Chapter Four: Data Collection and Methodology 4.1 Introduction
4.6 Issues in Transcription
4.6.1 Selection of data for transcription
Before describing the ways in which the video data has been transcribed it is first important to outline how the samples were selected for transcription. In the case of the video playback interviews with main participants the data was transcribed in its entirety. With the video of interpreted interaction, for reasons outlined below, it was necessary to select slices of interaction for further detailed analysis and it is this selection process which requires justification.
Early on in the current study I made the decision to transcribe the video data manually, without using any of the software options available. I believe that the complex mix of spoken and signed language data from my corpus would have
presented considerable challenges in terms of the limitations of the current software11,
the technology used to both capture and analyse the data, and my ability (and
11 With regards to signed language data, there are options such as SignStream, a (free) database tool
designed to enable linguistic and computer vision research on signed languages. SignStream allows users to annotate video data segments, assigning detailed information (e.g. manual signs and non- manual behaviours such as head nods, head shakes, eye gaze, raised or lowered eyebrows etc.) to distinct fields in order to produce a fine-grained, multi-level transcription (see Neidle 2000, Neidle et al. 2001).
available time) to master the techniques necessary to undertake a computer-aided analysis. The fact that the majority of my data is from interpreted mixed (signed and spoken) multi-party interaction would demand a software system that could take both the spoken and signed elements into account. Secondly, for reasons outlined earlier, all of my data were collected using a single video camera. As a result it was not possible to train the camera face-on to one specific participant (i.e. the deaf employee or the SLI), as I had to settle for a perspective that would encompass as many of the participants as possible, but predominantly include the deaf individual and the SLI. The lack of multiple views of the deaf participant made the use of a system such as SignStream difficult. Additionally, there were my own limitations to be taken into account in terms of my ability to learn and utilise a software system and the demands that this would place on my IT resources. I therefore decided to code the data manually, viewing and re-viewing the video tapes a number of times, a process which enabled me to identify patterns within the interaction.
Analysing data at a ‘local’ level, i.e. the turn-by-turn organisation of talk, involves close, repeated listening to recordings, or as in the current study, listening to and viewing video tapes (Wadensjö 2001). These can reveal previously ‘unnoted
recurring features of the organisation of talk’ (Silverman 2000: 150). The complex
nature of interaction and participant behaviour, with much of what takes place appearing on the surface rapid and mundane (Finlay et al. 2008), means that new aspects can often only become apparent after repeated viewings (Dobson et al. 2002) and so this was an essential part of the selection process.
In identifying patterns within the video data I was guided by Rampton’s (2007d) recommendations for investigating interactional data, beginning by looking at the types of activity occurring in the interaction. The main focus was how the participants were interacting, whether they were doing what was expected of them and how they managed the relationships between them, thus enabling me to gain an ‘initial orientation’ on what was happening (Rampton 2007d: 1). The selection of the data was therefore a somewhat organic process, i.e. I allowed the repeated viewings to naturally bring to the surface ‘points of interest’ where it appeared that the discourse process was not particularly smooth, or where the SLI’s management of the interaction suggested potential moments of cultural mismatch or discord. The process
was also informed in part by my own experience of interpreting team meetings, as well as anecdotal evidence arising from discussions with colleagues. That is not to say that any of the video data excluded from analysis was ‘uninteresting’. Transcribing the whole data set would have undoubtedly revealed relevant interesting aspects of interpreted workplace interaction, but given the time needed to manually transcribe such mixed-modal data this was beyond my resources. I therefore decided to select two ‘patterns’ for more detailed analysis. The first, the way in which the SLI managed overlapping talk in team meetings, is based on my observation that this form of talk appeared particularly challenging for the SLI. Furthermore, the difficulties posed by the nature of the overlapping interaction suggested that the SLI’s management of the discourse could be potentially exclusionary to the disadvantage of the deaf employee, thus warranting further exploration. Additionally I was aware that this aspect of meeting talk was problematic, evidenced from personal experience, the literature, and from SLI’s responses in the questionnaire and journal data (see Chapter Five).
The second pattern identified and selected for analysis was based on the unexpected finding of the frequent re-occurrence of humorous exchanges and small talk during the workplace meetings. I was aware that small talk was a feature of workplace discourse and had for this reason included a question on this aspect in the SLI questionnaire. However, I had not anticipated the way in which humour, sometimes interwoven with small talk episodes, appears to be a consistent feature of workplace meetings. The importance of small talk and humour in establishing collegial workplace relationships, along with the cultural differences between deaf and hearing people, particularly relating to humour, and the gap in the literature regarding interpreting humorous episodes thus led me to select this aspect for further exploration.
In total, five hours and 46 minutes of video data were collected (see Appendix L). Of this, approximately 60 minutes have been transcribed at a surface level of rough working transcription. From the surface transcriptions, slices of interaction were then identified for more detailed transcription and analysis. These excerpts have been selected to illustrate some of the issues facing SLIs in the workplace domain and thus form the core of Chapter Six. I should emphasise here that the excerpts of video data
selected for inclusion are representative of the interaction occurring across all of the data set. Although, for reasons of practicality, I have only highlighted relatively few examples, there were instances of relational talk and humour, as well as overlapping talk, throughout the video data corpus. I analysed a number of other examples of instances of small talk, humour and collaborative floor from the data set, but it has been impossible to include them all in this thesis. The selected excerpts therefore describe what I have seen commonly occurring in the workplace meeting discourse data.