• No results found

5.3

The next phase of the analysis was inductive, drawing on grounded theory (Glaser & Strauss, 1967) to engage with the interview data, and determine which topics the experts focused on during their interviews. This approach is appropriate for descriptive and theory-building research (Berg, 2009; Bogner, et al., 2009; Deming & Swaffield, 2011; Kvale & Brinkmann, 2009), as well as issues of rapid social change, where existing theory cannot keep pace with evolving knowledge and practice (Flick, 2014). Hence, its suitability within my research, where there was no evident pre-existing framework that I could apply to synthesise my data into a cohesive whole, what with its diverse disciplinary origins, personal and professional components (i.e. the ‘grey area’ discussed in 3.3.2 and 3.3.3), and my goal of identifying pragmatic pathways to creating socio-ecological change (itself a dynamic and evolving research area).

Grounded analysis is an interpretive process, involving identifying key words and concepts from the data, and sensitively developing categories and themes that reflect these (Suddaby, 2006). As discussed above, conducting the interviews and transcriptions myself was invaluable to the depth of understanding that I achieved and could bring to the analysis. This was furthered by my choice of manual data coding51 (as opposed to software based alternatives), which I discuss further below. By the time I had drafts of each transcription, I had already spent many months immersed in the data, and had a strong and nuanced grasp of what had emerged from the interviews, both individually and as a whole. Obviously, this continued to grow as I proceeded.

I used a ‘bricolage’ approach to coding the interview transcripts (Kvale & Brinkmann, 2009). This approach is prevalent in interpretive analyses for meaning-making research, and uses an informed “interplay” (Kvale, 1996, p.203) of techniques, as appropriate, to bring out connections in the data and generate structures of relevance to the research aim (Kvale & Brinkmann, 2009). As part of this, I applied a series of different analytical techniques, which I detail next.

There is no standardised method for such analyses, which are non-linear and require an iterative approach (Glaser & Strauss, 1967; Kvale, 1996). In this context, iteration involves constant comparison of the original data and research aim with the emergent codes and interpreted themes, which are refined accordingly (Suddaby, 2006; Tracy, 2013). In my first round of analysis, this involved repeated consideration of the interview data. In the second round, it was further complicated as I began the parallel development of the ecological habitus framework as a way to

theoretically conceptualise the interview data. Kvale and Brinkmann (2009, p.111) describe this expositively as “spiralling backwards”, where the novel insight generated through analysis compels revisitation of earlier analytical stages in a series of iterations. As described in previous chapters, this corresponds to the approach Bourdieu applied in his own research. In my case this occurred throughout the coding, mapping, and thematising stages of the analysis, and involved an iterative, non-linear (i.e. recursive ‘spiralling’) consideration of the interview data. Simultaneously, I was aligning the interview analysis with the theoretical development and its complexities. These dynamics are discussed more in the next sections.

Iterative analysis is a “deeply reflexive process” (Srivastava & Hopwood, 2009, p.77) consistent with the methodological foundation of this thesis. Reflexive iteration allows the more sophisticated understandings that only emerge in the later stages of the analysis (sometimes months or years into the research, as I discovered) to be captured within the analysis and correspondingly woven cohesively into the research results (Luker, 2008). This enables complex, insightful, and refined meanings to emerge from the data (Scott, 2004; Srivastava & Hopwood, 2009). Simultaneously, I was developing my capacity for and understandings of reflexivity, and specifically, the ecological strain of reflexivity discussed and developed in the next chapters. In addition to gradually focusing my analysis upon the themes that I ultimately sought (in response to my research aim), this added layers of richness to my understandings and interpretations as I proceeded with my analysis.

As highlighted earlier, conducting the research independently, using manual coding, and building on my understandings throughout these steps supported me in developing sensitised analysis and outcomes. This contrasts to more mechanical applications of coding, which often lack “the spark of creative insight” (Suddaby, 2006, p.638) sought in grounded research. My approach furthermore enabled a closeness to the data that is lost using coding software (Knight & Ruddock, 2008).

These approaches also presented some drawbacks. Qualitative research requires extensive data analysis, often necessitating a research team (Kvale, 1996; Witzel & Reiter, 2012), and my independent and manual approaches (in contrast to employing software and/or assistants), compounded the onerousness of the undertaking. Interview analyses are often “terminated because of time limits or exhaustion” (Kvale, 1996, p.188), before thematic or analytic saturation occurs or the desired level of understanding is achieved. The analysis certainly took much longer than I ever anticipated, and my understandings continue to grow despite nearing completion of my PhD.

There is also a converse risk, with manual techniques, of becoming too close to the data and losing sight of the “wood for the trees” (Johnston, 2006, p. 323, as cited in Knight & Ruddock, 2008). Coding and mapping the data visually (illustrated below) enabled a degree of metacognition (i.e. perception of the ‘bigger picture’), which was valuable to this end (Kvale, 1996). The use of reflexive iteration throughout the analysis further led to my considering the data from a multitude of

angles and at different stages of the research, both formally as reported here, and informally in my everyday life. Over a protracted period this provided me with what I view to be an ample perspective of the data, contextualised robustly within the larger research frame (i.e. in terms of theory and in terms of my aim), without necessitating mechanical abstraction, such as software-based coding.