Social categories and theoretical descriptions
Aggregates, collectives and sociolinguistic research
Age-based categories
Let us now consider the relevance of these two kinds of social category for research in sociolinguistics, taking age-based categories as a first example. One well-researched topic is the occurrence of distinctive features in the speech of adolescents (e.g. Cheshire 1982; Eckert 1988; Kerswill1996a; Romaine 1984). Studies on this topic usually start with the researcher identifying as salient the property of falling into a given age-range. From this perspective, 'adolescents' constitute an aggregate category, since there is no entailment (logical necessity) involved in belonging to the category 'adolescent' and engaging in any particular kind of behavior, including speech behavior. Aggregating people on the basis of a feature such as age is the choice of the researcher, and having that feature neither rests upon nor implies a common cultural understanding of what is involved in having it. In other words, people who just happen to be, say, 15 years old at a particular date are constituted as a group by the process of categorizing them as such. They do not have to understand what being an adolescent is in order to be categorized as an adolescent, but will become a member of the category involuntaristically. So they have nothing in common which is a direct product of being 'an adolescent', other than the contingent fact of being a certain age. This is the force of Chambers' argument, which Coupland summarizes as suggesting '... that age is not a truly social dimension at all, if by this we mean one that is socially negotiable' (Coupland 2001b: 188). As Coupland notes, however, this is a partial and unsatisfactory position for explanations of patterued linguistic behavior, and we suggest below how it might be modified in terms of the distinction between aggregates and collectives.
Research into the speech of members of the category 'adolescent' usually proceeds by analyzing variables at the levels of phonology, lexis or syntax, and establishing correlations between membership of the category and the tendency to use one of the variants. A finding that often emerges is of 'adolescence [as] the focal point for linguistic innovation and change' (Chambers 1995: 176). Kerswill (1996b), for instance, reports a tendency for adolescents to use ItI rather than I?I, even when style and context are accounted for. However, while the majority of speakers in the group studied display this tendency, there are invariably counter-examples. These findings often point to significant patterns, but they remain descriptions, as opposed to explanations, of the correlations found. We would suggest that it is at this point that the aggregate category 'adolescent' ceases to be useful in the process of looking for explanations of sociolinguistic phenomena, precisely because it is simply an aggregate of contingent features.
If we want to explain why many adolescents are in the vanguard of linguistic change, and why some are not, we need to introduce a strong notion of social agency- an acknowledgement that people have some degree of choice over what they do, including how they speak. This in turn requires the use of social categories that recognize the relevance of actors' own understandings. Those findings that appear at first to be anomalies in the statistical correlations become more explicable if the researcher can reconstitute the theoretical descriptions of social categories as collectives rather than aggregates. How 'adolescents' manage and negotiate an adolescent identity is quite another matter from how they get to be members of this category in its aggregate sense. If its members do behave in similar ways, as well they may, this is the result of cultural patterns and mores that influence 'doing being an adolescent', and not of their chronological age. Thus, especially if this social category is one which is meaningful to its members (if, for example, they have eagerly awaited the arrival of their thirteenth birthday and initiation into 'teenage' status, with the right to participate in activities deemed inappropriate for younger children, and so on), then they may behave as they think adolescents should. But any such similarities in behavior, values, or attitudes are not attributable to the 'common property that identifies the collection'- its aggregate status. Adolescents who subscribe to the value systems of either 'Jocks' or 'Burnouts', for example (Eckert 1988), do so on the basis of their own decisions (even if there is not an infinite range of options from which they can choose, and even if they see these as the only options available).
As research produces descriptions that are based on collective (as opposed to aggregate) social categories, it reveals properties of the interaction between structure, culture and agency. Thus, an investigation of why some young people use marked variants while others do not entails recognizing the role of the speakers themselves in this variation. However, moving away from an aggregative view of social categories presents the researcher with a different order of epistemological problem (which may be one reason why people frequently opt for an aggregative approach). The core of this problem is that the identification of collective categories involves the researcher in making judgements about who is 'party to' the norms and rules that constitute the category in question.
Such judgements will be less difficult in some cases than in others. To return to Greenwood's example, cited earlier, it is reasonable to assume that those people who are married have some understanding of what this description means (although, as we also pointed out earlier, this by no means entails that they will all have the same understanding). Constituting 'married people' as a collective category in this way does not preclude its deployment by researchers as a social aggregate category. Regardless of whatever meanings marriage has for the couples studied, once they become married, they are bound by the legislation applying to married people- including certain financial benefits or restrictions, for example. A researcher who is interested in a question such as whether there is any sort of correlation between being married and having a mortgage would use the category of 'married' as though it were an aggregate category; on the other hand, another researcher, interested perhaps in the causes of divorce, would be obliged to attend to marriage as a collective category, and ask people about their commitment to it as an institution. We are advocating methodological relativism here: which sort of category we employ will depend on what it is we are trying to find out.
Thus, it should be clear that in defining aggregate categories, less attention is paid to actors' own understandings. The feature defining membership of an aggregate category must be one that is contingent, external and measurable. You may see yourself as the youthful Adonis, but you've still been around for 50 years, and the researcher (or the insurance salesperson) will go by the birth certificate rather than rely on the personal PRo Analytically, then, with aggregate categories, priority is given to the analyst's own depiction of that category.
