Further sub-categorization allows for greater comparison and contrasting of different categories and can make the data sets more meaningful. Not all of these codes will be decided beforehand -- in fact, it can be more enriching for the final analysis to break down the data afterwards, to ensure that the lived experience of the subjects affects the coding process. Coding is often thought of in terms of word-based strategies of the subjects, and these can yield important assumptions about the ways individuals perceive their places in the world. Frequency of use of particular words, metaphors, analogies, and the use of local or regional phrases endemic to the area can all be flagged through coding and used to draw meaningful connections between apparently dissimilar sates of being (Gibbs 2010). By highlighting key words in transcripts, the researcher can physically have his or her eye drawn to meaningful bits of data. One argument for the use of linguistic coding is that if used properly it can offer more objective evidence, as the frequently-used words, transitions, connectors, idioms, and imagery are 'on the page' of transcripts for the observer.

Coding requires active, discriminatory analysis on the part of the recorder. For example, "if your respondent has been talking about the way her parents continued to give her financial support after she had left them and set up her own home, you can compare this with all the other ways that parents might support their children" (Gibbs 2010). By making this the salient feature...

But a different analytical approach, depending on the subject of the research, might be to code the situation as a symptom of the economic decline, and evidence of how the current generation of young graduates has had more trouble getting jobs than previous recent generations. This codes the behavior as economically- rather than psychologically-based, and again implies an analytical, discriminatory approach in 'labeling' behavior for further analysis.
Sorting requires the researcher to decide what is important and unimportant, and what behaviors can be connected. Some studies may connect apparently far-flung behaviors, like dependence on parents and depending on one's boss for direction at work, as evidence of a generation that is less self-starting than previous ones. Others might be narrower, and make use of broad-based, more obvious categories like 'leisure.' Regardless, the researcher must engage in constant self-scrutiny and questioning about what is suitable for categorizing and the codes used for the subjects: even when using very basic, exterior categories such as gender, the researcher must ask 'why is this code necessary to acknowledge in this specific situation?'

References

Gibbs, Graham. (2010, February 19). How and what to code. Online QDA. University of Huddersfield. Retrieved August 9, 2010 at http://onlineqda.hud.ac.uk/Intro_QDA/how_what_to_code.php