Analysing Interview Transcripts

Annette Kuhn

This is the third in a series of blog posts showcasing Cinema Memory Archive assets relating to the planning and conduct of Cinema Culture in 1930s Britain.[1] The previous post featured a 1995 discussion paper that looked at the potential role of computer applications in the project and explored the features of qualitative data analysis packages that were available at the time.

When the document that follows was drafted (in March 1996),[2] the decision had been made to work with NUD*IST (now NVivo). Training had been undertaken and transcription of interviews was close to completion. All core interviews had been listened to, and/or transcripts read, by project team members, and it was time to think about how the material was to be systematically analysed. A few months later CCINTB’s initial findings were presented at the Screen conference in Glasgow.[3]

Analysis of Interview Transcripts: A Discussion Paper

I have closely read and marked up two sets of transcripts: those of the interviews with Margaret Young and Mollie Stevenson (Glasgow) and Fred Curnick and Gwen Curnick (Harrow).  My comments are based largely on my reading of these transcripts, but also to some extent on my (less close) familiarity with the contents of other interviews, my overall sense of what the interviews as a whole contain, and also of their ‘structure of feeling’.

It seems to me that analysis of the interview material will be guided to some, as yet unspecifiable, degree by the questions we want to ask of the raw data. What sorts of things do we want to  get from the data, always bearing in mind that with qualitative analysis there is quite properly always some degree of open-ended exploration involved?  In some areas at least we will be working in a creatively analytic way with the data, and it would thus be wise to avoid preconceptions and remain open to what the material offers us.

It also seems to me that there are at least three levels at which qualitative data may be explored and analysed. In increasing order of openness and ‘exploratoriness’, these are

  1. base coding
  2. overt data level
  3. discursive or subtextual level

1. Base coding

This is the area VB [CCINTB Research Fellow Valentina Bold] calls ‘descriptive categories’ in her discussion paper[4]: they are the variables upon which we shall want to make comparisons between responses.  On what grounds do we want to explore differences between responses?  Do men talk about their cinemagoing in different ways than women?  Do people in different areas or with different social backgrounds have different preferences for stars or types of film? These categories are probably best kept simple, with perhaps even fewer  than those suggested by VB, for example:

  • Age/date of birth
  • Gender
  • Location
  • Education
  • Social class
  • Ethnicity

We shall need to make coding frames for some of these:, liaising with the questionnaire coder to establish consistency.[5]

2. Overt data level

This is more or less coterminous with the headings VB calls ‘thematic categories’, and refers to conscious, explicit, overt or factual responses:  ‘factual’ or descriptive information about cinema buildings and personnel, for example;  preferences as to films and stars; types of ‘fannish’ behaviour; other leisure pursuits and so on.  Significantly, these are exactly the sorts of areas explored in the questionnaire, and  all lend themselves to a straightforward content analysis type of reading and coding.  There is scope at this level for amalgamating and/or comparing interview and questionnaire responses, and it would be as well to bear this in mind when making decisions as to the coding both sets of data.

For NUD-*ST, the question of ‘trees’ and ‘nodes’ will immediately arise at this level.[6] How this is dealt with will depend on what we are likely to ask of the data (though one of the beauties of NUD-IST is that it’s possible to develop your thinking or change your mind about this as you go along). It is likely, for example, that we will be interested in preferences for Hollywood as against British stars, and male against female, and the STARS ‘tree’ can be constructed accordingly.

Areas/topics for coding at this level could include:

  • age/year of first cinema visit
  • age(s) of most frequent cinemagoing
  • frequency of cinemagoing
  • favourite cinema(s)
  • choice of film
  • went to cinema with…
  • home, work or school in geographical relation to cinema
  • mode of transport to cinema
  • favourite films
  • favourite stars
  • ‘fan’ behaviour
  • alternative entertainments

A number of more impressionistic, though still consciously articulated, items also arguably belong at this level. For example:

  • explicit reasons given for going to/enjoying the pictures;
  • explicitly articulated feelings about the pictures
  • courtship and the pictures; sex.

There is likely to be an area of overlap between this more qualitative end of the overt data level and the beginning of the next, deeper, level of analysis.

3. Discursive or subtextual level

This level of analysis calls for a particularly exploratory and open approach to the material, and for sensitivity to clues that go below or beyond the written or indeed the spoken word.  Most of my annotations on the transcripts I read relate to this level, at which I see emerging in particular issues relating to space, time and the narrativisation of memory.

There are different levels of analysis also within this category, beginning perhaps with

  • emotion or feeling tone–sometimes this comes through in the   transcript, sometimes only on tape; but in either case it is an important clue to the respondent’s feelings about what is being said.

Moving on from this, analysis might attend to:

  • changes in voice, tone or accent in relation to what is being said
  • items–turns of phrase, anecdotes, etc–that are repeated
  • slips of the tongue.

Other attributes of the account may also be revealing:

  • going ‘off the point’
  • self-contradictory statements.

Beyond this, the analyst can move into various forms of ‘symptomatic’ reading, always respecting the data.  One might wish, for example, to follow up a ‘hunch’ that emerges from what one knows about the respondents and how they speak.  For example,  there is something that comes up in many interviews about the ways in which people remember their outings to the pictures–the detailed accounts of local topography, for example, or of companions and activities surrounding actual picturegoing.

At another level, and central to the rationale of the project as a whole, is the question of memory, of how people construct memories or turn their memories into stories.  For example, I already note a difference between what might be called ‘repetitive’ memory (‘we always…’, ‘we used to…’) and singular memory (the recounting of one-off events, often anecdotes well-polished through many retellings); between memories in which the speaker involves herself in the telling, commonly collectively (‘we used to..’) and more impersonal remembering (‘there used to be…).

Another ‘symptom’  relates to the ways in which people talk about time:  an important form of memory narrativisation seems to be a ‘then/now’ motif, which carries across different contents of memory, and varies in value accorded to each term (the Curnicks are more positive about ‘then’, for instance, Young and Stevenson about ‘now’).

Conclusion

These various levels of analysis would seem to call for differing forms of attention to the data, reading/analytic skills, and so on.  Without having sufficient experience with NUD*IST to assess how feasible this is, it would seem on the face of it sensible to perform each level of analysis separately.

For example, it would seem good sense–and indeed NUD*IST provides for this–to go through and base code all transcripts before beginning analysis proper.  Moving then to the overt data level, the researchers can perform relatively straightforward indexing/coding procedures and at the same time get  used to working with the software.  It would perhaps make sense to engage with the discursive level later on in the project, and/or when particular sorts of ‘deep’ questions are being asked of the data,. For example: ‘how do our interviewees talk about the past as against the present, and how does this relate to their memories of going to the pictures?

Annette Kuhn, 1 March 1996

[1] We are able to to showcase these items here because non-participant materials in the Cinema Memory Archive  are now fully documented and accessioned.  The project’s primary focus has hitherto been the participant materials (interviews, letters, questionnaires, memorabilia etc) held in the CMA. While the bulk of these materials are readily accessible via the project website, non-participant items , including those featured in this series of blog posts, can currently only be accessed via Lancaster University Library Special Collections and Archives.

[2] Discussion Paper. ‘Analysis of Interview Transcripts’, 1 March 1995. CC-19-000NA012. Cinema Memory Archive, Lancaster University Library Special Collections.

[3] Annette Kuhn and Valentina Bold, ‘ Cinema Culture in 1930s Britain: Ethnohistory of a Popular Cultural Practice. Screen Studies Conference’, Glasgow, 29 June 1996.

[4] Discussion paper. Base coding suggestions for NUD*IST’, 19 February 1996.  CC-19-000NA008. Cinema Memory Archive. Lancaster University Library Special Collections.

[5] University of Glasgow Sociology student Callie Perks was responsible for coding CCINTB questionnaires, using quantitative analysis software SPSS. See Annette Kuhn, ‘Cinemagoing in Britain in the 1930s:  report of a questionnaire survey’, Historical Journal of Film, Radio and Television, vol.19, no.4 (1999).

[6] Workhop Manual for Q.S.R. NUD*IST qualitative data analysis software, 1995. CC-19-000NA020. Cinema Memory Archive, Lancaster University Library Special Collections.