In sociolinguistics, where identity tends to be our first explanatory resource, Peter Trudgill's claim that identity is “irrelevant” as a factor in his area of interest is particularly striking. There are at least three questions here. The first is Trudgill's direct concern: whether identity considerations impinged on the development of new national colonial varieties. The second is the underlying question of whether identity, in itself and in general, can stand as a motive for sociolinguistic action and change. The third is whether face-to-face linguistic accommodation, which Trudgill invokes as the core process through which new dialects come to be, can and does function in the absence of identity considerations.

I will not comment on the first question, other than to make one obvious point. Even if we reject the consolidated motive of “wanting to signal a national identity” as a plausible explanatory factor for new-dialect formation, this does not rule out other motivational and identity-related considerations from being relevant to the same processes of change. Trudgill's skepticism – asking how disparate people could possibly have formed an identity pact to represent themselves indexically as a community through a new way of speaking – certainly seems well founded. But to rule out all issues of identity, particularly in circumstances of demographic movement and cultural mixing, seems unnecessarily restrictive.

In relation to the second question, I would want to agree strongly with Trudgill that there are dangers in running too freely to causal explanations around identity. Sociolinguistics has suffered from reductive theorizing around identity, in simplistic purposive accounts of identity-as-motive (“people use variety x in order to mark their allegiance to community x”), in much the same way as biology has sometimes overinvested in functional explanations (“mammals developed wings in order to fly”). I would want to go further and argue that it is inadequate to construe identity as an independent variable that either does or does not determine language change in particular settings. Trudgill appears to accept the principle of causal determinism (that identity can cause change) when he argues vociferously that this isn't what is happening in the colonial case. But his argument that identity is likely to be as much a consequence as a cause of sociolinguistic practice is convincing, and I come back to this shortly.

So what about accommodation and identity? In 1986, when Trudgill theorized dialect contact and change in terms of interpersonal accommodation, I commented in a review that this was a valuable rapprochement between variationist sociolinguistics and sociopsychological approaches to interaction. It opened a perspective on human agency and relational processes underlying linguistic change. It acknowledged that it is people, not merely dialects, that are in contact, and that interpersonal and intersubjective dimensions of language use are where explanations for change must lie. In the present essay, Trudgill still finds an explanation for linguistic change in “how individual speakers behave linguistically in face-to-face interaction,” but, even more than in the 1986 book, he empties out the human and agentive values of interaction by arguing that interpersonal accommodation is automatic and mechanistic, and indeed genetically preprogrammed.

In contrast, since Howard Giles's first (1973) conceptualization of accommodation theory, human motivations and identities have been among its focal concerns. Meyerhoff 1998, for example, suggests that communication accommodation theory (CAT) mainly exists to address the interactional nature of identity construction, and this is closely echoed in other reviews (cf. Coupland & Jaworski 1997). We might even think that CAT has tended to be over-explicit about the motivations underlying accommodation and about the identity management strategies that are entailed. A typical formulation of CAT's core predictive claims is that people will converge to each other's speech characteristics if and when they want to improve communication effectiveness and/or to boost social attractiveness (cf. Giles, Coupland & Coupland 1991). This sort of rubric sets up categories of people and of circumstances where speakers' motives regarding social relationships vary, and mainly the contrast between people wanting to get on with others versus people wanting to symbolize their distinctiveness from others. Speakers linguistically “move closer” to their speaking partners if they want closer relationships, and so on. Identity is central to claims like these, for example because “boosting social attractiveness” means presenting a self that is more attractive to another person. Identity goals and relational goals are reflexes of the same intersubjective processes in social interaction, and Howard Giles developed accommodation theory very much in tandem with his theoretical work on social identity.

So CAT has certainly not assumed that accommodation (in the sense of mutual linguistic convergence) is automatic and mechanistic. When Trudgill writes that “accommodation is not only a subconscious but also a deeply automatic process,” he is challenging the fundamental claims of the model. Accommodation theory is a model of strategic operations around linguistic and communicative styling, committed to building predictive claims about the links between identity and relational aspirations/priorities and styles, and then between styles and social consequences. This is why we can call accommodation theory a “rational action” model (Coupland 2001:10). It invests social actors with a degree of control over their linguistic actions, and a degree of awareness and understanding of their probable social outcomes.

This is not to say that CAT hasn't considered the claim that accommodation is automatic. At least, there has been consistent recognition that there is a general “set” to converge – an overall propensity for people to “move closer.” But we don't have to interpret this tendency as mere automaticity, whatever that might mean. Literatures on social interaction have regularly formulated their own versions of the set to converge, from preference organization in CA (e.g., the preference for agreement) to Malinowski's phatic communion (talk in the fulfilment of a basic sociality) to Bell's audience design (stylistic adaptation for audiences). Mutuality and convergence in social interaction are – as a very gross generalization – the default design principle for social encounters, even though this design is resisted in many particular circumstances, which CAT labels “divergence.” But the very fact that we recognize the social impact of divergence, of dispreferred interactional moves, and so on, is sufficient to establish that even predictable and normative convergence is relationally meaningful. When the distinguished cognitivist Joseph Cappella writes that interactional synchrony meets the basic survival needs of bonding and comfort, he opens a window onto the strategic functioning of accommodation. He makes us realize that even when interactional stances and designs are “pervasive” and “relatively automatic,” they can nonetheless be designed around social outcomes. The apparent contrast between “automatic” and “strategic” practice is blurred.

All the same, Trudgill usefully shakes up our assumptions about rational action models and their handling of identity. There are certainly occasions when people purposively and rationally target identities for themselves in talk, and when people target specific relational outcomes of the sort that CAT deals with. Dialect is a rich resource for stylistic operations that can work toward these ends (Coupland 2007). But identity is often less coherent, less rationalized, more elusive, more negotiated, and more emergent than this. Identities are known to be often multiple and contingent. Constructionist epistemology has it this way as a matter of dogma, although it might be better to distinguish analytically between cases of greater or lesser rationality, emergence, and so on, on a case-by-case basis. But we will also need to analyze identity in its multiple dimensions, which we might summarize as “knowing, feeling, and doing.” The current theoretical vogue is to interpret identity as practice – as a form of doing and discursive achievement – and this is echoed in Trudgill's comment that identity might be seen as a consequence of accommodation rather than its “driving force.” But we should not ignore the dimensions of “knowing how to be”/competence, and feeling/affect too. Ben Rampton, for example, runs with Raymond Williams's interpretation of class identity as a structure of feeling. Identity, as Frederick Barth said, is often more a matter of establishing boundaries – sensing and displaying who we are not, rather than who we are – and accommodation often works negatively, in the strategy of “avoiding being different.” And if (as I have argued elsewhere) our selves are largely relational, then what we hold on to as “our own identities” is not clearly separable from how we believe others construe us.

These and other complexities mean that the sociolinguistics of identity will need to be nuanced, to catch the delicacies of interactionally constructed selves and relationships. And by implication, identity will rarely if ever be irrelevant.