6.4 Data Collection Method
6.4.8 Data Analysis
Two key questions helped to guide the process of analysis. The first was a reflective question of ‘why am I reading this in this way?’ suggested by Gill (1996) in her research on black identity using discourse analysis. This question is designed to facilitate the critical distance required to identify discourses, so that the multiple
‘authors’ and ‘listeners’ in a text can be identified (Gavey 2011). The cultural framework that the researcher brings to the interview interacts with and draws from the researcher’s own cultural context and the discourses associated with it. The process of analysis aims at moving between two contextual categories, as outlined by (Terre Blanche et al. 2007b), the one focuses on the micro-context of the conversation and the other the macro-context of ideologies. As discussed in section 6.2.2.1 on my ideological position as the researcher, my own contextual framework is informed by my narrative and the cultural and societal references of a white woman who grew up in South Africa during the 1970’s and 1980’s. The ways in which I am positioned as similar and different to the research participants, in relation to social phenomena, such as culture, class, ‘race’, education and gender also was taken into account in my reading of the text, alongside an attention to their context (Fine 1994; Jorgenson 1991; Wilkinson &
Kitzinger 1996). It is important, according to Terre Blanche et al. (2007b) that since researchers are also a part of the text’s context, they need to account for their role relative to the text. This is acknowledged both in sections 6.2.2.1 on the ideological position of the researcher and 6.5 on trustworthiness in the research.
160
The second question that guided the research process and was critical in the rigour of analysis, was identifying the effect of the text, by asking the question ‘what does the text do?’ (Terre Blanche et al. 2007b). As constructionism is not interested in identifying the ‘truth’ behind the text, its aim is to link accounts to actions and patterns in the way the text functions to consequences (McMullen in Wertz et al. 2011).
Edwards (2006: 46) in his journal article on discourse, cognition and social practices, claims that “Whatever people say is always action-oriented, specific to its occasion, performative on and for its occasion, selected from an indefinite range of options”.
Feminist researcher, Butler (2005), supports this performative notion of language.
Identifying the aims of the text are at times more obvious or explicit than others and texts may do a number of things at the same time. In his analysis of narratives, Gregg (2006) highlights the power of using structural analysis for analysing self-representation and the interplay between deep and surface structures as a means of identifying implicit structures of identity. In order to convey accounts of who and what we are, Silverman (2010) argues that constructionist researchers need to move from ‘what’ and ‘how’ to
‘why’ questions to understand the varied contexts out of which we draw from experience. Discourse analysis involves a close scrutiny of language to examine the ways in which certain themes and topics are discussed, allowing some ways of thinking, and undermining and excluding others (Burck 2005; Parker 2014). Discourse analysts ask questions about language which try to establish the following (Wetherell et al.
2001):
What actions the conversation performs
What accounts individuals are trying to create in interaction with each other How accounts change as the context changes
In attempting to build on the established tradition of discourse analysis, Antaki, Bilig, Edwards and Potter’s (2003) six key shortcomings of discourse analysis were avoided.
These describe how researchers often fall into the trap of under-analysis through:
Summary rather than analysis
Spotting features in the text rather than discourses Over-quotation or isolated quotation
Taking-sides in the debate
161
Circular identification of discourses and mental constructs where the analysis returns to what is stated in the text
False survey where the data is treated as if it is true of all members in the category
These are elaborated on in Antaki et al.’s (2003) journal article and the researcher’s approach in discounting them are discussed below. In the initial phase of the research while discourse clusters were being identified within and across interviews, summaries were helpful as preparation for analysis and consolidating insights, but did not constitute the analysis itself. ‘Spotting’ features rather than discourses was avoided by following the techniques recommended by Terre Blanche et al. (2007b) as discussed earlier in section 6.4.7 on organising data, which correspond with the criteria recommened by Parker (2014). The identification of binary opposites; recurrent themes and metaphors; and subjects within the text helped to identify the discourses. Thereafter the ‘effects’ of the text and ‘how’ the conversation was being constructed as opposed to simply ‘what’ was being said became the focus of analysis. In the discussion of the results, significant extracts are presented, followed by in-depth analysis and care has been taken not to over quote or allow quotes to ‘speak for themselves’.
Another potential pitfall of discourse analysis is to take sides by offering sympathy or scolding for a particular position in the discourse. Gavey (2011) emphasises that feminist discourse analysis allows for the complicated and conflicting desires and motivations expressed within a text and it is not the role of the analyst to debate these.
While the aim of the research is to critically analyse emerging models of power, the conclusions drawn will critique these models in light of transformation efforts in organisations within a feminist framework. While the researcher may be interested in transformation of traditional patriarchal structures as an outcome of the research, the critique focusses on the implications of the discourses for organisations rather than on the discourses themselves or the positions of the participants.
To avoid circular debate, which results in an observation from the text simply going back to the original assertion, attempts have been made to enhance the observations with accumulated insights relevant to previous data. By analysing each participant’s text independently and discussing them in relation to one another, rather than simply
162
reducing them to commonalities, care was taken not to treat the data as if it is true of all members of the category. This would result in false survey and is another potential shortcoming in the practice of discourse analysis. A heuristic approach, as recommended by Professor Merriam in her seminar on Hands-On Data Analysis at the University of KwaZulu-Natal (2008) was used in this process. A heuristic approach refers to the process of learning from experience and being able to apply that learning back to the process of analysis. This meant that as the techniques became more familiar, the process was refined and revised.