Analysing interviews

As explored elsewhere (see Chapter 14 in this volume and Davies, 2011), interviews can vary in style and format. I had adopted an approach that used semi-structured interviews to allow for a naturalistic discussion that could be steered at times largely by the respondent. This method of interview can lead to lengthy discussions and conversations that lean towards storytelling. In interviews of this nature, respondents slip into a narrative that is guided by their own recollection and experience. The interview elicits responses that are grounded in their own language, or, as in my own interviews, the recollection of language as used by others. Thus, the strength of

interviews is their flexibility and their potential for prompting interviewees to speak freely. My respondents relayed their perceptions of who is affected by child sexual abuse, how they are affected and what their needs are, and what services/support were available. I thus gained insight into the wider impact of CSA on non-abusing family members.

I was the principal investigator (PI), the author and owner of the project. I also conducted the research myself. I searched for and gathered the literature and docu- ments myself, negotiated and secured ethical permissions and access and carried out the interviews in person. By doing the interviews myself, I had a significant head start in terms of knowing my data. In other research projects, I had been only one of those engaged in fieldwork. Reading transcriptions of interviews as conducted by others and listening to audio-recordings of interviews carried out by members of your research team give you less of a first-hand feel for the data. Grasping the content and potential themes is rendered a more distanced exercise. Whilst some might argue this demands a greater level of transparency about how analysis is conducted because there has to be a clearly articulated strategy for team members to follow, which pro- vides for a more objective analysis, for others it introduces a mere veneer of objectivity, greater distance from the rich data and greater likelihood of ‘meaning’ being diluted, misinterpreted or lost. Though there are ways of safeguarding against the latter tendencies, for example by the use of memos, diary or field notes which allow you to capture impressions, thoughts and ideas, such notes also need to be understood by your fellow research team members and they too are susceptible to different interpretations. In qualitative research, these aide memoires are perhaps best adapted for the purpose of recording reflections that may be helpful in ensuring the transparency of how the analysis unfolded in a way that is intimately connected to the raw data (see the section below on reflexivity and critical reflection). In Box 12. 3, an example is provided of a research project that involved several researchers and where data analysis included several inductive phases.

(Continued)

with CSA histories perceive as barriers to disclosure? The analysis focused on data related to disclosure barriers. Conventional content analysis took place over a nine- month period. This approach is appropriate when research domains (such as barriers to disclosure of CSA for men) lack an extensive body of literature or sub- stantial theory development. Throughout the process, descriptive codes were developed that represented the main ideas pertinent to the research question. The analysis included several inductive phases:

1. Multiple readings (immersion) and initial coding of four transcripts to allow the researchers to develop an overall sense of participants’ perspectives. Independent line-by-line analysis and weekly meetings where researchers read and analysed transcriptions aloud, compared notes and emergent codes (disclosure barriers, disclosure facilitators, treatment strategies, help-seeking resources, mental health impacts) in order to reach a consensus.

2. First-author coding of the remaining transcripts. Revisions of code list (new addi- tions, collapsing of codes, refinement of definitions). Validity checks by the team, review of analytic decisions, development of codebook (4 categories, 12 codes, 25 subcodes).

3. Researchers independently reapplied the same code set to all data. Coding revised to nine and categorization decisions resulted in three categories being identified and defined: