Hanna de Jong-Markus

Chapter 2 44 the analysis started by listening to the audio files and reading the complete data set in order to get the whole picture of the phenomenon, while checking the transcripts. The insights and understandings that began to emerge were written down as reflections (cf. Elliot & Timulak, 2005). I subsequently edited the data by omitting obvious redundancies, repetitions and unimportant digressions (cf. Elliot & Timulak, 2005). Also, the data were uploaded to Atlas.ti24, the software programme used for coding. Next, the data were divided into distinctive meaning units, namely parts of the data that stand out of the context and could communicate sufficient information to provide a piece of meaning to the reader (Elliot & Timulak, 2005). This was followed by organizing the data, by assigning headings on the basis of the meaning units’ objects (Elliot & Timulak, 2005). Some headings were categories from the interview guide (e.g. ‘Tolerance’ and ‘Religious others’), others were new (e.g. ‘Refugees’). The meaning units with the same headings together were called a domain. The headings were defined in a code book (see Appendix VII). The definitions were continuously re-adjusted and compared to the meaning units, in order to reflect the content of the meaning units and domains clearly and comprehensively. Like all further coding, the coding was done in Dutch, because there would be a loss of meaning if the codes and concepts were translated into English in this phase already (cf. Van Nes et al., 2010). The domains and the definitions of the headings gave a general overview of the contents in the data. After several discussions among the research team, based on the reflections of the researcher it was defined which domains were up for more elaborate analysis in subsequent phases and from which perspectives. In this process, the overall research question was the guide to see which were important topics that showed up at several points in the data, like one recognises a musical theme. Based on the domains, the original research questions were adjusted and the themes of the various sub-studies were delineated. For example, originally we did not foresee that the differences between the pedagogical agents would be such a relevant issue. At this phase it was also decided that the scope would be widened towards religious diversity instead of focusing on religious tolerance, as participants elaborated much more on the broader area of religious diversity instead of focusing on tolerance only. Eventually, almost all domains were involved in one of these delineated sub-studies. For each sub-study, the meaning units of the specific domains were thematically coded one by one (cf. Braun & Clarke, 2006; Elliot & Timulak, 2005). This process was like a dialogue of the researcher with the data (cf. Elliot & Timulak, 2005): on the one hand I tried to use codes close to the 24 www.atlasti.com (initially version 7, later on version 8)