Axial coding
Axial, or theoretical, coding is about linking together the categories developed through the process of open coding. Glaser and Strauss, the begetters of grounded to be taken when trying to establish these relationships. Strauss and Corbin (1998) work within an exclusively interactionist paradigm, where axial coding is viewed as leading to an understanding of the central phenomenon in the data in terms of its context, the conditions which gave rise to it, the action and interaction strategies by which it is dealt with, and their consequences. Glaser (1992) takes a more purist grounded line. He argues that the axial codes, and the form that they take, should emerge from the data rather than being forced into any particular pre-determined format.
Whichever line is taken, axial coding is about putting together again in some way the data which have been effectively split apart into categories by open coding. As Mertens (1998) puts it:
During this phase, you build a model of the phenomena that includes the conditions under which it occurs (or does not occur), the context in which it occurs, the action and interactional strategies that describe the phenomena, and the consequences of these actions. You continue to ask questions of the data; however, now the questions focus on relationships between the categories. (p. 352)
If you are simply concerned with exploring or describing the phenomena being studied, this completes the analysis. However, grounded theory, as the term suggests, seeks to go further. For this, you need to go on to selective coding.
Selective coding