Social categories and theoretical descriptions
It's all a matter of variables •••
Correlating linguistic variation as the dependent variable with independent variables such as linguistic environment, style or social categories. is the primary empirical task of sociolinguistics.
(Chambers 1995: 17)
[S]ocial variables generally, including ethnicity, are usually untheorized or undertheorized and used in whatever way is convenient for the linguistic analytic purpose at hand.
(Milroy 2001: 237)
This claim by Chambers, articulates a well-established position in sociolinguistics, and there are two aspects of it that we want to address. Firstly, we wish to argue that the constitution of independent variables based on social categories is itself a task of theoretical description. Thus, the identification of such categories as 'independent' becomes immediately problematic. Secondly, and in line with the view of the stratified social ontology that we have elaborated in earlier, it implies that the linear variable correlations often adduced in sociolinguistic research provide an insufficient description of social reality. Where such models are useful, we want to suggest, is in indicating empirically the 'traces' of causal relations. We shall explore this idea in the context of beginning to suggest how a realist epistemology can inform methodology in sociolinguistic research.
Of the 'independent variables' listed by Chambers, it is the group identified as 'social categories' that will be our main concern. As any introductory text on sociolinguistics will explain, the social categories which are most often the concern of sociolinguistic research include age, sex (or gender), social class and ethnic or racial group. These categories, not surprisingly, coincide with those that are used in many other areas of research across the social sciences, but in recent years social scientists have begun to question their validity. One line of critique has come from those writing within a postmodernist tradition, where the challenge to a notion of objective social categories is part of a larger attack on any concept of objective truth and scientific rationality. Another source of objection is from those writers, influenced by social constructionism, who are wary of accepting a priori categories in the social world, preferring to explore how it- and everything in it- is constructed in texts and talk (e.g. Potter 2000).
The approach we are proposing may perhaps be seen as steering a middle course between the empiricist ontology that underlies many conventional quantitative studies- if only implicitly- and the strong relativism implied in full-blown constructionism. While we do wish to defend a modest notion of objectivity in social research, we also recognize the limitations of empirical research based on correlations of variables, and we are critical of its tendency to hypostasize social categories, and of its linear view of variation. Our perspective is consistent with that of a number of social scientists researching a range of issues in the social world, who have turned to realism and complexity theory to provide answers to the question of how we can know that the variables we name and use represent reality (Archer 1995; Layder 1997, 1990; Sayer 1992; Williams 2000).
Social categories such as 'class' or 'ethnic group' are not, of course, brute facts (Searle 1995). Take, for example, the categories listed on the 'equal opportunities' forms issued by employers to prospective employees. These categories are relatively recent: they did not exist until they were devised by particular people for specific purposes. This is one reason why debates about ethnic monitoring have proved so durable (and inconclusive): they are about who has the power to define whom and for what purposes, and the resulting categories carry within them a network of implicit propositions about what is salient for these purposes. For instance, the ethnic monitoring forms currently in widespread use by public service employers in England today distinguish ethnicity in terms of heterogeneous criteria: 'White', 'Black other' (skin color); 'Indian', 'Pakistani', 'Bangladeshi', 'Chinese' (belonging to a nation-state- although it is unclear whether this is meant to imply affiliation to, birth in, or parents' birth in, one of these states); 'Black African', 'Black Caribbean' (skin color plus compound geographical regions).
The process of monitoring employment patterns in relation to the responses selected from this list presents a number of problems: Is membership of these groups ascribed by others or self-chosen? If people are given the opportunity not to answer, how valid can the exercise be as a measure of patterns of job selection? Since the categories are not of like kind, how can they serve as the basis for a single categorization? We shall return below to the theoretical incoherence of ethnic categories for the purposes of social research, but we introduce some of these problems here because they provide a particularly striking example of the fact that social categories cannot be regarded as given, independent variables to be deployed in the way that correlative models imply.
Of course, it is not only social categories based on ethnicity that are permeated by interconnected- and often implicit- propositions. Another widely used social category in both research and policy is age. Legislation and social practices pertaining to the rights and responsibilities of people at specific ages vary widely, historically and geographically. Although there is an unavoidable physical, biological dimension to the ageing process, the significance attached to this is not determined by it. Attaining the age of 50 means many different things in different parts of the world and does not automatically entail any of them. In other words, all categories rest on propositions. In the instances cited, for example, these include implicit claims about correlations between geographical origin or skin color with way of life, or cultural attachments; and between chronological age, emotional maturity and responsibility.
All social categories, then, are forms of description. Those employed by social scientists need to be 'theoretical descriptions' (Layder 1990; Pawson 1989). Thus, the challenge facing social researchers is, at the least, to be conscious of the claims embedded in the social categories they use, and, preferably, to form their categories on the basis of a coherent theory. When the theoretical basis of social categories is made explicit in this way, they can be seen as 'theoretical descriptions'.
