remember going back
um// talking about language um/ uh/ my accent like they found it very hard
to comprehend when I was talking Hindi to my family/ yeah/ which is quite
interesting like the first time I went back/ the first time after we moved/ yeah/
and it’s- it’s uh become worse <laughs>/ […] they couldn’t really understand
me my- my grandparents wanted me to talk slower <laughs> and// uh// it
was just because now I had that accent when I talked in Hindi and it sounded
different to what it did before/ so/ yeah/ um/ […] well I really felt a little bit
out of place there <laughs>/ because I was like/ hey/ <laughs>/ I’m one of you
<laughs>/ you should understand me <laughs>/ but um// yeah it did bug me
for a little bit like oh you know a little bit out of place like/ yeah/
(NZ-Fji-2, female, DV)
you shouldn’t really isolate yourselves from like the New Zealanders and Kiwis/ so I wouldn’t
really want to send my children to like uhm a primary school where it’s just Indian children
because it kind of affects how they uhm socialise with non-Indian people in New Zealand/ like
I thought then/ uhm because then like you may as well just be living in India or in Fiji if you
kind of do that kind of thing/ like/ yeah” (NZ-Fji-2, female, SiN).
32. DV also comments on the ease with which she learnt English after the family had moved to
New Zealand: “like I can speak Hindi really well but I found English quite easy to pick up/ and
um/ I was just through I mean I like I started primary school here so/ that was a- an advantage
I guess” (NZ-Fji-2, female, DV).
164 Marianne Hundt
5. Conclusion
The quantitative results of my study have shown that article omission decreases
with the growing “distance” from the original diaspora setting, thus confirming
hypothesis 1. However, on closer inspection, there was considerable fluctuation
within speaker groups, most likely reflecting different degrees of English language
proficiency. This is more likely the outcome of the informants’ educational back-
ground than the degree of contact with native speakers of English: the first-gen-
eration informant with the highest rate of article omission is married to a New
Zealander.
With respect to language-internal constraints, my study confirms those
found in Sharma’s (2005a; 2005b) studies, namely that (a) indefinite articles show
a strong tendency to be omitted in non-specific contexts and (b) both kinds of
article are more likely to be omitted in modified NPs than in bare NPs. The results
of my study thus also confirm hypothesis 2. Finally, substitution of definite arti-
cles with demonstrative pronouns or one for the indefinite article is not a feature
found in the speech of second-generation migrants in New Zealand but only in
language used by first-generation speakers where it is a more subtle (but infre-
quent) indication of substrate influence than article omission.
The qualitative analysis shows that variable article use in the secondary di-
aspora does not always meet the expected patterns with respect to identity con-
struction or psycholinguistic theory. We saw that for the two first-generation
speakers, self-identification is at odds with the ratio of article omission that the
speakers exhibit. Article omission appears to correspond with different degrees
of proficiency in English rather than identity construction as a New Zealander of
Fiji Indian background vs. Fiji Indian. For the two second-generation informants,
article omission did not correspond to differences in the age at which the speak-
ers moved to an English-dominant speaking surrounding (recall that informants
who moved to New Zealand at a very young age are also included among my sec-
ond-generation sample). Instead, maintenance of transnational ties and attitudes
towards their community language appear to be more important factors that will
have to be taken into account in future research (see also Sharma, this volume).33
On the whole, article omission is extremely infrequent in the language of
the second-generation migrants in New Zealand. My study on variable article
use thus lends support to Hoffman and Walker’s (2010, 59) results: investigating
33. We are aiming to operationalize measures for the degree of maintenance of transnational
ties and thus make them amenable to statistical analyses that correlate the use of morphosyn-
tactic features with identity construction in an ongoing project based on first- and second-gen-
eration Indians in London (see Hundt & Staicov, in preparation).
different features, they also found that substrate influence did not persist into
the second-generation Italian- and Chinese-Canadians in Toronto. More detailed
analyses of other features than article omission will have to show whether, never-
theless, a focused ethnic variety is emerging amongst the second-generation New
Zealand Fiji Indians. This is a likely outcome in the light of research conducted by
Trudgill (2004) and Sharma and Sankaran (2011).
References
Agnihotri, Rama Kant, Khanna, Amrit Lal & Mukherjee, Aditi. 1994. The use of articles in
Indian English: Errors and pedagogical implications. In Second Language Acquisition: So-
ciocultural and Linguistic Aspects of English in India, Rama Kant Agnihotri & Amrit Lal
Khanna (eds), 178–198. New Delhi: Sage.
Algeo, John. 2006. British or American English? A Handbook of Word and Grammar Patterns.
Cambridge: CUP. DOI: 10.1017/CBO9780511607240
Biewer, Carolin. Forthcoming. South Pacific Englishes – A Sociolinguistic and Morphosyntac-
tic Profile of Fiji English, Samoan English and Cook Islands English. Amsterdam: John
Benjamins.
Bolinger, Dwight. 1996. Oddments of English. Journal of English Linguistics 24: 5–24.
DOI: 10.1177/007542429602400102
Filppula, Markku, Klemola, Juhani & Paulasto, Heli. 2009. Digging for roots: Universals and
contacts in regional varieties of English. In Vernacular Universals and Language Contact:
Evidence from Varieties of English and Beyond, Markku Filppula, Juhani Klemola & Heli
Paulasto (eds), 231–264. London: Routledge.
Friesen, Wardlow & Kearns, Robin A. 2008. Indian diaspora in New Zealand: History, identity
and cultural landscapes. In Tracing an Indian Diaspora. Contexts, Memories, Representa-
tions, Parvati Raghuram, Ajaya Kumar Sahoo, Brij Maharaj & Dave Sangha (eds), 210–
236. New Delhi: Sage. DOI: 10.4135/9788132100393.n12
Geraghty, Paul. 1984. Language policy in Fiji and Rotuma. In Duivosavosa: Fiji’s Languages:
Their Use and Their Future [Bulletin of the Fiji Museum 8]. Suva: Fiji Museum.
Hoffman, Michol F. & Walker, James A. 2010. Ethnolects and the city: Ethnic orientation and
linguistic variation in Toronto English. Language Variation and Change 22(1): 37–67.
DOI: 10.1017/S0954394509990238
Hoffmann, Sebastian, Hundt, Marianne & Mukherjee, Joybrato. 2011. Indian English – An
emerging epicentre? A pilot study on light-verbs in web-derived corpora of South Asian
Englishes. Anglia 129(3–4): 258–280. DOI: 10.1515/angl.2011.083
Hundt, Marianne. Forthcoming a. Who is the/a/Ø professor at your university? A construc-
tion-grammar view on changing article use with single role predicates in American Eng-
lish. To appear in María José López-Couso et al. (eds), English Corpus Linguistics on the
Move: Applications and Implications. Proceedings from the 34th ICAME conference, San-
tiago de Compostela. Amsterdam: Rodopi.
Hundt, Marianne. Forthcoming b. Variable article usage with institutional nouns. An ‘oddment’
of English? In The Noun Phrase in English, William van der Wurff & Alex Ho-Cheong Le-
ung (eds). Amsterdam: John Benjamins.
166 Marianne Hundt
Hundt, Marianne & Staicov, Adina. In preparation. Expressing and negotiating identity in the
London Indian diaspora. (Paper presented at the Sociolinguistics Symposium 19, Berlin,
2012).
Kjellmer, Göran. 2002. The Britain: An unexpected case of article usage in present-day Eng-
lish. In From the COLT’s Mouth… and Others’: Language Corpora Studies in Honour of
Anna-Brita Stenström, Leiv Egil Breivik & Angela Hasselgren (eds), 167–180. Amsterdam:
Rodopi.
Ilson, Robert. 1995. A(n)-dropping. English Today 11,1(41): 42–43.
DOI: 10.1017/S0266078400008075
Lal, Braj V. (ed). 2006. The Encyclopedia of the Indian Diaspora. Singapore: Singapore University
Press.
Mesthrie, Rajend. 1992. English in Language Shift. Cambridge: CUP.
Mugler, France. 1998. South Indian languages in Fiji: Language contact and attrition. In Pro-
ceedings of the Second International Conference on Oceanic Linguistics, Jan Tent & France
Mugler (eds), 95–108. Canberra: Pacific Linguistics.
Mugler, France & Tent, Jan. 2008. Fiji English: Morphology and syntax. In Varieties of English.
The Pacific and Australasia, Kate Burridge & Bernd Kortmann (eds), 546–567. Berlin: De
Gruyter.
Platt, John, Weber, Heidi & Ho Mian Lian. 1984. The New Englishes. London: Routledge.
Prasad, Rajendra. In preparation. The Grammar of Fiji Hindi. MA thesis, University of the
South Pacific.
Rastall, Paul. 1995. Definite article or no definite article. English Today 11,2(42): 37–39.
DOI: 10.1017/S0266078400008257
Sand, Andrea. 2004. Shared morpho-syntactic features in contact varieties of English: Article
use. World Englishes 23(2): 281–298. DOI: 10.1111/j.0883-2919.2004.00352.x
Sedlatschek, Andreas. 2009. Contemporary Indian English: Variation and Change [Varieties of
English around the World G38]. Amsterdam: John Benjamins. DOI: 10.1075/veaw.g38
Schneider, Edgar W. 2007. Postcolonial English: Varieties Around the World. Cambridge: CUP.
DOI: 10.1017/CBO9780511618901
Shameem, Nikhat. 1995. Hamai log ke boli (Our Language). Language Shift in an Immi-
grant Community: The Wellington Indo-Fijians. PhD dissertation, Wellington: Victoria
University.
Sharma, Devyani. 2005a. Dialect stabilization and speaker awareness in non-native varieties of
Sharma, Devyani. 2005b. Language transfer and discourse universals in Indian English article
use. Studies in Second Language Acquisition 27(4): 535–566.
DOI: 10.1017/S0272263105050242
Sharma, Devyani. 2012a. English in India. In Historical Linguistics of English, Vol. II [HSK
34.2], Alexander Bergs & Laurel Brinton (eds), 2077–2091. Berlin: De Gruyter.
Sharma, Devyani. 2012b. Shared features in New Englishes. In Areal Features of the Anglophone
World, Ray Hickey (ed.), 211–232. Berlin: De Gruyter.
ual change in London Asian Speech. Language Variation and Change 23: 399–428.
DOI: 10.1017/S0954394511000159
Siegel, Jeff. 1987. Language Contact in a Plantation Environment: A Sociolinguistic History of
Fiji. Cambridge: CUP.
Srebrnik, Henry. 2008. Indo-Fijians: Marooned without land and power in a South Pacific ar-
chipelago? In Tracing an Indian Diaspora. Contexts, Memories, Representations, Parvati
Raghuram, Ajaya Kumar Sahoo, Brij Maharaj & Dave Sangha (eds), 75–95. New Delhi:
Sage. DOI: 10.4135/9788132100393.n5
Tent, Jan & Mugler, France. 2008. Fiji English: Phonology. In Varieties of English. The Pacific and
Australasia, Kate Burridge & Bernd Kortmann (eds), 234–266. Berlin: De Gruyter.
Trudgill, Peter. 2004. New-dialect Formation: The Inevitability of Colonial Englishes. Edinburgh:
EUP.
Vertovec, Steven. 1999. Conceiving and researching transnationalism. Ethnic and Racial Studies
22(2): 447–462. DOI: 10.1080/014198799329558
Williams, Jessica. 1987. Non-native varieties of English: A special case of language acquisition.
English World-Wide 8(2): 161–199.
Zipp, Lena. 2014. Educated Fiji English: Lexico-grammar and Variety Status [Varieties of English
around the World G47]. Amsterdam: John Benjamins. DOI: 10.1075/veaw.g47
Appendix – Additional tables34
Bare Mod Quant Total % Zero

MS: N = 133 definite â•⁄47 â•⁄16 â•⁄1 â•⁄64
zero-def â•⁄â•⁄5 â•⁄â•⁄8 â•⁄2 â•⁄15 19.0
indefinite â•⁄19 â•⁄11 â•⁄4 â•⁄34
zero-indef â•⁄â•⁄5 â•⁄â•⁄4 â•⁄0 â•⁄â•⁄9 18.6
Total â•⁄76 â•⁄39 â•⁄7 122 18.9
PM: N = 223 definite â•⁄56 â•⁄16 â•⁄9 â•⁄81
zero-def â•⁄â•⁄9 â•⁄â•⁄8 â•⁄0 â•⁄17 17.3
indefinite â•⁄54 â•⁄39 â•⁄9 102
zero-indef â•⁄â•⁄4 â•⁄â•⁄5 10 â•⁄19 17.0
Total 123 â•⁄68 28 219 16.4
SN: N = 328 definite â•⁄88 â•⁄40 11 139
zero-def â•⁄â•⁄8 â•⁄11 â•⁄2 â•⁄21 13.1
zero-indef â•⁄â•⁄3 â•⁄11 â•⁄0 â•⁄14 â•⁄8.7
Total 179 â•⁄98 33 310 11.0
SaS: N = 282 definite â•⁄77 â•⁄54 â•⁄9 140
zero-def â•⁄â•⁄1 â•⁄â•⁄4 â•⁄0 â•⁄â•⁄5 â•⁄3.4
zero-indef â•⁄â•⁄1 â•⁄â•⁄3 â•⁄0 â•⁄â•⁄4 â•⁄3.1
Total 129 122 22 273 â•⁄3.3
34. Note that the total number (N) of variable articles is sometimes higher than the one given
in the tables because it includes instances of unclear contexts as well as hypercorrect use of
articles.
168 Marianne Hundt
Bare Mod Quant Total

NK: N = 99 definite 21 â•⁄7 â•⁄9 â•⁄37
zero-def â•⁄1 â•⁄1 â•⁄1 â•⁄â•⁄3
indefinite 21 22 â•⁄8 â•⁄51
zero-indef â•⁄2 â•⁄6 â•⁄0 â•⁄â•⁄8
Total 45 36 17 â•⁄99
SS: N = 46 definite 12 â•⁄4 â•⁄1 â•⁄17
zero-def â•⁄0 â•⁄1 â•⁄0 â•⁄â•⁄1
indefinite 14 â•⁄9 â•⁄3 â•⁄23
Total 26 16 â•⁄4 â•⁄46
SiN: N = 130 definite 28 â•⁄9 â•⁄1 â•⁄38
zero-def â•⁄1 â•⁄1 â•⁄0 â•⁄â•⁄2
Total 53 51 26 130
NL: N = 53 definite â•⁄0 â•⁄1 â•⁄1 â•⁄â•⁄2
zero-def â•⁄0 â•⁄1 â•⁄0 â•⁄â•⁄1
indefinite 21 17 â•⁄6 â•⁄44
Total 21 20 â•⁄7 â•⁄48
DV: N = 92 definite â•⁄1 â•⁄1 â•⁄3 â•⁄â•⁄5
zero-def â•⁄0 â•⁄0 â•⁄0 â•⁄â•⁄0
Total 36 36 18 â•⁄90
S1A-014 speaker A: N = 39 definite 12 â•⁄4 â•⁄2 â•⁄18
zero-def â•⁄2 â•⁄1 â•⁄2 â•⁄â•⁄5
indefinite â•⁄2 â•⁄3 â•⁄0 â•⁄â•⁄5
Total 18 â•⁄9 â•⁄5 â•⁄32
S1A-014 speaker B: N = 12 definite â•⁄4 â•⁄2 â•⁄1 â•⁄â•⁄7
zero-def â•⁄0 â•⁄0 â•⁄1 â•⁄â•⁄1
indefinite â•⁄1 â•⁄0 â•⁄0 â•⁄â•⁄1
Total â•⁄6 â•⁄4 â•⁄2 â•⁄12
S1A-020 speaker A: N = 42 definite 14 â•⁄5 â•⁄1 â•⁄20
zero-def â•⁄2 â•⁄6 â•⁄2 â•⁄10
indefinite â•⁄3 â•⁄5 â•⁄2 â•⁄10
Total 20 17 â•⁄5 â•⁄42

S1A-020 speaker B: N = 26 definite â•⁄6 â•⁄3 1 10
zero-def â•⁄0 â•⁄4 0 â•⁄4
indefinite â•⁄2 â•⁄3 1 â•⁄6
zero-indef â•⁄1 â•⁄3 0 â•⁄4
Total â•⁄9 13 2 24
S1A-021 speaker A: N = 46 definite 12 â•⁄1 1 14
Total 29 12 5 45
S1A-021 speaker B: N = 26 definite â•⁄5 â•⁄2 1 â•⁄8
Total 15 â•⁄6 2 23
S1A-030 speaker A: N = 23 definite â•⁄9 â•⁄2 0 11
Total 16 â•⁄4 0 21
S1A-030 speaker B: N = 42 definite 11 â•⁄4 0 15
indefinite 10 â•⁄5 0 15
Total 26 11 0 37
S1A-031 speaker A: N = 58 definite 33 â•⁄7 2 42
Indefinite â•⁄7 â•⁄3 1 11
Total 43 11 3 57
Total 17 10 1 28
S1A-044 speaker A: N = 50 definite 16 12 3 31
Total 27 20 3 50
170 Marianne Hundt

Total 15 10 0 25
Table A.â•‡ Variable article usage in first-generation Fiji Indians in New Zealand
sample mean and standard deviation – adult migrants only
Mean
definite 67 31.5 â•⁄7.5
zero-definite â•⁄5.75 â•⁄7.75 â•⁄1
indefinite 50.75 36.75 11.5
zero-indefinite â•⁄3.25 â•⁄5.75 â•⁄2.5
Standard Deviation
definite 18.81488772 18.78829423 4.434711565
zero-definite â•⁄3.593976442 â•⁄2.872281323 1.154700538
indefinite 24.99833328 20.46745384 6.757711644
zero-indefinite â•⁄1.707825128 â•⁄3.593976442 5
Table B.â•‡ Variable article usage in first-generation Fiji Indians in New Zealand
sample mean and standard deviation – all first-generation migrants
Mean
definite 291 133.6666667 39.16666667
zero-definite â•⁄24 â•⁄32.16666667 â•⁄5
indefinite 226.3333333 170.5 54.5
zero-indefinite â•⁄15 â•⁄29.33333333 10
Standard Deviation
definite 30.00944296 19.82338686 4.457203907
zero-definite â•⁄3.898717738 â•⁄4.135214626 0.98319208
indefinite 25.97434632 19.71463078 6.284902545
zero-indefinite â•⁄1.870828693 â•⁄3.188521078 4.082482905
Chapter 8
A lesser globalisation
A sociolexical study of Indian Englishes
in diaspora, with a primary focus on South Africa
Rajend Mesthrie
This paper makes a case for the comparative study of Indian English vocabu-
lary in diaspora. Comparing characteristic usage of Indian terms in their new
environments reveals socio-cultural niceties that may not always be possible
in studies of syntax and phonetics. The domains of food, music, kinship and
clothing are particularly rich in showing cultural retentions as well as degrees of
change, adaptation and hybridity. The change and adaptation is not only due to
‘host community – diasporic community’ contact dynamics, but to the relative
proportions of migrants from different parts of India, their regiono-cultural
differences and mutual influences in diaspora. The main examples come from
selected food, music, kinship and clothing terms drawn from the Dictionary of
South African Indian English (Mesthrie 2010), with some preliminary compar-
isons with Indian communities elsewhere. Attention is also paid to notions of
scale now prominent in the globalisation literature (Blommaert 2010), as cer-
tain terms are upscaled or downscaled in a new social environment.
Keywords: borrowing, Dictionary of South African Indian English, hybridity,

indenture, retentions, scaling, South African Indian English
1. Introduction
Given the increasing interest in transnationalism and diaspora, it is important that

sociolinguists highlight the linguistic dimensions of the study of cultural change,
adaptation, survival, retention and innovation in diaspora. At the same time the
influence of migrants on the language practices of host societies is worthy of con-
sideration. This paper focuses on the potential of comparative studies of the lexis
172 Rajend Mesthrie
associated with India in such diasporic contexts.1 Vocabulary studies have been
largely neglected in sociolinguistic studies, except in the field of regional dialectol-
ogy. Yet, as anthropologically-oriented linguists know, their importance in matters
of culture and identity hardly needs emphasising. The Indian diaspora has contrib-
uted enormously to new cultural formations and cultural enrichment of territories
on every inhabited continent. Their influence in matters of cuisine, dress, enter-
tainment, literature and even politics is great, though perhaps more so in some ter-
ritories than others. I would like to propose that a kind of modern lexicostatistics
is possible, a measure of the differential survival of key Indian terms in different
territories. Such a lexicostatistics will require careful preparatory groundwork. In
this chapter I examine the background to such an enterprise, by focussing largely
on the Indian diaspora in relation to English. My exemplification will come from
South African Indian English, since detailed lexical studies exist for the Indian
community there, culminating in A Dictionary of South African English (Mesthrie
2010). However, the main frame for this paper is an international one.
A prefatory note on spellings is necessary. Since the aim of this paper is to
record English usage in diaspora, the spelling forms used are of a hybrid nature
that follow certain desiderata: (a) a reasonably accurate indication of pronunci-
ation, and (b) a reasonably reader-friendly system. Spellings show the main pro-
nunciations in a particular part of the diaspora, in which vowel length is marked
without the misleading diacritical use of ‘r’, ‘w’, or ‘h’ to indicate length. The use of
a macron above the vowel is useful in showing length and – mostly as a by-prod-
uct – stress. Since all final vowels of English are slightly lengthened, this conven-
tion will not be used with final vowels. Use of the macron avoids doubling vowels
for length, another source of possible confusion. Thus ‘ee’ suggests long [e:] in
the English transliteration of Indian languages, but in ordinary English suggests
[i:] as in keen. However, where an established English spelling of an Indian word
exists (e.g. chutney) it is desirable to keep that spelling. There is thus a trade-off
between consistency and convention which seems the best way forward for Eng-
lish as a globalising language.
2. Historical considerations: Layers of diaspora history
In Mesthrie (2008) I proposed a tentative schema for characterising three identi-

fiably different periods and types of diaspora. The first diaspora comprises move-
ments out of India/ South Asia prior to the 17th century. This is admittedly a loose
1. I use the term ‘India’ in its modern demarcation, though for the purposes of the project
proposed here it could just as well apply to greater India of the British period.
Chapter 8.â•‡ A lesser globalisation 173
concept covering a wide period and is largely defined in terms of differences from
the 2nd and 3rd diasporas. In the first diaspora migration followed upon trade
and settlement by sailing ship in Southeast Asia and East Africa. There was at this
time also a religious impulse that took Hinduism and Buddhism out of India to
other parts of Asia.2
The second diaspora is more cohesively defined and mainly covers the forced
migration of slaves and indentured workers from India during the period of Eu-
ropean (mainly British) Imperialism. Places as far apart as Kabul, Cape Town
and Port Louis (Mauritius) were the recipients of large numbers of Indian slaves
in the 17th and 18th centuries. The movement of indentured workers in the 19th
and early 20th centuries is perhaps better known for its Indian communities in
the Caribbean, Mauritius, South Africa and Fiji. The term indenture refers to the
system where an employee was bound by contract to work for a fixed period for a
colonial employer who paid his (or her) passage to the colony. Similar migrations
took Indians to Singapore, Malaysia and other parts within Asia. Allied to the
movement of labourers in this period was the voluntary movement of traders,
especially Gujaratis, to the colonies that had a large presence of Indian inden-
tured labourers. The era of indenture is probably the best studied part of Indian
diasporic studies, and much is known of the sociolinguistics of Indian languages
taken to Fiji (Siegel 1987), Mauritius (Domingue 1971), South Africa (Mesthrie
1991, 1992), Guyana (Gambhir 1981), Suriname (Damsteegt 1988) and Trinidad
(Mohan 1978). Prominent among indentured labourers were the following lan-
guages: Bhojpuri, Awadhi, Hindi, Urdu, Tamil, Telugu and some Marathi (in e.g.
Mauritius). This set contrasted with the languages of the trading class Indians:
amongst which Gujarati was the most numerous, with smaller numbers of speak-
ers of Kacchi, Marathi, Konkani, Panjabi etc.
The third diaspora is a consequence of post-colonialism and globalisation – see
Sridhar’s (2008) sociolinguistic account of migrants to the UK and USA, though
other terrains like Australia, New Zealand, Canada would also apply). These in-
cluded professionals, and reasonably well-educated urban dwellers in contrast to
the second diaspora (see Sridhar 2008). Links with the home country were made
stronger in an era of air travel and satellite communications. However, third dias-
pora migrants were not assured of building the close-knit communities character-
istic of indenture in small coastal and island communities (see further Mesthrie
2008, 498–499). There are factors that complicate this simple schema – notably the
existence of “double diasporas” as people of Indian origin moved in large groups
from one terrain to another, often because of political uncertainties (see Hundt
2. One should not forget the Gypsy migrations out of India between the 9th and 12th
centuries.
174 Rajend Mesthrie
and Rathore in this volume). There are identifiable groups of East African Asians
in Britain, Indo-Guyanese in Canada, and Indo-Fijians in Australia.
3. Themes in comparative studies of second diaspora lexis
By ‘diaspora lexis’ I refer to words and concepts emanating from the home country
that have survived, perhaps with adaptation, in new terrains. This lexis is “from
below”, taken by migrants, rather than adopted by British or other European colo-
nials from days of empire. The usage of the latter is, of course, of interest in its own
right and follows an old tradition of dialect lexicography of the British Raj (start-
ing with Yule & Burnell 1886). Moreover, a rigorous separation of British usage
of Anglo-Indian terms versus that of their subjects is probably impossible. The
first diaspora is too distant to warrant a sociolinguistic account of relevance to
contemporary Indian English, even though some terms originating in this period
eventually entered the English lexicon, and pass as normal usage today: Buddha
and Buddhist, nirvana, karma etc. A sociolinguistic picture emerges in the second
diaspora, with the coining of a large number of common terms pertaining to the
experience of indenture. While this term is not of Indian origin by any means, the
concept was bestowed with Indian overtones in a number of colonies. The word
girmit first entered Bhojpuri and related languages of the indentured diaspora,
before passing into the English of Indian communities in places like South Africa
(Mesthrie 2010), Fiji (Siegel 1987), Guyana (Gambhir 1981), Trinidad (Mohan
1978), etc., where it is used with the same semantics. The word is based on the
British term agreement, and today has historical rather than colloquial signifi-
cance. It is kept alive by a historical sense that still celebrates the first arrival of
Indians in the different colonies. South Africa had its burst of commemorations
in 2010 of 150 years of the arrival of the first ship Truro (in 1860). This con-
nectedness with history has spawned many volumes of tales, recollections and
community histories in which the word was given a new lease of life. Girrmit
Tales (Govender 2008) was the title of one collection of short stories exploring
indenture and its aftermath in South Africa. The Bhojpuri agentive noun girmitya
‘one who signs a girmit, indentured worker’ also passed into English.3 Thus Gir-
mitiyas is a notable book by the Fijian/Australian historian Brij Lal (2004), with
the subtitle ‘The origins of the Fiji Indians’. The back cover blurb of Govender’s
book (2008, cited above) mentions the girrmitya experience (showing the use of
this noun as adjective) and of the Natal girrmityas (showing its English status by
3. By contrast, the Tamil equivalent girmit karan ‘agreement man, indentured worker’ did not
pass into English in this full form.
accepting a plural suffix). The term is so central to Indian identities and ideolo-
gies in the different terrains that it deserves a full lexicographical entry (slightly
adapted from Mesthrie 2010):
GIRMIT n.
1. A contract under indenture (q.v.).
2. Service under indenture.
3. Place of work under indenture. Older speakers. Historical. Also spelt
girrmet, girrmit. [Bhojpuri girmiṭ, ultimately from English agreement].
Indentured labourers, on the other hand, were just armed with a flimsy agree-
ment (girrmit), 10 shillings and rice per month – P. Naidoo, Preface Girrmit
Tales, xi.
Murugan did not work that day, the first time ever since he had come to work
under the girrmet to Mr Rutherford at Mount Edgecombe – P. Poovalingam
Anand, 128.
GIRMITYA / GIRMITIYA n.
An indentured worker – Older speakers. Historical. Sometimes pejorative,
denoting one from an impoverished background, as opposed to a merchant
Indian (q.v.) [Bhojpuri girmiṭ, plus agent suffix -yā].
The indentured labourers were those who went to Natal on an agreement for
five years, and came to be known there as girmityas from girmit, which was a
corrupt form of the English word ‘agreement’ – M.K. Gandhi Autobiography,
99.
Two more examples will be used to show the widespread use of terms from the
era of indenture. Indentured workers in the colonies often originated from a vast
array of villages, districts, provinces and castes, and hence speaking a variety of
regional and social dialects and languages. However, new identities were forged
from the outset in diaspora. In many colonies this realignment was expressed in
terms of a new and basic distinction according to the port of embarkation, the
place where one was held in a depot to await the next ship. The port thus deter-
mined the geographical provenance of the people one would be closely associat-
ed with for weeks on board ship and beyond. The terms Calcuttia and Madrāsi
carried a symbolic significance for over a century in the colonies. Today these
terms also have mainly historical significance, as the entries from South Africa
(Mesthrie 2010) show:
MADRĀSI / MADRĀJI / MANDRĀJI n.
Person of South Indian extraction, usually Tamil (q.v.) or Telugu-speaking.
So-called because they were transported from the port of Madras (q.v.) to
Durban by the British government as indentured (q.v.) labourers. Sometimes
176 Rajend Mesthrie
pejorative in the context of early rivalry between North and South Indians liv-
ing together in Natal, especially the Bhojpuri-Hindi pronunciation Mandrāji.
Older speakers. ¶ The number of people shipped from Madras to Durban
between 1860 and 1911 was higher than those from Calcutta or Bombay.
See Calcuttia, indenture, passenger Indian, Truro. [Gujarati, Hindi, Telugu,
Urdu Madrāsi, Bhojpuri Mandrājī, from Madras, the port from which South
Indian indentured workers were shipped out to the colonies, now known as
Chennai].
… caste was beginning to mutate into two very broad regional identities,
Kalkatia and Madrasi, which reflected origin from the Aryan North or
Dravidian South India – A. Desai & G. Vahed Inside Indenture, 174.
Never mind worrying about caste, they’re now busy chasing Tamils and
Muslims and making fools of themselves. Which nice boy will want to marry
a girl that’s been with a Madraji? – K. Pillay Looking for Muruga, 30.
CALCUTTIA / KALKATHIA n.
A Hindi-speaking person of North Indian extraction, or descendant of such.
Sometimes pejorative when used in connection with early minor rivalries
between North and South Indians living together in Natal. Older speakers. ¶
The term indicates the importance of the port cities in forming a new identity
among Indian migrants. It is a misnomer in terms of places of origins of the
North Indian migrants, very few of whom came from the city of Calcutta. See
Madrāsi. [Bhojpuri kalkatyā ‘one who embarked ship at Calcutta’ (for Natal
or other colonies)].
Lurking in the shadows, though less pronounced than the Kalkatia and
Madrasi distinction was the Tamil/ Telugu division – A. Desai & G. Vahed
Inside Indenture, 177.
Nithin consoled himself with the thought that Tony’s indifference towards
him was nothing personal – Calcuttans were normally taciturn people and
not given to long conversation – A. Hassim Lotus People, 297.
The above entries and the citations speak eloquently of new identity construction
according to broadly regional origins in India. The difference between Indians
form the north and south of India is perhaps exaggerated in the texts cited, since
at certain levels there was unity among all people of Indian origin, partly be-
cause of their ascribed status within the South African political and social sys-
tem. For the first hundred years since the initial migrations social relations were
good among all Indians, though there was a strong tendency not to marry across
(former) regional (Gujarati – Tamil – Bhojpuri) or religious (Hindu, Muslim,
Christian) lines. A hundred and fifty years later this arrangement has changed,
with less stigma attached to marriage across regional lines, and less stigma at-
tached to Hindu-Christian marriages among Indians. The salient sub-identities
that gelled in South Africa a century ago are thus less strong today, but still in ex-
istence, as witnessed by cultural associations like the Andhra Maha Sabha (Telugu
Cultural Association), the Gujarati Parishad (Gujarati Association) etc.
One term from the period of indenture that has achieved a (living) colloquial
status, at least in South Africa is lathās, an adjective meaning ‘shoddy, haphaz-
ard’. The term originated in Mauritian Creole, but was robust enough to pass into
South African and Fijian usage, and evokes linkages between the different sugar
colonies. Indians in the different sugar colonies were not entirely cut off from one
another. Mauritian Indian workers were brought to Natal to start the first railway
line in Africa (1875) and as forerunners in the establishment of Natal’s sugar in-
dustry. The historical records also speak of indentured workers returning to India
from one colony and later re-indenting to another. A small number of people in
this way had experience of Mauritius, the Caribbean, and/or Fiji. The entry below
is from South African Indian English (henceforth SAIE), as recorded in Mesthrie
(2010):
LATHĀS adj./n.
A shoddy job, a job hastily and badly done. Hence a lathās job, to make
latās, do latās. [South African Bhojpuri latās mār ‘to do a shoddy job’, from
French la tâche ‘the task’, via Mauritian Creole which drew a distinction
between lathas ‘a job to be done properly’ and tap latas ‘do a piece of work
quickly and shoddily’, from tap v. ‘to bang, knock, bump into’ – P. Baker &
V. Hookoomsing, Dictionary of Mauritian Creole].
E.g. Take your time – don’t make latās here.
Lathās in our days referred to the task for the day in the sugar-cane fields, like
planting, ploughing or weeding. Today the word means something different –
Sezela resident, June 2010.
4. Cultural aspects of lexical retention and variation
In this section I discuss a selection of key terms that is of interest from a cultural
point of view, and that of linguistic and regional origins. Some of these terms re-
late to notions of scaling and rescaling emphasised in current studies of linguistic
globalisation. Blommaert (2010, 34–36) refers to scale as a notion that provides
an understanding of the dynamics between local and translocal forces. In this
he emphasises participants’ experiences, perceptions and practices. Slembrouck
(2011, 159) cites Baynham (2009) in showing how the use of scale and scaling can
be ‘extended to address a theoretical concern with the effect of de/reterritorializa-
tion and dis/re-location on the construal of place, time and situational reference
178 Rajend Mesthrie
as they are interpretatively and interactionally ‘pinned down’ or exported to a new

context of expression’.
The terms to be discussed below also call into question the pertinence of the
conventional term borrowing: these are really retentions from Indian languages
frequently under conditions of language shift. The discussion is of a preliminary
nature and hopefully will spawn a more systematic larger-scale international
investigation. The focus will be on the most obvious semantic fields relating to
Indian culture in diaspora, viz. cuisine, kinship, dress, music and religion and
applies to second and third diaspora communities. Emphasis falls on everyday
usage, rather than the more formal terms found in dictionaries of Indian languag-
es, restaurant menus etc., though the latter would no doubt make an interesting
comparative study in its own right. Finally, the discussion should be based on the
terms that have survived in the English of the generations born in diaspora, rath-
er than the immigrating generation or the terms current in India (though again,
these could be of comparative interest too).
4.1 Cuisine
The basics of Indian cooking vary enormously according to region in India, but
may be said to include at least dhāl ‘split lentil soup’, bhāt ‘rice’, shāk ‘vegetables’
and roti ‘unleavened bread’.4 We may add a plethora of details regarding type of
lentils, rice, vegetables, rotis, meat dishes, and accompaniments like chutneys
(relishes), pāpadums (fried rice or lentil wafers), achārs (pickles), miṭhai (Indian
sweets), drinks etc. The possibilities for a comparative international project are
mouth-watering. But what is the linguistic interest of such work? A socio-lexical
analysis would have to pay attention to structural, semantic, etymological and
social issues like the following:
a. Existence of Indian vs. (local) English term

b. Changes in pronunciation
c. Changes in semantics (sense, reference, hyponomy)
d. Source language(s) and global networks
e. Hybridisation of terms
f. Variation and selection of variants
4. In a strictly linguistic transliteration of Indian languages like Hindi or Gujarati the spelling
dāl is normal; in a more practical dictionary orthography, I use <dh> generally for dental stops
of SAIE, and <d> for alveolar/post-alveolar/retroflex stops. The <h> is thus not intended to be
taken as an aspirate, rather <dh> is a voiced dialect counterpart of <th> spellings of English.
Attention to these aspects will illuminate the nature of social contacts within
specific diaspora settings: which subgroups were more influential than others in
contributing to the new ethnolect of English; to what extent are the terms and
cultures shared or distinctive; and are there ideologies that promote linguistic
retention of terms as against hybridity or calquing (translating the Indian term
into a dominant language like English)? In regard to (a) above, the persistence of
an Indian term as opposed to replacement by an English equivalent or near- or
even not-so-near equivalent, I focus on one linguistic item in detail and make
reference to related items of comparative interest. The term dhāl [d̪ɑ:l] is one that
survives robustly in diaspora, and the closest English equivalent is a phrase (‘split
lentil soup’) that hardly forms a viable colloquial alternative (since dhāl isn’t gen-
erally drunk as a soup, but is an accompaniment to a rice dish). It therefore forms
a useful diagnostic for lexical retention in diaspora. It is also a good candidate
for spread beyond the Indian community, as is the case in South Africa. In most
territories familiar to me the pronunciation has not changed, with a dental [d̪] at
the beginning. However, it wouldn’t be surprising to learn of the dental stop being
alveolarised under pressure from mainstream English. In SAIE many speakers
have adopted the pronunciation of colonial English speakers: [dɔ:l], with an al-
veolar [d] and a back rounded vowel instead of [ɑ:].5 (The alveolar pronunciation
also occurs among non-Asian British people). The rounded vowel quality dates
to 19th century British India, judging from a pamphlet from the Acting Protector
of Immigrants in 1874 entitled Notice to Coolies Intending to Emigrate to Natal.
The rations promised to emigrants are ‘dholl, …salt fish… ghee or oil…’ Under
category (c) one could ask whether there are any changes in semantics and/or hy-
ponymic relations. Prototypically dhāl is both a generic term (for a range of pulses
or lentils including mung, masūr etc.) and a specific term (in the unmarked case it
stands for either a yellow, split-lentil preparation whose closest English equivalent
is soup or the split lentil as an ingredient). It is also a contrasting term to curry
(dhāl typically uses little or no curry spice and is used to complement a spicy
curry and rice dish). In SAIE there is variability here. In colloquial speech most
speakers use the word as a superordinate term (hypernym), and hence may speak
of mūng dhāl and masūr dhāl. They also use it colloquially as the specific proto-
typical preparation mentioned above (referred to simply as dhāl). But some usage
suggests that the dhāl versus curry opposition doesn’t always hold: in particular
the colloquial witticism to get dōl curry and rice ‘to experience lean times, to un-
dergo a humdrum time’. Regarding networks, there is little to be said apart from
the direct influence from India independently upon different overseas territories.
5. Occasionally SAIE speakers use a retroflex in place of alveolar [d] here.

180 Rajend Mesthrie
An exception might be the hybrid term dōl-curry in SAIE, which isn’t accepted in
India (or by all SAIE speakers).6 And finally under (f) ‘selection of variants’ one
could ask whether the term dhāl or dōl prevails in different places.7
Rather than going through the same procedure for a variety of food terms, I
mention a few salient items that raise interesting analytic issues or highlight ques-
tions of diversity and layers of diasporic usage. The term roti is a generic term for
‘round, flat, Indian bread’. In South Africa it has become a specific term for one
particular kind of roti, corresponding to what is known elsewhere, as chapāti. The
latter term was known in South Africa, but is no longer part of colloquial usage.
Likewise there are two kinds of naans in South Africa. One is a soft, fluffy, round
leavened bread, the other the more usual thick unleavened Indian bread baked in
a clay oven. The latter type was introduced from India in the more recent age of
global Indian restaurants. To differentiate between the two the hybrid compound
naan-bread is increasingly used in shops for the former, since it has similarities
with western bread. The clay oven (n. tandoor, adj. tandoori) is also a late entry
in South Africa, under global influence, possibly via the U.K. Likewise the term
tikka (especially in chicken tikka ‘pieces of chicken’) is a late entry in South Africa,
coming not from India or South Asia generally, but via the U.K. in the 1980s. The
term refers in Hindi and related languages to ‘pieces of meat’. In South Africa a
related Bhojpuri-Hindi term with the same meaning, boṭhi has undergone se-
mantic shift to refer specifically to ‘tripe’. This might be a euphemism, as tripe was
associated with meat for the poor.
Another food term relevant to international comparative work is bhāji. In
India and territories like South Africa it refers to leafy vegetables which are lightly
cooked in oil. More common in SAIE is the term herbs which refers specifically
to ‘herbs for the pot’, rather than dried leaves used as flavouring or seasoning.
Terms for different type of edible herbs are sugar cane herbs (growing as weeds be-
tween rows of cane), red herbs (having red or velvet leaves), green herbs, mustard
herbs, drumstick herbs (growing on a tree, the Moringa pterygosperma), sour herbs
and so forth. In India bhāji may be used for snack food, e.g. pāū bhāji famous in
Mumbai. In the U.K. the term appears to refer to a spicy fried snack of flour and
6. The term curry can for synchronic analyses be taken as mainstream English, though its
ultimate origins lie in India.
7. As an aside, under the topic of lentils, a useful global dialectological diagnostic item would
be mūng versus mag [mʌg]; the former perhaps more widespread and based on Hindi, though
mag is a competing form in SAIE, from Gujarati. The linguistic role of Gujaratis in the Indian
English diaspora is worth monitoring, as the group that was frequently involved in importing
goods from India into different territories.
vegetables: hence the title of the 1993 U.K. movie Bhaji on the Beach.8 The equiv-
alent term among Indians in South Africa is bhajia, and in the wider society, chilli
bites. Particularly common in Gujarati households are special subtypes known
by hybrid terms like onion bhajia, potato bhajia, chilli bhajia etc. The last item
regarding food that illustrates differential regional effects in diaspora is dhālpūri.
This is a regional specialty in India which is widely known in the second diaspora
territories, via Bhojpuri and Awadhi. The Trinidadian author, Sam Selvon (1979)
writes of the difficulties he had as a Trinidadian in London of the 1960s in finding
this specialty in the Asian restaurants there. The term is well-known in South
Africa, Mauritius, Guyana and Fiji. And thanks to the subsequent influence of the
double diaspora (possibly from Mauritius) the dish is now a part of London’s fare
(Philip Baker p.c. c1992). There is some variation in the composition of the actual
product: the Mauritian dhālpūri is a sweet (deep fried) pūri with dhāl stuffing; the
South African variety is not sweet and more like a large parātha (a thicker type
of roti) made on a griddle lightly stuffed with dhāl and dhanya (coriander) and
jeera/jīra (cumin). But both senses are now known, thanks to recent Mauritian
migrants selling food on the streets of Durban.
4.2 Kinship
This is a semantic field which is quite durable in diaspora, though more so in do-
mestic rather than public contexts. Of particular interest are those terms that have
managed to spread across Indian groups. This does not generally happen. Thus
in South Africa terms for ‘paternal uncle’ (father’s brother) still vary according
to language group: kāka (Bhojpuri, Gujarati), chācha (Urdu), chicha (Konkani)
and perinaina (Tamil).9 But some terms show signs of gradual convergence and
rescaling. The Gujarati practice of suffixing the respectful terms bhai ‘brother’
and bēn ‘sister’ to a proper name (e.g. Rajend-bhai) is spreading to the English of
speakers of Bhojpuri and Tamil.10 This upscaling (in terms of increase in users
and the added prestige that goes with the term) contrasts with the downscaling
(in terms of semantic derogation) of two similar terms from Tamil. The noun
thambi is a neutral term for ‘younger brother’ and remains so in South African
8. The more usual term appears to be pakora in the U.K., and this use of bhaji/bajji may be
South Indian in origin.
9. By contrast the term māma ‘maternal uncle’ (mother’s brother) occurs in this from across a
range of Indic and Dravidian languages.
10. Although the practice of using bhai in this way occurs in other languages like Hindi, in
South Africa it was specifically the usage by Gujaratis that formed the model for wider usage.
182 Rajend Mesthrie
Tamil. But in passing into colloquial English, especially of other Indian speakers,
the term has acquired class connotations to denote a working-class, young male
of indentured, especially Tamil, background. A similar change is evident for the
term naina from Telugu, where it is a neutral or even respectful term for ‘father’ or
(paradoxically) affectionate term for ‘son’. In SAIE informal usage it has become
a nickname, especially in the playfully hybrid term Nine-boy, which would be the
affectionate term for a young boy with the nickname Nain(a). Another interest-
ing kinship usage is based on the Gujarati practice in South Africa of referring
to elder neighbours or acquaintances as follows. The male is referred to by name
plus the respectful title kāka ‘paternal uncle’. However, his wife is referred to by
her name plus the title māsi ‘maternal aunt’. This is a subtle way of indicating that
these respected acquaintances are not true kin, since a paternal uncle and his wife
would be referred to as kāka and kāki respectively and a maternal aunt and her
husband as māsi and māsa respectively. Finally, under kinship the extent to which
the Indian English terms cousin-brother and cousin-sister are retained in diaspora
are of interest. These terms are still robust in South Africa, even though they are
avoided in formal, public, and educated speech. There is still uncertainty as to the
exact referents. Informally asking people to state concretely who qualify as their
cousin-brothers and -sisters and who do not has not yielded conclusive results.
Some people use the terms for any male or female cousin. This usage would co-
incide with the explanation given by Nihalani et al. (2004, 58) for Indian English
“most languages in India indicate sex in the word itself, and ‘cousin brother’ is an
attempt to do this in English”. An alternate explanation is that cousin-brother is an
intermediate level between the Indian language system (where male cousins are
simply labelled bhai – i.e. ‘brother’) and contemporary Standard English (where
cousins are never labelled ‘brothers’). In this instance the gender specification is
not primary, but reflects the standard English distinction between brother and
sister. The Indian English usage then emphasises the phenomenon of ‘cousin as
brother’ or ‘cousin as sister’. Some SAIE speakers appear to draw a distinction be-
tween first-cousins (who are cousin-brothers or -sisters) and more-distant cousins
(who are not). This would indicate that kinship distance and gender play an equal
role. For these few speakers there is thus the possibility of a three-way distinction
between brother, cousin-brother and cousin. This kinship conundrum clearly de-
serves further scrutiny via an imaginative fieldwork procedure.
4.3 Dress, music, religion
For reasons of space I will again explore one term from each of these semantic
fields which continue the theme of variability of terms in diaspora, and the need
for undertaking comparative work. Regarding dress a good diagnostic might be

salwār kameez, the term for loose cotton trousers (salwār) and long shirt typically
with slits on either side (kamīz) worn over it.11 According to Hankin (1994, 183)
this item of clothing originated in the Panjab among male or female artisans and
similar classes in pre-independence India. There are phonetic variants like shal-
wār kameez and sholwār kameez based on different regions and languages in In-
dia. Whether diasporic communities have retained the term or evolved English
equivalents would be interesting to monitor. In South Africa the term is widely
used (with all the above variants used in advertisements, but the main term being
salwār kameez). However, in colloquial speech as well as written advertisements
the main term is the neologism, Panjābi. Visitors from India and departing con-
sul-generals have been known to remark on the incongruity of the term in their
farewell speeches reflecting on South African Indian customs. In India Panjābi on
its own would refer unambiguously to a person from Panjāb, while in adjectival
function before a noun it would mean ‘pertaining to the Panjab’. Hence the dress
in question would at a pinch be expressed in Indian English as Panjābi dress or a
Panjābi set.12
Regarding music in diaspora I choose the term chutney (literally ‘spicy relish
or sauce’) for an upbeat, short folk-song in Bhojpuri or related dialect. The term
gained international prominence via Trinidad, from where this form of music
was popularised in the 1970s, based on Indo-Caribbean adaptations of tradition-
al Bhojpuri and related language folk and religious songs from India. Diasporic
chutney songs may include English lyrics as part of that adaptation. In South Af-
rica the hybrid terms chutney-song and chutney-number exist and since the pop-
ularity of chutney music has passed into other languages, the term Tamil-chutney
also exists. A parallel can be found in bhangra music and dance which gained
prominence in the diaspora and in fact more broadly internationally 1990s via its
UK fusion of traditional Panjabi bhangra music and dance with Western and Reg-
gae influences. The hybrid term bhangra bash for a young peoples’ party remains
in vogue in South Africa and no doubt elsewhere.
Under religion it would be instructive to ascertain to what extent terms from
the different religions have been retained or have had to adapt to an English envi-
ronment (e.g. temple versus mandir (Hindi, Gujarati), alayam (Tamil), kovil (Ta-
mil), gurudwār (Panjabi) etc., or mosque versus masjid (Urdu, Arabic). In South
11. The term kamījz is related ultimately to the European term that gives rise to French chemise,
as a borrowing into Hindi-Urdu via Perso-Arabic from Latin. The etymologically related term
simīj also occurs in languages like Bhojpuri, as a borrowing from Portuguese.
12. However, since at least one visiting academic ventured to me that the South African usage
of Panjabi has a parallel in Bengal, the matter is certainly worth investigating further.
184 Rajend Mesthrie
Africa the term gurudwār was unknown until the recent (post 2000s) arrival of
Sikh immigrants and computer-experts on short term contracts. The Cape Town
community shows a refreshing pragmatism and tolerance in allowing its new
gurudwār to be housed within the same building complex as its older Gujarati
mandir. My illustration of a religious term comes from the term puttu/pittu used
in the South African Tamil community.
PUTTU / PITTU n.
A mound of earth as in an anthill or snake’s dwelling. Treated as a shrine, the
cobra being sacred to followers of Shiva. Hence puttu-house. [Prob. Kannada
puṭṭu, Telugu puṭṭa ‘anthill, snake’s hole, heap’, Tamil purru, South African
Tamil puṭṭu].
According to legend, a Puthu, a mound similar to an anthill in appearance
and which is regarded as home to the sacred Snake Goddess, materialized
on the present temple site [in Mt. Edgecombe]. For believers this was a sure
sign that the temple was located on an auspicious spot. The Puthu has been
growing each year, and stands at over two metres – A. Desai & G. Vahed Inside
Indenture, 234.
5. Conclusion
When people speak of linguistic globalisation it is often in terms of the West,

especially the Anglophone West, flooding the planet with languages like English,
Spanish and French. But important counter-flows exist: from Asia, e.g. the Indian
and Chinese diasporas, as well as the African diaspora in the era of slavery and
current post-colonial migrations. This paper has suggested that comparative lex-
ical analysis has much to contribute to the study of the Indian diaspora. It gives a
sense of social history from below, detailing the cultural retentions from different
parts of India and the adaptation of some practices arising from culture contact in
new environments, as well as the give-and-take between fellow Indians in diaspo-
ra with whom one might not have had much contact in the vast regions of India.
This is another kind of geographical rescaling. There is also a layering of history
according to early and later migratory flows to the same territories. The study of
lexical retentions, semantic shifts, adaptations, borrowings and neologisms is of
considerable importance in this approach. Such a study will not be without its
problems. The focus I have presented is on retention of Indian words when speak-
ing English, especially among the second generation. This of course would still be
subject to immense variability, depending on factors like: (a) difficulties of sepa-
rating languages in bilingual contexts, (b) influences from new immigrants and
of return trips to the Indian homeland; (c) separating knowledge of Indian terms
from actual colloquial usage; and (d) obtaining reliable information about actual
usage from interviewees. But these difficulties are not insurmountable as shown
by the Dictionary of South African Indian English. Admittedly colonial and apart-
heid practice in South Africa, which prevented a continuous flow of Indians into
the country after 1911 and an ensuing language shift, made it easier to demarcate
which lexis was ‘in’ and which not. Despite the potential difficulties, a study of the
vocabulary of Indian English in diaspora is crucial to diaspora cultural studies
and studies of English in a global era.
References
Baker, Philip & Hookoomsing, Vinesh Y. 1987. Dictionary of Mauritian Creole. Paris:
L’Harmattan.
Baynham, Mike. 2009. Just one day like today: Scale and the analysis of space/time orientation
in narratives of displacement. In Globalization and Language in Contact: Scale, Migration
and Communicative Practices, James Collins, Stef Slembrouck & Mike Baynham (eds),
130–147. London: Continuum.
Blommaert, Jan. 2010. The Sociolinguistics of Globalisation. Cambridge: CUP.
DOI: 10.1017/CBO9780511845307
Damsteegt, Theo. 1988. Sarnami: A living language. In Language Transplanted: the Develop-
ment of Overseas Hindi, Richard K. Barz & Jeff Siegel (eds), 95–120. Wiesbaden: Otto
Harrassowitz.
Desai, Ashwin & Vahed, Goolam. 2007. Inside Indenture. Durban: Madiba Publishers.
Domingue, Nicole. 1971. Bhojpuri and Creole in Mauritius: A Study of Linguistic Interference
and its Consequences in Regard to Synchronic Variation and Language Change. PhD dis-
sertation, University of Texas.
Gambhir, Surendra. 1981. The East Indian Speech Community in Guyana: A Sociolinguis-
tic Study with Special Reference to Koine Formation. PhD dissertation, University of
Pennsylvania.
Gandhi, Mohandas K. 1927. An Autobiography or The Story of my Experiments with Truth.
Ahmedabad: Navajivan.
Govender, Neelan. 2008. Girrmit Tales. Durban: Rebel Rabble.
Hankin, Nigel B. 1994. Hanklyn-Janklin, 2nd edn. New Delhi: Banyan Books.
Hassim, Aziz. 2002. The Lotus People. Durban: Madiba Publishers.
Lal, Brij V. 2004. Girmitiyas – The Origins of the Fiji Indians. Fiji: Fiji Institute of Applied Studies.
Mesthrie, Rajend. 1991. Language in Indenture: A Sociolinguistic History of Bhojpuri-Hindi in
Mesthrie, Rajend. 1992. English in Language Shift: The History, Structure and Sociolinguistics of
South African Indian English. Cambridge: CUP.
Mesthrie, Rajend. 2008. South Asian languages in the second diaspora. In Language in South
Asia, Braj B. Kachru, Yamuna Kachru & S.N. Sridhar (eds), 497–514. Cambridge: CUP.
DOI: 10.1017/CBO9780511619069.028
186 Rajend Mesthrie
Mesthrie, Rajend. 2010. A Dictionary of South African Indian English. Cape Town: UCT Press.
Mohan, Peggy Ramesar. 1978. Trinidad Bhojpuri: A Morphological Study. PhD dissertation,
University of Michigan.
Nihalani, Paroo, Tongue, Ray K., Hosali, Priya & Crowther, Jonathan. 2004. Indian and British
English – A Handbook of Usage and Pronunciation, 2nd edn. New Delhi: OUP.
Pillay, Kriben. 1995. Looking for Muruga. Durban: Asoka Theatre Publications.
Poovalingam, Pat. 2003. Anand. Durban: Madiba Publishers.
Selvon, Samuel. 1979. Three into One Can’t Go – East Indian, Trinidadian, West Indian. Open-
ing Address in Proceedings of the Second Conference on East Indians in the Caribbean. St.
Augustine: University of the West Indies.
Siegel, Jeff. 1987. Language Contact in a Plantation Environment – A Sociolinguistic History of
Fiji. Cambridge: CUP.
Slembrouck, Stef. 2011. Globalization theory and migration. In Sage Handbook of Sociolinguis-
tics, Ruth Wodak, Barbara Johnstone & Paul Kerswill (eds), 153–164. London: Sage.
DOI: 10.4135/9781446200957.n12
Sridhar, Kamal K. 2008. South Asian diaspora in Europe and the United States. In Language in
South Asia, Braj B. Kachru, Yamuna Kachru & S.N. Sridhar (eds), 515–533. Cambridge:
CUP. DOI: 10.1017/CBO9780511619069.029
Yule, H. & Burnell, A.C. 1886. [2nd edn. 1985.] Hobson-Jobson: A Glossary of Colloquial An-
glo-Indian Words and Phrases. London: Routledge & Kegan Paul.
Chapter 9
Indo-Fijian English
Linguistic diaspora or endonormative stabilization?*
Lena Zipp
This paper first traces the history and sociolinguistic situation of Fiji’s well-es-
tablished and influential Indian diaspora. It then weighs support for the
development of endocentric norms in the national second language variety
of English, Fiji English, against diasporic influence by combining linguistic
evidence from previous studies with data from an attitude survey conducted
among Fiji students in the country’s capital. The status of Fiji English as a na-
tional variety is investigated by evaluating statements of language attitudes and
reported usage based on a keyword analysis of responses to direct, open-ended
survey questions. The results support the claim of progressing endonormative
stabilization whereas indications for diasporic links to India in the use of Eng-
lish remain weak.
Keywords: Fiji English, attitudes, language use, variety status
1. Introduction
While Indian English spoken on the subcontinent itself has been well-researched
for many years, the Indian diaspora is a more recent area of study in which a
multitude of varied settings invites comparative views of language contact and the
linguistic effects of transnational network maintenance. In any national context,
however, a linguistic study of the Indian diaspora will have to determine wheth-
er and to what extent there are diasporic influences on the respective variety of
* Funding for the fieldwork was provided by the University of Zurich, project F-62420-01-01.
I would like to thank Marianne Hundt and an anonymous reviewer for their helpful comments
on an earlier version of this paper. I am very much indebted to University of the South Pacific
academic staff Ana Kitolelei, Asinate Koroi, Bhavna Vithal and Premila Devi for help in dis-
tributing the questionnaires, and to University of Zurich research assistant Adina Staicov for
digitization of the responses.
188 Lena Zipp
English spoken in the country, or whether these influences are evened out by the
country’s development towards endonormative stabilization of a national second
language variety of English (ESL). In order to address these issues with regard to
the Indian part of the population of Fiji, this paper presents an informative review
of linguistic findings in connection with Indo-Fijian English and links them to
a current snapshot of attitudes towards the emerging local national variety, Fiji
English.
One of the main defining characteristics of diasporic societies is the degree
of their transnational network maintenance. This is especially true of the Indi-
an diaspora, which is dispersed across over 70 countries. Sociologists have taken
to making a distinction “between the older diaspora who migrated primarily in
the nineteenth century (for instance to Fiji, Trinidad and Tobago, Guyana and
Mauritius) and the new diaspora who migrated primarily in the twentieth cen-
tury (for instance to UK, United States of America […] and Canada)” (Barn
2008, 192). According to this distinction, the older diasporas are clearly linked to
the legacy of British colonialism and the history of indentured labour, as in the
case of Fiji. Subsequent decolonization marked a major change for each of these
older diasporas, as influences of the local postcolonial governments grew strong-
er and questions of citizenship and belonging had to be redefined (Raghuram &
Sahoo 2008, 8–9). As a consequence, the links that early migrants have with the
Indian subcontinent are described as weak:
Many of these people […] have tenuous links with India. When faced with racial-
ised targeting they have migrated again, but often not to India […].
(Raghuram & Sahoo 2008, 9)
The Fiji Indian community presents a prominent, if not prototypical, example of

this kind of postcolonial diaspora. It has endured repeated political pressure (see
Section 2.1) and has seen continuous emigration to countries like New Zealand,
Australia, the USA, and Canada, rather than “back” to the mother country (see
Hundt, this volume). When assessing the linguistic consequences of this volatile
diaspora situation, it is therefore vital to weigh the current status of local variety
formation against linguistic links to the place of origin, India. For both of these
processes – the emergence of new varieties of English on the one hand and di-
aspora (re-)formation on the other – identity construction is one of the main
agents, as the following two authors have claimed:
[…] I propose that to a considerable extent the emergence of PCEs [Postcolonial
Englishes] is an identity-driven process of linguistic convergence […].
(Schneider 2007, 30)
Chapter 9.â•‡ Indo-Fijian English 189
The assumption of a shared identity that unites people living dispersed in trans-
national space thereby becomes the central defining feature of diasporas.
(Sökefeld 2006, 280)
Schneider (2007), in particular, presents a powerful dynamic model that defines

the process of variety evolution in postcolonial settings as a “sequence of charac-
teristic stages of identity rewritings and associated linguistic changes” (Schneider
2007, 29). He suggests investigating a country’s socio-political background and
sociolinguistic conditions, and proposes that identity constructions and linguis-
tic effects indicate progression through a succession of stages that mark variety
development. The model allows for more than the typical two agent groups in
colonization settings, introducing the label of “adstrates” (ADS) for immigrants
and contract labourers such as the Indian population of Fiji.1 For these groups,
Schneider (2007, 60) describes the following course of events:
Depending on group sizes and their cultural distinctness and cohesiveness, such
ADS strand groups may maintain a considerable degree of ethnic and linguistic
distinctiveness for a long while, sometimes for many generations. In the long run,
however, […] they are bound to take their share in the process of adopting and
nativizing their target language, local English.
As a consequence, progressing linguistic homogenization in the course of nation-

al variety formation might entail the loss of cultural and linguistic distinctiveness
for diasporic communities.
This paper will attempt to locate the Fiji Indian diaspora within the on-go-
ing “identity-driven process of linguistic convergence” (Schneider 2007, 30). On
the one hand, it will examine whether there are systematic linguistic differences
between the sub-varieties of English used by the Indo-Fijian and the Fijian part
of the population and whether these differences are owed to linguistic influenc-
es from the mother country on the local Indian diaspora. On the other hand,
the paper will shed light on characteristic identity constructions, which “mani-
fest themselves in […] sociolinguistic determinants of the contact setting (condi-
tions of language contact, language use, and language attitudes) […]” (Schneider
2007, 31). In particular, attitudes towards the emerging variety of English will be
investigated in order to locate Fiji English in phase four of Schneider’s 2007 mod-
el – two stages further along the cycle than previously acknowledged (Schneider
2007, 144–118). For it is in this stage that “[…] the acceptance of a new, indigenous
identity result in the gradual adoption and acceptance of local forms of English as
1. The main two participant groups of the variety formation process are the colonizers or
settlers (STL), and the colonized, or indigenous population (IDG, see Schneider 2007, 31–33).
190 Lena Zipp
a means of expression of that new identity, a new, locally rooted self-confidence”

(Schneider 2007, 49). Two indicators of this linguistic self-confidence are firstly,
that the new local norm is also “accepted as adequate in formal usage” (Schneider
2007, 50), and secondly, that references to “English in Fiji” are being replaced by
the label “Fiji English”, which signals a different conceptualization of the status of
the variety (see Schneider 2007, 50).
In sum, the paper attempts to prove that Fiji English has moved farther along
the developmental cycle of post-colonial Englishes by:
a. tracing the history and sociolinguistic setting of Indians in Fiji (Sections 2.1
and 2.2),
b. summarizing the status of research on the local variety of English in Fiji and
potential linguistic links that the Indian diaspora has to the mother country
(Section 2.3),
c. probing the attitudinal capital of the label Fiji English among a group of young
English speakers of two ethnicities in Fiji tertiary education, and
d. assessing the reported usage of Fiji English by these speakers across registers
(Section 4).
In the next part of this paper, I will give a short overview of the linguistic and so-
cio-cultural situation in Fiji and present a summary of previous linguistic findings
linked to Indo-Fijian English. In part three, I will introduce the data sample and
summarize methodological support for the use of open questionnaire items in
direct attitude studies. Part four of the paper will present and discuss the results
of the attitude survey.
2. Indians in Fiji – socio-history and sociolinguistic situation
2.1 Socio-history
Most of Fiji’s Indians came to the islands in the South Pacific between 1879 and
1916 as indentured labourers on five- to ten-year contracts. This indenture system
for Fiji’s new sugarcane plantations was initiated by Fiji’s British governor, Sir
Arthur Hamilton Gordon, following the examples set by Mauritius and Trinidad.
Workers were primarily recruited from northern India, with 80 per cent stem-
ming from Uttar Pradesh, another 13 per cent from Bihar and Bengal, and the rest
from southern India (see Gillion 1962; Srebrnik 2008, 77). The indenture system
or girmit2 has since been criticized for its dismal living and working conditions,
high mortality rates, “excessive discipline and repressive legislation; and […] gen-
eral unwillingness on the part of the government to guard the rights of the inden-
tured workers” (Lal 2006, 372). After the end of their indenture contract, workers
could choose between their right to free passage to India or staying in Fiji and,
by 1920, about 61,000 Indians had permanently settled in Fiji. As neither civil
and political rights nor educational or mission assistance were granted to them at
first, Fiji’s Indians took to political and educational self-organization in the early
decades of the diaspora formation. Due to demographic factors, the Indian pop-
ulation grew from approximately 38 per cent after the end of indenture in 1920
to 51 per cent in 1966 (Srebrnik 2008, 93), outnumbering ethnic Fijians for many
decades until the late 1980s (Lal 2006, 375, 379). In spite of the fact that most
Indo-Fijians started out as workers in the agricultural sector, inalienable indige-
nous Fijian landownership rights restrict them to leased farmland to the present
day and have ultimately lead to urbanisation, business entrepreneurship, and a
new Indo-Fijian middle class with considerable economic power. Both the rising
numbers of Indo-Fijians and their increasing economic and political influence
can be seen as triggers for growing ethnic Fijian nationalist feelings in the second
half of the 20th century, which culminated in political coups d’état in 1987, 2000,
2006, and a prolonged “constitutional crisis” in 2009.
Such recent developments have had their effect on the Indo-Fijians: Whereas
at independence in 1970, 98 per cent of Indo-Fijians opted to become Fiji citizens
in a movement for “a new emphasis on being fully Fijian” (Kelly 1995, 65, quoted
from Srebrnik 2008, 79), the years 1987 to the present have seen continuous emi-
gration of Fiji Indians to New Zealand, Australia, the United States, and Canada.
Today, of Fiji’s population of approximately 827,900, 57 per cent are ethnic Fijian,
and 38 per cent are of Indian descent (2007 census).
The majority of Fiji’s Indians are descendants of the girmitiyas, or indentured
labourers, and form part of the older Indian diaspora who were transported to Fiji
in the late 19th to early 20th century. However, a smaller group of Indo-Â�Fijians
consists of more recent immigrants, mostly Gujaratis and Sikhs from the Pun-
jab, who have become extremely influential in the trade and transport sectors
of Fiji’s economy. Srebrnik (2008, 79) points out that “[t]hese two groups, who
arrived later and retained closer cultural and personal ties with their communities
of origin in India, some even owning property in their ancestral villages, formed
separate segments within the Indo-Fijian community”. For the older diaspora,
2. Girmit is coined after the word agreement, and the term girmitiyas denotes the contract
labourers under the indenture system, which are held in high regards by the Indo-Fijian
community.
192 Lena Zipp
however, most of the modes that have been recognized as means of self-identifi-
cation of overseas Indians (Barn 2008, 192) have lost their importance: Firstly, the
Indo-Fijian community is casteless; secondly, religion has not been instrumental
as a structuring factor in Fiji’s society;3 and lastly, regions and languages of origin
have lost their importance for Indo-Fijian self-identification since the koinéisa-
tion of the national lingua franca, Fiji Hindi, on the early sugarcane plantations.
The next section will consider the sociolinguistic situation in Fiji and its influence
on the construction of Indo-Fijian identity.
2.2 Sociolinguistic situation, language use and attitudes
Fiji’s 1997 constitution recognizes English, Fijian, and Hindi as official languages
with equal status (Fiji Constitution (Amendment) Act 1997), but English is spoken
as a first language only by an estimated minority of 1–3 per cent of the popula-
tion (see e.g. Mangubhai & Mugler 2003, 371). The two main native languages
of about 95 per cent of Fiji Islanders are either Fijian, an Austronesian language,
or Fiji Hindi, an immigrant koiné that developed mainly from the north-eastern
Indian dialects Awadhi and Bhojpuri spoken by the indentured labourers (see
Siegel 1987, 136–147).4 Fiji Hindi also includes borrowings from Fijian and Eng-
lish, and written representation is usually in romanized form. However, Fiji Hindi
remains an uncodified language in that is almost exclusively used in spoken reg-
isters, acting as the low variety in an in-diglossia in which Standard Hindi is the
high variety used in writing, reading, and broadcasting, as well as for formal or
religious registers.5 Standard Hindi is also the official medium of instruction for
the first three years in Fiji’s Indian schools, as the Fiji education system follows
3. The majority of Indo-Fijians are Hindu, about a fifth are Muslim, and the rest are Sikhs or
Christians. While local religious practices are an important part of identity construction e.g. for
the Hindu majority, religious organisations are not involved in official functions other than in
the educational sector.
4. Other contributors to the koiné are “Hindustani, the lingua franca of North India – particu-
larly in its simplified form, Bazaar Hindustani – and Khariboli, […] Bihari sub-dialects such
as Magahi and Maithili, and a few other Western Hindi dialects such as Braj” (Mangubhai &
Mugler 2003, 377), besides Dravidian languages from Southern India.
5. Fiji’s linguistic situation has also been described as ‘out-diglossia’, in which English is in-
creasingly taking over the functions of the high variety (Siegel 1987; 2003). Moreover, ongoing
research indicates that Fiji Hindi might be in the process of standardizing, thus developing
towards a written variety (Prasad, in preparation).
an immersion bilingual education scheme.6 This education system has been crit-
icised for putting a considerable cognitive burden on Indo-Fijian students from
different heritage backgrounds:
The effect of using Hindi as the medium of instruction is that most, if not all,
Indian children who arrive in Grade 1 are taught in a language that is either a
second language […] or a second dialect for those who grow up with Fiji Hindi as
their first language. All Indian children also have to learn an Indian script, and all
the minority Indian languages have their own scripts (Tamil, Telugu, Malayalam,
Gujarati, Panjabi and Urdu), which are different from the Devanagari of Stand-
ard Hindi. […] The learning of the Roman alphabet, for English, makes it three
scripts […]. […]. Partly for this reason, Indian students generally do less well in
vernacular than in English examinations […]. (Mangubhai & Mugler 2003, 398)
However, while it has been claimed that the bilingual education scheme has re-
sulted in limited literacy in the students’ first languages (Geraghty 2005), the ver-
naculars in Fiji have been shown in a number of studies to be in no danger of
extinction (May 1990; Mugler & Tent 1998; Tent 2001; Siegel 2003). Moreover, the
use of the vernaculars for inter-ethnic communication (also termed “vernacular
bilingualism”) has been reported to be “significant, although it is often overlooked
because of the perception that English in Fiji is the lingua franca” (Mangubhai &
Mugler 2003, 385–386; also see White 1971; Mugler & Tent 1998; Tent 2001).
From the very beginning of its introduction into the school system by coloni-
al, later New Zealand, authorities in the 1920s, English was seen as the language
of national unity, “a ‘neutral’ lingua franca [that] was needed to allow Fijians and
Indo-Fijians to live together harmoniously, now that it was plain most girmitiyas
would not be returning to India. English was seen as the appropriate, if not the
only, language to fulfil that role” (Tent 2001, 246). It is the main language used
in the typical ESL domains of education, media, administration, legislation, and
economy but also provides the majority of loan vocabulary for local vernaculars
and is extensively used in code-switching particularly in Fiji Hindi contexts (Siegel
1987, 208–210). Several studies have investigated English in Fiji with regard to
language use and attitudes (White 1971; May 1990; Mugler & Tent 1998; Tent
2001; Mugler 2002); most have reported a general trend towards English bi- or
multilingualism but also increasing use of English in Indo-Fijian speech commu-
nity interactions and “a steady shift from standard Hindi to English in the do-
mains traditionally reserved for standard Hindi” (Tent 2001, 261; also see Siegel
6. The transitional bilingual programme entails basic literacy instruction in the local languag-
es (Fijian and Standard Hindi) for years one to three, and English-medium instruction from
grade 4 onwards for all students. However, it has been noted that local vernaculars are used
throughout classrooms in Fiji, particularly in the early years of English-medium instruction.
194 Lena Zipp
1987; 2003). Across ethnicities, the importance of English for education and em-
ployment is recognized (White 1971; Mugler & Tent 1998; Tent 2001), and a num-
ber of studies document the perceived bridging function of English as language
for wider communication and intergroup interaction (White 1971; Tent 2001).
However, attitudes towards English tend to be more positive among Indo-Fijians,
particularly at “an affective/aesthetic level” (Tent 2001, 261; also see Mugler & Tent
1998, 132). Attitudes towards exonormative model accents (e.g. North American,
English English, Australian, or New Zealand accents) showed a shift away from
English English as the traditional prestige accent in Fiji when tested on a student
focus group; instead, North American and Australian accents were ranked highest
(Mugler 2002, 83). Mugler also notes that “the local variety of English is not con-
sidered standard and […] [t]he local accent is downgraded by its own speakers in
comparison with British, American, and Australian varieties even on solidarity
dimensions” (Mugler 2002, 83), thus alluding to the “cultural cringe” observed
across post-colonial societies world-wide (e.g. Bayard 2000).
2.3 Previous research on Fiji English
While recognition of the local variety of English in the population itself might be
slow, it is one of the best-studied ESL varieties in the South Pacific. It is generally
acknowledged that Fiji English constitutes a continuum ranging from basilectal
varieties at the lower end of the proficiency scale to acrolectal varieties spoken by
educated speakers and local native speakers of English. Most of the distinctive
features of Fiji English are found in the basilect, which was described as “Pure
Fiji English” in detail by Tent and Mugler (2008) and related sources. It is also in
the basilect that ethnic sub-varieties are discernible, particularly with regard to
the phonology of the variety. Nevertheless, Fijian and Indo-Fijian English share
a considerable number of characteristic features such as unaspirated voiceless
stops, fortis or devoiced realization of lenis sibilants, clear post-vocalic /l/, flapped
or trilled variants of pre-vocalic /r/, deletion or reduction of final consonant clus-
ters (Tent & Mugler 2008, 252–253), absence of {-ed} suffix and third pers. sing.
pres. /-s/, and yod-dropping (Tent & Mugler 2008, 259–262; Tent 2001). Morpho-
logical and syntactic features have not been shown to differ across ethnic groups
in the basilect, with two tentative exceptions:
[A]bsence of definite articles and invariant tag isn’t it? seem more prevalent
among native speakers of Fiji Hindi, while the use of us-two and the èh tag are
more common among native speakers of Fijian. Yet some features which can be
traced to one of the substratum languages have spread to the general population.
(Mugler & Tent 2008, 565)
For the acrolectal end of the Fiji English continuum, a lexico-grammatical study
of prepositions and prepositional complementation patterns on the basis of the
International Corpus of English, Fiji component, supports this finding with regard
to prepositional collocations and prepositional -ing colligation patterns (Zipp
2014). There are striking similarities between the two ethnic sub-varieties of Fiji
English, which “suggest that, at least to some extent, a national variety of English
appears to be emerging in Fiji” (Zipp 2014, 189). On the other hand, parallels
to prepositional verb phrase patterns in comparable registers of Indian English
were shown to be rare, and are more likely to be attributable to first language
interference from Hindi rather than language contact with Indian English. Along
the same lines, Tent and Mugler (2008, 250, 259) list a number of phonological
characteristics of the basilectal Fiji English spoken by Indo-Fijians that are not
shared by the “typical” Indian English of the sub-continent: It is non-rhotic, it
has monophthongised diphthongs, the intonation contours are very different, and
“[t]he realisation of alveolars as retroflexes is much less common in Indo-Fijian
English, though some speakers […] do exhibit this characteristic” (Tent & Mugler
2008, 250). In sum, there is conflicting evidence as to potential exonormative
models across previous studies, and little to no information on how parallel struc-
tures developed diachronically. On the other hand, endonormative tendencies in
empirical studies are but a first step towards observing emerging local norms, as
“[…] corpus-based research only allows us to observe the frequency-based habit
formation of a local linguistic norm. Attitudinal data are needed to test for the
‘salience’ of these linguistic ‘habits’” (Hundt 2013, 195).
2.4 Interim discussion
The three subsections of Section 2 above shed light on the sociolinguistic situation
of English in Fiji, its use and attitudes towards it by the different ethnic groups of
Fiji’s population, and its structural characteristics and relation to Indian English.
There are a number of factors that challenge the notion of strong diasporic ties to
India and according linguistic influences on the Indian diaspora in Fiji:
1. the historically tenuous nature of Fiji’s links with India after the indenture
system
2. the emigration pattern to English-speaking countries after political pressures
on the Fiji Indian community
3. the fact that English in Fiji was introduced by colonial education authorities
and not transported from India
4. weak structural ties between the two English varieties in India and Fiji, and
196 Lena Zipp
5. increasing evidence for structural nativization and a move towards endonor-

mative stabilization of a national variety of standard Fiji English.
However, when weighing linguistic diaspora status against variety status, a set
of criteria needs to be considered that goes beyond sociolinguistic setting and
nativization of linguistic form. In particular, factors such as functional expan-
sion and institutionalization of the local variety have been brought forward as
requirements of an endonormative standard (Mollin 2006, 45–52). While English
in Fiji has seen functional expansion through its use in the domains of education,
administration, media, creative writing etc., this paper will add to the picture by
locating Fiji English in the realm of private communication in the following anaÂ�
lysis of statements of reported usage. Institutionalization, on the other hand, is an
attitudinal criterion linked to acceptance and official recognition of the local va-
riety. Beginning codification of Fiji English is clearly evident in the publication of
the Macquarie Dictionary of English for the Fiji Islands, and in its foreword by then
Prime Minister, Honourable Laisenia Qarase: “Like many countries, from Canada
to South Africa, from Jamaica to India, Fiji now has its own variety of Standard
English” (Geraghty et al. 2006, vii). The following attitudinal analysis will inves-
tigate whether acceptance of the local variety and its label Fiji English has gained
ground in the young and educated part of the population since this major step in
the institutionalization of the local variety (Section 4).
3. Data and methodology
The original data for this study was collected during a field work trip to the Univer-
sity of the South Pacific in Suva, the capital of Fiji, in 2010. A written questionnaire
was distributed to university students both in and outside of classroom contexts.
The questionnaire bore the header “Questionnaire on the use of English. What is
your opinion?” and contained items asking for personal and linguistic background
information as well as open questions about language attitude and use, followed
by a general comment section. The total number of returns was 149, of which two
ethnic sub-samples of 67 Indo-Fijians and 63 Fijians were selected on the basis of a
questionnaire item that asked for self-identification with an ethnicity label.
In the questionnaire, a written response mode to open, direct attitude ques-
tions was employed. The advantages of direct attitude measures, if questions are
asked carefully – avoiding hypothetical, strongly slanted or multiple questions –
are that there are no skewing factors in the stimulus material (such as speaker id-
iosyncrasies, prosodic or paralinguistic features in language stimuli, see Edwards
1999). Moreover, written response procedures allow for “more anonymity and
uniformity, the avoidance of interviewer effects, and the possibility of collect-

ing data from a greater number of respondents in a shorter time” (Garrett 2004,
1255). The focus of the present paper lies on the evaluation of one questionnaire
item and the immediately adjoining comment section. The questionnaire item
read, “In which contexts do you prefer to use Fiji English?”, and was originally
designed in order to elicit reactions to the label Fiji English, as well as to tap re-
ported usage of the local ESL variety of English. Open-ended questions such as
these invite the respondent to answer in the form of a few words, sentences or a
paragraph of free writing, thus eliciting idiosyncratic responses. Respondents find
these questions relatively easy to answer (Geer 1988; Haddock & Zanna 1998,
134–135; Esses & Maio 2002, 78), but they are more time-consuming to evaluate.
However, this less restrictive mode allows for unexpected components to emerge
and might also give an indication of the salience of the three components that
attitudes are traditionally understood to consist of: the cognitive, the affective,
and the conative component (Lasagabaster 2004, 400). The emerging picture of
language attitudes is accordingly more complex:
Open-ended items […] rest on a more discursive model of attitudes, in which
attitudes are considered to be best accessed in a more contextualized form, al-
lowing the explicit weighing up of contrasting viewpoints, and the expression of
modality. (Garrett et al. 2003, 37)
It has been shown that all three attitude components can successfully be elicited by
open-ended measures (Haddock & Zanna 1998); in particular, they are well-suit-
ed to assess the cognitive and affective components separately (Esses & Maio 2002,
80–82). It has also been suggested that open-ended items are less likely to pro-
duce on-line judgements, i.e. “non-attitudes” (see Haddock & Zanna 1998, 135).
However, wording strongly influences the set of components that is elicited in
open-ended questions (Sutton et al. 2003). While open-ended items are claimed
to be more prone to elicit socially desirable responses, the open answering format
can also provide clues as to this bias in the wording of the answers (Garrett et
al. 2003, 36, 38; see e.g. examples (18) and (19) below). While the questionnaire
item “In which contexts do you prefer to use Fiji English?” is certainly uncon-
ventional in terms of traditional open attitude questions (as it contains implica-
tions such as that there is a variety of Fiji English and that the respondent can or
does use it), its content analysis proved to shed light on domains of usage as well
as attitudes towards this emerging variety in a multilingual society. After all, the
way that direct, open-ended questions lend themselves to cross-cultural studies
is another methodological advantage: “[…] open-ended measures are not sam-
ple- or culture-specific […], because participants themselves provide their own
list of relevant dimensions for evaluation” (Esses & Maio 2002, 79). Evaluation of
198 Lena Zipp
the responses usually follows a bottom-up approach known as content analysis,

“a research method for the subjective interpretation of the content of text data
through the systematic classification process of coding and identifying themes
or patterns” (Hsieh & Shannon 2005, 1278), i.e. inductive category development.
Very few previous studies of attitudes towards non-standard varieties of English
have employed direct, open-ended questions and qualitative content analysis.
Winford (1976), Romaine (1999), Mühleisen (2001), and Akande and Salami
(2010) administered open-ended questions in their studies of English-based Cre-
oles; however, responses were used impressionistically to support evidence from
quantitative parts of the attitude studies rather than analysed in a structured, qual-
itative manner. A comparatively dated study by Hofman (1975) and a more recent
one by Smit (1996), both on multilingual African settings, serve as examples of a
direct, open-ended, interview-based approach; both employ a coding procedure
and present a number of examples to support classification decisions.
In the particular case of the present study, a number of key concepts were
identified from existing theory and research about language attitudes in the data,
such as well-established domains, modes of use, or functions. New categories
were created for responses that did not fit previous theories (see Hsieh & Shannon
2005), as the method of “directed content analysis” allows to define codes both be-
fore and during data analysis. Answers to the above mentioned question and the
immediately following comment section of the questionnaire were combined as
one item per person for the bottom-up text-based coding procedure. Categories
were established on the basis of keywords and semantic fields, and each item was
assigned to at least one category (see Carley 1993). Code-based content analysis
has been criticised amongst others for reliance on researcher-driven classification
and non-exhaustive coding schemes (Jackson & Trochim 2002, 311); in this study,
great care was taken with regard to the actual wordings: all keyword lists contain
the full set of keywords that each code is based on, and potential sub-classifica-
tions within and across codes are being taken into account in the discussion of
data. This renders the analysis comparable to word-based methods. The results
are presented according to analytical category with examples, and also shown as a
comparison of rank order of frequency according to code in the following section
(see Hsieh & Shannon 2005, 1282–1283).
4. Analysis and results
The open-ended questionnaire item used in this study produced answers on a

number of relevant levels to the question of acceptance of the variety label. They
were grouped into reports of usage overall, mode of usage, domains of use, func-
tions, and a more general discussion of attitudinal data from the comment section
of the questionnaire, as discussed in the subsections below.
4.1 Reported usage
One of the main objectives of this study was to probe the attitudinal capital of the
label Fiji English, as the acceptance of the new variety label is believed to indicate
progression towards endocentric norms in variety formation. The first results are
thus concerned with purely positive or negative usage self-reports (see Table 9.1).
Table 9.1 shows that a generally positive answer to the question In which con-
texts do you use Fiji English – indicated by keywords such as all, normal, any,
everyday, everywhere – was given by 15 out of 130 respondents, i.e. 12 per cent of
the sample. While the positive responses are divided equally across the two ethnic
subsamples, the 7 per cent negative responses were given mainly by Indo-Fijian
respondents, indicated by keywords such as none, no idea or open rejection of the
variety label (see examples (1)–(3)7):
(1) I don’t think Fiji English exists
(2) I don’t think there is anything like Fiji English
(3) Fiji does not have its own English […]
It has to be noted, though, that this category of negative attitudes groups together
two different beliefs about Fiji English: on the one hand, the belief that Fiji Eng-
lish exists but is not used by the respondent; and on the other hand, the belief
that Fiji English does not exist (which was stated by three Indo-Fijians and one
Fijian). The answers of two Indo-Fijians clearly show that reported usage does not
have to entail acceptance of the variety label; the respondent who answered “Fiji
does not have its own English” claimed to use Fiji English “in informal talks, in
Table 9.1â•‡ Reported usage of Fiji English: positive versus negative

Code (keywords) Indo-Fijian Fijian Total
(N = 67) (N = 63) (N = 130)
All (all, normal, any, everyday, everywhere) 8 (12%) 7 (11%) 15 (12%)
None (none, hardly anywhere, no idea, no Fiji English) 8 (12%) 1 (2%) â•⁄ 9 (7%)
Other (American, British, no answer) 4 (3%) 3 (5%) â•⁄ 7 (5%)
7. All examples are quoted from the questionnaires as they appear; omissions are indicated by
[…].
200 Lena Zipp
chatting or texting”. Another respondent stated that he or she uses Fiji English in
“conversations, but I don’t really think Fiji English exists”. The mostly Indo-Fijian
respondents who rejected the variety label thus exhibit awareness of standardiza-
tion processes, language status, and prestige; the fact that Fijians tend to have less
negative attitudes towards label and usage could indicate greater loyalty to the
national variety.8
4.2 Modes of usage
The most frequently named mode for Fiji English in both ethnicities is that of
spoken language (see Table 9.2). Both the realm of spoken language and informal-
ity seem to be salient and closely associated with the use of Fiji English, which ties
in with previous research that sees Fiji English as “the variety spoken colloquially”
(Mangubhai & Mugler 2003, 384). Interestingly enough, almost all responses that
were coded as “written mode” go into two particular directions: the domain of
mobile or computer-mediated communication, and writing in private domains
(see examples (4) and (5)).
(4) Emails to friends, chatting on facebook, txting
(5) spoken and informal writing
As to computer-mediated communication, there are nine mentions of chatting (in

connection with modern media), seven mentions of email, and three mentions
of texting, which indicates that the global trend of linking electronic means of
communication with the English language continues in Fiji. However, there is
Table 9.2â•‡ Modes of usage: spoken versus written; register: informal

(N = 67) (N = 63) (N = 130)
Mode: Spoken (spoken, conversation, talking, speech, 32 39 71
communicating, discussions, verbal, orally)
Mode: Written (emails, letters, texting, chatting, on 10 â•⁄3 13
facebook, writing, reading, on the internet, essays)
Register: Informal (informal, casual, except formal) â•⁄7 10 17
8. Note though that this could also be an issue of labeling; while the label Fiji English is
well-established in linguistic theory, Indo-Fijians might have responded more positively to a
more ethnically aware (but made-up) label such as ‘Fiji Indian English’.
a clear disparity between the ethnic sub-samples with regard to the more infor-
mal registers of written communication: of the ten Indo-Fijian respondents who
comment on “mode”, eight state that they use English for emails, chatting, tex-
ting, and writing letters to friends, whereas none of the three Fijian respondents
mention either letters or emails. This is likely due to the sociolinguistic situation
in Fiji (see Section 2.2) and the fact that the Fiji education system introduces
vernacular literacy in Standard Hindi and the Devanagari script (or Urdu and the
Arabic script) for Indo-Fijian children, because there is no standardized spelling
system for the local koiné, Fiji Hindi. The figures above support the claim that
English replaces Hindi particularly in informal written contexts, a trend that was
observed already by May in 1990. He reported that the language of (presumably
private) letter-writing in Fiji was 48 per cent English, 16 per cent Fijian, but only
3.7 per cent Hindi, and that Indians “would probably prefer to do most writing in
English” (May 1990).
4.3 Domains of use
A number of the responses given to the questionnaire of this study include do-
mains known from previous research on multilingual societies (see Table 9.3).
The comparatively frequent mention of private domains for the use of Fiji English
is first and foremost proof of the fact that the question was correctly understood
to evaluate usage of the non-standard variety, and not just of “English in Fiji” as
opposed to indigenous languages – in which case public domains from the typ-
ical ESL categories such as education, administration, or economy would have
probably featured more prominently than they did (see Table 9.3). The private
domain of “friends” is the second most frequent code overall, and this pattern
seems to be salient across ethnicities, supporting previous research by Mugler
and Tent (1998, 118), which also recorded highest use of English in the domain
“with friends”. A trend observable in both private domains is that Indo-Fijians do
not report equally high usage of Fiji English as Fijians in this study. This is on the
Table 9.3â•‡ Domains of use

(N = 67) (N = 63) (N = 130)
Private: Family (family, home, relatives) â•⁄ 3 (5%) 18 (29%) 21 (16%)
Private: Friends (friends, peers, socialising, mates, 21 (31%) 35 (52%) 56 (43%)
people of the same age, sport gatherings)
Public (school, lecturer, university, teacher, essays, â•⁄ 6 (9%) â•⁄ 7 (11%) 13 (10%)
academics, at work, public)
202 Lena Zipp
one hand proof of the fact that Fiji Hindi is a vibrant language in no danger of
extinction. On the other hand, reluctance to use the local variety could also hint
at the more international orientation of Indo-Fijians brought about by emigration
after the military coups of the past and continuing political instability. The Indo-Â�
Fijian community is known to maintain close links to family members that have
migrated mostly to English-speaking countries like Australia or New Zealand
(see Hundt, this volume), which might result in greater awareness of different
varieties of English. However, the high reported use of English in Fijian homes is
consistent with the questionnaire biodata, in which only five Indo-Fijians entered
English as their first language as opposed to nine Fijians. Moreover, eight Fijian
mothers and six Fijian fathers reputedly had English as their first language, but no
Indo-Fijian parents, a fact that reflects societal changes in the Fijian parts of the
urban population of Fiji. Higher figures for Fijians’ English language use in the
home were noted before by Mugler and Tent (1998, 118), with 8.1 per cent versus
4.4 per cent Indo-Fijian usage. However, the authors advise to view the reported
figures in their study with caution: “[…] speaking English at home is commonly
viewed as a sign of sophistication and education. The relatively high incidence of
reported English use would be more likely then to be a reflection of this attitude
than an accurate picture of behaviour” (Mugler & Tent 1998, 118). On the other
hand, increasing multilingualism at home is indicative of phase 4, endonormative
stabilization, of Schneider’s (2007) dynamic model, and implies that “the role of
ethnicity, and ethnic boundaries themselves, will tend to be redefined and regard-
ed as increasingly less important” (Schneider 2007, 49). Tent and Mugler (2008)
observed the phenomenon of collapsing ethnic boundaries between Fijians and
part-Europeans from a linguistic point of view:
The Pure Fiji English spoken by part-Europeans and Fijians is essentially a sin-
gle variety. This is not surprising, since part-Europeans usually identify social-
ly, culturally and ethnically with the Fijian community. Since Independence,
part-Europeans have shifted away from their historical identification with co-
lonial European heritage and have moved towards reclaiming their Fijian roots.
(Tent & Mugler 2008, 237)
Multicultural heritage thus seems to be a societal phenomenon more represent-

ative of urban Fijians and part-Europeans than of Indo-Fijians;9 it could also be
9. Mixed marriages between Indo-Fijians and Fijians are rare, and ethnic boundaries between
Indo-Fijians and Fijians are largely impenetrable; this is apparent from data collected for the
2007 census, in which the following ethnic categories were used: iTaukei [Fijian], Indian, Eu-
ropean, Part-European, Chinese, All Others (Fiji Islands Bureau of Statistics, June 2012). In the
ethnic self-categorization item of the questionnaire to the present study, part-Europeans were
argued that a great number of those Indo-Fijians who could be the equivalent
urban, modern part of the population have left the country in the last decades of
brain-drain emigration ensuing from the situation of political unrest.
The list of domains discussed in this section shows that self-reported language
use of the local variety Fiji English is to be located in the private realm of family
and friends. However, on a methodological level, the response data can also be
seen as indicative of the salience of various domains in the minds of the respond-
ents – the more often a domain is named in connection with English, the stronger
the psychological association between variety and domain (also see Geer 1991).
The list of salient domains derived from data-driven analysis in this study can
thus also complement any set of domains developed for a particular multilingual
setting. According to Fishman (1972, 441), there is no invariant set of domains
that is suitable to be applied to all settings, because sociocultural patterns change
across speech communities. Fishman points to the fact that domains can be de-
fined on the level of sociopsychological analysis (with domains like informal), on
the societal-institutional level (e.g. family, friendship, academics), or in connec-
tion with locales (home, school, workplace). In all cases, as Fishman emphasizes,
domains are extrapolated from original speech data, by the “integrative intuition
of the investigator” (Fishman 1972, 451). This study shows that it is also possible
to infer domains from meta-data, namely from language use reports, by conduct-
ing qualitative content analyses based on inductive category development.
4.4 Functions of Fiji English
Two notions have traditionally been associated with English in Fiji in the course
of its history. The first is related to debates about the status of official language.
This issue was first addressed in the 1997 Constitution, which states that all three
languages – Fijian, Hindi and English – have equal status. In the present study,
five Fijian respondents state that the use of Fiji English is a particularly local affair
(see Table 9.4). Indo-Fijian respondents, on the other hand, do not mention the
local variety of English in connection with Fiji, which might again be indicative
of a reluctance to adopt a less internationally accepted variety (or a potentially
non-inclusive ethnic variety label). Moving on to the second focus of Table 9.4,
both ethnic sub-samples refer to the classic function of English in Fiji, i.e. the rea-
son why English was originally promoted by the colonial authorities: the bridging
classified together with Europeans, which might have resulted in part-European respondents
with strong Fijian identification and English as a first language to select the Fijian ethnicity
label.
204 Lena Zipp
Table 9.4â•‡ Functions of Fiji English

(N = 67) (N = 63) (N = 130)
National language (Fiji, Fijian, local) – 5 (8%) â•⁄ 5 (4%)
Wider communication (someone who doesn’t 8 (12%) 4 (6%) 12 (9%)
understand my language, all religions, other people,
other races, when we meet new people, communication
with people, different ethnicity)
function, according to which English is used for wider communication.10 A num-

ber of scenarios are mentioned in the responses (see examples (6)–(10)).
(6) while having conversation with friends or someone who doesn’t understand
my language
(7) conversation with other people of different ethnicity
(8) […] communicate with other races eg. Chinese, Islanders
(9) usually to communicate with the pacific people of other races
(10) English as a language is compulsory for all religions and race because it helps
in communicating with each other
The scenarios range from quite generic (“someone who doesn’t understand my
language”) to addressing ethnicities and religious background. Examples (8)
and (9) specifically name other nationalities, Pacific Islanders, which is clearly
influenced by the fact that the questionnaire was conducted at the University of
the South Pacific, a regional university jointly owned by the governments of 12
member countries (Cook Islands, Fiji, Kiribati, Marshall Islands, Nauru, Niue,
Solomon Islands, Tokelau, Tonga, Tuvalu, Vanuatu, and Samoa); the university’s
main campus in Suva, Fiji, is characterised by its mix of international and Pacific
Islander students and staff. It is striking that comparatively few respondents men-
tion this bridging function and that no-one addresses communication between
the two major ethnic groups in Fiji, whereas the ethnic minority of Chinese is
named in example (8). This finding questions the motivation to enforce English as
single lingua franca and supports previous studies that report the use of Fiji Hindi
and Fijian as lingua francas in situations of inter-ethnic communication (White
1971; Mugler & Tent 1998; Tent 2001; Mangubhai & Mugler 2003).
10. It has to be noted that the function that has been termed ‘wider communication’ here is
not the same as the language function described for multilingual settings by Stewart (1968,
540–542), in which a language of wider communication is not the same as an official or provin-
cial language.
4.5 Attitudes towards Fiji English
After probing the self-reported usage of Fiji English in a directed content analysis
with results ranging from modes to domains and functions, I will now focus on
more attitudinal data mostly collected from the last section of the questionnaire.
This open-ended item, entitled “Comments”, provided space for a short para-
graph of writing; the opportunity of adding comments was seized by 27 out of
63 Fijian respondents and 30 out of 67 Indo-Fijian respondents. The comments
allow a glimpse into attitudes towards and stereotypes of Fiji English, ranging
from negativist to positivist attitudes and beliefs, with an unexpected instrumen-
tal dimension uncovered in the course of the investigation.
Starting with negativist views of Fiji English, only five respondents provided
one-dimensionally negative comments (examples (11)–(15)).
(11) If you referring to broken English which to me is mixture of English + Fijian
(12) Fiji English would mean “broken” English where Fijian words are substituted,
as well as, Fijian grammar is substituted for English grammar
(13) In Fiji English, we use a lot of slangs and a mixture of American and UK
English
(14) English has been widely used and I feel that if people in Fiji use the correct
terms and structure of English then it could improve
(15) English is the well known language in the world and we still need good teach-
ers to help us to speak English in a proper way
Comments (11) to (13) pick up on the controversial nature of the label Fiji English
that was pointed out by previous studies, namely that “[…] the label ‘Fiji English’
is normally used only for the basilectal end of the speech continuum and consid-
ered sub-standard […]” (Mugler 2002, 83); also see Lynch and Mugler (1999, 11):
“It is worth noting that within Fiji, the term “Fiji English” tends to refer only to the
basilectal end of the continuum, perhaps because only that lect is recognised as
distinctive”. However, notions about a local variety like the ones in examples (11)
to (15) have been summarized under the label of “complaint tradition”, and were
linked by Schneider to phase 3, nativization, of variety development. It is in this
phase that “an awareness of the deviance of some local linguistic usage from old
norms of correctness grows” (Schneider 2007, 43), which is even more evident in
examples (16) to (19).
(16) I prefer to use the proper English language when in class at school but when
among my friends, I prefer the Fijian English language
206 Lena Zipp
(17) Fiji English is informal so I use it for informal contexts. However, English
proper/formal in Formal English
(18) We learn proper English in school and I can speak proper English but during
conversations I automatically speak Fiji English because I do not know some
words in Fijian – unfortunately
(19) Although I think I use British English during teaching and writing I am aware
of lapses that occur due to the influence of my mother tongue and “movie
language”
These responses reveal a considerable awareness of stylistic variation and the ex-
istence of conflicting linguistic norms, as Schneider (2007, 43) notes: “[…] is the
old, metropolitan norm still the only ‘correct’ one, as conservative circles tend to
hold, or can local usage really be accepted as correct simply on account of being
used by a significant proportion of the population, including educated speakers?”.
The use of Fiji English might be, but is not necessarily negatively connoted in the
opposition of “proper/formal” and “informal” in examples (16) and (17), maybe
more so in the words “unfortunately” in example (18) and “lapses” in example
(19). However, awareness of differing norms necessarily spreads gradually, as one
respondent who claimed to prefer using Fiji English “among my local friends and
my family members (older family members)” stated:
(20) Wow, I have never thought about English from this perspective before. It
really makes me consider my influences and I’ve realised that I speak different
“versions” of English depending on who I’m speaking to.
Quite typically for meta-level discussions in countries with emerging national

varieties of English, former or existing exonormative prestige models are called
upon for authorization (examples (21)–(26)).
(21) I like American english the most but I mostly use Fiji english
(22) while speaking we use Fiji english i.e. British english because we learnt these
english since our primary school
(23) British English is the best English because I find it pure and it is pleasant to
hear
(24) well, i would say that British Eng. are better and sounds good in nature
(25) Most of the time I use british language. But I am quiet intrested to listen to
the U.S. English
(26) I think American English is quite popular, or should I say used more in Fiji
While all of these six comments could be classified as exhibiting “(post-) colonial
cringe”, examples (21) and (22) still do not deny the existence of a local variety.
Examples (23) and (24) show a strong belief in the superiority of the historically
established colonial prestige model, British English, whereas examples (21), (24)
and (26) betray the growing global influence of American English. As opposed
to this influence, there are three Indo-Fijian respondents who claim to be “not
familiar with English used outside Fiji/English spoken all over the world”, because
they “have never been out of Fiji”. Only one respondent chooses to express their
preference of a local vernacular over English in English-dominated domains (ex-
ample (27)), but significantly, the answer takes the form of (rhetorical) questions:
(27) Possible to communicate in Fijian in universities? Speak, read and write?
Instead of applying a second language?
Moving from negative to positive attitudes, there are a number of comments that
more or less explicitly approve of the local variety of English (examples (28)–(33)).
(28) English spoken in different countries is unique so we should have Fiji English
as well.
(29) I reckon the usage of English in our country is a very good means of com-
munication. And so that’s how we communicate here more often. We all have
different means of speaking this language.
(30) No english is perfect. Every country have different pronunciation of english
therefore I find Fiji English much better because it is quite clear.
(31) Due to a highly diversified society, we tend to speak in a sort of english which
includes jargons and vernacular languages.
(32) I believe there is not many slangs and fancy jargons.
(33) Even though it is a second language, it is interesting to study.
While examples (28), (29) and (30) are openly in favour of the local variety of
Fiji, examples (31), (32), and (33) are generally positive and more concerned with
the structure and status of Fiji English in a linguistically informed manner. All
these examples fall in the realm of endonormative stabilization in which “[t]he
existence of a new language form is recognized, and this form has lost its former
stigma and is positively evaluated” (Schneider 2007, 50). It has to be noted that
both examples (13) and (32) above contain the word slangs, a very distinctly Fiji
English count noun that was found to be one of the most pervasive and distinctive
morpho-syntactic features of Pure Fiji English by Mugler and Tent (2008, 546):
“There are a number of distinctive count nouns, preceded by an article in the
singular and with a suffixed {-s} in the plural, […]. Two of the most noticeable
such nouns are slang ‘a slang expression/word’, and swear ‘a swearword’ […]”. It is
very revealing of the degree of endonormative stabilization to see these instances
208 Lena Zipp
of morpho-syntactic and lexico-grammatical nativization being used in meta-lin-

guistic comments.
In the course of the investigation, one unanticipated dimension to the atti-
tudes and beliefs surrounding Fiji English came to the fore in the comment sec-
tion. It was pointed out before that question-wording in open questions might
elicit instrumental rather than affective dimensions of attitudes and beliefs, if af-
fective dimensions are not explicitly addressed (Sutton et al. 2003). The same ef-
fect was apparently achieved by the “comment” section in the present study. There
is a surprisingly unanimous instrumental topos running through 24 out of the 57
total comments given by the respondents, as shown selectively in examples (34) to
(42) (Indo-Fijian respondents) and (43) to (50) (Fijian respondents).
(34) The difference in the English language would not be of much importance to
Fiji situation. More emphasis and interest is on how words are pronounced.
(35) English should be spoken in a manner that is grammatically correct and easily
understood by all concerned.
(36) something which has 0 [zero]standard and which could be understandable
(37) it is simple
(38) normal conversation does not need perfection, so it is easy to talk and
understand
(39) prefer using informal speeches because people can easily understand
(40) the reason because we will be able to understand others point of view clearly
and what they understand about particular situations
(41) easily understood by all since we are taught from the school how to speak
english
(42) because it is understand by us and people around us well
(43) Fiji English is best because it is easy to understand
(44) It will be easier to use Fiji English in academics because it will be easier to
understand
(45) Easier to get our message to be understood
(46) I find the Fijian English and American English easy to understand in both
conversation
(47) the pronunciation is fine
(48) Pronunciation used is very clear and also easy for audience to understand
(49) Because Fijian people will understand the basis of Fiji English
(50) Fiji English is more understandable esp. if one want to know exactly, or
understand what another, is trying to say. This is considering second language
speakers.
This list of examples shows the pervasiveness of the dimension of “ease of under-
standing”, an instrumental rather than affective component of language attitudes,
indicated in the data by the following three recurring keywords: pronunciation,
understand, and easy. Moreover, the general evaluation of local pronunciation
patterns is positive throughout and clearly points towards acceptance of the local
variety of English across both ethnic subsamples. Pronunciation, as Schneider
notes, is an obvious case in point of linguistic nativization: “[…] speakers will
consistently show a marked local accent, which frequently can be identified as
transfer phenomena from the phonology of indigenous languages” (2007, 44).
However, the reasons for adopting local accents are multidimensional:
Communicative effectiveness, the greater likelihood of being understood, may be
one factor promoting the use of such features, but certainly that is not all. Local-
isms are unavoidably also a display of an increasingly locally based identity […].
(Schneider 2007, 42–43)
5. Conclusion
The aim of this paper was to weigh diasporic (i.e. exonormative) effects against
endonormative stabilization of a local variety of Fiji English for Fiji citizens of
Indian descent. In order to do so, I traced the socio-history and sociolinguistic
situation of Indians in Fiji and their diasporic relation to the country of origin.
When summarizing the status of research on the local variety of English in Fiji,
it was shown that there are only few potential linguistic links between the Indo-Â�
Fijian sub-variety of English and Indian English. Moreover, it is nearly impossible
to distinguish between similarity of forms caused by either diachronic evolution,
language contact, or parallel first-language influence. On the other hand, there is
strong evidence for an emerging national standard common to both major ethnic
groups in Fiji, suggesting that Fiji English has moved farther along the develop-
mental cycle of post-colonial Englishes than hitherto acknowledged (but already
hinted at in Schneider’s (2007, 118) comment that “[c]learly, the potential for pro-
gress toward phase 3 is visible […]” in Fiji). In the analytical part of the paper, the
usage of the local variety was assessed based on open-ended self-reports from a
sample group of young Fiji citizens in tertiary education. The general trend across
ethnicities was one of acceptance of the variety label Fiji English, which presents
a strong argument for endonormative stabilization according to Schneider’s
210 Lena Zipp
(2007, 50) model and disproves his claim that the use of English for intra-nation-
al communication in Fiji “does not appear to have affected identity construc-
tions, which […] are still predominantly ethnicity- rather than nation-based”
(Schneider 2007, 116). Usage patterns clustered around the mode of spoken lan-
guage; the domains of friendship, family and education/work were shown to be
salient, as well as the traditional bridging function of English in a multilingual
country. Indo-Fijian respondents reported overall lower usage of the local variety
of English than Fijian respondents, but it cannot be concluded whether or not
this might be linked to different values attached to varieties of English with re-
gard to international acceptance in the face of emigration and political pressures.
Attitudes towards Fiji English comprised instrumental rather than affective com-
ponents, and ranged from negative (“colonial cringe”) to explicitly positive. None
of the Indo-Fijian respondents’ statements concerned diasporic connections, and
none of the responses showed any kind of “enregisterment”, or recognition of an
Indian style of English in Fiji, which again questions the notion of an Indian di-
aspora in the minds of the Fiji Indian population. In sum, endonormative variety
stabilization appears to be a stronger influence on the English spoken by Fiji’s
Indians than diasporic influence.
There are certain shortcomings to this study, among them certainly the un-
orthodox question design and the limited number of respondents. A further
loophole is the fact that it cannot be ruled out that some participants might have
understood “in which contexts do you prefer to use English in Fiji, in contrast
to the vernacular languages?”, instead of “in which contexts do you prefer to use
Fiji English?”. Two comments seem to betray this misconception (examples (51)
and (52)).
(51) I prefer speaking in english because it our best language
(52) The English language has become a big part of my life and most of the time I
forget to speak my own mother tongue.
Apart from the shortcomings that the questionnaire design presented, all attitude
studies are prone to social desirability bias and acquiescence bias (although less so
when conducted in written form). The latter, however, could be a serious skewing
factor in this investigation in the following way: By presenting the respondents
with the term Fiji English in print on the questionnaire sheet, more respondents
might have adopted it in their answers, and might have shown greater tendencies
to admit using this variety. This is an issue of the present study that could not be
resolved. However, research on language attitudes towards ESL varieties is a very
recent topic (see e.g. Tan & Tan 2008; Sand 2011) and is well worth investigating
in future studies on Fiji English on a greater scale – either using indirect methods
such as the matched-guise technique or larger-scale open-ended methods that

could be evaluated for example in a multi-rater cluster analysis (“concept map-
ping”, see Jackson & Trochim 2002).
References
Akande, Akinmade T. & Salami, L. Oladipo. 2010. Use and attitudes towards Nigerian Pidg-
in English among Nigerian university students. Marginal Dialects: Scotland, Ireland and
Beyond. Aberdeen: Forum for Research on the Languages of Scotland and Ireland, Robert
Millar (ed.), 1–79. Aberdeen: University of Aberdeen.
Barn, Ravinder. 2008. Indian diaspora in the United Kingdom: Second-generation parents’
views and experiences on heritage language transmission. In Tracing an Indian Diaspora.
Contexts, Memories, Representations, Parvati Raghuram, Ajaya Kumar Sahoo, Brij Maharaj
& Dave Sangha (eds), 191–209. New Delhi: Sage. DOI: 10.4135/9788132100393.n11
Bayard, Donn. 2000. The cultural cringe revisited: Changes through time in Kiwi attitudes to-
wards accents. In New Zealand English [Varieties of English around the World Gf25], Allan
Bell & Koenraad Kuiper (eds), 297–324. Amsterdam: John Benjamins.
DOI: 10.1075/veaw.g25.16bay
Carley, Kathleen. 1993. Coding choices for textual analysis: A comparison of content analysis
and map analysis. Sociological Methodology 23: 75–126. DOI: 10.2307/271007
Edwards, John. 1999. Refining our understanding of language attitudes. Journal of Language
and Social Psychology 18(1): 101–110. DOI: 10.1177/0261927X99018001007
Esses, Victoria M. & Maio, Gregory R. 2002. Expanding the assessment of attitude components
and structure: The benefits of open-ended measures. European Review of Social Psychology
12(1): 71–101. DOI: 10.1080/14792772143000021
Fiji Islands Bureau of Statistics. The Official Statistical Agency for the Fiji Government. <www.
statsfiji.gov.fj/Social/popn_summary.htm> (15 October 2012).
Fiji Constitution (Amendment) Act 1997. <www.paclii.org/fj/legis/num_act/ca1997268/> (14
June 2013).
Fishman, Joshua. 1972. Domains and the relationship between micro- and macrosociolinguis-
tics. In Directions in Sociolinguistics. The Ethnography of Communication, John J. Gumperz
& Dell Hymes, 435–453. New York NY: Holt, Rinehart and Winston.
Garrett, Peter. 2004. Attitude measurements / Messung von Einstellungen. In Sociolinguistics /
Soziolinguistik. An International Handbook of the Science of Language and Society / Ein
internationales Handbuch zur Wissenschaft von Sprache und Gesellschaft, Vol. 1, Ulrich
Ammon, Norbert Dittmar, Klaus Mattheier & Peter Trudgill (eds), 1251–1260. Berlin: De
Gruyter.
Garrett, Peter, Coupland, Nikolas & Williams, Angie. 2003. Investigating Language Attitudes.
Social Meanings of Dialect, Ethnicity and Performance. Cardiff: University of Wales Press.
Geer, John G. 1988. What do open-ended questions measure? Public Opinion Quarterly 52(3):
365–367. DOI: 10.1086/269113
Geer, John G. 1991. Do open-ended questions measure ‘salient’ issues? Public Opinion Quarter-
ly 55(3): 360–370. DOI: 10.1086/269268
Geraghty, Paul. 2005. Literacy and the media in the Fiji Islands. Pacific Journalism Review 11(1):
48–57.
212 Lena Zipp
Geraghty, Paul, Mugler, France & Tent, Jan (eds). 2006. Macquarie Dictionary of English for the
Fiji Islands. Sydney: The Macquarie Library of the Macquarie University.
Gillion, Kenneth L. 1962. Fiji’s Indian Migrants. A History to the End of Indenture in 1920.
Melbourne: OUP.
Haddock, Geoffrey & Zanna, Mark P. 1998. On the use of open-ended measures to assess atti-
tudinal components. British Journal of Social Psychology 37: 129–149.
DOI: 10.1111/j.2044-8309.1998.tb01161.x
Hsieh, Hsiu-Fang & Shannon, Sarah E. 2005. Three approaches to qualitative content analysis.
Qualitative Health Research 15(9): 1277–1288. DOI: 10.1177/1049732305276687
Hofman, John E. 1975. Language attitudes in Rhodesia. In The Spread of English. The Sociology
of English as an Additional Language, Joshua A. Fishman, Robert L. Cooper & Andrew W.
Conrad (eds), 277–301. Rowley MA: Newbury.
Hundt, Marianne. 2013. The diversification of English: Old, new and emerging epicentres.
In English as a Contact Language, Marianne Hundt & Daniel Schreier (eds), 182–203.
Cambridge: CUP.
Jackson, Kristin M. & Trochim, William M.K. 2002. Concept mapping as an alternative ap-
proach for the analysis of open-ended survey responses. Organizational Research Methods
5(4): 307-336. DOI: 10.1177/109442802237114
Kelly, John D. 1995. Bhakti and postcolonial politics: Hindu missions to Fiji. In Nation and Mi-
gration: The Politics of Space in the South Asian Diaspora, Peter van der Veer (ed.), 43–72.
Philadelphia PA: University of Pennsylvania Press.
Lal, Brij V. 2006. Fiji. In The Encyclopedia of the Indian Diaspora, Brij V. Lal (ed.), 370–382.
Honolulu HI: University of Hawai’i Press.
Lasagabaster, David. 2004. Attitude/Einstellung. In Sociolinguistics. An International Handbook
of the Science of Language and Society, Ulrich Ammon, Norbert Dittmar, Klaus J. Mattheier
& Peter Trudgill (eds), 399–405. Berlin: De Gruyter.
Lynch, John & Mugler, France. 1999. English in the South Pacific. <www.vanuatu.usp.ac.fj/
paclangunit/English_South_Paci-fic.htm> ( 11 February 2005).
Mangubhai, Francis & Mugler, France. 2003. The language situation in Fiji. Current Issues in
Language Planning 4(3–4): 367–456. DOI: 10.1080/14664200308668058
May, Thor. 1990. Language in Suva. Language use and literacy in an urban Pacific community.
Paper for the conference of the Australian Linguistic Society at Macquarie University in Syd-
ney, NSW. <thormay.net/lxesl/tech7.html> ( 15 July 2012).
Mollin, Sandra. 2006. Euro-English. Assessing Variety Status. Tübingen: Gunter Narr.
Mugler, France. 2002. Attitudes to accents in English: A Pacific study. Te Reo: Journal of the
Linguistic Society of New Zealand 45: 65–89.
Mugler, France & Tent, Jan. 1998. Some aspects of language use and attitudes in Fiji. In SICOL:
Proceedings of the Second International Conference on Oceanic Linguistics: Vol. 1, Language
Contact, Jan Tent & France Mugler (eds), 109–134. Canberra: Pacific Linguistics, Research
School of Pacific and Asian Studies, Australian National University.
Mugler, France & Tent, Jan. 2008. Fiji English: Morphology and syntax. In Varieties of English,
3: The Pacific and Australasia, Kate Burridge, Bernd Kortmann & Edgar W. Schneider
(eds), 546–567. Berlin: Mouton de Gruyter.
Mühleisen, Susanne. 2001. Is ‘bad English’ dying out? A comparative diachronic study on at-
titudes towards creole versus standard English in Trinidad. Philologie im Netz 15: 43–78.
Prasad, Rajendra. In preparation. The Grammar of Fiji Hindi. MA dissertation, University of
the South Pacific.
Raghuram, Parvati & Sahoo, Ajaya Kumar. 2008. Thinking ‘Indian diaspora’ for our times.
In Tracing an Indian Diaspora. Contexts, Memories, Representations, Parvati Raghuram,
Ajaya Kumar Sahoo, Brij Maharaj & Dave Sangha (eds), 1–20. New Delhi: Sage.
Romaine, Suzanne. 1999. Changing attitudes to Hawai’i Creole English. Fo’ find one good job,
you gotta know how fo’ talk like one haole. In Creole Genesis, Attitudes and Discourse. Stud-
ies Celebrating Charlene J. Sato [Creole Language Library 20], John R. Rickford & Suzanne
Romaine (eds), 287–301. Amsterdam: John Benjamins. DOI: 10.1075/cll.20.20rom
Sand, Andrea. 2011. Language attitudes and linguistic awareness in Jamaican English. In Var-
iation in the Caribbean. From Creole Continua to Individual Agency [Creole Language
Library 37], Lars Hinrichs & Joseph T. Farquharson (eds), 163–187. Amsterdam: John
Benjamins. DOI: 10.1075/cll.37.10san
Schneider, Edgar W. 2007. Postcolonial English: Varieties Around the World. Cambridge: CUP.
DOI: 10.1017/CBO9780511618901
Siegel, Jeff. 1987. Language Contact in a Plantation Environment. A Sociolinguistic History of Fiji
[Studies in the Social and Cultural Foundations of Language 5]. Cambridge: CUP.
Siegel, Jeff. 2003. Indian languages in Fiji: Past, present and future. In Indian Diaspora: In Search
of Identity, Rekha Sharma & E. Annamalai (eds), 105–136. Mysore: Central Institute of
Indian Languages.
Smit, Ute. 1996. A New English for a New South Africa? Language Attitudes, Language Planning
and Education. Wien: Braumüller.
Sökefeld, Martin. 2006. Mobilizing in transnational space: A social movement approach to the for-
mation of diaspora. Global Networks 6(3): 265–284. DOI: 10.1111/j.1471-0374.2006.00144.x
Srebrnik, Henry. 2008. Indo-Fijians: Marooned without land and power in a South Pacific ar-
chipelago? In Tracing an Indian Diaspora. Contexts, Memories, Representations, Parvati
Raghuram, Ajaya Kumar Sahoo, Brij Maharaj & Dave Sangha (eds), 75–95. New Delhi:
Sage. DOI: 10.4135/9788132100393.n5
Stewart, William A. 1968. A sociolinguistic typology for describing national multilingualism.
In Readings in the Sociology of Language, Joshua A. Fishman (ed.), 531–545. The Hague:
Mouton. DOI: 10.1515/9783110805376.531
Sutton, Stephen, French, David P., Hennings, Susie J., Mitchell, Jo, Wareham, Nicholas J.,
Griffin, Simon, Hardeman, Wendy & Kinmonth, Ann Louise. 2003. Eliciting salient beliefs
in research on the theory of planned behaviour: The effect of question wording. Current
Psychology 22(3): 234–251. DOI: 10.1007/s12144-003-1019-1
Tan, Peter K.W. & Tan, Daniel K.H. 2008. Attitudes towards non-standard English in Singa-
pore. World Englishes 27(3-4): 465–479. DOI: 10.1111/j.1467-971X.2008.00578.x
Tent, Jan. 2001. The current status of English in Fiji. In Who’s Centric Now? The Present State of
Post-colonial Englishes, Bruce Moore (ed.), 241–268. Oxford: OUP.
Tent, Jan & Mugler, France. 2008. Fiji English: Phonology. In Varieties of English, 3: The Pacific
and Australasia, Kate Burridge, Bernd Kortmann, & Edgar W. Schneider (ed.), 234–266.
Berlin: De Gruyter.
White, Ronald V. 1971. Language use in a South Pacific urban community. Anthropological
Linguistics 13: 361–384.
Winford, Donald. 1976. Teacher attitudes toward language varieties in a Creole community.
International Journal of the Sociology of Language 8: 47–75.
Zipp, Lena. 2014. Educated Fiji English. Lexico-grammar and Variety Status [Varieties of English
around the World G47]. Amsterdam: John Benjamins. DOI: 10.1075/veaw.g47
Chapter 10
Transnational flows, language variation,

and ideology
Devyani Sharma
Queen Mary University of London
Migrant groups maintain ties to their region of origin to different degrees, but
this variable factor has not been consistently integrated into sociolinguistic
studies of diaspora communities. A balanced consideration of transnational
interaction, as well as the social valuation of such ties, is needed in order to
understand the development of diasporic varieties. This article examines the
role of ‘material’ and ‘ideological’ conduits of transnational influence: Does
a person’s transnational activity influence their language use? And can their
ideologies of distant varieties, in particular Indian English (IndE), influence
their language use as well? Focusing on the use of an IndE accent feature among
second generation members of a British Punjabi community, I first develop an
exploratory metric to track the impact of transnational activity on language be-
haviour. The analysis finds decreasing transnational activity over time as well as,
independently, decreasing correlation with transnational factors. A qualitative
consideration of the participants’ interviews points to positive alignment with
educated IndE as a relatively prestigious variety, supporting not just a material
but also an ideological basis for maintaining selected IndE accent features.
Keywords: transnationalism, ideology, British Asian diaspora, Indian English
1. Introduction
Communities that have arisen as a result of migration are often described in

terms such as ‘immigrant’, ‘diasporic’, and ‘transnational’. Although often used in-
terchangeably, these terms tend to correspond to differences in how much ties are
maintained with the region of origin. In an age of unprecedented international
travel and long-distance communication, these social practices may play an im-
portant role in how the linguistic and cultural systems of migrant communities
develop. In particular, they may affect whether a distinctive variety in the new
216 Devyani Sharma
group emerges and whether a heritage language is maintained. Yet these factors
are still relatively new and under-theorized in sociolinguistic research.
Sustained transnational ties might directly influence the direction and rate of
development of a new variety, and might counteract assimilation into the majority
variety. Alternatively, they might show no influence at all, for instance if the local
peer environment in early language acquisition is a stronger factor. An analysis
of transnational activity can thus clarify the relative balance of local acquisitional
and global social forces in the process of dialect change. The first part of this study
examines the degree of influence over time that an individual’s transnational ties
exert on the variety of English they use.
If transnational activity does influence language variation, are actual ‘physi-
cal’ network links the only conduit for this influence? Or can ideology, independ-
ent of the presence or absence of actual ties, be strong enough to influence usage
as well? Transnational ties connect not just individuals but speech varieties too,
so in principle transnational dynamics could give rise to new language ideologies
and changing valuations of, and thus relative influence of, varieties (Heller 2007).
The second part of this study therefore turns from ‘face-to-face’ transmission
effects to ideological effects. In particular, it assesses whether British Asians asso-
ciate Indian English (IndE) with low status, deriving from a perceived non-native
status or regional stereotyping, or higher status, deriving from an increasingly
stable and prestigious pan-Anglophone status for Standard IndE (Hoffmann,
Hundt & Mukherjee 2011), and whether these valuations seem to relate to lin-
guistic practices.
The diasporic social effects to be explored can thus be framed as two questions:
Question 1 (Transnational ties):
Does a person’s transnational activity influence their language use?
Question 2 (Language ideology):
What ideologies are ascribed to distant varieties, in particular IndE, and can
these influence language use as well?
These questions will be examined in relation to a Punjabi community in West

London. Members of this community vary significantly in whether and how they
maintain transnational ties, depending on social, historical, and economic factors.
So who in the community should we study? I focus here on second-genera-
tion members of the community, following Vertovec (2009, 75–76):
High among questions and criticisms regarding the transnational lens on mi-
gration are issues as to how members of second and subsequent generations
are affected by transnationalism. There is one common view that transnational
Chapter 10.â•‡ Transnational flows, language variation, and ideology 217
practices among second-generation youth are currently minimal and likely to

dwindle further in the course of time. However, another view – and one in line
with an understanding of how a transnational habitus is shaped and acted upon –
suggests that there exist ‘strong influences in the transnational social fields in
which the second generation is embedded. This view stresses the importance of
the sending-country individuals, resources, and ideas that are a constant pres-
ence in the lives of the second generation and holds that even selective, periodic
transnational practices can add up’ (Levitt & Waters 2002, 4).
Vertovec goes on to note that intensive transnational exchange can occur at par-
ticular life-stages among the second generation and considerable variety can exist
across different second-generation groups. This diversity derives not only from
social diversity within communities, but from the fact that migration history does
not stop once a local generation is born. Members of the second generation are
born at different points of the community’s continuing development, a point ob-
scured by the focus in many studies on younger local-born individuals (note this
blurring in the phrase “second-generation youth” in the quote above).
Focusing on second-generation people of different ages, this study will show
that being born earlier or later in the history of the migrant community has a
profound impact on social activity, language variation, and ideology (see Sharma
2011 and Sharma & Sankaran 2011 for further evidence of age differences).
In responding to Question 1, the analysis first notes some methodological ob-
stacles in quantifying and interpreting the relative strength of transnational fac-
tors. Despite these, an exploratory correlational analysis of transnational activity
and language use does suggest that a person’s level of transnational social activity
influences their use of an Asian-derived dialect feature in English. However, these
effects are strongest in the earliest second generation. Later stages show continued
use of Asian dialect style, but with new local meanings that are much less clearly
tied to personal transnational activity.
In relation to Question 2, a qualitative analysis of meta-linguistic commen-
taries indicates that the urban middle class variety of Standard IndE has emerged
from being a stigmatized variety to being a stable and valued reference point in
the diaspora, even accruing prestige associations and possibly influencing the
choice of variables maintained in diasporic varieties.
These transnational effects, both linguistic and ideological, may account in
part for the lasting presence of Asian speech styles among later generations of
British Asians.
218 Devyani Sharma
2. Transnational communities
A vast literature has arisen around the themes of transnationalism and diaspora.
Not all descriptions of these states and processes can be applied straightforwardly
to the narrower sociolinguistic processes of language acquisition, variation, and
change. This section identifies a few core themes from the wider literature on
transnationalism that are relevant to sociolinguistics, and then briefly outlines
social and historical details of the West London Punjabi community.
2.1 Transnationalism and sociolinguistics
Transnationalism has been described as “a set of sustained long-distance, bor-

der-crossing connections” (Vertovec 2004, 3). In wider social theory and social
sciences, transnationalism has attracted interest for some decades now because
it has forced analysts to rethink certain traditional notions of social structure. Of
particular interest to sociolinguistics are notions of community, norms, acqui-
sition, ideology, and transmission across generations. I focus more narrowly on
these in this section. I address the more practical question of how transnational
activity is measured later in Section 3.
Early on in theorizing transnationalism, Gupta and Ferguson (1992, 9) ob-
served that “[s]omething like a transnational public sphere has certainly rendered
any strictly bounded sense of community or locality obsolete. At the same time, it
has enabled the creation of forms of solidarity and identity that do not rest on an
appropriation of space where contiguity and face-to-face contact are paramount.”
But linguists have long recognised the centrality of face-to-face interaction
for language acquisition. How easily can Gupta and Ferguson’s descriptions of
fundamentally transformed social worlds apply to the minutiae of accent and
dialect acquisition and use? How readily does their dismantling of ‘community’
apply to ‘speech community’?
First, the expansion of communities beyond the ‘local’ need not render face-
to-face interaction less relevant; as we will see in the discussion of Question 1
later, despite extraordinary distances, many forms of transnational activity act
precisely to ensure regular face-to-face interaction with non-local groups.
Second, much sociolinguistic research has clarified that, beyond face-to-face
interaction, an important role must be given to more agentive ‘acts of identity’
(Le Page & Tabouret-Keller 1985) and ‘initiative’ design in speech (Bell 2001),
such that language variation can be influenced by allegiance to more removed
groups, even ‘imagined communities’ (Anderson 1991), in which a direct or local
personal network may not be a core property. The notion of imagined commu-
nities has been invoked in sociolinguistics to account for the social dynamics of
numerous scenarios, including language learning (Pavlenko & Norton 2007), lan-
guage contact (Omoniye 2004), regional dialects (Llamas 2006; Mendoza-Denton
2010), and, most relevant to the present work, diasporic Asian speech communi-
ties (Shankar 2008; Bhatt 2010). This ideological mediation of language variation,
sustained through some transnational activity but also through ideologies that
develop through those experiences, applies in particular to the later discussion
of Question 2.
A number of ethnographic and sociolinguistic studies have reflected on the
role of transnational flows and ties. I describe a sample here, first more qualitative
and then more quantitative, to highlight shared concerns and concepts across the
field.
An important qualitative study of a comparable South Asian diasporic com-
munity is Shankar’s (2008) anthropological study of second-generation South
Asian teenagers in California. She observes distinct migration and class histories
within the South Asian community (e.g. differences in the realisation of an ideal-
ised American dream of upward mobility), which in turn correspond to distinct
social and transnational dynamics.
In the context of Britain, Harris (1996) noted a gap in the study of ethnicity
in British cultural studies, one that lies at the heart of the present research. He
argues that, perhaps due to an inordinate focus on youth subcultural frameworks,
insufficient attention has been paid to the older populations in ethnic minority
communities, particularly in the context of British Caribbean groups (though this
holds for most ethnic minority groups). He proposes that we turn our attention
to older historical agents within these stories. The findings of the present study
highlight the importance of doing so, as a binding force between South Asia and
younger British Asians is revealed to be the earliest (oldest) British-born group,
often overlooked in sociolinguistic studies of minority groups.
Rampton (2011; 2013) and Sharma (2012) indicate important interactions
in the indexing of migration status and ethnicity in the community over time,
such that British class structure increasingly intersects with originally migra-
tion-linked forms. Finally, Harris, Leung, and Rampton (2002) raise an impor-
tant further point, namely that not all diasporic individuals present in the British
scene are committed to permanent migration. I return to implications of this
point in Section 5.
A number of studies have also attempted to estimate the influence of dias-
poric links in quantitative terms.
Meyerhoff and Walker (2007) investigated the effect on grammatical varia-
tion of periodic, sometimes extended, travel abroad from the island of Bequia, in
the Grenadines, by groups they termed ‘urban sojourners’. As with Harris et al.’s
(2002) description above, these can be long-term but not permanent migrants.
220 Devyani Sharma
Their importance for sociolinguistics lies in offering some insight into the ques-
tion of lifespan change. Meyerhoff and Walker found that although these indi-
viduals may sound different, their underlying grammars are not substantially
restructured through extended transnational contact, at least not for a variable
with low social awareness. They note, importantly, that the real social networks of
such sojourners must be measured, not simply the overall time spent away from
the island. If the individual “taps into and embeds himself/herself in dense social
networks of expatriate speakers” while away, the new contact situation might be
expected to exert a weaker effect (Meyerhoff & Walker 2007, 360).
Li Wei (1994), investigating a comparable community in Britain, the Chinese
community in Newcastle, used Milardo’s (1988) tripartite distinction of active,
passive, and exchange ties, where passive ties in particular can include affective
bonds based on more irregular or infrequent interaction. His ‘Chinese ethnic
index’ specifically measured, within each category, the proportion of Chinese
pre-migration links within the network, and he showed that code-switching prac-
tices correspond to differences in network. Wong (2007) adapts this methodology
in a study of Chinese-Americans, but includes a close quantification of cultural
lifestyle choices as well, and finds that selected phonetic variables correlate with a
combination of these factors. In related work, Matsumoto and Britain (2009) draw
a parallel between migration and post-colonial scenarios, suggesting that whereas
pre-migration links may be important in a migrant community, in a post-colonial
context (in their study, Palau), pre-colonial links appear to serve a similar func-
tion, and they use a similar ‘Japanese ethnic index’ to measure such links.
Many of the articles in the present volume also address the question of trans-
national effects. Alam and Stuart-Smith (this volume) note dense networks and
sustained transnational practices in the Pakistani community in Glasgow; it is
conceivable that these might vary systematically with the communities of prac-
tice they describe. And Hundt (this volume) points to the interesting case of two
younger women whose article use suggest a stronger influence of transnational
and attitudinal factors than age of migration.
In sum, although transnational ties inevitably complicate the analysis of mi-
grant groups, a balanced consideration of local and supralocal ties, as well as the
ideological status ascribed to those ties, is certainly possible. Indeed, the inclusion
of such considerations is increasingly desirable for a more complete understand-
ing of change in diasporic varieties as well as in original ‘source’ varieties.
2.2 The present community1
South Asians, at 7.5% of the total population in 2011, are the largest ethnic mi-
nority in the UK. 35.9% of this ethnic group lives in London, and major concen-
trations of Asians reside in the West London boroughs of Ealing and Hounslow.
The data for this study were collected from a small town in Ealing called Southall.
Since the mid-20th century, Southall has attracted Punjabi speakers from India,
Pakistan, and East Africa and it is still considered the historic heart of the Pun-
jabi, and in particular Sikh, community.2 Demographic estimates vary: The 2011
Census found that out a total population of 98,000, 69.5% were classified as non-
White ethnicity, and 51% as Asian. Some neighbourhoods in the Borough have
very high concentrations of Asians, even by official Census figures, e.g. 80.7% in
the ward of Southall Broadway. Other sources (DMAG 2006; Ealing JSNA 2010)
estimate higher statistics for minority ethnicities and Asians in Southall. Tak-
ing into account undocumented residents, it is reasonable to estimate that the
Asian population exceeds 60% and the overall ethnic minority population may
exceed 80%.
The population of Southall includes both India-born (Gen 1) and British-born
(Gen 2 and Gen 3) residents. Members of the Gen 1 group were born in South
Asia and migrated to the UK in adulthood; this group is continually renewed
through ongoing migration. Due to their continuous arrival, the Gen 2 group –
born in West London of India-born parents – has a very diverse age range. The
earliest British-born are now in their 50s, even 60s, but children being born now
to recent India-born migrants are also technically Gen 2 individuals.
This study focuses on the second generation but from a historical perspective,
dividing this group into older and younger members. All the individuals in the
present study were raised in West London and all have first-generation parents
who grew up in South Asia.
The recent history of Southall is of central importance to an analysis of trans-
national activity and language use. Although transnational flows have been main-
tained for over half a century by a continually replenished Gen 1 presence, the
1. Background descriptions of the community (Section 2.2) and of the coding of the phonetic
feature (Section 3.2) are based on descriptions first presented in Sharma and Sankaran (2011).
2. Other South Asian languages are concentrated in different London neighborhoods, e.g.
Gujarati in Wembley and Bengali in East London. The participants in this study are over-
whelmingly of Indian Punjabi heritage; one family is of Pakistani Punjabi heritage, but the
foreign-born members of this family were born before the creation of Pakistan. In terms of
religion, the participants are also overwhelmingly Sikh, with a minority of Muslim and Hindu
individuals. A strong shared Punjabi identification cross-cuts these differences.
222 Devyani Sharma
wider history of the community can be divided into two broad phases, during
which social practices have transformed gradually leading to incremental chang-
es in transnational practices in Gen 2.
Phase I (late 1940s-late 1980s): In the post-war period, the UK faced severe
labor shortages and encouraged labor migration from former colonies through
the British Nationality Act of 1948, which permitted ‘Commonwealth citizens’
to enter the UK without restriction. The Asian population grew substantially be-
tween 1948 and 1971, before a series of immigration acts began to limit numbers.
By the late 1970s, 30% of the population of Southall was Asian – still a minority
but a large and highly visible one. The economic climate and British public opin-
ion had shifted, and Southall became a lightning rod for racial tension (CARF
1981, 43; Oates 2002, 107). Far-right, anti-immigration parties held rallies in the
town, leading to violent riots and racially motivated deaths (CARF 1981). These
experiences were formative for older Gen 2 participants, all of whom repeatedly
alluded to them in their interviews.
Our interview data indicate that early Asian migrants during this period
maintained many traditional Asian practices with their Gen 2 children, including
transnational arranged marriages (i.e. spouses from India), greater social restric-
tions on girls, and an expectation that boys would enter their fathers’ businesses,
which tended to be transnational in nature.
Phase II (late 1980s-present): Although racial tension persisted through the
1990s, the second phase of Asian history in Southall is characterized by a striking
reduction of overt hostilities. It is no coincidence that this change in race rela-
tions corresponds to a shift in Southall demographics, such that the Common-
wealth heritage population, mostly South Asians, became the majority and the
white community the minority (Oates 2002, 107; Meads 1983). Southall schools
became dramatically more multi-racial, with the proportion of minority ethnic
origin students in Ealing schools now ranging from 40% to 99% (Ealing JSNA
2010, 18). Today, many public signs in Southall are in English and Punjabi (even
at the local pub) and the town’s lively Punjabi atmosphere – bhangra music, Indi-
an restaurants, clothing and jewellery shops – is well-known in London.3
The content of our interview data reveals a stark difference in lived experience
and social practices among the younger Gen 2 people raised during this phase.
These children grew up in a climate in which wider British society accepted an in-
creasingly visible, legitimated, even celebrated, middle class British Asian culture
(Herbert 2009; Sharma 2011). These participants rarely volunteer narratives of
3. Several participants even suggested that the high point of Punjabi culture in Southall may
have passed, with an upwardly mobile movement of South Asians out of the area and new, non-
Asian migrants coming in.
racial tension. More importantly for the present study, a consistent shift emerg-
es such that traditional Punjabi social practices are less unquestioned: arranged
marriages are less common, transnational arranged marriages are extremely rare,
it is unusual for a boy to go into his father’s business, employment is overwhelm-
ingly in British-based industries, and girls are far better educated, marry later, and
have more professional employment than older Gen 2 women.
Sharma (2011) and Sharma and Sankaran (2011) found relatively sustained
use of Punjabi-influenced dialect features as well as bilingual use of Punjabi; how-
ever, they also found that growing up during these very distinct historical phases
corresponded to differences within the dialect grammars and style repertoires of
these groups.
The quantitative segment of this study will explore: (i) whether the different
social practices during different phases led to differences in degree of transnation-
al activity (Section 4), (ii) whether transnational activity exerts an influence on
language use, and (iii) whether this influence is steady or declining (Section 5).
Moffatt and Milroy (1992, 139), in a study of British Punjabi children’s
code-switching, offer the intriguing speculation that Punjabi and Bengali com-
munities may not be experiencing “such a sharp inter-generational disjunction
as the Chinese community”, suggesting that the nature of interpersonal ties in
these communities may differ. However, their comments overlook a potentially
important feature that distinguishes South Asian diasporic communities from
many others. This is the presence of IndE and stable, widespread English-Hindi
and English-Punjabi code-switching registers in India (unlike China), which can
form an important cultural and norm-setting reference point.
Such transnational sociolinguistic reference points must be taken into ac-
count. In the case of London, an additional reference point is the steady rise in
upper middle class and upper class South Asian ‘sojourners’ (Cohen 1997), or
temporary transnational residents, who count Standard IndE and code-switched
English among their primary varieties. In Section 6 – the qualitative analysis of
language ideologies circulating in the community – I elaborate on the potential
influence of this group within the wider South Asian London social space.
3. Methodology: Measuring transnational activity and language variation
Semi-ethnographic fieldwork was conducted in the community over a period of

nine months. Participants were approached through a number of points of entry
into the community, including local establishments such as restaurants and shops,
a local radio station, and friend-of-a-friend recommendations. In the wider pro-
ject, a total of 75 participants were recorded twice for a total of approximately
224 Devyani Sharma
120 hours.4 The first recording was a sociolinguistic interview, lasting 1–2 hours;
these recordings were the source of the phonetic data used here. The second re-
cording had the explicit goal of collecting detailed information on biography,
network, bilingualism, and cultural preferences (e.g. music, TV, cinema, leisure
activities); these recordings were 0.5–2 hours long and were used to calculate the
transnationalism index in the present study. A subset of participants conducted
self-recordings in diverse speech situations in the absence of either researcher;
these 38 recordings are not examined in detail in this article (see Sharma 2011),
but do inform a wider understanding of individual social practices.
The present study reports on Gen 2 participants, and only on the subset of 23
for whom recordings are fully transcribed and coded. The older Gen 2 group are
aged 36–55, while the younger group are aged 18–35. As noted, the older group
grew up during Phase I, outlined earlier, and the younger group grew up during
Phase II.
Methodological decisions for the quantitative analysis of Question 1 are out-
lined in this section. For the qualitative discussion of ideologies in Section 5, in-
terview content that related to questions about India and IndE was extracted and
analysed.
3.1 How should transnational activity be measured?
In addressing the practical challenge of measuring transnationalism, an imme-

diate question that arises is whether we should measure only ‘real’ social interac-
tions, virtual and physical, or also cultural practices that signal close identification
with the country of origin. Vertovec (2005, 3–4) distinguishes between transna-
tionalism and diaspora along these lines:
When actual exchanges of resources or information, or marriages or visits, take
place across borders between members of a diaspora themselves or with people
in the homeland, we can say these are transnational activities; to be transnational
means to belong to two or more societies at the same time. At that moment, the
diaspora functions as a transnational community. When such exchanges do not
take place (sometimes over many generations), but people maintain identifica-
tion with the homeland and co-ethnics elsewhere, there is only a diaspora. In this
way, not all diasporas are transnational communities, but transnational commu-
nities arise within diasporas.
4. Light-weight portable recorders (Sony MZ-RH1 and M-Audio Microtrack 24/96) were
used with lapel microphones in interviews; Zoom H2 recorders with lapel microphones were
used in self-recordings.
Thus, ‘actual’ or direct exchanges can be contrasted with more indirect forms of
transnational contact. The circulation of more indirect cultural presences and
practices are of great interest and may show some links with language use in ethnic
minority communities (Hoffman & Walker 2010; Sharma & Sankaran 2011, fn. 7).
However, as they involve more diffuse diasporic cultural flows, and our focus is on
transnationalism, the analysis will focus on direct transnational exchange such as
travel to India, long-distance communication, and marriage.
The three broad alternatives for classifying Gen 2 (or any) individuals ac-
cording to transnational activity is to develop a continuous index, a set of ordinal
(ranked) categories, or a set of nominal (qualitatively different and unranked)
categories.
The four main categories that derive the index used in the present work are:
an individual’s self-reported frequency of travel to South Asia, their self-reported
frequency of communication, the national origin and upbringing of their spouse,
and whether their job requires direct engagement with India.5 All four favour one
of the first two options above, namely a continuous or ordinal measure of degree
of activity rather than nominal categories.
A transnational index was derived from the four main measures listed above.
Individuals could receive one of three values for their frequency of travel to India:
‘0’ if they have travelled to India 0–5 times in their lifetime, ‘1’ if they have trav-
elled more frequently but still less than annually, and ‘2’ if they travel annually or
more.6 Similarly, for frequency of communication, they receive values of ‘0’ for
describing their interaction as minimal to occasional, e.g. less than weekly, and ‘1’
for describing it as regular to frequent. Individuals receive ‘1’ if they have a spouse
born and raised in India, and ‘0’ for either a British Asian or a non-Asian spouse.
And finally, they receive ‘0’ for work that involves no particular involvement with
India as opposed to ‘1’ for work that requires business trips or main business
partners in India.
Technically this constitutes a continuous index. However, because the total
value possible is 5, and the values allocated are whole numbers, the resulting val-
ues only include whole numbers ranging from 0 to 5. So the index effectively
consists of a 6-category ordinal scale.
5. With only two exceptions involving Pakistan, all South Asian ties involve India, so I refer to
India throughout.
6. A finer measure could build in length and type of stay as well, which can be a very impor-
tant factor in the quality of interactions involved. Similarly, more ethnographic detail in the
measure of work-related travel might factor in a measure of the balance of use of IndE and of
other languages in interaction.
226 Devyani Sharma
Network ties are an obvious element of transnational influence, as highlight-

ed in earlier work (Li Wei 1994, Milroy & Li Wei 1995, Wong 2007). In particular,
a further form of Asian contact takes the form of family and friends in the UK
who belong to the first generation, i.e. who were raised in South Asia. This repre-
sents a type of connection to South Asia, but falls partly in the category of indirect,
or what Vertovec classifies above as diasporic rather than transnational, influence.
Furthermore, to some extent these ties are not as much a matter of choice as other
types of transnational activity (e.g. all Gen 2 have India-raised parents). More op-
tional South Asian (non-British Asian) network ties, in the form of marriage and
work relationships, are incorporated into the four measures listed above. Com-
plete network information, including full measures of kin and friendship ties as
well as wider cultural practices, has been coded in detail for this dataset and is be-
ing assessed in ongoing research (Sharma 2013). Given the considerations above
and space limitations, fine details of network distinctions are excluded from the
present index, but broad network effects are captured.
Finally, as noted earlier, other related cultural choices and practices, such as
taking language classes, participating in community activities, and taste in mu-
sic, TV, and cinema, were also coded but not included in the transnational in-
dex, which more narrowly measures active interaction with India. (Sharma and
Sankaran suggest that these cultural practices are so closely tied to linguistic prac-
tices, that they are effectively endogenous variables that should not be treated as
independent factors.)
3.2 Coding postalveolar articulation of /t/
The phonetic feature examined in relation to the transnational index, as part of

Question 1, is postalveolar articulation of /t/ in the English spoken by partici-
pants. Although the focus is on just one phonetic feature, other analyses from the
wider project indicate strongly that both further English accent features (Sharma
2011) and maintenance of Punjabi use (Sharma 2013) show some correspond-
ence to the patterns noted in the present analysis.
Postalveolar articulation of /t/ is one of the most salient properties of IndE.
Retroflexion – a contrastive obstruent series in most Indic languages (Bhatia
1993) – is an articulation of consonants involving retraction of the tongue tip in
the postalveolar region, with considerable variation in point of articulation and
tongue shape (Masica 1993; Heselwood & McChrystal 2000). Punjabi and Hindi
have a contrast between retroflex and dental stops absent in British English (BrE),
which only has an alveolar stop. Conversely, Punjabi and Hindi do not have an
alveolar stop, and alveolar /t/ and /d/ are commonly replaced by retroflex or re-
tracted variants in IndE (Pingali 2009).
Sharma and Sankaran (2011) studied this feature in depth, but only from the
point of view of broad social and linguistic factors. The study discovered system-
atic inter-generational changes in the use of the feature, but with no consideration
of whether transnational activity correlated with its use.
As the 2330 tokens are subjected to auditory analysis only, I include all vari-
ants within the range of postalveolar retraction beyond British alveolar /t/.7 The
coding thus encompasses a range of retroflex and retracted forms. Heselwood
and McChrystal (2000) also group ‘retroflex/postalveolar’ in their analysis, and
the presence of a similar range within the source languages supports this group-
ing. However, recent research has shown that finer phonetic distinctions can be
measured between retroflex and postalveolar stops among British Asians, and
can have important social correlates (Alam & Stuart-Smith, this volume; Lambert
et al. 2007; Alam 2007; Kirkham 2011). These distinct articulations are almost
certainly distributed differently across the generations examined here, and so
closer analysis at the phonetic level is very much warranted in future work on the
present data. Nevertheless, the whole range is associated with South Asianness
(Heselwood & McChrystal 2000; Lambert et al. 2007), and so for the present anal-
ysis we look at this entire range in relation to South Asian network ties.
Postalveolar /t/ was selected for several reasons. First, it does not occur as a
variant of /t/ in indigenous BrE varieties and so is clearly an exogenous element
in a British context, with a consistent indexical value linked to Asianness. Sec-
ond, the feature is highly salient in the community (Rampton 1995; Alam 2007;
Lambert et al. 2007; Chun 2007), and is cited by many participants as typical
of Asian speech, so is above the level of consciousness to some degree. This not
only facilitates reliable auditory coding, but can also make a feature more readily
adopted or discarded (Trudgill 1986). Third, the feature has been studied in pre-
vious work on British Asian communities, but only with respect to younger UK-
born speakers; the present work can help to contextualize those findings within
change over time in the community.
Articulation of /t/ was coded in three positions: syllable-initial, word-medial,
and word-final. A minimum of 50 instances (maximum 100, where possible) of
/t/ were coded for each speaker in each of the 3 positions, so each speaker was
coded for 150–300 tokens. No more than 5 tokens per type were coded. The var-
iants for syllable-initial /t/ were [t] and [ʈ], where the latter IPA symbol includes
a broad phonetic range. Five variants were coded word-medially and word-final-
ly: [t], [ʈ], [ɾ], ø, [ʔ]. Internal (linguistic) factors were coded differently for each
context, due to different potential conditioning factors in different positions. In
7. In order to check the reliability of our auditory coding, 5% of the data were coded blind by
both coders, resulting in an inter-rater reliability of 90%.
228 Devyani Sharma
the earlier study, these internal factors and a range of social factors (excluding
transnational effects) were examined; see Sharma and Sankaran (2011) for details.
Here, only overall rates of use of the Asian variant are considered.
4. Transnational activity in the present community
Let us first set aside language use and simply examine patterns of transnational
activity among the four main demographic sub-groups of the second generation
of British Asians in Southall: older men, older women, younger men, and younger
women. It is important to remember that all groups are second generation: they
were all raised in the UK by first-generation parents.
Tables 10.1–10.4 present the four dimensions of transnational activity out-
lined in Section 3 for each individual under analysis. The final column provides
the cumulative transnational index value for each person.8
The patterns of transnational activity within Gen 2 in Tables 10.1–10.4 illus-
trate a number of systematic changes over time.
Table 10.1â•‡ Transnational activity among older Gen 2 men

Name Visits to India Communication Spouse/ India-linked Transnational
with India partner work index
Anwar 1x per year regular for work South Asian yes 5
Naseem 1x per year regular for work South Asian yes 5
Sharan 3x per year regular for work Non-Asian yes 4
Satpal 3 little British Asian no 0
Table 10.2â•‡ Transnational activity among older Gen 2 women

Rani every 1–2 years little South Asian yes 4
Nimmi 1–4x per year often British Asian yes 4
Amrita 1–2x per year moderate Unmarried no 3
Jaipreet every 2–3 yrs little South Asian no 2
Isha every year till 1995 little British Asian no 1
Simran 1 visit since 11 little British Asian no 0
8. All names are anonymised.

Table 10.3â•‡ Transnational activity among younger Gen 2 men

Ravinder every 2 years moderate None no 1
(one 4-year stay)
Anand every 2 years little None no 1
Vikas every 4–5 years little British Asian no 1
Prashant 3 moderate British Asian no 0
Sameer 1 none British Asian no 0
Kirpal 1 little None no 0
Rohan 3 none None no 0
Table 10.4â•‡ Transnational activity among younger Gen 2 women

Deepti 4 none Non-Asian earlier yes, 2
(one 2-year stay) now no
Guddi 6 little British Asian no 1
Renu 6 none None no 1
Rita 5 little British Asian no 1
Preeti 1 little Non-Asian no 0
Namrita 0 little Non-Asian no 0
A decrease in transnational travel over time is noticeable. Older Gen 2 men

and women favour annual or near-annual visits. By contrast, many younger Gen 2
men and women have only visited a couple of times in their life. Three younger
Gen 2 individuals have visited only once and one has never visited. Even taking
into account the difference in age between the older and younger groups, travel is
less frequent in the younger group.
Transnational travel could be class-linked, as it is costly to travel between the
UK and Asia. Sameer and Simran mention cost as a constraint. However, Amy,
who is also working class and among the poorest of the participants in the project,
spent much of her earnings on near-annual travel, in keeping with the typical
practice of her generation. Furthermore, social mobility means that younger in-
dividuals such as Namrita and Preeti have greater financial resources than some
others, yet rarely travel to India. So a generational shift in social practices is prob-
ably a stronger factor in the decline of transnational travel than income per se.
A preference for ethnically Asian partners is fairly consistent across all groups,
although we see a clear shift from South Asian to British Asian partners with
230 Devyani Sharma
age. We also see that younger Gen 2 women are the only group in which several
individuals have non-Asian partners (this supports Sharma’s 2011 discussion of
gendered changes in community orientation and network types). A change that
underlies these shifts is a moderate decrease in arranged marriage practices. All
the younger Gen 2 individuals are in their twenties and unmarried, with the ex-
ception of Deepti, who was 35 and divorced. In all younger Gen 2 cases, arranged
marriage was still a possibility, but almost certainly with a British Asian partner.
By contrast, a common practice among parents of older Gen 2 individuals was to
find a South Asian partner. This, combined with a shift to more nuclear family
structures and practices, is another source of decreased transnational communi-
cation among the younger Gen 2 group.
Finally, systematic changes in employment patterns also influence transna-
tional activity. Older Gen 2 men typically went into their father’s businesses (in
Table 10.1, all did; however, Satpal was unhappy and changed careers). This meant
in all cases a high level of travel to Asia for work-related business and dense Asian
ties. By the younger Gen 2, almost no individuals went into their parents’ busi-
nesses, with correspondingly little work-related travel to Asia.
The transnational index in Tables 10.1 and 10.2 show that the older groups
exhibit the greatest amount of diversity in transnational activity, including some
of the highest rates. The index values in Tables 10.3 and 10.4, in contrast, show
lower transnational activity and lower variation in the younger groups.
Recall that all four groups presented here are second-generation individuals,
born to first-generation migrant parents. So the clear decline from high to low
values does not correspond to migration stage in generational terms, but rath-
er to changing social practices over time within the second generation in this
community.
In the next section, I evaluate two aspects of this measure of transnational
activity: Does an individual’s level of transnational activity correlate with their use
of postalveolar /t/? And if so, is this effect steady over time?
5. Transnational activity and language variation
Tables 10.5–10.8 repeat the transnational index value for each individual, along-
side their rate of use of postalveolar /t/ variants.
I do not present statistical measures for the data for two reasons. First, space
limitations prevent a full consideration of other social factors, most of which
(aside from transnational activity) were examined in the detailed regression anal-
ysis presented in Sharma and Sankaran (2011). Second, even though Spearman’s
correlation coefficient is appropriate for small sample sizes of below 20 (Neave
Table 10.5â•‡ Transnational activity and language use (older Gen 2 men)
Name Transnational index % retracted /t/ in interview
Anwar 5 37.8
Naseem 5 14.9
Sharan 4 â•⁄0.9
Satpal 0 â•⁄1.1
Table 10.6â•‡ Transnational activity and language use (older Gen 2 women)
Rani 4 27.6
Nimmi 4 13.7
Amrita 3 31.3
Jaipreet 2 â•⁄6.0
Isha 1 â•⁄9.4
Simran 0 10.5
Table 10.7â•‡ Transnational activity and language use (younger Gen 2 men)
Ravinder 1 â•⁄9.2
Anand 1 19.5
Vikas 1 â•⁄0.0
Prashant 0 16.7
Sameer 0 20.8
Kirpal 0 22.0
Rohan 0 27.3
Table 10.8â•‡ Transnational activity and language use (younger Gen 2 women)
Deepti 2 0.0
Guddi 1 2.1
Renu 1 1.3
Rita 1 1.0
Preeti 0 2.0
Namrita 0 1.0
232 Devyani Sharma
& Worthington 1988), breaking the groups down by age as well as gender cre-
ates groups towards the lower bound of even correlation analysis. Fortunately, the
groups are small enough to permit an informal visual examination of the predict-
ed correlations in the vertical dimension, namely that numbers declining in the
transnational index column will correlate with a similar decline in the phonetic
variation column.
In Table 10.5 – older Gen 2 men – we see a close correspondence between /t/
retraction and individuals’ degree of transnational activity. Older Gen 2 wom-
en (Table 10.6) similarly show an observable correlation between the two. Two
apparent exceptions to the pattern are Sharan in Table 10.5 and Nimmi in Ta-
ble 10.6, who both show low rates of /t/ retraction despite high transnational in-
dices. In both cases, these individuals are less exceptional than they seem, as both
in fact do show robust (nearly bidialectal) shifts to use of retracted /t/ and oth-
er Indian English variants in non-interview settings with certain kinds of Asian
interlocutors.
When we turn to younger groups in Tables 10.7 and 10.8, we see two differ-
ences from the older groups. First, as we know, levels of transnational activity
have dropped by this generation. The second difference is notable: Even taking
this decline into account, we can see that the little range of transnational activity
that remains does not correlate at all with use of retracted variants of /t/.
It would be conceivable that even with a decline in transnational activity, the
remaining range of activity (e.g. ranging from ‘2’ to ‘0’ in Table 10.8) would still
show a correlation with the accent feature, just corresponding to slightly lower
levels of use. However, Table 10.8 shows no correlation with use of postalveolar
variants of /t/, which are uniformly low in this group and so do not correlate with
transnational flows. In Table 10.7, younger Gen 2 men show a similar, but more
dramatic lack of correlation. Young men have, overall, a much higher average use
of postalveolar /t/ than the women in Table 10.8, but the three highest users are
the three men with the lowest transnational index values.
In sum, two distinct patterns of decline are found: (i) the level of transnational
activity and (ii) the degree to which transnational activity, if present, influences
language variation. Even when there is a strong correlation, it is important to re-
member that correlation need not imply causation. More likely, a linked cluster of
social and linguistic practices weakens over time.
This may occur due to the generational changes in social practices described
earlier. At least half of the older group exhibit or report bidialectal ability, arising
from early and heavily transnational social practices, whereas none of the younger
group do. This ability naturally makes a wider range of variables available, and
subject to, finer control in the older group and, as noted elsewhere (Sharma &
Rampton, forthcoming), this group engages continually in complex ethnopolitical
and bicultural positioning in discourse. By the time we reach the later phase of
the community, i.e. the younger group, dialect use has focused further such that
younger speakers, especially men, use their Asian-influenced variety as a ‘default’
and vary much less according to interlocutor and situational effects (Sharma &
Sankaran 2011; Sharma 2011; Sharma & Rampton, forthcoming). By this stage,
Asian-derived forms may have developed much more local, rather than transna-
tional, meanings, causing the loss of correlation. This mirrors the findings of Alam
and Stuart-Smith (this volume) and Mesthrie and Chevalier (this volume), both of
which show the emergence of very local meanings for originally heritage forms. It
may also be the case that when transnational activity drops below a certain thresh-
old, it is no longer able to have as direct an impact on an individual’s repertoire.
From the point of view of wider dynamics of diaspora and comparative anal-
ysis, it is worth noting that this new Asian style among young British Asians, fre-
quently commented on in British public discourse, is not one that has necessarily
developed across all Indian diaspora communities globally. For instance, second
and third generation Asian Americans appear not to have developed as clearly
distinctive an Asian speech style. These global divergences may point to deeper
differences in how diaspora communities come to be embedded within systems of
class, ethnicity, and social mobility in different receiving cultures.
A final question that arises from this analysis is whether we can generalise
from just one accent feature to more general age differences in language use in the
community. Sharma (2011) provides some support that the patterns presented
here may not be restricted to postalveolar /t/ alone: that study shows that retract-
ed /t/ clusters with other Asian phonetic features in the repertoires of these four
age/gender groups in the second generation. Sharma (2013) indicates that Punja-
bi use may also parallel the English accent findings presented here to some extent,
i.e. it corresponds more closely to transnational activity for older second-gener-
ation individuals.
So are ‘real’ face-to-face ties the only important mechanism for the transmis-
sion of language practices (e.g. Trudgill 2004; 2008)? Or is it possible that ideolo-
gies about speech variants and speech varieties influence language practices too,
even at extraordinary distances? The final section turns to this question.
6. Transnational ideologies and language variation
In recent decades, India’s status as a global presence has transformed. Although

massive crises of development continue, the simple association of India with pov-
erty in the West has shifted steadily with its economic growth, following mar-
ket-based reforms and economic liberalisation after 1991. With its new place in
234 Devyani Sharma
the global economy as part of the BRICS group of rapidly developing large econ-
omies, India has come to be equally associated with economic growth and com-
petitive performance in numerous sectors, particularly information technology,
telecommunications, industrial sectors, and services.
Nowhere has this transformation been more socially tangible than in the
United Kingdom. Spectacular displays of wealth have become associated with In-
dians. Lakshmi Mittal, an Indian national and CEO of the largest steel company
in the world, was ranked the richest man in the UK in 2007; his home in Central
London was ranked the world’s most expensive residence in 2004. Jaguar Land
Rover, an iconic British automotive firm, was acquired and subsequently massive-
ly expanded by an Indian firm in 2008. In 2013, The Economist (July 27) reported
on the substantial and visible presence of India’s super-rich elite in London’s most
exclusive neighbourhood, Mayfair. This includes professionals, industrialists, film
stars, and sports stars, all of whom have long favoured London as an overseas des-
tination. And all of these groups are familiar with Southall, an established cultural
reference point for the London South Asian diaspora.
This highly visible and influential group often does not involve permanent
migration. They have second homes in London or live in London for limited peri-
ods of time, practicing what Harris, Leung and Rampton (2002) term ‘sojourning’,
following Cohen (1997):
new classes of people educated in a whole range of modern skills are now pre-
pared to migrate or re-migrate and respond to the pull of centres of power and
wealth and the new opportunities in trade and industry… these people are artic-
ulate, politically sensitive and choose their new homes carefully…
(Cohen 1997, 164–165, cited by Harris et al. 2002, 36)
Harris et al. describe this as a prime example of “how globalisation encourages a

reconceptualisation of ethnic relations, both in England and elsewhere” (p. 36).
Can we see evidence of such a reconceptualisation within language ideolo-
gies, even language practices, in diasporic groups? Has this increased presence in
London of high status South Asians and their locally (in South Asia) high status
variety of IndE, associated with wealth and transnational power, affected the per-
ception and use of Indian forms among British Asians?
In this section, I examine related themes in the content of the recorded inter-
views with second-generation Southall community members: perceptions of the
status of India, Indians, and IndE. In closing, I indicate briefly how these affilia-
tions and orientations may affect real linguistic practices – in particular, invento-
ries of variation – that we find in the group.
All participants were asked what they thought the perception of India was
in the UK. Without exception, all second generation individuals reported either
positive or improved perceptions of India, citing association such as information
technology, becoming a global power, Lakshmi Mittal, and Bollywood. Prashant,
one of the younger second-generation men, simply said: “Gore [white people] like
India now.”9
Participants did convey much more nuanced ideologies and assessments of
attitudes to South Asians in other parts of their interviews, such as comments
on the perception or behaviour of “freshies” (recent arrivals) or “pehndus”
(villagers), both terms used widely with pejorative meaning. See, for instance,
Rampton’s (2011; 2013) discussion of complexities and tensions surrounding
‘freshie’ in the present data set. However, characterisations of these categories
within the community were often not monolithic in their negative valuation, and
tended to be informed by personal experiences and positionings. Among the sec-
ond generation, we see surprisingly few seriously derogatory uses of these terms,
with participants more often using the terms to address the experiences or per-
ceptions of such groups. More than one participant expressed covert prestige af-
filiations with the category of “freshie”, as in (1):
(1) Rita, a 19-year-old second-generation woman, is commenting on what “freshie”
means. Basma is a close friend, also a second-generation British Asian, present
during the interview. Romi is Rita’s boyfriend, also second generation; he was
not present.
Rita: He [her brother] probably is the- probably the most Indian Indian you
can come across. He’s under- he calls himself undercover freshie. I call Basma
a freshie sometimes. Because um she’s just very freshie. Hehe. She’s freshie
because she’s fresh. No just because um Basma’s very traditional, as in she
will have her headscarf and all that freshiness. So she’s very freshie compared
to me. I can be freshie. For example Romi. HE’s freshie. He’s so freshie, it’s
9. A playful element of the interviews included conducting the infamous ‘Tebbit test’ or ‘crick-
et test’. In 1990, a Conservative politician, Norman Tebbit, suggested in an interview that Brit-
ons of South Asian heritage who do not express support for the England cricket team are not
culturally integrated. Plenty of politicians and members of the British public have noted that
this view reveals a worrying ignorance of identity, history, migration, even sport. Out of over
70 participants in the wider project, spanning three generations, only 5 (all younger second
generation individuals) claimed that they would support England over India. Given that the
sample of British Asians all expressed moderate and broadly integrationist attitudes (e.g. uni-
formly believing that immigrants should learn English), all identified with being British, and
included upwardly mobile professionals and successful businesspeople, the results both sup-
port the positive perceptions of India in this group and the baselessness of simplistic cultural
measures of integration such as Tebbit’s.
236 Devyani Sharma
unbelievable. He- that boy sings Punjabi songs like no man’s business. And he
always, whenever he talks to me, he will talk to me in Punjabi. And um I’m
like ‘babe i understand English’ he’s like °no you don’t°, right we’re Indians,
right, stick to it… What else does he do? Um when he’s on the phone to his
friends.. his- his er tone of his voice suddenly becomes so freshie it’s unbeliev-
able. his accent- it’s just like- he’s talking in English, but he’s still like got that
freshiness in him. like that accent… Like Basma over here walks in with her
traditional salwar kamiz, that’s why I call her a freshie. And she’s proud of it.
I on the other hand need [phone rings] need an occasion [answers phone].
Rita’s description of the term ‘freshie’ here builds a cluster of associations with
social practices and language choices. She links two language features – Asian
English accent features and the use of Punjabi – with freshiness. Yet Rita and her
friends and siblings exhibit a strong ideological affinity with Punjabi “freshiness”,
if to varying degrees. Rita uses Punjabi in teasing side-interactions with Basma
elsewhere in the interview, supporting her assertion that these language forms
still have positive social value within younger second-generation interactions, if
of a more local type.
Deriving from these broadly positive associations with Indianness are more
specific evaluations of IndE in particular. One set of individuals, from a single
family, fondly describe the IndE spoken by first generation migrants with lower
English ability as ‘bud-bud’, specifically enregistering retroflexion in the /d/.
(2) a. Naseem (older Gen 2 man): The younger generation, yeah, they say that
to eh you know he came he had a bud bud accent you know (xx). I think
it’s nice I think it’s very very attractive bud bud accent you know. It’s a
lovely warm accent.
b. Naseem: But then when I’m speaking to a person… you know he’s not-
English is not his first language, then you tend to, I mean (xxx) my my
daughter was saying to me ‘you spoke to that person like bud-bud’ you
know. You know, but I- because that’s the response I was getting.
c. Anwar (older Gen 2 man): My son says ‘Dad, why do you speak bud-bud
English?’ I said, ‘yaar if I speak properly to this guy… I need to get closer
to him.’10
These older second-generation men cast the younger generation as policing this
boundary between ‘bud-bud’ and more acceptable Englishes – which complicates
the rosy picture painted by Rita in (1) – but they themselves readily express fond-
ness for this speech and both describe convergent accommodation in (2b) and
(2c) (see Sharma 2011 for evidence of active use of this lectal range by older men).
10. Yaar: Hindi ‘friend’; functions similarly to the English discourse marker ‘man’.
Interestingly, participants seem very aware that ‘bud-bud’ is not the only kind
of IndE, and implicitly or explicitly distinguish it from higher status IndE. Na-
seem in (2b) describes ‘bud-bud’ as the English of a second language speaker, in
stark contrast to the educated IndE described, for instance, by Namrita in (3).
(3) Namrita (young Gen 2 woman): I think people who’ve been educated e- er- in
English b-back- from back home, their English, their English is grammatically
better. Umm but then it’s their accent or their knowing of you know, like,
obviously if you go to America certain things and culture, culturally speaking
how you use words and language in- and obviously that differs. So you can’t
share a joke in the same way as you might share it here you know, that kind
of difference.
Notice Namrita’s ready comparison of educated IndE to stable varieties such as

American English (AmE), fully accepting its independent norms and not casting
this type of IndE as a learner variety. Indeed, Namrita articulates a widespread
ideology among diasporic Indians of educated IndE as ‘grammatically better’ than
British and American varieties (see Sharma 2005 for similar ideologies among
Indians in the United States).
How does this awareness of an educated, ‘high’ variety of IndE correspond to
usage? Descriptions of this variety are in several cases accompanied by acknowl-
edgements of style-shifting to accommodate to it, as in (4).
(4) a. Nimmi (older Gen 2 woman): Yeah so when you’re when you’re in
India you talk completely different to when you are here. You still speak
English, and their command of the English language is far better than the
command of the English language in this country [the UK]. Far better.
But obviously they have an accent and they have a dialect, you know,
which they use. And you and you pick that up when you are there.
b. Shilpa (younger Gen 2 woman): (Interviewer: when you speak to some-
one from like India do speak to them in English with an Indian accent
maybe? Or do you think that you speak to them like you normally speak,
like to me?) No it automatically changes. Like if you- when I was in India
and my cousin-sister, she would speak English with me? Because she
speaks respectful, you think ‘Come on Shilpa, you know, don’t speak the
slang words because, you know, you have to be good you know show
them you know, yeah, you learnt proper you know’! Because, yeah,
because they know proper English, like from the book yeah. What they
read yeah. And we would like change words? It’s really bad English but
I’ll be like try to be decent respectful English you know. (Do you speak
like with an Indian accent with her?) Yeah with a Indian accent you know,
how they speak. Like American, when they speak English, they speak
English like American American accent?
238 Devyani Sharma
Note that both Nimmi and Shilpa repeat Namrita’s view that educated IndE is
“better”, in a prescriptive sense, than BrE, and Shilpa repeats Namrita’s compari-
son to AmE, rather than to learner varieties, to explain the nature of differences
between the two varieties. The overt prestige and stable norms accorded to edu-
cated IndE seem very clear.
From the point of view of the impact of this ideology on actual use, (4a) and
(4b) show both older and younger women suggesting that they accommodate to
this variety. Sharma (2011) showed substantial style-shifting among older men
towards this variety in interactions as well.
Most strikingly, older middle-class transnational British Asians, e.g. Anwar,
Naseem, Nimmi, and Sharan, who have had sustained transnational exposure and
ties to educated IndE, show evidence of this influence in the kinds of Asian traits
they themselves choose to use. In their own British Asian styles, these individuals
show an absence or very selective use of marked Punjabi and typically learner
traits and a clear retention of more ‘high status’ IndE features. Sharma and Tusha
(2012) show that low stigma IndE forms such as quotative choice or absence of
definite article allophony, an established feature of educated IndE, are much more
common in second generation usage than typical learner forms such as absence
of indefinite article allomorphy or variable subject agreement.
7. Conclusions
In this study, the impact of transnational relationships in a diasporic Punjabi com-

munity in West London has been examined from the perspective of language use
and language ideology, using quantitative and qualitative analysis respectively.
A pattern of decreasing transnational ties was observed from older to young-
er second generation British Asians. Importantly, an independent pattern showed
that the correlation between transnational activity and the use of retracted /t/
disappeared in the younger group, with the Asian English phonetic trait being
retained but losing its clear link to transnational ties as was found among older
individuals.
This may suggest either that below a certain threshold (as younger individuals
all have less dense and multiplex ties to India), transnational ties lose their im-
pact, or that with the passage of time, more local meanings come to be associated
with forms, and so their use is driven by local rather than transnational indexical
potential. This would likely correlate with shifts in phonetic quality as well (Alam
& Stuart-Smith, this volume; Kirkham 2011).
In the analysis of ideological orientations, evidence of overt prestige for ed-
ucated or Standard IndE emerged in both age groups, with some indication that
these ideological alignments have an active effect on selective use of certain Asian
traits more than others, particularly among those (mainly older individuals) who
have sufficient contact with and command of IndE to draw such fine distinctions.
These ideologies and imaginings of Indians and their speech are notable for
the ways in which participants, even younger British Asians, find myriad points
of positive affiliation. Even with diminishing personal ties to South Asia, sec-
ond-generation discourses, ideologies, and activities are infused with positive
affiliations, and the variety is not simply reserved for voicing cultural distance or
mocking ‘parent’ or outsider positions. The findings also support claims of the
emergence of IndE as an increasingly stable global, or at least ‘epicentre’, variety
(Hoffmann, Hundt & Mukherjee 2011).
Acknowledgements
The generous support of the UK Economic and Social Research Council (ESRC
Standard Grant RES-062-23-06-04) is gratefully acknowledged. I am also indebt-
ed to Marianne Hundt, Ben Rampton, and an anonymous reviewer for useful
feedback on this chapter.
References
Alam, Farhana. 2007. Language and identity in ‘Glaswasian’ adolescents. MLitt dissertation.
University of Glasgow.
Anderson, Benedict. 1991. Imagined Communities: Reflections on the Origin and Spread of Na-
tionalism, rev. edn. New York NY: Verso.
Bell, Allan. 2001. Back in style: Reworking audience design. In Style and Sociolinguistic Varia-
tion, Penelope Eckert & John Rickford (eds), 139–169. Cambridge: CUP.
Bhatia, Tej K. 1993. Punjabi: A Cognitive-descriptive Grammar. London: Routledge.
Bhatt, Rakesh. 2010. Unraveling post-colonial identity through language. In The Handbook of
Language and Globalization, Nicholas Coupland (ed.), 520–539. Oxford: Blackwell.
DOI: 10.1002/9781444324068.ch23
CARF (Campaign against Racism and Fascism). 1981. Southall: The Birth of a Black Communi-
ty. London: Institute of Race Relations.
Chun, Elaine. 2007. The Meaning of Mocking: Stylizations of Asians and Preps at a U.S. High
School. PhD dissertation, University of Texas, Austin.
Cohen, Robin. 1997. Global Diasporas: An Introduction. London: UCL Press.
DOI: 10.4324/9780203228920
DMAG (Data Management and Analysis Group, Greater London Authority) Briefing, 2006.
<www.london.gov.uk> ( 3 November 2010).
Ealing JSNA (Joint Strategic Needs Assessment) report. 2010. <www.ealingpct.nhs.uk/Publica-
tions/needs-assessment.asp> ( 3 November 2010).
240 Devyani Sharma
Gupta, Akhil & Ferguson, James. 1992. Beyond culture: Space, identity and the politics of dif-
ference. Cultural Anthropology 7(1): 6–23. DOI: 10.1525/can.1992.7.1.02a00020
Harris, Roxy. 1996. Openings, absences and omissions: Aspects of the treatment of race, culture
and ethnicity in British cultural studies. Cultural Studies 10(2): 334–344.
DOI: 10.1080/09502389600490201
Harris, Roxy, Leung, Constant & Rampton, Ben. 2002. Globalisation, diaspora and language
education in England. In Globalisation and Language Teaching, David Block & Deborah
Cameron (eds), 29–46. London: Routledge.
Heller, Monica. 2007. Multilingualism and transnationalism. In Handbook of Multilingualism
and Multilingual Communication, Peter Auer & Li Wei (eds), 539–553. Berlin: Mouton de
Gruyter.
Herbert, Joanna. 2009. Oral histories of the Ugandan Asians in Britain: Gendered identities in
the diaspora. Contemporary South Asia 17(1): 21–32. DOI: 10.1080/09584930802624646
Heselwood, Barry & McChrystal, Louise. 2000. Gender, accent featured and voicing in Panja-
bi-English bilingual children. Leeds Working Papers in Linguistics and Phonetics 8: 45–70.
Hoffman, Michol & Walker, James. 2010. Ethnolects and the city: Ethnic orientation and lin-
guistic variation in Toronto English. Language Variation and Change 22: 37–67.
DOI: 10.1017/S0954394509990238
Hoffmann, Sebastian, Hundt, Marianne & Mukherjee, Joybrato. 2011. Indian English – An
emerging epicentre? A pilot study on light verbs in web-derived corpora of South Asian
Englishes. Anglia 129(3–4): 258–280. DOI: 10.1515/angl.2011.083
Kirkham, Sam. 2011. The acoustics of coronal stops in British Asian English. In Proceedings of
the 17th International Congress of Phonetics Sciences (ICPhS).
Lambert, Kirsten, Alam, Farhana & Stuart-Smith, Jane. 2007. Investigating British Asian ac-
cents: Studies from Glasgow. In Proceedings of the 16th International Congress of Phonetics
Sciences (ICPhS), Jürgen Trouvain & William Barry (eds), 1509–1511. Saarbrücken.
Le Page, Robert & Tabouret-Keller, Andrée. 1985. Acts of Identity: Creole-based Approaches to
Language and Ethnicity. Cambridge: CUP.
Levitt, Peggy & Waters, Mary C. 2002. The Changing Face of Home: The Transnational Lives of
the Second Generation. New York NY: Russell Sage Foundation.
Li, Wei. 1994. Three Generations, Two Languages, One Family: Language Choice and Language
Shift in a Chinese Community in Britain. Clevedon: Multilingual Matters.
Llamas, Carmen. 2006. Shifting identities and orientations in a border town. In The Sociolin-
guistics of Identity, Tope Omoniyi & Goodith White (eds), 92–112. London: Continuum.
Masica, Colin. 1993. The Indo-Aryan Languages. Cambridge: CUP.
Matsumoto, Kazuko & Britain, David. 2009. The role of social networks in understanding lan-
guage maintenance and shift in post-colonial multilingual communities – The case of the
republic of Palau in the Western Pacific. Essex Research Reports in Linguistics 58: 2.
Meads, Richard J. 1983. Southall 830–1982. London: Merlin.
Mendoza-Denton, Norma. 2010. Individuals and communities. In The Sage Handbook of Socio-
linguistics, Ruth Wodak, Barbara Johnstone & Paul Kerswill (eds), 181–191. London: Sage.
Meyerhoff, Miriam & Walker, James A. 2007. The persistence of variation in individual gram-
mars: Copula absence in ‘urban sojourners’ and their stay-at-home peers, Bequia (St. Vin-
cent and the Grenadines). Journal of Sociolinguistics 11(3): 346–366.
DOI: 10.1111/j.1467-9841.2007.00327.x
Milardo, Robert M. 1988. Families and social networks: An overview of theory and methodol-
ogy. In Families and Social Networks, Robert M. Milardo (ed.), 13–47. Newbury Park CA:
Sage.
Milroy, Lesley & Wei, Li. 1995. A social network approach to code-switching. In One Speaker,
Two Languages: Cross-disciplinary Perspectives on Code-switching, Lesley Milroy & Pieter
Muysken (eds), 136-157. Cambridge: CUP. DOI: 10.1017/CBO9780511620867.007
Moffatt, Suzanne & Milroy, Lesley. 1992. Panjabi/English language alternation in the early
school years. Multilingua 11(4): 355–385.
Neave, Henry R. & Worthington, P.L. 1988. Distribution-Free Tests. London: Unwin Hyman.
Oates, Jonathan. 2002. Southall and Hanwell. London: The History Press.
Omoniyi, Tope. 2004. The Sociolinguistics of Borderland: Two Nations, One Community.
Trenton NJ: Africa World Press.
Pavlenko, Aneta & Norton, Bonny. 2007. Imagined communities, identity, and English lan-
guage learning. In Kluwer Handbook of English Language Teaching, Jim Cummins & Chris
Davison (eds), 669–680. Dordrecht: Springer. DOI: 10.1007/978-0-387-46301-8_43
Pingali Sailaja. 2009. Indian English. Edinburgh: EUP.
Rampton, Ben. 1995. Crossing: Language and Ethnicity Among Adolescents. London: Longman.
Rampton, Ben. 2011. From ‘multi-ethnic adolescent heteroglossia’ to ‘contemporary urban ver-
naculars’. Language & Communication 31(4): 276–294.
DOI: 10.1016/j.langcom.2011.01.001
Rampton, Ben. 2013. Styling in a language learned later in life. Modern Language Journal 97(2):
360–382. DOI: 10.1111/j.1540-4781.2013.12010.x
Shankar, Shalini. 2008. Desi Land: Teen Culture, Class, and Success in Silicon Valley. Durham
NC: Duke University Press. DOI: 10.1215/9780822389231
Sharma, Devyani. 2005. Dialect stabilization and speaker awareness in non-native varieties of
English. Journal of Sociolinguistics 9(2): 194–225.
Sharma, Devyani. 2012. Multi-scale network effects in British Asian English. Paper presented at
NWAV 42, Pittsburgh, USA, October.
Sharma, Devyani. 2013. Multi-scale network effects: A case study of British Asian English. Pres-
entation at ICLaVE 7, Trondheim, Norway.
Sharma, Devyani & Rampton, Ben. Forthcoming. Lectal focusing in interaction: A new meth-
odology for the study of style variation. Journal of English Linguistics.
Sharma, Devyani & Sankaran, Lavanya. 2011. Cognitive and social factors in dialect shift:
Gradual change in London Asian Speech. Language Variation and Change 23: 399–428.
DOI: 10.1017/S0954394511000159
Sharma, Devyani & Tusha, Aurela. 2012. Abrupt vs. gradual change in immigrant groups:
Evidence from morphosyntax. Paper presented at Sociolinguistics Symposium 19, Berlin,
Germany.
Trudgill, Peter. 1986. Dialects in Contact. Oxford: Blackwell.
Trudgill, Peter. 2004. New-dialect Formation: The Inevitability of Colonial Englishes. Edinburgh:
EUP.
Trudgill, Peter. 2008. Colonial dialect contact in the history of European languages: On the
irrelevance of identity to new-dialect formation. Language in Society 37: 241–254.
242 Devyani Sharma
Vertovec, Steven. 2004. Trends and impacts of transnationalism. Centre on Migration, Policy
and Society Working Paper No. 3, University of Oxford.
Vertovec, Steven. 2005. The political importance of diasporas. Centre on Migration, Policy and
Society Working Paper No. 13, University of Oxford.
Vertovec, Steven. 2009. Transnationalism. London: Routledge.
Wong, Amy. 2007. Two vernacular features in the English of four American-born Chinese.
University of Pennsylvania Working Papers in Linguistics 13(2): article 17.
Index
/h/â•… 88, 93 double diasporaâ•… 57, 79–80, 181 grammaticality judgementâ•… 126
/v/â•… 88, 93 dressâ•… 150, 172, 178, 182–183 Gujerati (Gujarati)â•… 30, 52,
/w/â•… 44, 47, 88, 93 dual identityâ•… 138 59, 61–64, 74–75, 79, 81, 83,
89, 92, 108, 113, 136, 142, 173,
A E 176–178, 180–184, 193, 221
adolescenceâ•…155 East African English (EAfE)â•…
aspirationâ•… 49–50, 85, 88, 91–92, 57–58, 62–63, 65, 74, 76, 78, 82 H
102–103 East African Indianâ•… 55, 57–58, habitualâ•… 115–116, 125
attitude (attitudes, attitudinal) â•… 82 Hindiâ•… 1, 7, 10, 12, 25, 108, 113,
2, 4–8, 25–26, 64, 76, 86, 118, East Midlands English (EMidE)â•… 120–121, 124–126, 133–134, 139,
135, 139, 157, 160, 162, 164, 57–58, 62–63, 65, 74, 76–80 141–144, 151, 153, 158, 161–163,
187–189, 190, 192–200, 202, endonormative stabilizationâ•… 2, 166, 173, 176, 178, 180–181, 183,
205, 207–213, 220, 235 5, 187–188, 196, 202, 207, 209 185, 192–195, 201–204, 212,
ethnicityâ•… 3, 7, 9–11, 13–27, 29– 226, 236
B 30, 40, 48, 51–52, 82, 104, 109, hybridisationâ•… 138, 178
Bhojpuriâ•… 1, 10–12, 24–26, 113–114, 137, 196, 202–204, 211, hybridityâ•… 34, 50, 53, 83, 171, 179
173–177, 181, 183, 185–186, 192 219, 221, 233, 240–241
borrowingâ•… 4, 35, 136, 171, 178, ethnographyâ•… 29, 31, 37–38, 211 I
183 identity constructionâ•… 4, 6, 138,
British Asianâ•… 26, 30, 49, 52–53, F 160, 162, 164, 176, 188, 192
55–56, 61, 77–82, 104, 215, 222, F0 (fundamental frequency)â•… ideologyâ•… 6, 215–218, 237–238
225, 227, 229–230, 235, 238, 9–10, 13–23 imperfectiveâ•… 114–121, 124–126
240–241 face-to-face interactionâ•… 218 indenture (indentured,
Fiji Hindiâ•… 139, 141–142, 144, indentureship)â•… 7, 10, 24,
C 161–162, 166, 192–194, 201– 58–59, 89, 103, 108, 136, 139,
Communities of Practiceâ•… 3, 26, 202, 204, 212 171, 173–177, 182, 184–185, 188,
29, 31, 36–38, 51, 220 first generation(s) (first- 190–192, 195, 212
consonantsâ•… 29, 36, 44, 88, generation(s))â•… 3–4, 30, 32, Indian English (IndE)â•… 2–3, 5,
90–91, 226 35, 55–58, 61–68, 70–80, 95, 7–8, 35, 55, 57–58, 62–63, 65,
cuisineâ•… 172, 178 131–133, 135, 143, 145, 154–161, 74–83, 85, 90, 92, 103–104,
164, 170, 221, 226, 228, 230, 106, 115, 119–120, 124–126,
D 236 131–133, 136, 142, 146, 149–150,
diaspora consciousnessâ•… 138 160, 165–166, 171–172, 174, 177,
Dictionary of South African G 180, 182–183, 185–187, 195, 209,
Indian Englishâ•… 171, 185–186 genderâ•… 4, 6, 14–15, 17, 24, 26, 215–217, 223–226, 232, 234,
direct exchanges (direct)â•… 41, 30, 52, 65, 70–73, 75, 77, 81, 236–241
49, 57, 60–61, 73, 78–80, 127, 85, 88, 92, 95, 99–103, 118, 121, indirect exchanges (indirect)â•…
135, 179, 187, 190, 196–198, 218, 134, 182, 232–233, 240 210, 225–226
225, 233 generationâ•… see first generation Indo-Fijian (Fiji Indian)â•… 62, 74,
domain (domains)â•… 6, 12, 34, 39, and second generation 131–132, 135–139, 141–142, 149,
76, 111, 171, 193, 196–203, 205, girmit (girmitiyas)â•… 136, 150, 154, 160–161, 164, 187–195,
207, 210–211 174–175, 185, 191–193 199–203, 205, 207–210
244 English in the Indian Diaspora
interdental fricative(s)â•… 85, 92 P Singaporeâ•… 2, 4, 7, 105–111,

interspeaker variationâ•… 14, 16, Pakistaniâ•… 3, 29–34, 36–38, 46, 113–116, 119–121, 124–130, 166,
18, 20 49, 51, 53, 56, 220–221 173, 213
pan-Indian identityâ•… 110, Singlishâ•…111
K 138–140 sociophoneticsâ•… 29, 53, 55,
kinshipâ•… 12, 171, 178, 181–182 perceptionâ•… 9, 13–15, 22–23, 26, 85–86, 95–96, 104
koinéâ•… 141, 192, 201 40, 52, 193, 234–235 South African Indian English
phonationâ•… 9, 13–14, 16, 18–26 (SAIE)â•… 2–4, 7–8, 62, 74,
L phoneticsâ•… 2, 25, 50–53, 81, 86, 85, 88–99, 102–104, 171–172,
Leicesterâ•… 3, 55–58, 60–61, 63, 90, 95, 104, 106, 171, 240 177–180, 182, 185–186
75–76, 79–82 pitchâ•… 3, 9–10, 12–13, 15–18, 20, Speak Good English Movement
lingua francaâ•… 25, 112, 142, 22–24, 93 (SGEM)â•…111–112
192–193, 204 postalveolar /t/â•… 5, 49–51, 227, specific (specific NP(s), specific
Little India, Singaporeâ•… 111 230, 232–233 context(s))â•… 134, 158
postvocalic /r/â•… 41, 53, 55, 57–58, Standard Hindiâ•… 141–142,
M 62–68, 70–80, 88, 93 192–193, 201
Malayâ•… 4, 105–107, 110, 112–113, proficiencyâ•… 4, 35, 61, 64, 133– stativeâ•… 115–116, 118, 124–126,
116, 118–121, 124, 126, 128–130 134, 142, 164, 194 129
musicâ•… 138, 171, 178, 182–183, Punjabiâ•… 5, 30, 34–36, 39, 49–51, stereotypeâ•… 9, 13–14, 24, 35, 133
222, 224, 226 59, 63, 75, 79, 136, 140, 142,
Muslim (Islam, Islamic)â•… 29–31, 215–216, 218, 221–223, 226, T
33–34, 36–38, 46, 51, 111, 113, 233, 236, 238–239 Tamilâ•… 4, 10, 83, 89, 91–94, 103,
176, 192, 221 105–106, 108–110, 112–113, 116,
R 118–121, 123–127, 129, 173–176,
N religionâ•… 12, 30, 37, 59, 111, 113, 181–184, 193
nativizationâ•… 196, 205, 208–209 128, 130, 178, 182–183, 192, 221 transnational indexâ•… 225–226,
networkâ•… 2, 5, 76, 187–188, 216, retentionsâ•… 93, 171, 178, 184 228, 230, 232
218, 220, 224, 226–227, 230, retroflex consonant(s)â•… 36, transnational tiesâ•… 4–5, 11, 32,
241 226–227 61, 110, 135, 141, 157, 164, 216,
New Zealandâ•… 2, 4, 13, 26, 55, rhoticityâ•… 3, 58, 62, 64, 67–68, 220, 238
57, 60, 80, 88, 129, 131–135, 70, 74–77, 80 transnationalismâ•… 138, 167, 171,
137–140, 142–143, 148, 154– 215–216, 218, 224–225, 240,
165, 170, 173, 188, 191, 193–194, S 242
202, 211–212 Scottish Englishâ•… 41, 48, 53 twice migrantâ•… 55, 57, 80
non-specific (non-specific second generation(s) (second-
NP(s), non-specific generation(s))â•… 3–5, 12, 30, U
context(s))â•… 134, 158, 164 36, 52, 55–57, 61, 63–69, 77–81, Urduâ•… 30, 34–36, 49–50, 89,
NURSE vowelâ•… 3, 85, 88, 95, 99 86, 131–133, 135, 137, 140–141, 92, 113, 153, 173, 176, 181, 183,
143–144, 154, 158–160, 162, 193, 201
O 164–165, 184, 211, 215–217,
open-ended question(s)â•… 197– 219, 221, 228, 230, 233–236, V
198, 211 238–240 vowel systemâ•… 12, 48
secondary diasporaâ•… 2–4, 55,
80, 131–132, 135, 141–145,
154–156, 159–160, 164

English in The Indian Diaspora

Uploaded by

Copyright:

Available Formats

English in The Indian Diaspora

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

English in The Indian Diaspora

Uploaded by

Copyright:

Available Formats

English in the Indian Diaspora

Varieties of English Around the World (VEAW)

John Benjamins Publishing Company

the American National Standard for Information Sciences – Permanence

CIP data is available from the Library of Congress

Varieties of English Around the World, issn 0172-7362 ; v. G50

© 2014 – John Benjamins B.V.

List of tables vii

Table 2.1 Mean F0 and F0 range for men and women

Marianne Hundt and Devyani Sharma

According to India’s Ministry of External Affairs, more than 20 million Indians

1. To what extent have the different diasporic situations resulted in different

2. Hallett (2010) compares code switching in newspaper language in English-medium publi-

model, that of endonormative stabilization, but instead of comparing language

Glenda Alicia Leung and Dagmar Deuber

Keywords: Indo- versus Afro-Trinidadians, ethnicity, perception, F0

At the southernmost point of the Caribbean islands, Trinidad forms a two-island

characterized by a distinctive type of pitch, but there is little empirical evidence

2. Historical, social, and linguistic background of the Indo-Trinidadian

Presbyterian missionaries (Samaroo 1975), were increasingly made use of and

3. Previous studies of Indo-Trinidadian speech

Although Indians are linguistically well-integrated into society, it is not implausi-

2. Transnational ties to India are not generally maintained by Indo-Trinidadians today.

groups who originally used another language or distinctive variety”. A natural

(L+H*). Furthermore, Drayton (2006) notes that “other Afro-Trinidadians who

4. An investigation into a popular stereotype surrounding pitch

As mentioned before, there is a commonly held folk belief that Indo-Trinidadians

notable attention in relation to gendered ethnicity (Stuart-Smith 1999; Alim 2004;

H1: Indo-Trinidadians use a wider F0 range than Afro-Trinidadians.

After a detailed description of the method in Section 5, the results relating to

In order to examine the aforementioned hypotheses, an experiment was designed

using Audacity, an audio editing software. F0 was manipulated by semitones since

1. The Indo-Trinidadian female voices were lowered by –1.25 semitones.

This study was originally designed to test F0 as an ethnicity assessment cue.

6.1 Hypothesis 1: Difference in F0 range between ethnicities

As a point of comparison, we referred to Gut’s (2009) values for mean F0 (av-

Table 2.1â•‡ Mean F0 and F0 range for men and women

6.2 Hypothesis 2: Response to modified Indo-Trinidadian voices

Fem-Indo Male-Indo Fem- Afro Male-Afro

6.3 Hypothesis 3: Response to modified Afro-Trinidadian voices

Hypothesis 3 predicted how listeners would perceive the Afro-Trinidadian voic-

6.4 Interspeaker variation – Trinidadian women

Thus far, it is evident that F0 is a fairly salient cue in ethnicity identification.

6.5 Interspeaker variation – Trinidadian men

In this section we examine individual speakers in relation to what we assume to

8. Conclusion and outlook

We wish to thank Kathy-Ann Drayton, Jason Leach, Dionne Ligoure, Shelley

1. Identify the ethnicity of the speaker:

2. How sure are you?

I used these cues/clues to distinguish I used these cues/clues to distinguish

AFRO-Trini MALE INDO-Trini MALE

Identity, ethnicity and fine phonetic detail

Farhana Alam and Jane Stuart-Smith

Keywords: sociophonetics, identity, ethnicity, consonants, Scottish

1.1 The Glasgow Asian community

1.1.1 Historical background

1.1.3 Sociopolitical situation

1.1.4 Linguistic context

1.2 The realization of /t/ in Glasgow Asian

the American National Standard for Information Sciences – Permanence

List of tables vii