It is achieved by carrying out three kinds of coding:
1. Open coding to find the categories;
2. Axial coding to interconnect them; and
3. Selective coding to establish the core category or categories (strauss and Corbin, 1998).
Open coding
Here data (interview transcripts, field notes, documents, etc.) are split into discrete parts. The size of the part chosen is whatever seems to be a unit in the data: perhaps a sentence, or an utterance, or a paragraph. The question asked is: ‘What is this piece of data an example of ?’ The code applied is a label. It is provisional and may be changed. A piece of data may have several codes (labels), i.e. it may be considered to fall within more than one conceptual category. Labels can be of whatever kind that seems appropriate, including descriptive (e.g. ‘accepting advice’), ‘in vivo’ (i.e. a direct quotation from the data), or more inferential.
These conceptual categories arise from the data. Using pre-determined coding categories and seeking to fit data into such categories is against the spirit of grounded theory. However, this distinction is somewhat metaphysical as the ‘conceptual baggage’ you bring to your data (whether derived from a pre-existing theory or from analysis of data collected earlier) will inevitably have some influence on what you are likely to ‘see’ in the data.