3. Methodology and theoretical framework
3.3 Methods for conducting a Frame analysis
3.3.2 My approach to data analysis
In accordance with these principles, the method of analysis used in this study is explained step- by-step below. My initial focus changed to some degree throughout the study as assumptions were tested and discarded. Key among these was the experience of stigma from mental health professionals which has been reported by mental health service users in other studies. Initially, I was interested in exploring whether these experiences could be better explained in the light of frame disputes. During the analysis, however, it became clear that there was little evidence of psychiatrists behaving in ways that could be easily identified as stigmatising or patients reacting in ways that suggested they felt stigmatised. Such an approach would have required the creation of a definition of ‘stigmatising’ behaviours in the interaction which would have lacked validity without access to either participant to assess intention and reaction. Whether because of the nature of these particular psychiatrists (see further discussion on the data
below) or because of the fundamental challenges of attributing attitudes to behaviours, it became clear that this focus could not be supported by the data.
Two broad questions did remain throughout the study. The first focused on participants’ – particularly patients’ – construction of identity and role within the clinical setting and how these related to influencing decisions and performing credibility. The second concerned family members’ roles in the interaction and the effect that they had on each participant’s
performance. The analysis was oriented to these questions throughout, though my
understanding of what they encompassed changed as more data was considered. As well as identifying frames, I explored the data for techniques and actions which appeared significant in addressing these performances.
Step 1: Identifying frames inductively
Working through the data from the earlier consultations to the later ones, I applied an
inductive analytic approach to identify possible frames and techniques. The choice of where to begin with the data was not theoretically driven. Instead I began with the earlier recorded consultations, supplemented with a sub-set of 21 consultations in which a third party was present since this was a particular interest of the research. I started from the assumption that all the spoken interaction (and non-verbal interaction, though due to the nature of the data, this could not be analysed in this study) was oriented to one or more frames, though some of these might be of more interest in relation to the study questions than others.
As each new consultation was analysed, codes were added or amended. Identified frames were labelled and key characteristics of these frames were attributed. I initially used printed data for this process, which allowed me more freedom to adjust and expand on the codes. Once a number of core categories started to take shape, I moved to using the software NVivo to capture these frames, allowing a more systematic approach and easy retrieval and
Step 2: Validating frames against set criteria
In order to qualify as a valid frame, data had to demonstrate that participant roles, expectations and behaviours were influenced by the construction or shift of frame; frame shifts must be shown to alter the understanding of what is going on for the participants in the
interaction. This is not as straightforward as merely noting the actual response of the hearer
to a shift in frame. These responses may be the result of mis-framing or a frame dispute, or the influence of another factor such as an external interruption. It is for the analyst to ascertain how the framing shifts expectations, informed by the actual response, the first speaker’s response to that response, and my own communicative competency to hear these changes. Having identified a series of potential frames from the data I revisited each frame and tested it against four key criteria to assess the usefulness and consistency of the coding. By laying out these criteria, my definition of frames should be transparent, though they may still be open to challenge and critique. The four criteria listed below for frames to be considered valid were drawn from Goffman’s (1974) descriptions of how they may be seen to govern social interaction:
a) The frame should alter the expectations of how the hearer(s) should react to the speaker
By framing, Goffman argues, the speaker places obligations on the other participants. For example, in relation to everyday conversation:
“What the individual spends most of his spoken moments doing is providing evidence for the fairness or unfairness of his current situation and other grounds for sympathy, approval, exoneration, understanding or amusement. And what his listeners are primarily obliged to do is to show some kind of audience appreciation. They are to be stirred not to take action but to exhibit signs that they have been stirred.” (Goffman,
One way of evidencing a frame shift is to ascertain that there is a change in the expectation of how the other participant will respond. For example, they may be expected to offer advice or to listen; to answer or to acknowledge; to sympathise or to challenge. Expectations are detected through a combination of the participant’s actual response, the first speaker’s acceptance or non-acceptance of this response, and my own reactions drawn from my
communicative competence. Evidence of expectations are discussed more fully below (see, for example, section 4.2.1, 120).
b) Shifts of frame and footing should change the way that participants are aligned to one another
As explained above, through footing, participants change the alignment they take up to one another. In doing so, relationships between the participants change. This is closely related to the idea of expectations. As Ribeiro notes:
“*Footing+ conveys a metamessage about what kind of relationship is being enacted at that point in the interaction. Thus in a standard psychiatric interview [...], the footing between participants may shift from a more personal conversation (a “woman-to- woman talk”) to a formal interaction in a professional stance.” (Ribeiro, 1994:55)
c) The frame should allow participants to emphasise and construct different identities and roles
Goffman relates his frame analysis back to his theory of interaction, in which participants express a face to others. Keying allows participants to shift roles and project different faces.
“The individual comes to doings as someone of particular biographical identity even while he appears in the trappings of a particular social role. [...]There is a relation between persons and role. But the relationship answers to the interactive system – to the frame – in which the role is performed and the self of the performer is glimpsed.”
(Goffman, 1974: 573)
If frame shifts influence communication between two people, it should be possible to identify what participants are doing in the interaction to initiate, negotiate or reject frames.
Identification of frame shifts requires a more conversation analytic approach to identify ways in which activity is bracketed.
This step of re-examining the identified frames also involved returning to the initially analysed transcripts and checking the refined model against the data. At this point, identified frames were merged, split or re-defined based on the criteria applied.
Step 3: Applying frames to more data
Having developed and checked the coding model in step 2, I then applied these codes across the whole dataset. Every transcript was coded in full using these frames. This allowed me to test whether the identified codes stood up to application across more cases or whether it required further development. Some further refinement was conducted at this stage, both to the definition of frames and to their content. For example, a loosely defined ‘administrative frame’ was clarified and changed at this stage to ‘stage direction’ (see section 4.2.3, page 153) which reflected a crystallising in my understanding of the data through seeing and comparing more examples. A full transcript with frames identified and applied is given in appendix B as an example. In this case, much of the transcript takes place in a Troubles-telling frame (see 4.2.1, page 120), but shifts between frames are indicated.
Step 4: Micro-level analysis of discourse techniques
Having identified a robust set of codes and coded all the consultations, I then used a micro- level analysis to identify how frames were shifted through participants’ talk. This stage resembled more closely a conversation analysis approach to those parts of the data which I had identified as containing a change of footing. The analysis focused on the use of
Step 5: Drawing links between techniques and identified frames
The final stage of the coding was to explore the distribution and relationship between codes, in particular between frames and the various participant techniques identified.