The analysis of the discourse data will follow principles laid out in my earlier work on narrative development and emotion talk, programmatically summarized in Bamberg (1999a, 2000) under the header of positioning analysis.

Central to positioning analysis is the assumption that we typically take up positions in interactive settings (traditionally discussed as ‘roles’) and make claims as to how we want to be understood (also discussed as "the discursive production of selves", cf. Hollway, 1984; Davies & Harré, 1990, 1999; Harré & Langenhove, 1999, Langenhove & Harré, 1999). In a number of recent contributions this concept has been put to use for the empirical analyses of specific discursive domains (Bamberg1997b, 1997c, 1999a, in press a, in press b; Carbaugh, 1999; Schrauf, in press). Positioning analysis, constructed along those lines, builds on the general ethnomethodological principle to start from where the speaker starts, and to explicate what they themselves take to be worth explicating. The accounts we will be eliciting from the participants in our study on the explicated list of discourse topics, will simultaneously elicit different "degrees" of retrospection and involvement (in writing, in the personal interview, and in the group discussions). These deliberative self-positionings and deliberate positionings of others in the four different discourse settings explicate identity positions that to some degree will be contradictive, while in other respects orienting toward some consistency and coherence with regard to particular culturally explanatory models. Eckert’s demonstration that particular sound variants "are directly related to key burnout [one of the key cultural categories for high-school students] cultural themes: alienation to school, restricted substances, trouble, fights, and disagreements" (Eckert, 2000, p. 218) shows that this type of analysis can be extended to the realm of topics relevant for adolescents and children. It is our aim to show similar tendencies for the group of subjects younger than hers, paying closer attention to within subject variability (and contradictions), and using talk about self and others, the way it is grammatically and lexically fashioned, rather than vocalic (sound) variables.

In previous work on how children between 4 and 12 years of age position themselves vis-à-vis self and others in explicating sadness and anger, I was able to plot out a developmental ‘process’ in the way these positions were linguistically fashioned and explicated (Bamberg, 1997c, in press b). One of my arguments in this work was that the practice of "doing" emotion talk in different discourse modes actually may function as a developmental "mechanism". These insights have been the instigators for turning from the construction of emotions and value stances to the broader issue of identity construction, and for trying out positioning analysis in discourses of pregnant women diagnosed with diabetes (Bamberg, 1997a, 1999a; Talbot, Bibace, Bokhour & Bamberg, 1996).

Within the framework for this proposal, we work from the assumption that developmental analysis will have to start with identity (or better: identities) as positional stances, occasioned in situations, and put to use for discursive purposes. Analyzing these stances requires the analysis of the situational context and the way the interaction is ‘performed’. This type of analysis is inherently socio-linguistic and proceeds by analyzing talk in terms of how contexts are linguistically indexed. Following the framework laid out in Gumperz (1982, 1992), elaborated by Schiffrin (1987, 1994) and Eckert (2000), and extended by Duranti (1992, 1997) and Ochs (1992, 1993, 1996), eliciting talk as data from participants yields access to "native" local ordering activities. Whether this talk is produced in "naturally" occurring situations or in "artificial" interview situations, the analysis proceeds from a participatory perspective, offering insights into language practices where content and context are in the process of being constructed. ‘Language practices’ are understood here not as symbolic means of "having" or "transporting" meaning in or between monads, but as joint productions of what talk is about (its content) and what talk is for (its purpose).

Central to this approach to language analysis is the assumption that speakers "have many alternate ways of saying ‘the same’ thing" (Labov 1972, p. 188). Speakers’ choices of particular construction types (intonation, lexical or grammatical) indicate claims to particular memberships as well as stances toward particular categories. However, rather than assuming that these choices are indicative of symbolic knowledge systems which in turn are implemented with the discursive effect "in mind", I would like to suggest that they are residues of language practices that have been practiced elsewhere and now are locally instantiated to claim positions vis-à-vis a particular category. Thus, while traditional psychological analysis suggests that social categories (or knowledge thereof) are the resources for drawing up stances and positions and to express them in communicative situations, I am suggesting to reverse this view: Linguistic variation is the resource for the construction of social categories by drawing up positions vis-à-vis categories. And it should be obvious that the process of construing positions is tied first of all and in a very profound way into local places and local people, from where broader cultural categories can emerge.