Journal of Sociolinguistics 11/1, 2007: 74–93
Ideologised values for British accents1
Nikolas Coupland and Hywel Bishop
Cardiff University , United Kingdom
We report quantitative results from a large online survey of 5010 U.K.
informants’ reactions to 34 different accents of English, based on simple
accent labels. Patterns of accent evaluation, in terms of adjudged levels of
prestige, social attractiveness and some other variables, in many regards
confirm broad findings from earlier research. Accent-types associated with
‘standard’ speech are, for example, strongly favoured in the prestige and
attractiveness dimensions. Several urban U.K. vernaculars, but not all,
are systematically downgraded. On the other hand, robust differences
emerge which have not been strongly evidenced previously – particularly
differences according to informant gender (with females regularly producing
more favourable evaluations) and region (with informants often favouring
their own and linked varieties). There are also some important effects
by informant age, for example with younger informants attributing less
prestige to ‘standard’ accents.We interpret the findings as indicating rather
persistent U.K. language-ideologies around accent difference that are being
reconstituted only gradually and in specific regards.
KEYWORDS: Accents of English, language attitudes, ideologies,
prestige, social attractiveness, gender differences, language loyalty
RESEARCHING SOCIAL MEANING
Sociolinguistic indexicalities – relationships between speech forms/varieties and
social meanings – can be accessed in different ways and against different
theoretical assumptions. Language ideology research assumes that, in particular
socio-cultural environments, certain beliefs about the value of sociolinguistic
features, styles and practices are structured into people’s everyday understanding
(Schieffelin, Woolard and Kroskrity 1998; see also Lippi-Green 1997; L.
Milroy 2004). For example, James Milroy (2001: 530) argues that speakers of
standardised languages live in ‘standard language cultures’where the legitimacy
of what is thought of as ‘the standard language’ is taken for granted and where
issues of ‘correctness’ in relation to it are taken to be natural concerns that
matter. Establishment ideologies have constructed ‘standard English’ as being
an intrinsically ‘authentic’ variety (Coupland 2003) – an assumption which
sociolinguists have generally been at pains to resist.
C_ The authors 2007
Journal compilation C_ Blackwell Publishing Ltd. 2007
9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden MA 02148, USA
IDEOLOGISED VALUES FOR BRITISH ACCENTS 75
The fields of language attitudes research and folklinguistics have, in
their own ways, contributed substantially to our understanding of indexical
relationships and language ideologies, although their approach has tended
more towards empirical analysis than towards staking theoretical claims. In
language attitudes research it is conventional to distinguish ‘direct’ from
‘indirect’ approaches (Garrett, Coupland andWilliams 2003: 24–81; Garrett in
preparation; Ryan, Giles and Hewstone 1988; Ryan, Giles and Sebastian 1982).
Direct approaches involve forms of overt questioning about the meanings of
linguistic varieties,while indirect approaches try to uncover tacit and (arguably)
more deeply held beliefs and predispositions. Tore Kristiansen has argued
that overtly expressed beliefs about language variation may bear little or no
explanatory value in relation to language change, while covert subjectivities
(which need to be accessed indirectly) have considerable explanatory value (see
Kristiansen, Garrett and Coupland 2005: 13ff.). This stance builds on William
Labov’s (1972) distinction between overt and covert footings for ‘prestige’,where
an appeal to covert beliefs helps to explain a social group’s persistence in using a
linguistic variety that they are prepared to describe (in overt accounts) as ‘ugly’,
‘common’, ‘coarse’, and so on.
Under the rubric of folklinguistics, on the other hand, Nancy Niedzielski and
DennisPreston(2003)stress theimportance of people’sovert representationsand
evaluations of language varieties. For example, asking informants to draw maps
of dialect zones in the U.S.A., then to label them and to characterise the zones’
distinctiveness in their ownwords, has proved to be very revealing. The technique
creates an opportunity for people to articulate their senses of where linguistic
boundaries exist and of the meaningful differences that linguistic variation
represents. Folklinguistics makes language ideologies visible, even though some
of its methods (like the map-labelling task) seem rudimentary. Preston in fact
argues (in a personal comment) that the very starkness of variety labels (such
as ‘Southern speech’ in the U.S.A. or ‘the Liverpool accent’ in Britain) invites a
‘purity’ of ideological response that is not possible when people engage with real
speakers or with particular instances of talk-in-action.
In the tradition of accessing patterned ideological responses to linguistic
variation – in this case variation across most of the principal accent varieties of
English in Britain, as they arecommonly labelled –we report the results of a recent
large-scale survey. Our aim is to capture wide-scale and contrastive patterns of
attributed social meaning when informants are asked to produce simple, scaled
responses to a very large number of target varieties, presented to them in the
form of labels alone. The method is usually referred to as ‘conceptual’ (Garrett,
Coupland andWilliams 2003: 79), in the sense that it invites people to explore the
meaning associations of a simple (and arguably ‘pure’) sociolinguistic concept.
We believe it allows us to explore, quantitatively in this instance, how ideological
beliefs around British accent variation are socially structured. Social structure
here includes the evaluative ranking of one labelled variety relative to many
others on particular dimensions of judgement, but also the regional structuring of
C_ The authors 2007
Journal compilation C_ Blackwell Publishing Ltd. 2007
76 COUPLAND AND BISHOP
ideological beliefs about accents. In this way (as Preston and others have argued)
we are suggesting that social diversity in the attribution of social meaning to
speech varieties is sociolinguistically significant, in much the same way as social
diversity in speech itself is sociolinguistically significant.
THE VOICES SURVEY
We report data from a large online survey that we helped to design and
implement in collaboration with the British Broadcasting Corporation (BBC) in
2004. In 2005 the BBC developed and ran a large web-based initiative under
the title Voices, a multi-faceted and interactive exploration and celebration of
language variation in Britain linked to a series of radio and TV broadcasts (see
.co.uk/voices/). The survey was preliminary to the main Voices
initiative. It was designed to provide findings on contemporary British language
attitudes which, when summarised in press releases and media broadcasts, might
trigger U.K.-wide interest and participation in the interactive web tasks being
launched.
In some respects it is a replication and extension of Howard Giles’ widely cited
(1970) study which collected data on people’s evaluations of several British
accents, presented firstly as audible spoken ‘guises’ and then (as in the Voices
survey) as conceptual labels (see also Giles and Powesland 1975: 28–32). A
comparison of Giles’ original results with results from the new survey is available
in Bishop, Coupland and Garrett (2005) where we dealt with a sub-set of the
available new data chosen to match Giles’s original voice-types as closely as
possible.2 The Voices survey, however, collected evaluative data on 34 different
accents from 5010 respondents. The sample was demographically diverse and
very widely distributed across geographical regions of the U.K. All respondents
were over 15 years of age, and all completed the entire online questionnaire. As
such the data provide a rare opportunity to survey general patterns of response
to many of the most conventionally labelled British accent varieties, and to assess
the extent of variation within a large judgement sample.
The online BBC survey was conducted between the 17th and the 26th
November 2004. A market research company, Greenfield Online, was contracted
by the BBC to administer the questionnaire. The company maintains a panel of
registered respondents from which its samples for different surveys are drawn.
In Britain this panel consists of approximately 140,000 individuals, all of whom
will have ‘double opted in’ to participate in online surveys. This means that they
must respond to a confirmation message after registration before they enter the
panel in question, ensuring that third parties can not fraudulently add data to
the results. Members are recruited to the panel on an ongoing basis via a range
of media sources.
In the Voices survey informants were asked a variety of questions about their
own language use and about their general preferences about linguistic diversity.
As the main task, they were asked to rate the 34 labelled accents of English,
C_ The authors 2007
Journal compilation C_ Blackwell Publishing Ltd. 2007
IDEOLOGISED VALUES FOR BRITISH ACCENTS 77
which included most major regional British accents, some accents associated
with other countries but which have some presence in, or relevance to, British
social life, and some major global varieties of English. The choice of variety
labels is of course crucial, and the results themselves sometimes allow us to
trace the effects of alternative labels (see below; also Bishop, Coupland and
Garrett 2005). In most cases we were able to select relatively unambiguous
regional or national accent labels. In the area of ‘standard English’, however,
we opted for the phrases A standard accent of English and Queen’s English.
These are not uncommon but certainly not neutral characterisations, and we
doubt that neutrality of representation is possible in conceptual research of this
sort. Informants communicated their judgements electronically by clicking on
numerical values of seven-point rating scales across a number of judgement
dimensions.
Below, we report informants’ responses to questions about the prestige and
pleasantness of the 34 accents,which were the survey’s main focus. The questions
were posed directly for each accent: ‘How much prestige do you think is associated
with this accent?’,and‘How pleasant do you think this accent sounds?’. Language
attitudes research has found that these are highly productive and inclusive
dimensions in the social evaluation of regional and social speech varieties (see for
example Garrett, Coupland andWilliams 2003; Zahn and Hopper 1985). A few
supplementary questions were asked about perceptions of ‘proper speech’ and
about pride in people’s own accents.
Working with outside agencies – the BBC and Greenfield – meant that we did
not have full control of the overall design and even the final content of the survey.
For example, not all of the accent categories that we wanted to include could
be accommodated, and the regional classification of respondents was carried
out in ways relevant to the regional organisation of the BBC itself. But the
survey collected a greater volume of data than is usually possible under the
normal constraints of academic research. The sample was not selected on strict
quota principles, for example to reflect gender, age, social class or geographical
demographics or to ensure perfect balance between regions. However, maximum
respondent numbers were set for all regions and the design ensured a good
geographical spread of responses (although there was a slight shortage of
respondents in some respondent cells from Northern Ireland). Table 1 gives
demographic details for the whole sample of respondents. There is a good gender
balance in the sample, but a slight imbalance in age distributions: the majority
of informants fall in the twomiddle-age categories, leaving the 25–64 age group
over-represented and the 15–24 and 65+age groups under-represented. Again,
however, the large overall sample size allows valid comparisons to be made (and
the statistical package in any case adjusts for unequal cell sizes). No social class
details were collected, as this was not a priority for the BBC.
Informants were asked to indicate where they currently lived, information
which we recoded into six U.K. regional categories: Wales, Scotland, Northern
Ireland, North of England/Midlands, South-East of England and South-West
C_ The authors 2007
Journal compilation C_ Blackwell Publishing Ltd. 2007
78 COUPLAND AND BISHOP
Table 1: Demographic characteristics of the whole sample
n %
Gender
Female 2544 50.8
Male 2466 49.2
Age
15–24 301 6.0
25–44 2301 45.9
45–64 2123 42.4
65 + 285 5.7
Region
Wales 279 5.6
Scotland 550 11.0
Northern Ireland 98 2.0
North/Mid England 1998 39.9
South-East England 1458 29.1
South-West England 575 11.5
Diversity
Low 319 6.4
Medium 2218 44.3
High 2473 49.3
of England. We asked for a scaled response to the statement ‘I like hearing
a range of accents’, which we took as a rough index of informants’ stances
towards sociolinguistic diversity. Responses to this diversity measure were then
grouped into three categories – high (those who answered either six or seven
on the diversity measure), medium (four or five), and low (one, two or three).
They arguably reflect more progressive to more conservative stances towards
sociolinguistic diversity. Table 1 shows that a large majority of informants
categorised themselves as moderately or very well disposed to diversity, but there
is variation which, as we will see, is systematically associated with patterns of
response to the different accents.
In the remainder of the paper we report results for all 34 accents, firstly
according to mean responses for the whole sample, for both prestige and social
attractiveness (Table 2). To arrive at more differentiated results, we carried
out eight separate MANOVA (multiple analysis of variance) analyses, four for
prestige and four for social attractiveness, with the 34 accents as dependent
variables, and with informant age, region and sex, plus the ‘diversity’ dimension,
as independent variables. These results are recorded in the two largeAppendices.
For the MANOVA using region as an independent variable, we included age, sex
and diversity as covariates in the analysis, and correspondingly for the other
analyses.
C_ The authors 2007
Journal compilation C_ Blackwell Publishing Ltd. 2007
IDEOLOGISED VALUES FOR BRITISH ACCENTS 79
Table 2: Mean ratings (whole sample) of 34 accents by social attractiveness
and prestige
Social attractiveness Prestige
1. Accent identical to own 4.87 (2) 4.14 (3)
2. Afro-Caribbean 3.72 (21) 2.90 (30)
3. Asian 3.21 (31) 2.74 (33)
4. Australian 4.04 (13) 3.51 (11)
5. Belfast 3.67 (23) 3.11 (27)
6. Birmingham 2.92 (34) 2.70 (34)
7. Black Country 3.16 (33) 2.81 (32)
8. Bristol 3.64 (25) 3.22 (21)
9. Cardiff 3.67 (24) 3.16 (25)
10. Cornish 4.22 (8) 3.38 (13)
11. Edinburgh 4.49 (5) 4.04 (4)
12. French 4.09 (11) 3.74 (9)
13. German 3.20 (32) 3.21 (23)
14. Glasgow 3.45 (29) 2.93 (29)
15. Lancashire 3.90 (15) 3.24 (20)
16. Leeds 3.73 (20) 3.15 (26)
17. Liverpool 3.40 (30) 2.82 (31)
18. London 3.70 (22) 3.89 (6)
19. Manchester 3.61 (27) 3.22 (21)
20. Newcastle 4.13 (10) 3.21 (23)
21. New Zealand 4.37 (6) 3.84 (7)
22. North American 3.90 (15) 3.80 (8)
23. Northern Irish 4.05 (12) 3.30 (17)
24. Norwich 3.81 (18) 3.38 (13)
25. Nottingham 3.78 (19) 3.39 (12)
26. Queen’s English 4.28 (7) 5.59 (1)
27. Scottish 4.52 (4) 3.98 (5)
28. South African 3.51 (28) 3.34 (16)
29. Southern Irish 4.68 (3) 3.63 (10)
30. Spanish 3.88 (17) 3.29 (18)
31. Standard English 4.96 (1) 5.44 (2)
32. Swansea 3.64 (25) 3.11 (27)
33. Welsh 3.95 (14) 3.29 (18)
34. West Country 4.16 (9) 3.36 (15)
MEAN RATINGS
Table 2 lists the 34 accents alphabetically, but also provides rank-orderings
(bracketed) for each semantic dimension.Onthe seven-point scale, themaximum
possible mean value is 7.0 and the minimum is 1.0, with 4.0 being the