3.11.1 The analysis process: theoretical links to CHAT
In my Review of the Literature relating to this research, I stated that social constructionist theories informed the project, and that I was situating the research within a Cultural and Historical Activity Theory (CHAT) framework, a theoretical perspective itself lying within socioculturalism. In order to demonstrate how this framework informed the data analysis process of my research project I will explore those aspects of CHAT which make it a relevant framework for the analysis.
137 In the Review of the Literature (Chapter 2) I noted, with reference to historical perspectives, how the use of CHAT also allows an examination of the educational objective of the fiction texts young people are encouraged to read. Issues of quality can be examined in terms of both a cultural and societal educational objective and also in terms of status.
Edwards (2011) is illuminating on how CHAT can be particularly valuable in educational research. She notes that researchers who work with CHAT see context as integral to the analysis (p.1). This was certainly true in the case of this study, since interpretations of quality in fiction for 11 – 16 year olds were formed against a background of curriculum and institutional expectations. Similarly Edwards sees CHAT as useful in analyses of workplace activity, citing classrooms as a particular example. It is of relevance to the choice of CHAT as an analytical framework for this study that Edwards cites its value in uncovering how people use tool and material. Her comment that analysis through CHAT can look at how individuals interact and also:
the purposes, values and knowledge to be found in the practices in the institutions or systems they inhabit. (p.2)
This, too, resonates with this study. Additionally Edwards sees CHAT as a means of capturing the voices of those who operate within systems. However she notes that one advantage of the CHAT approach is that it allows focus on what happens in practice, very much at the heart of this research. The benefits of CHAT as an analytical tool are well summarised by Brown and Cole (2012):
A special virtue of the use of activity as an adjunct to, or substitute for, the concept of context is that it both forces attention to the historical dimension of the context/activity in question and allows a means of identifying crucial constituents of the phenomenon being investigated as they relate to each other (p.4)
This forcing attention to the historical dimension whilst finding a way of identifying crucial constituents of the phenomenon is key to this research.
138 Thus CHAT provided an umbrella framework for analysing the data. Questionnaires and interviews were considered separately.
3.11.2 Questionnaire data analysis
Cohen, Manion and Morrison (2000) note that coding is the primary form of data reduction (p.265). This means classifying and grouping answers. There were three key elements to the questionnaires: the factual data relating to the teachers’ professional biographies; the books recommended and the attendant comments; and the two open-ended questions regarding criteria for choosing a book for 11 – 16 year olds and teachers’ perceptions of quality.
Cohen, Manion and Morrison (2000) detail seven stages of analysing qualitative data: 1. Establish units of analysis of the data; 2. Create a ‘domain’ analysis; 3. Establish relationships and linkages between the domains; 4. Make speculative inferences; 5. Seek negative and discrepant cases; 6. Seeking negative and discrepant cases; 7. Generate theory (p.148). These are helpful steps and give a framework for approaching qualitative data. Strauss and Corbin (1998) offer three stages of analysis for grounded theory approaches: open, axial and selective. These follow the same pattern as Cohen, Manion and Morrison in that they move from establishing codes (the open stage); to creating links and relationships, the narrative (axial stage); and finally the generation of theory (the selective stage).
In order to generate codes it is generally agreed (Cohen, Manion and Morrison, 2000; Creswell, 2007; Sapsford and Jupp, 1996) that the researcher needs to become close to and familiar with the data by frequent rereading – the immersion described by Strauss and Corbin (1998). It is only by the process of iteration and reiteration (Cohen, Manion and Morrison, 2000 p.149,) that codes emerge (Glaser and Strauss, 1967) and can be refined and modified.
Whilst acknowledging the influence of grounded theory on qualitative data analysis, the analysis of the data reported here uses the approaches of grounded theory, particularly the three stage approach of Strauss and Corbin (1998), rather
139 than being situated within a grounded theory paradigm. This meant that codes emerged from questionnaire answers and were refined by an iterative process. This framework helped to make sense of the inevitably messy, confusing and fundamentally non-linear process of coding (Marshall and Rossman, 1989, p.55). However the analysis was not linear but dependent on frequent iteration and revision of codes.
Each question in the questionnaire was coded separately and transcribed manually in tabular form (Appendices 12, 13, 14, and 15). The open questions to do with a) criteria in choice of books (What do you look for in a book for your students to read?) and on perceptions of quality in books for 11 – 16 year olds; and b) definitions of quality ( How would you define quality in relation to books written for students at KS3 and KS4?) were coded, as described above, with the key word in the questions (here criteria or quality) acting as the open stage – in which the emerging axial codes were situated.
3.11.3 Interviews data analysis
All the interviews were transcribed from digital audio recordings. It is important to realise that the transcriptions are in themselves interpretations. The transcript turns the flexibility of the live interview into a closed conversation; that is, a conversation in which new avenues can no longer be explored. Kvale (1998) calls the transcript a bastard (p.182), meaning it is a cross between a living conversation and written text. He warns against losing the original spoken conversation in the reduction of the written transcript into butchered fragmented quotes (ibid). Whilst Cohen, Manion and Morrison (2000) are more moderate in their comments, they also warn against completely literal interpretations of the transcript. They remind us that data can be lost in the transcribing process and note that the transcription itself is an interpreting process. They include the need to be open to, for example, tone of voice and inflection; silence; the mood of the interviewee; pauses and interruptions. It is also important to note interviewer bias at any point in the phrasing of questions or response to answers.
140 One way of addressing the hybrid nature of the transcript is to listen to the recordings in addition to reading transcripts. However at all times it is important that the researcher, who has been completely immersed in all the stages of the research and participant in it, must nevertheless attempt to take an objective stance and to be open to their own potential for biased interpretation. I tried to do this.
The interview transcripts were analysed using a computer programme QSR Nvivo-9. Creswell (2007) notes the advantages of using a computer programme to analyse qualitative data. Here he cites advantages as: a) the potential to store and organise data conveniently; b) the potential to locate easily passages of text relating to codes; c) the possibility of comparing code labels; d) the potential to sift, shift and reorganise codes hierarchically; e) the manifestation of a visual representation of the codes; f) the potential to annotate codes and transcripts with linked memos; and g) the flexibility for the researcher to create their own coding template to fit the analysis model chosen.
However, in line with a phenomenological approach, I was also mindful to put aside my own experiences in order to minimise my own impact on the data as the codes were created. At this point reflexivity was crucial as I sought to be objective and minimise what may be termed researcher impact. I was aware, as Finlay (2002) describes, that reflexivity is challenging and difficult but I aimed to be aware of the potential for bias in creating the codes and open about the process. Constantly revisiting the codes and being open and willing to adapt and change was part of the process.
I found the flexibility of the computer programme helpful to the organising of the data and particularly because of the flexibility offered in the creation and structuring of the codes. In the first instance I took quality as the key code. It became apparent that there were inevitably many versions of quality and Nvivo allowed for the creation of multiple and interrelated codes.
I have explained at the start of this section how CHAT related to this project in giving a framework for analysing an activity, here 11 – 16 year olds’ reading of
141 fiction, within a cultural and historical context. Once the initial, open or inductive, codes were in place, axial, deductive, codes were created and finally placed within a CHAT framework. In Chapter 4, Findings, p.170 I demonstrate how the codes were developed and how they link to a CHAT analytical framework. I have drawn on Twiselton (2004) for this, beginning with a CHAT framework used by her to analyse practice. I have developed this framework for the research reported here and I explain this in Chapter 4. In Chapter 5, The Discussion, I also show how the interpretations of the data were also developed though a further aspect of CHAT that is boundary objects (Star and Griesemer, 1989; Wenger, 1998; Engeström and Sannino, 2010; Gorodetsky and Barak, 2009) which proved significant in this research.