The categories, and underlying concepts, are likely in the early stages of analysis to be imprecise and relatively poorly defined. They may be pretty mundane and descriptive. Blumer (1954) distinguishes between such ' sensitizing concepts' and ' definitive concepts'. The latter are based on precise definitions and have a position within an explicit model. When you have some confidence in the set of categories as providing a description of the group, although not yet at the stage where a definitive status has been achieved, the next step is to explore their inter- relationships.
While grounded theory views this as a purely inductive process which is driven by the data, the more usual ethnographic stance is wider. Clearly the data, and your understanding of relationships within the data, constitute an invaluable resource. However, pre- existing theory, previous empirical research, your own expectations and hunches can also play a part. Hammersley and Atkinson (1995, p. 214) cite work by Bensman and Vidich ( 1960) where ten different theoretical positions were used to provide the focus for their analysis, each time asking the question : ' what in [ these] theories would permit us to comprehend the data?'
Realism, theory and ethnographic studies
As discussed in chapter 6, early ethnography adopted a ' naive' realist position which was subjected to lethal constructivist criticism (see p. 29). The more sophisticated ' subtle' advocated by Hammersley (1992, pp. 50-4), is largely in line with realism as discussed in some detail in chapter 2, and appears to provide a defensible approach to the analysis of ethnographic studies. Note, however, that, as pointed out by porter (1993, p. 595), there are differences between subtle and critical realism. In particular, the former is concerned with understanding the perspectives of others rather than judging them, whereas Bhaskar (1989) asserts that critical realism entails evaluation, which he sees as an essential part of social research.