Introduction

This paper will attempt to discuss some issues related to the way in which certain forms of discourse analysis – a broad term here, referring to the various ways in which social-scientific disciplines represent and analyze stretches of text – make, define, construct and/or represent data. The cover term I would like to use for the whole complex of data-related activities is ‘data formulation’. Data formulation often proceeds almost automatically, on the basis of received wisdom and as part of our perception of professional behavior. It is a custom rather than a method.

The target of my comments, and the starting point of a broader and more general inquiry into ‘data’, will be one specific practice, shared mainly among scholars in conversation analysis and commonly called ‘data sharing’: the circulation of more or less decontextualized corpora of ‘prepared’ (i.e. transcribed, sometimes annotated and/or translated) data among members of the scholarly peer group, thus making the data accessible for analysis by scholars other than the one(s) who recorded or elicited the data and put them in a conventional format for data representation (transcription). The suggestion, commonly used in data sharing, is that as long as data are correctly transcribed on the basis of a shared set of transcription conventions, these data ‘tell their own story’ and can be analyzed successfully by other scholars. This practice and its epistemological and methodological assumptions (or indeed its theoretical decisions) are, I believe, problematical, and I will attempt to explain why.

This reflection was prompted – apart from an incident in my own professional practice, to be discussed below – by participation in a workshop on Discourse Analysis in Amsterdam, March 1997, which used data sharing as one of its main structuring tools. All of the participants to this workshop were sent three transcripts of recorded fieldwork interviews on interethnic relations in New Zealand, gracefully made accessible by Margaret Wetherell.