Transcription is often seen as an essential part of qualitative research projects that generate data from discussions or other interactions with participants. There has long been debate about its nature and role, both in terms of analysis, and in relation to technologies for facilitating the process. Whether transcripts are required is one question. If they are, then the question becomes what form they should take. There is no one answer to these questions, because what is required depends on the project objectives, practicalities and the focus of the analysis.
As technologies and methods continue to develop new dimensions are brought into the discussion. Even though online data collection methods are as old as the internet itself, the Covid pandemic forced many researchers to consider these ways of generating data for the first time, out of necessity if not design. This brought new attention to online methods from many quarters (luckily there was already a broad literature in this field, and a great textbook by Janet Salmons). It also prompted an increase in automated transcription tools, both those embedded within video-conferencing platforms such as Zoom and Teams, and also what now appears to be a plethora of standalone products from which to choose.
Such developments provide new contexts in which to consider the old-age questions of whether and how to transcribe.
Verbatim transcription
It’s very common for qualitative researchers to transcribe verbatim, that is, fully write out everything that is said between interviewer and respondent, among participants in a focus-group discussion or a naturally-occurring encounter. This process has long been facilitated by transcription software, initially in the form of tools to make manual transcription easier - such as Express Scribe, and f4transkription - that allow, by means of keyboard short-cuts and foot pedals, recordings to be slowed down, looped and for time-stamps inserted and so on. These could be used by researchers themselves, or by professional transcriptionists who turn around transcription pretty quickly. Most researchers who transcribe verbatim value these tools, considering them to greatly facilitate the time consuming process of verbatim transcription. Where time-stamps are included verbatim transcripts can easily be synchronised with the corresponding audio or video file within a CAQDAS-package, allowing concurrent analysis of the two. Once synchronised transcripts are usually editable, allowing them to be added to as part of the analytic process within your chosen CAQDAS-package.
Incorporating characteristics of conversation and non-verbal interaction into transcripts
But verbatim transcripts are just one form and for many purposes they’re not detailed enough. For example, any form of analysis that focuses on interaction requires much more than a verbatim transcript of what is said. The transcription notation convention developed for conversation analysis by Gail Jefferson in the 1960s for example, captures characteristics of conversation including timing, speed, emphasis, pitch, and volume, and also ‘embodied elements’ of non-verbal interaction such as facial expressions and other gestures etc. (see this short article by Song Hee Park & Alexa Hepburn for an overview of the benefits of this type of transcription). Even when not doing a conversation analysis it is important for many projects to consider at least some non-verbal elements, because it's often as interesting to consider how something is said as what is said. These aspects can be incorporated into a transcript in a variety of additional ways, for example using standard text formatting to capture different foci.
Is a transcript always necessary?
Some projects consider transforming non-verbal cues into text, however complex the transcript, to constitute loss of fidelity and therefore prefer to work directly with audio or video materials. Many CAQDAS-packages can import visual materials thus allowing direct access to original recordings for analysis by means of marking, segmenting, annotating and coding. But it's not always necessary to make a choice between direct and indirect working, because of the possibilities in some CAQDAS-packages to combine and/or isolate the verbal representation (in the form of a transcript) and the non-verbal (in the form of the recording).
It's also worth noting other potential reasons for not using a transcript. When there's limited time or the analysis required is high-level or summative transcriptions may take the form of brief notes, such is common practice in market research. Most social research projects, in contrast, would usually see such notes as inadequate representations of an encounter, but 'gisted transcripts' as described by Paul Dempster and David Woods constitute a hybrid approach.
Transcription is an analytic act
Whatever the choices that are made, transcription is always an analytic act, in the sense that the choices we make about what to transcribe, how to format the transcripts, and whether to synchronise them with the audio or video recording greatly affect what is possible analytically. I've written elsewhere about transcription as one of the many 'moments of contact' we have with materials during a qualitative analysis and seeing transcripts as dynamic is an aspect of this. David Woods discusses these topics in detail in the webinar he delivered for the CAQDAS Networking Project, including the use of multiple transcripts. See also this article by David Woods and Paul Dempster.
The rise of automatic transcription
The availability of automatic transcription services have mushroomed recently. They receive a varied press, with some advocates but also many who question their value given the amount of time required to correct the errors made due to accents and the like. In addition are the ethics of uploading confidential recordings to potentially insecure servers. The pandemic arguably accelerated the normalisation of these technologies as increasing numbers of researchers turned to online data collection, and video conferencing tools such as Zoom and Teams integrated transcription into their tools. Nevertheless the security and ethical issues involved have prohibited many academic researchers from using these tools. That’s why the new transcription service launched by Quirkos software is worth checking out and watching as it develops, because along with accuracy and cost, it has security at its heart, designed to satisfy the requirements of IRBs, ethics boards, or data security reviews, in ways other services are not.
I invited Daniel Turner, founder of Quirkos software to deliver a free webinar for the CAQDAS Networking Project on 26th April 2023 and he’s going to use that space to open up discussion about the new possibilities that automatic transcription brings to qualitative research. One thing he’s going to raise is the “possibilities discussed include being able to quickly create transcripts of verbal field notes and reflexive journals so they can be fully integrated into an analysis with ease”. Reading this part of his webinar abstract reminded me of when I ran focus-groups and interviews with young people for my PhD research many years ago. I intentionally drove to the sessions rather than catch the train so I could record myself talking out my reflections immediately afterwards - which I felt was quicker and more spontaneous than writing them out in a notebook a few hours later. At the time (late 1990s) I had a cassette-based Dictaphone not a digital audio recording device or smartphone, and although I listened back to these reflections many times and pulled out some key elements as notes that made it formally into the analysis, I didn’t fully transcribe them. Being able to do so securely and quickly would have transformed the way I integrated my reflections into the analytic process.
There are so many dimensions to think about when transcribing. Join us to discuss some of them at the webinar, 26th April 2023, 4pm London time
Comentarios