Introducing Sociolinguistics-MIRIAM MEYERHOFF , CHAPTER 3 Variation and style

CHAPTER  3

Variation and  style








Key terms introduced in this chapter:
■   accents
■   dialects
■   variety
■   speech community
■   style-shifting
■   attention to speech
■   audience design
■   triangulation
■   sociolinguistic interviews
■   stratified




INTRODUCTION

■   monotonic
■   trend
■   rapid and anonymous
■   speech community
■   overt  prestige
■   covert prestige
■   observer’s paradox
■   participant observation
■   speaker design




Accent

Where speakers differ
(or vary) at the level of pronunciation only (phonetics  and/or phonology), they have different accents.
Their grammar may be wholly or largely the same. Accents  can index a speaker’s regional/geographic origin, or social factors such as level and type of education, or even
their attitude.

‘And what do you do for a living?’ her new acquaintance asked.  Feeling somewhat pained by the man’s inability to see  she wasn’t interested in talking to him, she replied tersely, ‘I’m a sociolinguist.’
He  doesn’t  get  the  message. ‘Oh, yes?’ (An ingratiating  smile.) ‘And what  does  a sociolinguist do?’
She pauses, then levels a steely look at him:  ‘It means I listen to the way people talk and
I judge them on it.’
In general, the judgements sociolinguists  make about other people’s speech are pretty innocuous. Some sociolinguists know a lot about what features typify the accents or dialects of speakers from different regions, and these sociolinguists  are pretty good at identifying speakers’ origins from the  way they speak.  When linguists talk about  accents, they are referring  only to how speakers pronounce words, whereas they use  dialect to refer  to distinctive features at the level of pronunciation  and vocabulary and sentence structure. So, for example, the  English used  by many Scots  would be considered a dialect  because it combines  recognisable features of pronunciation, e.g., a backed short /a/ sound  in words like trap or man, with constructions like This data  needs examined  . . . (i.e.,   ‘needs  to be examined’) and the use of the preposition  outwith (meaning  ‘beyond, outside’). Since all of

Dialect

A term widely applied to what are considered sub- varieties of a single language. Generally, dialect and accent are distinguished  by how much of the linguistic system differs. Dialects differ on
more than just pronunciation, i.e., on the basis of morphosyntactic
structure  and/or how semantic  relations
are mapped  into the syntax. (See also Variety.)


Variety

Relatively neutral term used to refer to languages and dialects. Avoids the problem of drawing a distinction between the two, and avoids negative attitudes often attached to
the term dialect.

Style-shifting

Variation in an individual’s speech correlating with differences in addressee,  social context, personal goals or externally imposed tasks.

Attention to speech
Labov proposed that the different distribution of forms in different styles was motivated by the amount of attention the speaker was paying to the act of speaking. In activities, such as reading aloud, reading word lists or minimal pairs, Labov argued that speakers are paying more attention to their speech than they are
in interviews and in interviews they pay more attention than when conversing with friends and family. Contrasts with accommodation- based accounts of style-shifting such as audience design. Also contrasts with more agentive theories  of style-shifting such
as acts of identity.
these features occur even in quite formal styles of speaking, they are quite reliable cues that the speaker comes  from or has lived a long time in Scotland.
Outside linguistics, the term dialect may have quite negative  connotations. These  may be revealed  implicitly rather  than  explicitly. For example, on Bequia (the Caribbean  island mentioned  in Chapter  2), people  speak  a variety of English which differs radically from the Standard English used  in North America, the  UK or Australasia. Bequians  generally call the variety they speak  Dialect. When researchers ask people  to describe the local variety, locals will often contrast Dialect with what they call proper or good English. The opposition between good and dialect forms of English implies that dialect is bad and is linked to all sorts of attitudes about the local variety, such as where and when it is appropriate to talk Bequian.
Many linguists avoid the term ‘dialect’ because of these complicated,  and sometimes negative, connotations in everyday speech. However, where  they do use  it, they intend it to be a neutral  description  or a cover-all term for a variety that differs systematically  from others  on the basis of pronunciation, grammar and vocabulary. (In Chapter  4 we will look at the  ways in which people’s perceptions and beliefs  about  different  varieties  can also be relevant  factors  in identifying different  dialects.) I will  often  simply use  the  term  variety because potentially it is less loaded.
Sometimes the kinds of judgements that sociolinguists make are about whether a person is speaking formally or informally, whether  they sound like they grew up in a working-class or a middle-class neighbourhood – many of the judgements non-linguists make all the time about the people they are talking to.
Sociolinguists differ from the average listener, though, in trying to develop an awareness of language that goes  below the level of social stereotypes. They are concerned with trying to determine how very subtle  patterns of variation provide a systematic basis  by which speakers can indicate or mark social cohesion and social difference. (We defined stereotypes, markers  and indicators in Chapter  2.)


STUDYING  VARIATION  IN  SPEAKERS’  STYLE

In Chapter 2 we looked at the methods and findings used to identify this systematic variability on Martha’s Vineyard. In this chapter  we begin by looking at the next study Labov undertook, a social dialect survey in the Lower East Side of Manhattan. We will focus on Labov’s finding that, for a number of variables, all speakers in the survey show the same  general  patterns in formal and informal styles. This apparent orientation to the same  norms became critical for what he defined  as a speech community. This consistency in the patterns between and within speakers across  different styles provides further evidence  against  the notion of free variation (thereby developing the arguments introduced  in Chapter  2).
We will look at three  possible  accounts for the consistency of style-shifting across individuals in a speech community. The first is the suggestion that people  pay more or less attention to their speech when they are engaged in different kinds of verbal tasks. The second is the idea that speakers have an audience in mind, and they design their speech to suit that audience.  The third is the idea that different  linguistic styles present different personas that the speaker identifies with. We will consider  the first two in most detail since they differ most radically and have been  subject  to the most careful empirical study.
After considering  Labov’s findings and the way he manipulated  style in the Lower East Side study, we will examine two other studies  that manipulated  and examined  speaker style somewhat differently. These  focus  on the speakers’ relationships with their interlocutors.

This audience design account of style-shifting emphasises the dual role played by language variation in use: reflecting  and constituting  social meaning (see  Chapter  2).
The Martha’s Vineyard study established some basic methods for social dialect research and  these remain  well-used  tools  in sociolinguists’  toolbox. However, Labov and  other sociolinguists  have subsequently added  other creative methods for gathering the kinds of sociolinguistic information we seek. In this section, we look at some of the other methods he devised for reliably identifying the patterns underlying language variation and we focus on ways a researcher can manipulate  data collection so as to elicit different styles of speech. We will consider  what variation across  different styles might tell us about the orderliness of language variation in a community of speakers.


THE  NEW  YORK  CITY  SOCIAL  DIALECT  SUR VEY

After the success of the Martha’s Vineyard study, Labov turned his attention  to variability in the speech of New Yorkers. New York City (NYC) is an interesting site for fieldwork because historically it is a dialect  pocket  on the  eastern coast  of the  United States;  that  is, it is surrounded by other varieties of US English from which it differs quite perceptibly. Generally speaking,  the NYC accent is highly stereotyped in the United States;  that is, residents and non-residents find the distinctive characteristics of the NYC accent highly salient and they are readily stereotyped (as defined  in Chapter  2).
Historically, one of the more salient features that sets  NYC speech apart from varieties spoken nearby (e.g., in New Jersey), and from the more general variety of Standard American, is that NYC has been r-less. This means that unless an orthographic ‘r’ occurs before a vowel, it is not pronounced as  a constricted ‘r’  – in this respect NYC speech differs from most northern  and western varieties  of North America. Like British English, the  post-colonial Englishes  of the  Pacific and southern Atlantic and some  varieties  of Caribbean  English, words like car, port, garden, and surprise (i.e., words where the ‘r’ is in what phonologists call the coda of a syllable) do not get pronounced with a constricted,  consonant [r.] This feature of the New York accent is widely stereotyped and is one that New Yorkers themselves may have quite negative  feelings  about  – some  of them  say they dislike it even if they, their families and friends are all r-less speakers.
Labov’s study of NYC was more ambitious than the study of Martha’s Vineyard, and he looked at a wider range  of variables. Some  of these,  like the (r) variable, were ones  that speakers were consciously  aware of; some of the others, though, were ones  that speakers were much less aware  of and seemed to be perceptible only to a trained linguist. (We will see  results for some of the other variables in subsequent chapters.)  However, even though speakers’ level of awareness differed for the variables Labov identified, he found that there were some consistent patterns in the way the variables patterned across  different groups of speakers and in different styles or activities.
Labov obtained  his data on (r) using several different methods.  The idea behind this is called triangulation and is basic to science. If, using different methods, you get results that are consistent with the same  analysis or conclusion, then your conclusion is much stronger than it is if you arrive at it using only one means of measurement.
Audience design

Derived from accommodation theory. Proposal that intraspeaker variation arises because speakers are paying attention to who they are addressing or who might be listening to
or overhearing them, and modify their speech accordingly.






















Triangulation

A researcher’s use of several independent tests  to confirm their results and aid in the interpretation  of their results. For example, use of data from sociolinguistic interviews and a rapid and anonymous study.

The  sociolinguistic  interview










Sociolinguistic interview
An interview, usually one on one, in which different tasks  or activities are used to elicit different styles of speech.  (You will sometimes hear it used simply to refer to a one-on-one interview lasting at
least an hour covering a range of topics.)
Labov extended the basic interview paradigm he had used on Martha’s Vineyard. He added several language tasks  that he would ask people to do during the interview, and in the free conversation part of the interview he separated out speech directed  at the interviewer and speech directed  at friends and family members.  A good interviewer can get several  hours of speech from a single speaker and because the interviews are almost always conducted in the  interviewee’s home  or somewhere they feel comfortable,  a skilled (and somewhat lucky) interviewer may also have the chance to record the interviewee talking to other people who pass  through while the interview is taking place.
Labov interviewed a random sample of people  from the Lower East Side in New York in their own homes.  The sociolinguistic  interviews consisted of four structured parts. The interviewee was asked  to:

(i)   read a list of minimal pairs (pairs of words that have different meanings but only differ from each  other in one sound);
(ii)  read a list of words in isolation (some of which contain the variables under investigation and some of which do not);
(iii)  read aloud a short narrative (carefully constructed to contain the variables in as many linguistic environments as possible);
(iv)  talk with the interviewer about their life, some of their beliefs, and their life experiences.

Labov was aware that for a variable like (r) there were clear differences in which variant was considered appropriate for formal and informal speech. But there  are problems  with investigating  the  spontaneous production  of different  styles  of speech. One problem is agreeing what constitute different ‘styles’ in the first place, another  is agreeing which ones are more or less formal, and even if those problems can be overcome there can be problems with recording  enough  people  using language in all those  styles to allow the researcher to make valid generalisations.
Labov tackled these problems in defining and working with style by proposing that the formality or informality of styles was a function of speakers’ attention  to their own speech: in more formal styles they pay more attention; in more casual styles they pay less attention.
The activities in (i)–(iv) were intended to elicit different speech styles: (i) and (ii) require the speaker to pay much more attention  to language,  while in (iv) a good interviewer will foster quite animated  and lively conversation between the interviewer and interviewee. This kind of speech can be called ‘informal’ speech, and Labov found that he had a lot of success in getting  informal conversation by asking  people  questions about  things  like fights, dangerous situations  the  speaker had  been  in, the  supernatural, their first girlfriend or boyfriend, and important events  in their childhood. Side conversations that the interviewee might have with friends or family during this part of the sociolinguistic interview are presumed to involve the least attention  to their speech, and these can be called ‘casual’ speech.
A minimal pair for the (r) variable would be god and guard – when guard is pronounced without a constricted /r/, it sounds just like god. Dock and dark are also minimal pairs in this variety of English because the vowel is the same, so only the presence or absence of an /r/ differentiates them. The layout of minimal pairs helps to focus the speaker on the form of the words. If you present someone with a card that looks like Figure 3.1 and ask them to read each  line carefully and clearly, the speaker will be concentrating considerable attention  on how they pronounce each  word.


Connections with  theory

Labov has never claimed that reading  aloud, especially reading  words in isolation or in minimal pairs, is related to conversational speech. He acknowledged that they are qual- itatively different activities from having a conversation. The activities in (i)–(iii) are artificial strategies that enable the researcher to control how much attention the speaker pays to their speech. They therefore allow the researcher to test  the hypothesis  that attention to speech is an important constraint  on variation.



Please  read across  each line carefully and clearly:

guard dock pin
sauce
god dark pen
source
exercise

Figure 3.1   Example of a presentation of minimal
… … pairs used to elicit most careful and attentive pronunciations.


The hypothesis  is that if a speaker focuses all their attention  on the pronunciation  of a word in a task  like this, then  here, of all places,  they will use  a constricted /r/ – even if they don’t usually do so in their casual  speech. The reading  list and the reading  passage tasks  were intended  to require somewhat less  attention  to the  form of individual words. Consequently, you would find progressively more of the local r-less variants in these activities.


Designing  materials  for  a  social  dialect  sur vey

It is not all that easy  to create a plausible  and  moderately interesting narrative that includes  a variable in a wide range  of linguistic environments, e.g., sentence-finally, before  a vowel, before  different kinds of consonants, in common  and in less  common words, etc. (Sociolinguistic interview activity (iii), discussed on p. 30.)
Pick a sound that shows variability in your own speech or the dominant  speech community  you live in and  try to construct a short  (100–200 word)  narrative  that showcases this variable in as many environments as possible.


facts



Interviewees can become  surprisingly oblivious of a microphone even


if they are attached to it, getting up to show you things, etc.
However, even someone you have got to know quite well may surprise you by showing they are paying attention to the recording, for instance  by saying something  like,
‘That’s enough for tonight’ when a tape audibly
comes  to its end.
Finally the distinction between informal and casual  speech in the interviews was also taken to reflect a (natural and spontaneous) difference in attention to speech. No matter how hard interviewers  work or how skilled they are, unless  they have invested  a considerable amount of time in getting to know the interviewees, we assume that their conversations will always be subject  to somewhat more attention  than  conversation with friends and family. Because of this, it was expected that they would also use the constricted variant of (r) less in casual conversation than in the informal part of the interview.
Labov’s findings supported his hypothesis. On average, everyone used  pronunciations with an  [r] more  when  they  were  reading  the  narrative  aloud  than  they  did in casual conversation. They used even more [r] when they were reading the word lists, and they were most likely to use [r] variants when they were reading minimal pairs.
This sensitivity to style shows up in all the variables Labov examined. This is illustrated in Figure 3.2, where the frequency of constricted variants for (r) is shown in the four different styles averaged across  everyone who was interviewed in the NYC study.
Figure 3.2 shows that the variants associated with these variables are stratified by style. This means that there  is a consistent order for the styles across  speakers. The rate  of [r] presence drops steadily as you go left to right along the bottom axis of this figure. We can also say that the relationship between style and the variants is monotonic, or that the data show  a trend,  i.e., a consistent tendency to use  less  of the  constricted variant as  the researcher has manipulated the formality of the talk or the speaker’s overt attention to speech.
The same  monotonic relationship between style and linguistic variation can be seen in Figure 3.3, which plots the frequency of vernacular, raised variants in two vowel variables short (a) and (oh) (the TRAP and CLOTH  vowels respectively – we’ll return to these variables again in Chapter  8).


70

Stratified 60 58
See Broad and Fine
stratification. 50
The systematic  and
consistent patterning
of a variant with 40 38
respect to some
independent factor 30

Monotonic 22
20
Percentage of constricted  [r]
A steady increase or 16
decrease in a feature
along the x-axis of 10
a graph. 6
0
Trend

Steady increase  or decrease in the frequency of a form across  a scale or set of measures.
Minimal pairs Word lists Reading Interview Casual
Style

Figure 3.2   Occurrence of constricted [r] in New York City English in five speech  styles. (Dashed line indicates the qualitatively different activities involving the use of unconnected speech.) (Source, Labov 1966: 221.)


Connections with  theory

The results from the NYC study are frequently used to illustrate the systematic patterning of different groups  of speakers from different social classes across  different styles. As we will see  in Chapter  8, the results  for style and social class  show patterns that are startlingly similar. The similarity of these trends  was central to the analysis of the social meaning of the (r) variable in NYC.



45

40
35 36
35

37 39

35

Index of raising
30 32
29
25 27

(eh) (oh)

20

15

10
Word lists Reading Interview Casual
Style

Figure 3.3   Raising index for short (a) and (oh) variants in New York City speech  in four styles (TRAP
and CLOTH vowels). Higher index means a more raised vowel. (Source, Labov 1966: 221.)


This monotonic relationship supports Labov’s conjecture that the three artificial linguistic tasks  that he included in the sociolinguistic interviews were, indeed, an effective means of modelling changes in stylistic formality. The trend that emerges between informal and casual speech in the free conversation part of the interview is continued  in the manner predicted  if style differences manifest a speaker’s different level of attention  to their speech.


A  note  on  graph  notation

Notice that the frequency with which the constricted tokens, [r], are used in each style in each activity is joined with a line. The line becomes a dashed line at the point where it is clear that there  is a qualitative break  between the  kinds of activity represented. It is nevertheless continuous. This is very important because it makes  a claim about the relationship between each of those  points. The use of a continuous line asserts that style is a continuum. In other

words, there  might be an infinite number of degrees of attention  to speech, of which these five activities or styles represent only a sample. If you were able to devise some  other task that required  a little bit more attention  to speech than informal speech, but a little bit less attention  than reading  out loud, you would expect  that the frequency  with which speakers use constricted [r] would fall somewhere on or very close to where the line in Figure 3.2 runs between these two styles.


Rapid  and  anonymous  surveys

A second method  that Labov pioneered for studying  variation was the  use  of rapid and anonymous  surveys. The rapid and  anonymous  survey of the  realisation  of (r) in three department stores is one of the most famous studies  in sociolinguistics. As well as trialling a novel and easily replicable methodology, the department store survey has made a significant theoretical  contribution  to variationist sociolinguistics.  Together  with the  sociolinguistic interviews in the Lower East Side, it demonstrates that quite different methods for gathering data, including quite different ways of manipulating attention  to speech, can produce  similar and mutually informative results.
We noted  that  one  advantage of the  interview format  for data  collection is that  it generates a large amount  of information for subsequent analysis. Moreover, because the interviewer spends a fair amount of time getting to know the interviewee, they can make more sensitive evaluations when they come to assessing variation in and across social groups (we will return  to the use  of sociolinguistic  interviews to gather  this kind of data  in Chapters
7–10). But one of the disadvantages of interviews is that they can take a long time to arrange and conduct. So it is often helpful to be able to complement them with methods that collect data more speedily.
Labov chose  three  department stores as the venue for some quick fieldwork. He tried to elicit as many tokens  of the phrase fourth floor as possible from staff working in the three stores. This phrase was a good one from a linguistic point of view because it has one token of the (r) variable before  a consonant and one token word and phrase finally. The decision to ask staff (and not, for instance,  customers) was practical – staff were more likely to be able to give the desired answer. He would find some item on the store directory boards that was sold on the fourth floor, e.g., lamps or shoes, and he would ask staff where lamps, or shoes, were. They would say, ‘Fourth floor’, and then he’d pretend  he hadn’t heard and ask them to repeat it. With slightly more care, the staff member would repeat ‘Fourth floor’. Labov would walk off and write down how often they had and had not used  a constricted [r], and a few basic social facts about the speaker (their occupation in the store, their sex, a rough estimate
of their age).



Connections with  theory

Labov chose  the three department stores according  to the general socioeconomic level of their target customers. Because of this, he was able to employ his rapid and anonymous data to also check  on the effects of social class found in the interviews. Social class is discussed further in Chapter  8.

This strategy  was extremely productive. It provided:

■      two casually uttered tokens  of the (r) variable;
■      two tokens  of the variable uttered more carefully;
■      tokens  of the variable in different linguistic contexts  (one preconsonantal environment,
fourth, and one at the end of a word); and
■      a speedy  source  of the information.

For these reasons the methodology  is known as a rapid and anonymous survey.
In all three department stores, speakers were more likely to use the Standard American pronunciation  with a constricted [r] when they repeated fourth floor carefully for a second time than  they were the first time they uttered it. Because the pattern  emerged across  a considerable number of staff in all three stores, this supported the empirical findings in the interviews, and widely held attitudes about the (r) variable. The use  of constricted variants with careful  speech was  robust  and  consistent across  several  measures. The parallels between the findings in the department store survey and the interview data can be seen in Figure 3.4.
Notice that in Figures 3.2–3.4 it is clear that even if style has a strong correlation with the use of [r], it doesn’t determine which variant will be used. The results from these studies tell us about the probability with which a particular variant will occur. It is more likely that a person  will use the constricted variant of (r) in a word list than in casual conversation. But in all styles, a speaker might use the other variant.
On the basis of the evidence  accumulated in the New York studies, Labov argued  that speakers could be considered co-members of a speech community if they share:
Rapid and anonymous study
A questionnaire used to gather data quickly in the public domain. (See also Sociolinguistic interview; Triangulation.)






Speech community



45

40 39
Percentage  constricted [r]
35 36

30

25

20

15

10

5

0







30
Sociolinguistic
interviews
Department store survey

11
Variously defined on subjective or objective criteria. Objective criteria would group speakers together in
a speech community if the distribution
of a variable was consistent with respect to other factors (e.g., style). Subjective criteria would group speakers as a speech community if they shared  a sense of
and belief in
co-membership.
More careful More casual
Styles

Figure 3.4   Frequency of constricted [r] in more careful styles of the sociolinguistic interview (reading passage,  word lists and minimal pairs combined) and more casual styles (informal and casual combined) compared with careful and casual styles in the department store survey. (Source, Labov 1966: 74, 221.)
exercise

■      the same  variants in their repertoire, and
■      the same, consciously or unconsciously  held, attitudes to those  variants.

In the case  of (r), consciously  held attitudes were reasonably accessible, but in the case  of other  variables, where  speakers did not have much or any conscious awareness of the variation, consistent patterns of stratification  in different  speech styles  provided crucial evidence  that speakers share  underlying attitudes about the variable.



Connections with  theory

Labov assumed that when he got department store staff to repeat, ‘Fourth floor’, that he was getting  them to pay more attention  to their speech. This analysis focuses on style from the perspective of the researcher.
The analysis can be turned around so that it takes  the perspective of the staff and focuses on the  addressee, Labov himself. In this case,  their variation might not be a function of an increase in the attention  they paid to their own speech, but an increase in
the attention they paid to their addressee. We will return to this alternative analysis shortly.




The  ef fect  of  topic  on  style

There  have  been some  efforts  to break  down  the  styles  within the  informal/casual speech parts  of the sociolinguistic interviews. This has been done  with some  success on the basis  of topic. It seems that when people are getting  a bit preachy (about  any topic) or when they are talking about ‘language’ itself, you elicit more careful styles than you do when the person is talking about, say, childhood memories.
Suppose you had been talking to someone for a few hours about  their life history, their beliefs and their personal hopes. Make a list of some of the topics you might expect to come  up. Would  you expect  some  topics  to elicit speech that is more careful and
some  to elicit speech that is more casual? Which ones? Why?



To sum  up, each  of the  methods Labov used  has  its advantages. The  rapid and anonymous  survey quickly generates a lot of tokens  of a restricted kind, and it is easily replicated. The interviews mean that the analysis can draw on more detailed facts about the interviewees and can be based on a wide range  of tokens. They also enable  the analyst to find out what kinds of social facts are important to the people  being interviewed and what their attitudes are to different variables. Although both methods have advantages, it was the combination  of the two that allowed Labov to forward a very strong  claim that a person’s attention  to speech was an important constraint  on the variable (r) in NYC.


Connections with  theory

Attention to speech postulates a single grammar, of which the variation is an integral component:  this is the  notion  of inherent  variability. This highlights  a fundamental difference between variationist sociolinguists and formal linguists. Sociolinguists consider intraspeaker variation to be evidence  of inherent  variability in a communal  grammar. Formal linguists generally prefer to explain intraspeaker variation in terms of alternations between different grammars (akin to a bilingual speaker alternating between languages).



PRESTIGE  OF  A  VARIABLE

It is tempting to think about the consistent orientation of speakers to a particular variant in more formal contexts,  or careful tasks  in terms of the relative prestige of the variants. But sociolinguists recognise prestige as a complex value that speakers orient to in different ways. In particular, we have to be careful that we do not define a prestige variant as ‘a variant which has more status’ if the only evidence  we have for the variant being statusful is that it is used more by members of a speech community who themselves have more status (e.g., the upper class). If that is our only evidence, then our definition of prestige is circular and we might more accurately say that we have a variant that marks higher social class rather than a variant that is prestigious. Prestige is not necessarily something speakers are consciously aware of, nor something that is associated with the highest social classes or more powerful speakers in a community. These observations have led sociolinguists to make a distinction between overt and covert prestige.


Over t  vs  cover t  pr estige

Overt prestige is understood to be the prestige associated with a variant that people  are highly aware of and which is associated more with the speech of higher-status speakers. This would include a variant like the  constricted [r] in NYC, which speakers will describe in evaluative terms as sounding  ‘better’ or ‘nicer’ than the r-less variant.
But it is equally apparent that speakers orient themselves to other kinds of prestige. So, for example, most speakers of German  have command  over at least  their local vernacular variety of German  and Standard German  (Hochdeutsch). They will have learnt their local vernacular  variety with its characteristic pronunciations and  lexical items  at home, and continue  to use it there, whereas Standard German is something they are taught  at school and they use when they are speaking or writing in public (situations  like this are discussed further in Chapter  6 in the section  on diglossia).
The local variety carries  strong  connotations of naturalness and straightforwardness because it is acquired  naturally and used  in people’s most  informal and intimate styles. Standard German, on the other hand, may sound stilted and distant because it is so seldom used  in intimate and friendly contexts.  Politicians and other professional salespeople make strategic use  of the positive connotations of different  varieties, for example, by using the vernacular  when they think it will be advantageous to sound  like an everyday person  and







Overt prestige

The prestige associated with a variant that speakers are aware of and can talk about in terms
of standardness, or aesthetic and moral evaluations like being
‘nicer’ or ‘better’. (See also Covert prestige.)

Covert prestige

A norm or target that is oriented to without the speaker even being aware that they are orienting to it. Evidence of covert prestige  can be found in mismatches between speakers’ self-report of using one variant and actual
use of another variant. Often used (wrongly) to refer to the value associated with
non-standard or
vernacular varieties.

the Standard when they want to sound authoritative. (We return to perceptions about different linguistic varieties, and the connotations associated with them, in Chapters 4 and 6.)
This kind of local prestige is sometimes referred  to as covert prestige,  but in fact the use of the term for the value speakers associated with local or working class pronunciations is often far from covert. They can explain it to you quite clearly: ‘I’d never talk like that with m’ mates, they’d think I was a tosser.’ Hence, there is often very little that is covert about the persistence of non-standard or highly regionalised  veracular varieties.
The term covert prestige more accurately  refers  to cases where  speakers’ positive evaluation of a variant is genuinely covert or hidden. Peter  Trudgill’s (1972) early work on Norwich English found that some speakers recognise and overtly talk about one variant as being ‘better’ than another; furthermore, they claim to use the ‘better’ form, but in fact do not.
In this particular  case,  many of the  men  (not so many of the  women)  that  Trudgill interviewed said that the Standard Southern British English pronunciations of tune and dune, i.e., [tjun] and [djun], were better  than the local Norwich variant without the [j], i.e., [tun] and [dun], and reported  to him that it was the form they used. That is, they were overtly oriented to the supra-local  prestige of [tjun] and recognised the [tun] variants as being distinctively local and non-standard. Nevertheless, they used the local variant most often in their interviews with Trudgill. So their orientation  to the local variant was covert: they believed  they were doing something different from what they actually were doing.
Trudgill suggested that this mismatch between what the speakers say they do and what they actually do should be considered evidence of covert prestige. Chapter 4 discusses other examples of mismatches between how people think they are talking and how they are actually talking, and when we consider the notion of language vitality in Chapter 6, we will see another
basis for distinguishing between institutionally based prestige and more locally based prestige.

Observer’s paradox
The double-bind researchers find themselves in when what they are interested in knowing is how people behave when they are not being observed; but the only way to find out how they behave is to observe them.


IN  SEARCH  OF THE UNKNOWABLE: THE OBSER VER’S  PARADOX

The paradox  every sociolinguist  faces  in trying to account accurately  for variation within a speech community is that they want to know what people say and do in their everyday lives, but as soon as they start to record them they change the dynamic even slightly. So what they want to know is, in one sense, unknowable. This has come to be known in sociolinguistics as  the  observer’s paradox. Labov’s rapid and anonymous  surveys were an attempt  to overcome this problem, as was his practice of trying to turn interview conversations to more lively and personal  topics.



Connections with  theory

The observer’s paradox makes the same kind of generalisation about studying language that Heisenberg’s uncertainty principle makes about studying particles. That is, we cannot observe something without changing it. One reason for the uncertainty principle in physics is that particles do not exist independently as things, they exist as sets  of relationships. Sociolinguists, too, are actually studying sets of relationships when they look at variables. This will become more apparent as we progress.

Some sociolinguists take a leaf out of anthropologists’ books in an attempt to overcome the observer’s paradox. They spend long periods of time working and/or living with the people whose  speech they are interested in, and they hope  that by doing this they will eventually achieve insider status themselves. This is known as participant observation.
It is also possible  to turn the observer’s paradox to good use. It has also been  pointed out that sometimes the  way people  talk when they are  aware  of being recorded can be sociolinguistically illuminating too. Natalie Schilling-Estes (1998a) noticed  that some  of the speakers she recorded from Ocracoke,  an island off the North Carolina coast, seemed to enjoy giving quite flamboyant  performances of stereotypes of the  local accent.  She found that these highly self-conscious performances of the  local accent didn’t produce anything that was inconsistent with what the same speaker produced  in less self-conscious conversation.  Moreover, the  relatively extreme  variants that one  speaker produced  in his performances provided  telling evidence  about  the  underlying  system  of phonological contrasts.
A widely recognised stereotype of Ocracoke speech is (ay), the diphthong  in PRICE, which has  a noticeably  raised  onset  in the  traditional Ocracoke brogue,  e.g., [  i]. This is commented on  by both  islanders  and  outsiders.  A comparison   of  one  man’s, Rex’s, performances of (ay) and tokens of the variable in conversation showed that raising certainly was one dimension  on which the (ay) variable contrasts with other  Ocracoke vowels and with the non-raised variants, e.g., [ɑi], more typical of non-Southern varieties of American English. However, Rex also shortened the nucleus of (ay) in his performance speech (1998a:
67). Schilling-Estes argues that this means that Rex is maintaining a contrast between the Ocracoke diphthong  and the  Southern monophthong,  e.g., [ɑ ], which is typical of North Carolina and other Southern varieties of English.
Schilling-Estes points out that this is interesting  because when Rex talks about  the local variant he only ever contrasts his performance with non-raised variants, and gives no indication that the monophthongs associated with Southern accents are a relevant contrast for him in this variable. In other words, performance styles can provide otherwise unknowable information about the social and linguistic significance of a variant.


Sur r eptitious  recor ding  and  other  ethics  issues

It is not possible for a sociolinguist to avoid the observer’s paradox by gathering their data in secret.  Surreptitious  recording,  i.e., using  a hidden  audio or video recorder  without the speakers’ knowledge, is not condoned by professional linguistic associations. It is also illegal in many parts of the world. Surreptitious  recording  is an abuse of the privacy of the people you are  recording.  You might think (as one  linguist I know did) that it’s OK to hang  the headphones of your personal  stereo around  your neck, and give the impression  that you have just paused in your listening, but in fact have the machine in record mode.
You might assume, as he did, that because you are just recording your family and friends, they would certainly give consent after the fact. But consent given after the fact is seldom free from some  degree of coercion, and it may be especially hard for friends and family to assert their rights or preferences in this respect because they may fear doing (further) damage to the relationship.
Jennifer  Coates,  one of Britain’s leading sociolinguists, has talked quite frankly about her attempt  to get retrospective consent from a group of close friends that she had secretly recorded.  She found that some of them were furious at her for the breach  of trust.



Participant observation
The practice of spending  longer periods of time with speakers observing how they use language, react to others’ use of it, and how language interacts  with and is embedded in other social practices  and ideologies. A means of gathering qualitative data rather
than quantitative data.

There are some forms of talk which it is generally agreed can be used  as data without getting  express permission  from the  speaker(s). This is talk that is already in the  public domain, such as media broadcasts, or oral history archives. Standards vary as to whether  or not recordings  made for one purpose can be used freely for another, or whether permission has to be sought again from the speakers. Most universities have research ethics committees and specific staff in departments who can advise on all these questions.


CHALLENGING STYLE AS  ATTENTION TO SPEECH

Many people were impressed by the consistency of the effects Labov had found across  the four different activities that he included in his sociolinguistic interviews. However, not everyone was persuaded that the differences observed  indicated that speakers were paying attention to their own speech. The British social psychologist Howard Giles had begun to look closely at the  role language plays in shaping  the  dynamics  of interaction  between groups  and between individuals.
Giles drew on principles that social psychologists had determined play a significant role in how people behave in intergroup and interpersonal interactions quite generally. Research had shown, for example, that people  tend to favour other members of their group (ingroup members) at the expense of members of another  group (outgroup  members),  especially in situations  that involve some form of competition.
Giles therefore suggested that Labov was wrong in attributing  speech differences across   different  styles  to  the  effect  of  speakers’ attention   to  their  own  speech. He argued  that social behaviour  is seldom  so egocentric,  and that interviewees would have interpreted their  sociolinguistic  interviews  as  intergroup  or interpersonal interactions. The distinction Labov made between informal speech (to the interviewer) and casual speech (to family and  friends)  was a move in the  right direction, but Giles argued  that  Labov’s paradigm did not fully grasp  or deal with the effects that our interlocutors  may have on the way we talk.
Giles suggested that all the stylistic variation was actually caused by speakers attuning, or accommodating, to the  norms associated with different  addressees. Attunement and accommodation will be explained in detail in Chapter 4. For our purposes at the moment, we can adopt a common-sense understanding of the terms; that is, speakers fine-tune the way they talk according  to the  situation  they find themselves in. And an important  factor  in determining how speakers make adjustments to their speech is who they are talking to. We are all aware  that we are expected to speak  differently when talking to friends than when talking to a teacher,  a judge, a call centre, etc. Learning to make the expected attunements to others  is part of the process of becoming  socialised  in a community of speakers, so it is very reasonable to assume that  such  processes might play a role in determining  how respondents speak  in sociolinguistic interviews.
Table 3.1 shows  how a change in the addressee alone may be associated with quite marked changes in the way a speaker talks. This shows how often a speaker of Bislama (the English-based creole spoken in Vanuatu) omitted subject pronouns when telling a story, first to his extended family after dinner one evening, and then to me only.
In Bislama, as in many languages, you don’t have to have a subject  in every sentence. Instead  of repeating a subject  with a pronoun like you do in English (1), when the subject stays the same  across  sentences, you can have a gap as in (2):

Table 3.1   Number of sentences in which Sale omitted a pronoun subject as a percentage of all clauses in the conversation (two different audiences).

Addressees



Extended family
MM only  

Speaker
Omitted pronoun
All clauses (%)
Omitted pronoun All clauses (%)  

Sale
N = 50
71
N = 40 62

(1)   The  captain told everyone to stay quiet. He waited until he thought it was safe. Then
he signalled an advance.
(2)   The captain told everyone to stay quiet. _Waited until _thought it was safe . . .

Table 3.1 shows  that Sale generally omitted subjects in both tellings of the story, but you can see  that he used more full subjects when he was telling me the story (even though it was, in fact, the second time he had told it to me). There are a number of reasons why he might use  more subject  pronouns  with me than he does  with his family. But one plausible factor is that by using more pronouns  he is providing a non-native speaker with more overt information about who he is referring to in any given sentence. That is, he is trying to make the story clearer and easier  to follow for the specific person  he is addressing.


Style  as  ‘attention  to  others’

Giles suggested that many of the effects observed in the New York studies might be caused by speakers attuning their speech to the more salient aspects of the context. These include the interviewer himself, a university-educated person  (conducting  a ‘study’, no less!). Giles noted that by inviting such a person  into their home to conduct a study, the interviewee had already established a willingness to help out and accommodate the needs of the researcher. It would therefore be  a small step  indeed  (in terms  of interpersonal relations)  for the interviewee to continue their accommodating manner and to attempt  to produce  the kind of speech they perceive to be most appropriate for the different tasks  of the interview.
This means that the difference between informal and casual speech can be seen quite simply as a function of who the speaker is addressing, rather than pushing this dynamic to one remove (as Labov did) by proposing  that a change in addressee changes how much attention  the  speaker pays  to her/himself. Furthermore,  when  asked  to perform  non- conversational tasks  like reading  aloud, an accommodating person  would be very likely to attune  their behaviour  to the  norms they have been  socialised  to associate with reading aloud; that is, careful, school, or testing environments.
An important aspect of this alternative view of the way speakers shift between styles is that it foregrounds the importance  of the speaker’s and addressee’s relationship  and their attitudes towards one another. It presents a picture of speakers in which they come across more as thinking agents with interpersonal goals and desires than they do in the attention to speech model. In subsequent chapters,  we’ll see  that an emphasis on agency  and the dynamic quality of group and personal identities has become very influential in sociolinguistics. Many sociolinguists  believe that people  are  really doing different  things  when  they use

different variants (see, in particular, discussions of communities  based on shared practices in Chapter  9).
However, Giles’s work – and most of the work following his lead – has been done within the experimental traditions of social psychology; that is, it has relied on data elicited under highly controlled circumstances rather than on the kind of naturally occurring speech favoured by sociolinguists. This is one reason why it has proved difficult to convince some  sociolin- guists that a speaker’s attitude  to and relationship  with their addressee can and should be incorporated into models of variation and change.
One of the best-known proponents of the Gilesian view of variation has been Allan Bell, the sociolinguist who refined Giles’s insight and tailored it more directly to the predictive and explanatory  interests of sociolinguists.  Bell believed  that  Giles’s arguments captured a very powerful dynamic in sociolinguistic variation. Moreover, he saw its applicability beyond face-to-face interaction, and he argued that it could even account for phonological variation in radio broadcasts where announcers have no single or immediate addressee, and instead must be speaking with some kind of Gestalt idea about their audience as a group. He called this broader framework for analysing variation audience design.


Audience  design

The term audience design both classifies the behaviour (the speaker is seen as proactively designing their speech to the needs of a particular audience) and encapsulates the presumed motive for the behaviour (who is the speaker’s audience).
Bell had recorded several newscasters working for the national radio news network in New Zealand. These  news  readers would read on two of the government-owned stations, one of which was a middle-of-the-road, popular music station, and one of which was the clas- sical station. The classical station generally attracted an audience from higher socioeconomic brackets, while the popular station attracted a broader range of listeners, including those from lower social classes.
Bell examined the occurrence of several variables, including the realisation of intervocalic
/t/. Between vowels, (t) can be realised  as either a stop or a flap in New Zealand  English. This means words like better  and city can either  be pronounced [bεtə]  or [bεɾə] (better or bedder)  and [siti] or [siɾi] (city or ciddy). Although the news was essentially  the same  on both stations,  and although  the newscasters were exactly the same  speakers, Bell found that there were more of the conservative, stop variants when the newscasters were reading on the classical  station  than when they were reading  on the popular one. He argued  that because the topics were held constant,  and because the activity was the same, a plausible way to account for the differences was to assume that the newscasters were attuning their speech to what they believed the norms were for the different radio audiences.
This rather  modest  assertion actually has  a very strong  theoretical  claim embedded within it. It claims that an individual’s style-shifting (intraspeaker variation) derives from the differences probabilistically associated with different  groups  of speakers (interspeaker variation). If this is in fact true, then it means we can make a very specific prediction about the newscasters. If we compare the frequency of the innovative flapped variant of (t) in a single newsreader’s speech on the popular and the classical shows, this intraspeaker variation will be less than the difference between the frequency of the innovative variant in the two target audiences. We will return to this prediction shortly and see what support has been found for it in empirical studies.

Dif fer ent  audience  types

Bell’s framework  made  another  helpful contribution  to the way sociolinguists  might apply principles  of accommodation and  convergence to sociolinguistic  variation. This was  to distinguish  between several kinds of audience that a speaker might be thinking about. He suggested that a person  we are directly talking to has the greatest impact on how we talk. This person  is our ‘addressee’. But we also have to take other listeners  into account when we are speaking,  and he proposed that we distinguish between ‘auditors’, ‘overhearers’ and
‘eavesdroppers’. Each of these other kinds of listeners would have progressively less and less
influence  on the way you speak, as is shown in Figure 3.5.






Eavesdropper

Addressee

Auditor

Speaker
Overhearer




Figure 3.5   The strength of the effect of different interlocutors (known, ratified and addressed)  on a speaker’s choice of variants and different styles. (This representation  is based on a figure in Hay et al. 1999, with kind permission of the authors.)


Bell proposed a system  for distinguishing between these different kinds of addressee by using three criteria, whether someone is known, ratified or addressed. This is summarised in Table 3.2. An addressee is known to be part of the speech context, ratified (that is, the speaker acknowledges their presence in the speech context) and is addressed (that is, ‘I’m talking to you’). If a teacher praises  a student in front of the  whole class  she  or he  is communicating with both the student and the rest of the class (for whom the student is being held up as a model). Both the student and the class  are the audience.  The student is the addressee (known, ratified and addressed) and the rest of the class are auditors (known and ratified, but not addressed).
So audience design  predicts  that  the  speaker will attune  their speech most  to an addressee, next to an auditor, and then to any overhearers who the speaker thinks might be lurking around. The speaker will attune their speech less to auditors, overhearers and eaves- droppers because the speaker’s relationship with them is more attenuated, and consequently the speaker has less clear relational goals. The speaker may also have much less detailed ideas  about what kinds of people  their auditors  and overhearers might be, and this in turn means that the speaker will have less specific ideas about how they might attune their speech.

Table 3.2   Different types of audience according to their relationship with the speaker. (Bell 1984: 160, table 3.)



Known
Ratified
Addressed  

Addressee Auditor Overhearer Eavesdropper
+
+
+
–
+
+
–
–
+
–
–
–

By the time we move to the effect of eavesdroppers, we are talking about an audience that the speaker can probably only conceptualise in very rudimentary ways, so their effects will be very superficial in linguistic terms, e.g., we might be careful about the general topic or we might try to avoid swearing, but we won’t alter our pronunciation or syntax much at all. Finally, Bell argued,  any effect  that the topic of conversation might have would also be extremely limited. Under this framework, topics would derive their effects from a speaker’s stereotypes about who they are likely to be talking to when a topic comes  up.



Connections with  theory

A related approach to the idea of style-shifting as audience design is Coupland’s (2001) suggestion that speakers use different styles to present themselves differently according to the context or who they are talking to. We might call this speaker design. The main differences between speaker and audience design  models  of style-shifting  lie in what kinds of motives or goals are ascribed  to the speaker and which are assumed to drive variation. For example, speaker design  is readily compatible  with the accentuate the positive and eliminate the negative motives (Chapter 2), but less so with the idea that one is testing  hypotheses about  others  (It’s a jungle out there).  Speaker design  is also compatible with style-shifting where there is no independent evidence of a change in the speaker’s attention  to their speech or their audience.


Speaker design

A further approach to analysing
style-shifting. Stresses the speaker’s  desire to represent her/himself in certain ways. (See also Acts of identity.)
Relationship between social and linguistic constraints

Figure 3.6 summarises how Bell conceptualised the relationship between social variation (that is, variation between groups  of speakers) and stylistic variation (that is, variation in a single speaker).  This figure shows how, according  to Bell, intraspeaker variability derives from the variability that differentiates social groups:

(variation between groups) > (variation in individuals)

Bell predicted  that because it derives from social group differences, the variation any one individual shows  in their speech will never be greater than  the  differences between the groups that their style-shifting is derived from. This may seem  somewhat odd, since a group


Range  of variation between speakers from different social groups  (interspeaker  variation)








Range  of variation in an individual speaker (intraspeaker variation)

Figure 3.6   Bell’s predicted relationship between  linguistic differences across  social groups and an individual’s variability across styles.

is made  up of many individuals, but one reason for it is that when a speaker attunes their speech to the  norms  of a group  that  they  do not personally  identify with, they  will be approximating  a target  that they have only limited firsthand  knowledge of. The variability triggered by topic alone should be even less than an individual’s stylistic variation since topic effects are supposed to derive from the speaker’s pre-existing  variation space.
A few linguists have tried to test the details of Bell’s audience design framework. Dennis Preston (1991) reviewed some of the data available to him using this framework and found that his data  generally concurred with Bell’s proposed ranking of constraints on variation. However, he also considered the relative strength of purely linguistic constraints on a variable. As is generally the case  with a linguistic variable, the linguistic constraints have the most powerful effect of all on a variable (we noted this, in passing, for the Martha’s Vineyard data in Chapter  2).
This means that, for instance, the nature of the following or preceding sound, or whether the variant is a subject  or an object, accounts for far more of the variation we can observe than any social factors, such as who a speaker is talking to. The effect of any social constraints defines a smaller range of variation than the linguistic constraints do:

(linguistic factors) > (variation between groups) > (variation in individuals)

However, overall, empirical support  for this ordering  of factors  is still rather  shaky. Other researchers have found evidence  that leads  them  to conclude  the  opposite  to Bell and Preston – namely, that for some  variables the range  of a speaker’s style-shifting  exceeds the differences between social groups. John Baugh’s work on African American Vernacular English in Los Angeles seemed to indicate that certain variables may rank social and individual factors  differently. Some  variables may show the  relationship  between interspeaker and intraspeaker variation that Bell suggested, but for some variables speakers’ style-shifting may outweigh even the effects of linguistic constraints (Baugh 1979, cited in Rickford and McNair- Knox 1994). Thus, for the variable use  of third-person singular -s and non-prevocalic  (r), Baugh found that:

(variation in individuals) > (variation between groups)
exercise
exercise

Baugh suggested that whether a variable shows greater effects for linguistic or non-linguistic constraints depends on whether it carries much semantic information. He suggested that the two variables showing this pattern  – (r) in non-prevocalic  environments (the same  variable as Labov studied in NYC) and third-person singular agreement on verbs (which may or may not be marked with -s, i.e., she walks or she walk_) – do not carry any crucial information.
For instance,  there  are  relatively few minimal pairs created by variable presence or absence of [r] in English, so they will seldom cause confusion in conversation. Third-person singular agreement on verbs is actually redundant in English because there  are very few contexts  in which we do not have a full noun or pronoun as the subject  of a sentence. So this, too, plays a very limited semantic role. Baugh  suggested that when a variable carries limited informational or semantic load, it might show more variation within the  speech of individuals than it does  between groups.
Note that by invoking linguistic factors  as a primary consideration in order to account for the  reversal  of the  non-linguistic  constraints on style-shifting, Baugh  implies a fuller ordering of factors than even Preston did:

(linguistic factors: high information/semantic load) > (variation between groups)
> (variation in individuals)

(linguistic factors: low information/semantic load) > (variation in individuals)
> (variation between groups)

We still lack a lot of the detailed  comparisons between individuals and groups  to test Bell’s framework  thoroughly. A lot more research needs to be done  in order to find out whether  Bell is right in suggesting that there is a systematic or predictable relationship that derives intraspeaker variation from interspeaker variation. In the next two sections we examine empirical studies  that have tried to test the predictions  of Bell’s framework.


Integrating  topic  shifts  in  audience  design

Many people have found that in the informal part of a sociolinguistic interview, speakers use  more of the standard or more conservative variants  when  they are talking about language than when they are talking about  the supernatural. How could you account for this in terms of Bell’s notion of audience design?
Do you think this is a more satisfactory account than one  based on attention to
speech?



‘Attention  to  speech’  vs  ‘audience  design’?

Labov sees style-shifting as a linguistic reflex of changes in the amount  of attention a speaker pays to their speech.
Do you think Bell’s audience-design approach simply restates style-shifting  in terms of how much attention speakers are paying to real or possible listeners? Or do you think the differences between the two approaches are more fundamental?

Recipr ocal  audience  design?

Before leaving the topic of audience design, it is worth noting that the interviewer may also be attuning  their speech to the audience.  Peter  Trudgill conducted a major social dialect study in his home town of Norwich and he found that, quite unconsciously, he had used more regional Norwich variants when he was talking to speakers who were themselves high users of those  variants than he had when he was interviewing Norwich speakers who were low users  of those  variants (Trudgill 1986).
This raises the question of whether people consciously make these kinds of attunements or accommodations, and how much control they have over the success of their attunement. We will come  back to these questions in the next chapter,  where  we examine  some  case studies  showing that what people do and what they think they are doing may be somewhat at odds.


Attunement to dif fer ent addr essees: Foxy  Boston’s case  study

The most serious test of Bell’s audience design has been undertaken by John Rickford and Faye McNair-Knox in a case  study involving repeated interviews over a period of years with a single speaker, a young woman they call Foxy Boston. Rickford and McNair-Knox’s (1994) study represents an important milestone  in the study of variation, and of stylistic variation in particular, because it draws  both on detailed  knowledge of an individual and quantitative information on the  linguistic behaviour  of the  relevant  social groups  that  the  speaker’s addressees represent.
Rickford and McNair-Knox decided  the use  the interviews that had been  conducted with Foxy Boston as the basis for a rigorous test of Bell’s proposals. They compared Foxy’s use  of a number  of sociolinguistic  variables  when  she  was talking to Faye, an African- American, and when she was being interviewed (on very similar topics) by an Anglo-American interviewer, Beth. Rickford and McNair-Knox looked at several variables that are well known to favour one  variant in African-American English and a complementary variant in White American English. Audience  design  predicts  that  with variables  like these,  Foxy would accommodate to the norms of the different groups when her addressee is a member of one or other of these groups. That is to say, she would use more of the variants that are typically used by African-American speakers (as a group) when she was talking to Faye, whereas she would use more of the variants typical of White American English when she was talking to Beth (and therefore fewer of the African-American variants). This follows from Bell’s idea that style-shifting derives from social differences.
Two of the variables they looked at are shown in Table 3.3. They are the presence or absence of the verb BE and the presence or absence of third-person –s agreement on verbs. Examples of the variables and the variants that are statistically more likely to realise them in African American  Vernacular  English (AAVE) and  White American  Vernacular  English (WAVE) are shown in Table 3.3.
There is a considerable body of work on AAVE showing that absence of third-person agreement and absence of BE are reliable markers of African American Vernacular English, differentiating speakers of varieties of AAVE from vernacular varieties of White Americans. Because the  differences between AAVE and WAVE are  so extreme  for these variables, Rickford and McNair-Knox had to determine that Foxy wasn’t just switching between different

Table 3.3   Two variables studied in Foxy Boston’s speech, showing the variants statistically more likely to be used by speakers of African-American Vernacular English and speakers of White American Vernacular English.



Variant more common in:  

Variable
African American Vernacular
White American Vernacular  
English English  

Copula and auxiliary BE
BE absent
BE present  
He _ a teacher. He is (He’s) a teacher.  
She _ going downtown. She is (She’s) going downtown.  
You _ crazy. You are (You’re) crazy.  
Third-person –s agreement Subject agreement  absent Subject agreement  present  
She think you don’t care. She thinks you don’t care.


grammars (one AAVE syntax and one WAVE syntax) when she was talking to Faye and Beth (see  Chapter  6 for more on code-switching).  They took a lot of care to make sure that they really were comparing the same  variables in the two interviews.
They did this by determining  that when Foxy did use third-person singular agreement or BE that she did so in the same  linguistic contexts  in both interviews. Having determined that they were in fact comparing  Foxy’s use  of the same  variables in both interviews, they examined  the frequency  of the variants associated with AAVE norms; that is, absence of agreement and absence of BE. They found that Foxy did, indeed, have more absence of BE and –s agreement when she was talking to Faye than when she was talking to Beth. They interpreted these results as support for the notion of attunement through audience design.
But audience design  made a stronger  claim than simple attunement. Bell argued  that the range of variation covered by a single speaker’s accommodation to different addressees will be less than the range of variation between the groups that those addressees represent. Rickford and McNair-Knox found support for this crucial component of the audience design framework. The difference between how often Foxy omitted BE and –s agreement when she was talking to Faye compared to when she was talking to Beth was less than the differences between the frequency of these variants in African and White American Vernaculars taken
from larger social dialect studies. This is shown in Table 3.4, where the final column shows


Table 3.4   Differences in frequency of copula and -s agreement  absence for Foxy Boston talking to African-American and White American Vernacular English-speaking addressee. (Source, Rickford and McNair-Knox 1994, table 10.2 [modified].)







Speakers  of AAVE
Speakers  of WAVE
Difference
(in % points)  

Foxy Boston’s averages for absence with different  
copula 70% (N = 283) 40% (N = 176) 30 points  
-s agreement 73% (83/114) 37% (46/124) 36 points

addressee:
exercise

the difference (in percentage points) between when Foxy was addressing a speaker of AAVE
and a speaker of WAVE.
Data from these variables in the  vernacular  speech of White Americans  shows  that rates  for third-person singular -s absence are very low, and the rates  of copula deletion are virtually zero. If we sample across a group of speakers, AAVE has a rate of about 65 per cent third-person singular  -s absence and a rate  of about  38  per cent  copula  deletion  (these figures are taken from studies reported by Rickford and McNair-Knox, and in Labov 1972b). This means that the difference between the group averages for AAVE and WAVE speakers is greater than the range Foxy shows in her speech to an African-American and then a White American addressee.
There was less support  for Bell’s hypothesis  about the effect  of topic changes on an individual’s style-shifting. We noted  earlier that Bell proposed that the  influence  of topic would not exceed audience-based shifts. In fact, the frequency  with which Foxy used  the AAVE variants of BE  and -s absence varied more within each  interview (that is, depending on what topic she  was talking about)  than they did between the interviews overall (that is, depending on who she was talking to). So this result did not support  Bell’s hypothesis. The
role that topic plays in style-shifting still needs a good deal of study.



Testing  assumptions  about  topic-induced  shift

Sumittra  Suraratdecha tried to test  audience design’s predictions about  the effect of topic in the speech of Thai university students living in the United States. She recorded them with different interlocutors and compared how often they switched back and forth between Thai and English during their group conversations.
She found that code-switching from English into Thai happened most often when the students were speaking about the supernatural, and switching from Thai into English happened more often when they were talking about  travel and what they wanted to do after university.
Why do you think the code-switching might have gone  in that direction  for those
topics?



So Rickford and McNair-Knox’s study with Foxy provided a fine example of the ways in which quantitative studies of variation in groups and detailed analyses of one person’s stylistic repertoire  could complement each  other. As far as testing audience design is concerned, it provided mixed support  for the specific hypotheses of the framework. In addition, Rickford and McNair-Knox were cautious  about claiming that their data from Foxy showed  that she was attuning her speech for her addressee because they realised they had not controlled all social variables in their study.
The interviewers, Faye and Beth, certainly differed from each other in their ethnicity, but they also differed in how well they knew Foxy. Faye and Foxy had met before and knew each other pretty well by the time they had the conversation on which Rickford and McNair-Knox based their study. However, Beth  and Foxy had never met before  their interview. So, as Rickford and McNair-Knox point out, it is possible  that differences in Foxy’s speech in the two interviews were  due  to differences in how familiar her addressee was, rather  than

differences in their ethnicity. In the next section,  we will see  how sociolinguists  might go about teasing these factors apart.


Attunement  to  groups  or  familiarity?

A lengthy and ongoing  study of a speech community in a small town in Texas  provided Patricia Cukor-Avila and Guy Bailey (2001) with exactly the opportunity to attempt  to tackle the  outstanding problem from the  Foxy Boston  study. Cukor-Avila and Bailey have been working in a small community, which they call Springville, since 1988. Over the years they have developed  close  friendship  networks  in the  community and are  now recognised by many locals as insiders themselves. They started out collecting data in recordings  similar to the classic Labovian sociolinguistic interview, but over the years their recordings have become increasingly informal and are often controlled by the residents themselves. Cukor-Avila and Bailey realised that they were almost uniquely placed to try and resolve the ambiguities left from Rickford and McNair-Knox’s study. Most of the residents of Springville are African- American, and this meant  that they could examine  the  same  AAVE variables in the  Foxy Boston  study. But their lengthy involvement with Springville meant  they could also control for speech with familiar and unfamiliar WAVE-speaking addressees. This was important because the  closer  they could match  the  social and linguistic variables of Rickford and McNair-Knox’s study, the more valid and the more informative the comparison  between the two sets  of data would be.
They decided  to focus on the speech styles used  by two African-American teenagers and one elderly African-American woman when talking to other African-Americans and with a  White  fieldworker. The  teenagers, Samantha and  Lashonda,  had  known  both  their interviewers for some time (and the interviews span a period a several years) so their familiarity with the two interviewers was high. Audrey, the older woman, was interviewed by her nephew and then also by two White fieldworkers that she had not met before. The differences in social factors characterising their interviews and Foxy’s interviews are summarised in Table 3.5. For completeness, Cukor-Avila and Bailey also indicate whether the speakers share community membership,  and whether  other  people  were present during the  interview, in case  these non-linguistic factors might be relevant.
Cukor-Avila and Bailey found that there were minimal differences between Lashonda’s and Samantha’s styles  in the  interviews conducted by an African-American and a White interviewer, both of whom had known them for a long time. In other  words, there  was no evidence that, when talking to a White addressee, they attuned their speech so as to use more of the variants typical of White varieties of American English. The familiarity of the interviewer appears to have overridden  any effect  that her ethnic  identity might have had. However, Audrey did exhibit style-shifting similar to Foxy’s. Since they only found the same addressee effects that had been  found with Foxy in Audrey’s interviews, Cukor-Avila and Bailey sug- gested that this indicates that Foxy’s lack of familiarity with Beth was more likely to have been the salient factor constraining  Foxy’s style-shifting than Beth’s ethnicity.
As stated originally, Bell’s audience design  framework requires  speakers to be aware of variation in the linguistic behaviour of different social groups  in the larger speech com- munity. An addressee is then perceived and categorised as a representative of one of those groups, and the speaker’s understanding of the characteristic features of the group’s speech is then used as the basis for designing a style appropriate to that addressee. This formulation of the framework gives primacy to social groups. However, the Springville data indicates that

Table 3.5   Comparison of social features characterising the interviews with African-American interviewees conducted by Rickford and McNair-Knox and Cukor-Avila and Bailey. (Source, Cukor-Avila and Bailey 2001: 256, 263, tables 1–3.)

Foxy’s interviews Samantha and Lashonda’s      Audrey’s interviews interviews



Faye
Beth
Local teen
Fieldworker
Audrey’s nephew
Fieldworkers  

Race of interviewer
Afr.-Am.
White
Afr.-Am.
White
Afr.-Am.
White  
Familiar? Yes No Yes Yes Yes No  
Community member? Yes No Yes No
(but 7 years’
association) Yes No  
Other people present? Yes No Yes Yes Yes Yes

speakers may principally perceive  their audience in interpersonal terms  (e.g., as a familiar friend  or not), and  that  this  kind of personal  information  trumps  group  information  in constraining  intraspeaker variation.
A lot more work is needed testing  and replicating  these kinds of studies  in order to determine whether  their findings are robust and whether  the conclusions drawn from them are generalisable across  speakers and communities. Many sociolinguists  today believe that style-shifting  is part of ongoing  interpersonal negotiations, and it is these interpersonal negotiations that ultimately give reality and meaning  to group identities  and group mem- berships. The studies  with Foxy and in Springville provide us with a basis for evaluating the relative importance  of group factors and interpersonal factors for a sociolinguistic theory of style. The difficulty that they demonstrate in controlling for all the non-linguistic factors that might be relevant shows that this kind of careful comparative work often requires a long-term commitment  to research in a community of speakers.
In turn, this is a good reminder  to us of the importance  of having sound  information about  the  social backgrounds of our speakers, and  an understanding of how they see themselves in relation to others  in the larger social matrix. Because Cukor-Avila and Bailey have years  of friendships  and experiences in Springville based on extensive  participant observation, they can draw on this information to add social and interpersonal meaning  to their analysis of variation.


CHAPTER  SUMMARY

This chapter has introduced stylistic variation – that is, variation within the speech of a single speaker – as a locus of sociolinguistic enquiry. It began  by looking at the way in which style has been  operationalised in social dialect surveys, including both the rapid and anonymous surveys and the different tasks  built into a longer sociolinguistic survey. The fact of stylistic variation is undisputed in sociolinguistics,  but there  are  some  disagreements about  its

underlying causes and therefore the  way in which this kind of variation should  be char- acterised. Intraspeaker, or stylistic, variation can be characterised as the amount of attention the speaker is paying to their speech, or as the speaker’s desire  to attune  their speech to their addressee’s perceived  norms. We have reviewed some of the more significant studies that reflect  both frameworks.  In doing so, we have highlighted differences in the methods required to test the two perspectives, but we have also drawn attention  to the fundamental difference in the role of the speaker.  The attention  to speech framework presents a picture of the speaker that is fairly egocentric, while the attunement and audience design frameworks see  speakers as co-participants in social and conversational interactions.
These  different  views of the speaker underpin  a tension  in sociolinguistics  between generalisations made across large social groups such as social class or age (as is associated with the study of a speech community), and generalisations relevant only to much smaller, and sometimes quite idiosyncratic, communities  that are  constituted through  members’ shared practices (these ‘communities of practice’ are defined  and discussed in more detail in Chapter  9). As Rickford and McNair-Knox showed, patterns that differentiate  groups can be used  to inform the details of an individual speaker’s performance. The potential for such complementarity will recur again in later chapters.
This chapter has devoted considerable space to explaining the methods used to analyse style-shifting. These  different methods are very important because they have implications beyond  the  study  of style alone. The methods associated with the  attention  to speech accounts of style-shifting model variation as something that reflects non-linguistic information (e.g., how much attention  the speaker is paying to their speech).  The methods associated with audience (or speaker) design, on the other hand, treat variation as constitutive of non- linguistic factors  (e.g., as selecting an intended  audience,  or attempting  to stress similarity and identity with an audience). These differences are central to ongoing debates about how sociolinguistics should develop. This is why we have started this introductory text by examining research on style; several  of the methodological and theoretical  issues raised  here  have echoes in later chapters,  where  we will turn our attention  to other  social constraints on variation.
In the next chapter, we develop further the notion of accommodation that was introduced in the discussion of audience design.  Although the focus  in Chapter  4 is more solidly on attitudes that speakers have to other groups of language users and the varieties of language that they associate with those  groups, discussions of multilingualism and code-switching in Chapter  6 again show how speakers balance considerations about the immediate needs of their conversation with their acquired  knowledge about what is typical or expected across society as a whole.


FURTHER  READING
In addition to the references provided in this chapter,  you may find the following specific readings helpful:

Biber (1995) – a corpus linguistics approach to register and style; uses principal components analysis.
Bell (1991) – a more wide-ranging  linguistic analysis of the forms and styles used  in the news media.
Giles and Coupland (1991) – on linguistic accommodation and attunement.

Coupland  (1984) –  a  study  looking  at  one  speaker’s stylistic variation with different addressees.
Schiffrin et al. (2001) – a range  of articles  providing different  perspectives on discourse analysis. Chapters in section  III deal with styles of discourse in different  domains  or contexts.