Author: Sharon A. Bong
School of Arts and Social Sciences, Monash University, Malaysia

Working with data

'Transcription is theory' (Ochs 1979). By transcribing in-depth interviews, a process both tedious but fulfilling, you begin theorising. You not only familiarise yourself with the data by revisiting the scene as it were, but you also begin the first steps of data analysis as you begin to 'see' links between or among theoretical categories that you began your project with, in this case, 'sexuality' and 'religion'.

I find it problematic when I hear from both novice and seasoned researchers alike that they only transcribe what they feel is 'important'. As a grounded theorist (Glaser and Strauss 1974, Strauss and Corbin 1990, Charmaz 2000, 2006) this approach to generating data risks forcing data to fit theory (deductive) rather than building theory from data (inductive). Transcribing is akin to the production of texts as opposed to working with extant (already available) texts. It is therefore vital not only to offer a faithful transcription of the conversation shared between interviewer-interviewee but also to listen not only to voices but also noises in the data during the analysis process.

The textual data generated is huge-an interview of 1½-hour's duration generates almost 30 pages of text (Times New Roman, 12-size font, 1.15 line spacing). So an integral aspect to working with the data is managing the data; organising and making sense of it. To students whom I have supervised, I encourage them to use a form of CAQDAS (Computer-Assisted Qualitative Data Analysis Software), ATLAS.ti specifically, to more effectively manage their data that comprises the dual-process of de-contextualization and re-contextualization (Tesch 1990: 115-127). De-contextualising data serves the purpose of breaking down data into manageable bite sizes for researchers to chew on. This is done principally through coding which is the smallest unit of meaning for each quotation or extract of the textual data. When all 30 interviews are coded, what emerges is a list of family codes-codes, i.e. religion-Muslim, religion-Christian, religion-Tibetan Buddhist, religion-Hindu, religion-New Age spiritualist and religion-non-believer. This list is comparable to a compression of hundreds of pages (30 pages per interview multiplied by 30 interviews) of textual data. Data is de-contextualised as I study this apart from the textual data (interview transcripts), for a while, in order to map out what constitutes research findings of the project, in other words, the story that is waiting to be told.

Data must then be re-contextualized within their primary source which is the interview transcript and secondly, with other family codes-codes towards theory building (where theory is the relationship between or among categories). I apply this by studying the list of family codes-codes in tandem with their respective list of quotations (with its source fully cited by ATLAS.ti's code-and-retrieve function, i.e. which primary document/ interview it comes from, which quotation number and lines).

The researcher drives the interpretative qualitative analysis of his/her research. Coding that is enabled and made fun through the use of ATLAS.ti in my experience, serves as a 'heuristic device' (Coffey, Holbrook and Atkinson 1996, paragraph 7.7., Seidel and Kelle 1995: 52-53). I have elsewhere (Bong 2007) expounded on debunking myths in CAQDAS use and coding in qualitative data analysis.

Back to Project Home Page