Using Nvivo Audio-Coding: Practical, Sensorial and Epistemological Considerations
Total Page:16
File Type:pdf, Size:1020Kb
Issue 60: Autumn 2010 social researchUpdate Using NVivo Audio-Coding: Practical, Sensorial and Epistemological Considerations • Social researchers have depended on verbatim transcripts Megan of audio-recorded interviews in order to apply rigorous and systematic analyses. With software programs such as NVivo8, Wainwright transcription is no longer a necessity. and Andrew • This paper describes NVivo8 audio-coding for open-ended interviews, based on the narrative-ethnographic study of an Russell outpatient clinic. Megan Wainwright, a PhD student • While thematic narrative analysis was the approach applied in Anthropology, is the Van to the interviews, the process described and issues raised are Mildert College Trust Research relevant to any form of analysis undertaken using programs Scholar at Durham University, similar to NVivo8. undertaking ethnographic research • The use of this technology raises questions about the on Chronic Obstructive Pulmonary epistemological and sensorial differences between visual and Disease (COPD) in Uruguay. She has researched and published in audio interpretation and analysis. the areas of medical education, palliative care, end-stage renal The semi-structured or unstructured is a laborious process and it is disease and COPD at McGill open-ended interview is a seminal increasingly common that transcriber University. The research on chronic method in qualitative social research. and analyst (and sometimes even constipation discussed here won In many research fields, the interview the interviewer) are different people the RAI Arthur Maurice Hocart offers access to individuals’ own (Mishler 1984). Producing a transcript Prize 2009 and the Dryburn interpretations, perceptions, can be seen as the first step in the Research Prize 2010. <megan@ experiences and practices. Transcripts, analytical process, assuming that meganwainwright.ca> the verbatim write-up of audio- the researcher and transcriber are recorded interviews, are a respected the same person. When they are and in many ways fetishised form not, at the very least the researcher Dr Andrew Russell, a Senior of data, in part because they lend should check transcripts against Lecturer in the Department themselves to systematic content- audio recordings to correct any of Anthropology at Durham coding and analysis. transcription errors, which is a University, is Director of the lengthy process in itself. University’s Community Outreach However, voluminous transcripts are and Engagement programme. produced as something of a ‘knee- Here we demonstrate how one <[email protected]> jerk’ response to data creation, with can work directly with audio data most of the material they contain throughout the analytical process; i.e. left to gather dust in the researcher’s without using a written transcription. drawer. Furthermore, transcription Above all, this Update calls for http://sru.soc.surrey.ac.uk/ 1 social research UPDATE critical reflection on the centrality fieldwork was of five months’ topics or themes created by the user of transcripts and the relationship duration and it would have been to which data-chunks (i.e. parts of between mode of engagement with unfeasible for the researcher to an audio recording or sections of a data and analytical thought about transcribe all the interviews in that written document) can be coded. An them. time-frame in addition to analysis imported audio-file will appear as and writing-up. Because of the an Audio Wave Form which can be The Project participatory nature of the research listened to and divided up into audio- NVivo8 is a qualitative data-analysis it was considered important that excerpts. software program released in 2008 each interview be given equal An audio-excerpt is assigned a that enables the coding and analysis analytical attention (i.e. we did not ‘Timespan’, notes can be typed of text, image, audio and video want to select only a sub-sample for into the ‘Content’ column (to help data. The specific features of NVivo transcription and analysis). Secondly, identify the excerpt) and then is and how to use them have been the narrative interviews elicited assigned to a Node. For example, discussed elsewhere (Auld et al 2007; very intimate details of participants’ 1.55 min to 2.30 min (Timespan) Bringer et al 2006). However, the lives and it was felt inappropriate was coded as ‘Family Relationships’ ability to directly code audio data is to outsource transcription to a third (Node) with the note “discussing the a new feature of NVivo8 and has not party. yet been discussed in the literature. effect the illness has on marriage” While programs such as ATLAS.Ti Using Audio Coding – an (Content). The creation of Nodes (Hwang 2008), Transana (specific Example is dynamic: as more interviews are coded, Nodes can be added, to video and audio) and MAXQDA Audio-coding in NVivo8, like changed, collapsed or deleted. The also support audio data, this Update document (text) or video coding, second step is to create a ‘Memo’ - a focuses only on NVivo. requires the creation of ‘Internal short summary and initial analytical Source’ files within the ‘Sources’ tab. One aim of the MSc research on reflections linked to that particular For example, in this study eleven which this paper is based was to ‘Internal Source’. One can re-listen audio ‘Internal Source’ files were test whether the analytical process to the excerpts coded under each created by importing the interview of thematic coding, previously used Node for emerging themes. The way recordings (WMA format) into in the analysis of text-based data, one navigates the various Nodes NVivo8 and labelling them according could be replicated by working is the same as for textual data and to participant code. Once the exclusively from audio-recordings is explained in Bazeley (2007). ‘Internal Sources’ have been created, using the applications available However, one noteworthy feature a number of applications become in NVivo8. Eleven unstructured is that from within each Node a available such as Nodes, Memos, narrative interviews (lasting hyperlink opens the corresponding Sets, Queries, Models, Links and between one and three hours) interview and highlights the relevant Classifications. Bazeley (2007) gives exploring the illness experience sections along the Audio Wave Form, a detailed description of these and and healthcare relationships of making it easy to return and listen to other features of the NVivo program. people receiving specialist care for audio-excerpts. chronic constipation were used. The Nodes are essentially codes – the theoretical orientation of the project was that of grounded theory, and analysis was informed by two similar approaches to narrative-analysis: the holistic-content perspective (Lieblich, Tuval-Mashiach and Zilber 1998) and thematic narrative analysis (Riessman 2008). As discussed elsewhere, qualitative data analysis software programs lend themselves to grounded theory (Bringer et al., 2006; Peters and Wester 2007), although MacMillian and Koenig (2004) offer a thought-provoking critique of this association. Exclusive use of audio-coding offered significant advantages. Firstly, the 2 http://sru.soc.surrey.ac.uk/ social research UPDATE Practical, Sensorial and requires a good deal of practice. analytical process as the researcher Epistemological Considerations One could argue that annotating is “sensorially” (Nichter 2008) closer Audio features offer time and is a form of transcription, but in to the data. The researcher can think budgetary savings; advantages our experience the most difficult analytically about the data while software companies certainly process is to annotate briefly so being immersed in the flow of the advertise. We believe that using the that the researcher is drawn back to recorded interview, attending to audio features to engage in audio- listening to the excerpts as opposed utterances, silences, emotions and analysis is also attractive for analytical to reading a summary of what was the interactive dialectic between reasons and raises interesting said. For instance, if one is coding interviewer and interviewee in ways questions concerning the relationship thematically one can type in the that are difficult when reading even between a researcher and her data content column what is happening detailed transcriptions. Of course, an that need to be explored further. or what the participant is “doing” in interviewer returning to an interview The software facilitates working that period of time in the interview transcript may find it elicits sensorial with audio according to the same rather than writing out everything memory. The phenomenon of sub- processes we are accustomed to with that was said (e.g. “participant vocalization – when we hear in our transcripts: systematic and rigorous describing relationship with general minds the words we read – might analysis, enabling one to work practitioners”). By then coding all play a part in bringing us back to through an entire interview without such passages across interviews that interview, that house, that day neglecting sections, annotating and as “healthcare relationships” the etc., assuming we were there in the memo-ing along the way etc. It researcher then has to re-listen to first place. However, this may not be also has the advantage of enabling each excerpt under that Node to as effective as hearing the recorded the researcher to move swiftly begin to analyze across interviews. voices all over again while engaged between codes and audio excerpts, Admittedly we often