Social categories and theoretical descriptions
Aggregates, collectives and sociolinguistic research
'Ethnicity'-based categories
Whilst some social categories can therefore be constituted by researchers as both social aggregates and social collectives, others are more troublesome. This is particularly the case when we are dealing with aspects of social reality that do not lend themselves readily to empirical measurement and where, moreover, actors' interpretations are critical. Adolescents may be defined aggregatively by there searcher as all individuals between the ages of13 and 18. Researchers may take the decision to focus on this age-group because of a belief (derived from previous research) that adolescents tend to use language in particular ways. However, using language in these ways is not simply an irresistible stage in 'sociolinguistic maturation' (Kerswill 1996b); it involves a degree of choice. This is why actors' understandings are a central element in the theoretical description of social collectives. 'Jocks' and 'Burnouts' are descriptions that rely on social actors' adoption, and maintenance, of distinct linguistic features. Their adoption of this form of social action suggests that the individual is not only 'party to' the norms and rules relevant to being a member of the group, but has also made some assessment of the potential benefits and losses attaching to group membership.
It is considerably more difficult to disentangle aggregate and collective categories when we use descriptions such as 'racial' or 'ethnic' groups. On the whole, 'ethnic' is preferred to 'racial' as the term for these kinds of categories when they are used in sociolinguistic studies, although some appear to use the terms interchangeably (e.g. Rampton 1995b). Milroy (2001) draws some distinction between the two by an appeal to the learned, cultural components of ethnicity and the physical characteristics of race. However, the distinction is elided by her claim that race denotes those physical characteristics 'which are treated as ethnically significant' (p.237), and she is quick to note the salience of discrimination by majority groups against minorities for the identification of both racial and ethnic groups. Similarly. Fought (2002) suggests that'... ethnicity is not about what one is, but rather about what one does', and points out that'... the category of race itself has historically been socially constructed, and is extremely difficult to delimit scientifically' (p.444). Nevertheless, we find her use of the concept of ethnicity ambiguous and inconsistent, and the argument we develop here is equally applicable to 'race' and 'ethnicity' as a basis for a researcher's social classifications (d. Carter 2000).
How are people to be placed in 'ethnic' categories when ethnicity is not an empirically observable feature of human beings? If this is done by skin color (which is problematic, as it is not a categorical variable), or place of parents' birth, 'ethnic group' is being employed as an aggregate category. It therefore implies nothing about commitment to sets of norms or conventions (of language, dress, custom or habit, although all or some of these may be observable amongst many members of a particular category). If people are classified into 'ethnic groups' with reference to some notion of identity or culture, then the term is being employed as a collective category. The researcher then faces the familiar problem of making judgements about who is party to which sets of norms and conventions, and who is not.
We would claim that this problem arises because itis not possible to distinguish between ethnic groups as aggregate categories and ethnic groups as collective categories, and that this inevitably leads to confusion about what it is that is actually being measured. Researchers are then unlikely to be able to operationalize concepts with the precision that they need. They may seek to side-step the issue by simply adopting commonsense terms and definitions, or by deploying the self-definitions used by their informants, but, as we argue below, such a move raises a number of problems for the analysis of any findings.
To take one recent example, Fought (1999), in her study of 'Chicano English', is concerned with a sound change which is taking place in both the 'minority' and 'majority' communities of California. Her article classifies speakers with reference to ethnic categories- as it must do, given her commitment to 'the need to analyze variation within the context of those social categories that are of particular significance to the specific community being studied' (p.5). These categories, like those from the UK discussed above, are of several kinds and are resistant to empirical verification. They include: 'speakers of Anglo ethnicity', 'Anglos', 'Anglo speakers', 'the majority Anglo community', 'the California Anglo community', 'Puerto Ricans', 'Puerto Rican speakers', 'a Latino community', 'Latino young adults', 'Los Angeles Chicanos', 'the minority Mexican-American community', 'speakers of other ethnicities', 'minority communities', 'minority groups', 'minority speakers', 'African-Americans', 'African-American local communities', 'the black and white communities'.
Our disagreement with Fought is not about her study as a whole, but rather with the use of ethnic categories in this kind of research, of which her study is but one example. In order to establish correlations between use of a specific linguistic feature (in this case, lui-fronting) and membership of a social group, the researcher identifies the group and those groups with which it contrasts- a priori. Since the membership qualifications of 'ethnic' categories are notoriously difficult to pin down, as we have argued, the boundaries of the categories to be compared are inherently unstable.
For speakers themselves, this analytical problem may be of very little significance. In order to go about one's social business on a daily basis, it is quite satisfactory to define oneself and others- when it is relevant for particular types of interaction- in terms such as 'Anglo', 'black', 'Puerto Rican' or 'Latino'. And indeed such self-definitions are important for the analyst in interpreting such phenomena as strong identification with one social group- and the language associated with membership of it- rather than another. However, from our perspective, it is quite clear that there can be no logical necessity between belonging to such an imprecise category as 'Anglo' or 'Chicano' and using language in particular ways. Firstly, the category is inherently imprecise. It is not clear whose definition of it is to be authoritative, nor which attribute(s) of the category are to count as definitive. Secondly, speakers are not bound by their membership of any social category to use one set of language features rather than another. Coming from a Puerto Rican background may predispose speakers to use one pronunciation rather than another, but there is no logical or necessary entailment for them to do so. For us, the concept of 'ethnicity' is conflationary, in that it fails to distinguish between culture, structure and agency.
