Data analysis

Once the transcriptions were checked for accuracy, the data analysis focused on the transcribed words, the written text.

Jenny Horseman and Mary Hamilton (2003, personal conversations) have commented on the difficulty of describing the data analysis process especially when teaching how to do it. One could describe the process of coding, and finding themes, but in reality, they argued, there is a point where the process is like a "black hole." One has to go through it to understand what it is that needs to happen, and the more times one does it, the better one gets at doing it even if not at describing it. Perhaps this is why some authors complain about researchers not reflecting enough on the interpreting process and its difficulties, pretending that the process is smooth and without contradictions (Fontana and Frey, 2000).

As I finished each interview, I reviewed the tapes and took notes about particular comments that seemed to be important. I would then incorporate questions or probing comments about these concepts into the interview protocol. In this way, I made sure I collected data about the topics that were emerging as relevant from the participants' viewpoints. This was the beginning of the coding process. For example, one interviewee articulated the notion that many community-based researchers do the same research they would do if they were based in the university. I was intrigued by this notion because it shifted the emphasis from the organizational location of the interviewee to their understanding of research. In subsequent interviews I mentioned this understanding to interviewees and gathered their thoughts about it. Later on, this became an important theme in my writing.

Once I had all the transcripts I read and re-read the interviews, coding participants' words line-by-line (Charmaz, 2000; Ryan and Bernard, 2000). This process allowed me to stay attuned to participants' ideas and expressions (Charmaz, 2000) while starting to interpret their words.

The initial codes allowed me to engage in a rigorous process using the constant comparative method, which, following Charmaz (2000), includes five aspects. I compared each code as it appeared in each interview, looking for similarities and differences and trying to make sense of what these meant. I also compared ideas from the same individual to look for inconsistencies. I made sure that I balanced the words of each interviewee when using parts of the text to represent her or his ideas⁸. As some themes and categories started to emerge I compared each code to the categories and compared the categories with each other.

⁸ For example, one interviewee referred to the "world of academia" and the "world of teachers" in dichotomous terms, emphasizing that one is for thinkers; the other one for doers. Although he dedicated quite a lot of time to describing these opposed worlds, he also argued that this was an artificial construct. Had I not taken into account his ideas as a whole and used only the coded segments, I could have argued that he had a dichotomous analysis of the relationships between academia and society.