Complicating transcription (3)

[For previous discussions on complicating transcription, see my posts here and here]

A recent tweet about transcription by Professor Trisha Greenhalgh of the University of Oxford has caused considerable interest and reaction. As might be expected with my professional transcriber hat on, I have some thoughts!

Firstly, and possibly surprisingly, my first reaction wasn’t ‘this is the end of the world for my business’. Even since writing my first two posts about transcription, just over a year ago, technology has come on in leaps and bounds and I have to face the fact that there are some really good AI programmes out there which are making automated transcription much more accurate than it used to be. These will undoubtedly be gamechangers for cash-strapped students and researchers. Even for people who want, and are able, to pay someone like me to transcribe their sound files, there are surely ways to make the process more efficient. There will always be bits of redundant speech in any interview, and parts that are more or less relevant, and they can’t always be identified in advance, but if there is time for the interviewer to listen to at least the beginning and end of the recording before sending it to the transcriber, then a note to say ‘start from 2 mins 50 secs in’ and ‘don’t bother with the last 5 minutes’ is no bother, it saves me a bit of time, and you a bit of money, and the chitchat at the beginning and end which is never going to be used can be left out with no harm done to anyone.

Secondly, I was very struck by the phrase ‘mindlessly transcribing’. Anyone who’s spent any time at all in undertaking qualitative research can probably relate to it – even as a professional transcriber, there are times when it can be pretty mind-numbing work, and it will be even more so if you are a slow typist and don’t have decent equipment. But I’m not sure that’s what Prof. Greenhalgh meant, exactly – it’s not just that it’s, for many people, a boring and time-consuming (and thus resource-intensive) thing to have to do, but that it’s been seen as essential for demonstrating high-quality research and analysis for so long that this is no longer questioned, and it is one of those ‘givens’ that you just have to do and then you’ll have reached the required standard. I think it’s great and important that this assumption is challenged, and there are no doubt a number of other ‘givens’ in research that could probably benefit from the same. Anything that helps researchers reflect on why they’re doing what they’re doing is surely a good thing – when we pride ourselves on rigorous and insightful analysis, ‘because we’ve always done it like this’ shouldn’t be acceptable in our research methods.

The other word in Prof. Greenhalgh’s tweet which leapt out at me was ‘necessarily’. She is right of course – the mere act of transcribing and entering the transcript into NVivo (or whatever other analysis software you are using) of course doesn’t ensure anything at all, other than you’ve spent the time doing it. However, I do worry a little bit about throwing out the baby with the bathwater. My experience in my PhD was that having to retranscribe every interview (because several of them had been initially transcribed inaccurately by someone else) and then going through them to code each transcript in NVivo meant that I was completely immersed back in my data in a way that I couldn’t be when working with my fieldnotes (although the fieldnotes were immensely helpful in conjunction with the transcript as a prompt for reflection and analysis). Of course as a PhD student I had the luxury of time that most employed academic researchers don’t have, but as a formative experience, that transcription work was incredibly useful to me to reflect on my performance as an interviewer, and the information that I was able to elicit (or not), which I’m sure positively influenced the researcher I became. Later on when I was a postdoc, my interviews were transcribed by somebody else (a sensible use of time and resources, much better than getting me to use my more expensive time doing it), but even so the coding work and going back to the written transcripts also made that immersion in the data more smooth. An additional factor in that particular study was that although we were working in a team of researchers, I was the only person conducting the interviews, so the written transcripts, cumbersome as they were, did allow my colleagues to access the data reasonably straightforwardly. I am aware that the sound files could be shared, as an alternative to sharing the written transcription, and time stamps given of relevant sections so that they could listen to the relevant section without having to plough through irrelevant chitchat, but in our experience when we were analysing the data, quite often colleagues picked up on different issues than I had, whereas if they’d just been working from my fieldnotes and suggestions of relevant passages, that richness of collaboration might have been less likely.

I do realise, of course, that a single tweet can’t cover all the nuances and issues involved in a complex issue like this. I welcome the honesty of the proposed sentence in her forthcoming paper, and particularly welcome the disruption of some of the methods section shorthand which is included unquestioningly in seemingly every paper these days (it’s not just transcription that this relates to; use of grounded theory is another obvious one!). I understand from a subsequent tweet that Prof. Greenhalgh is considering writing a paper leading on from the discussion her initial tweet engendered, and I for one can’t wait to read it. I would caution against throwing the baby (and let’s be honest, part of my business) out with the bathwater, but am absolutely all for reflection on why we do what we do, and considering how we can ensure it is still fit for purpose.

Complicating transcription (2)

Following on from my previous post, I want to use this post to think about some of the more practical issues researchers might need to think about when considering hiring a transcription service. Much of this I’ve learnt the hard way in this first year or so of my business, and I hope that people find it helpful.

DIY, or get somebody in?

A few times I have been contacted by researchers who have undertaken some interviews, started transcribing themselves, and then got completely bogged down in the task. As raised in Julie Mooney Somers’ blog post (here), sometimes grant funders may be reluctant to release a lot of money for transcription, and it may be tempting to see transcription as something that you can just do yourself, to save costs. However, I think it is worth having a think about that. As mentioned in my previous post, I charge £15 per hour of my time, and would expect to take between 3-4 hours (usually around 3.5) per hour of sound recording. I usually quote 4-6 hours, just to cover myself if the recording quality is particularly poor, or if it is a focus group with lots of different, not-always-identified speakers, but I only charge for the time I take. If you are planning on doing (say) 40 interviews of up to an hour long, this would mean requesting up to £3,600 (if I’ve got my maths right!). I actually think that’s a pretty reasonable amount, but some funders might question the need for it, or ask for it to be reduced. It might be tempting to pay a student the national minimum wage (currently £7.83 per hour for over-25s) to do the work instead, but if you decide to go this route, then it is worth thinking about how long will it take them? Most students, whilst they can find their way round a keyboard, are not 80+wpm typists. They may not have access to good transcription equipment, such as foot pedals, so the stopping and starting the recording could be slower. Therefore, they may take considerably longer to do the same amount of work, and so the savings to your budget start to shrink. In addition, if they are taking 6 or 8 hours to do a piece of work that would have taken me 3.5 hours, then as well as costing you £46.98-62.64 (instead of the £52.50 I would have charged) that is in effect a day that they are not able to work on their own research. If you happen to have a student who is an amazing touch-typist and who can do the work quickly and accurately, and has the time for it, then it can be really beneficial for both parties, and probably the ideal solution, but don’t assume that employing students will necessarily always lead to a huge cost-saving.

Also, don’t forget that if you decide to do it yourself, you are in effect taking away your time from other important tasks (analysis, writing, applying for grants, teaching), and that your time too is costly (and presumably more than £15 per hour, if you have a substantive contract).

Finally on this point, if it seems too good to be true, it probably is! There are automated transcription tools out there, some of them free, which can do a job up to a point. However, I have been told by a few researchers who have started off using them that they ended up being just as costly and time-consuming as they required so much editing and checking. They are great for people who simply cannot afford services like mine (particularly, as mentioned in the previous post, student researchers), but are not the magic free solution they are often touted as.

Who to choose? Agency or freelancer?

If you decide to go down the route of employing a transcriber, you will need to check with your university how much choice you have. Some friends of mine have regretfully said that they can’t use my services because the university they work for only have one preferred supplier (usually an agency – issues around these and the amounts paid to the people who work for them are discussed in my previous post and in Julie’s blog post). If you do have the freedom (and make that choice) to go with a freelance transcriber such as me, there are a number of things to think about.

* Recommendations. Although I do occasionally try to ‘advertise’ my services on social media, most of the work I get these days is repeat business from people who know my work, or recommendations from others who see a discussion about transcription and say “I used Jackie for my transcription and she was great”. I absolutely recognise that taking a chance on someone whose work you don’t know is a risk, particularly when public money is involved, so if you can find someone who comes recommended then that is obviously ideal. As a researcher in the past I mainly went this route, but occasionally did use people who had cold-contacted us – this was only if we were overrun with files that needed transcribing and our usual transcribers didn’t have enough capacity. If you do need to use someone whose work you’re not familiar with, I’d recommend a trial period first – so maybe give them just one or two files to start with, and see how they go with that, before making the decision to take them on for the rest of the project. I’ve had a bit of work that way too, and am always happy to do the odd one-off job as well as work for my ‘regulars’.

* Are they registered with your university as a supplier? This is a HUGE issue for freelancers. In these days of GDPR etc, universities don’t just pay on receipt of invoices. Many have a purchase order system, so they need to have provided the transcriber with a purchase order before they can send in an invoice, and they won’t do that until they have registered that person as a supplier. To make things even more complicated, each university has different requirements as to what makes a supplier GDPR-compliant, and often require several forms to be completed and returned to them before you can get set up as a supplier on their payment system. One thing I have discovered over the past year or so is that, depending on the finance department, this can take a LONG time. Practically, this has led me to having no work one month (despite the researcher having the sound files ready and waiting for me to start on them), because the process of filling in forms and getting them approved took so long. It has also led in one case to me not being paid for months and months (most places do get there eventually, but it is time-consuming and stressful to have to keep chasing). This has led to many researchers being very embarrassed and angry on my behalf, but feeling quite powerless. Therefore, something that you can do practically as a researcher is start the process of getting your preferred freelancer set up on the university finance system early, in the hope that that will mitigate against payment delays in the future. If you have already got the work done and you’re trying to chase up payment for your transcriber, be aware that it can take a long time, so please keep your transcriber in the loop and let them know what is happening. If you can find someone efficient and helpful in the finance department, cultivate them! (and keep badgering them!)

Funding

The other big issue! My advice and plea to researchers would be to always include a budget for transcription in your applications for grants for qualitative research. This might sound really obvious, but I’m amazed how much research is started without this, with the assumption that the researcher themselves or a research assistant will do it as part of their job, and then it just gets more and more delayed, or important things aren’t done, because so much time is being taken up doing the transcription that could have been farmed out. Sometimes in desperation the transcription then will get farmed out a bit further down the line, but without a budget for it the researcher might well end up paying the transcriber from their own pocket.

I hope in these two blog posts that I have been able to show that as well as a valuable task, transcription is a skill and a good transcriber represents really good value to research.

Complicating transcription (1)

Last week my attention was brought to this article on the ethics of transcribing qualitative interviews by Julie Mooney Somers, an Australian academic. She raises some really important issues, which I wanted to both reiterate and build on, from the perspective of (a) my freelance business as a transcriber, and (b) my experience as a qualitative researcher. I’ll break this down into two posts – this one looking at the ethical, political and intellectual issues around transcription, and the next one will give some practical things to think about for researchers who are thinking about employing a transcriber (costs, contracts, alternatives, etc).

I have to say I agree with pretty much everything Julie raises in her post. It is really easy to see transcription as a repetitive, dull, and onerous task – and therefore not particularly skilled or intellectual, and pretty much values-neutral. That’s certainly how I first thought of it, when I was doing my PhD back in the day, and it wasn’t until I came to review the transcripts of my interviews that I realised how much more there was to transcription. I have just gone back to my thesis, and reproduce in full here a section from my methodology chapter where I talk about the transcription process:

I made the decision that my interviews would be fully transcribed so that I could return later to everything that was said and decide during post-fieldwork analysis what issues were particularly important from the interviews. As I had conducted the interviews in a language that was not my mother tongue I arranged for the transcriptions to be done by native speakers, primarily due to time constraints, but also because although my spoken Romanian is good I was not sure that without the visual, non-verbal cues that take place in a face-to-face interaction I would be able to capture everything as accurately as a nonnative speaker. I had specified that I wished the transcriptions to be word-for-word, so that I could make the decisions on the interpretation and significance of the data; however on listening to the interviews once they had been transcribed it soon became clear that (particularly in the case of many of the Moldovan interviews) they had been ‘tidied up’ considerably. Sometimes it was just that one word had been substituted for another of the same or similar meaning; at other times points that had been repeated in the interview were only transcribed once; and more seriously in many cases entire sections had been considerably paraphrased. In one case the very important point made by the respondent had been paraphrased in such a way that the meaning in the transcribed Romanian was exactly the opposite of what had actually been said. I had not appreciated before the event the extent to which “[r]esearchers who delegate transcription work to others become distanced from this piece of the process and often are not aware of the decisions made on their behalf” (Tilley, 2003:758). After discovering the extent of the problem, with the original transcriptions to hand, I completely retranscribed the interviews myself, a task that took a considerable amount of time and effort but which was ultimately a rewarding way to immerse myself in the data and remind myself of issues which had arisen. During this process I also noticed other themes emerging which perhaps had not been apparent during the actual interviews themselves, so although it was a laborious process it was ultimately useful. As Silverman (2001:164) highlights, the process of transcription is not merely “a technical detail prior to the main business of the analysis” – the very process of transcribing the interviews involved detailed listening, thinking and reflection which ultimately benefited my analysis enormously. It also meant that, as the researcher who had conducted the interviews, I was able to make “informed interpretive decisions [and] decide whether the transcript was representative of the taped piece transcribed” (Tilley, 2003:762).

It is also clear that, as with translation and the use of a second language to undertake the research, the process of transcription is not neutral or apolitical. I had initially approached the issue of transcription as “mundane, technical, [and] unproblematic” (Lapadat and Lindsay, 1999:67), but it soon became clear that the very act of transcription is an interpretive one in and of itself (ibid, p.81; Tilley, 2003) and the finished text is a representation of an event rather than the event itself (Lapadat and Lindsay, 1999).

Because during this research I was working abroad, using a language that was not my first language, I had had to make some decisions about transcription that – if I had really thought about it at the time – would have been my first indication that there’s more to this transcription lark than necessarily meets the eye. Firstly, I was aware that a couple of my respondents were known to the people who were doing the transcription, and I therefore made the decision to not give those recordings to the local transcribers, but wait until I was back in the UK and try and find a native Romanian speaker who could help me out. There was also one interview where an issue being discussed (which was very emotive, and contrary to the beliefs of the person doing most of my transcription) could have caused some backlash to my respondent, and so I held that one back too until I was in the UK and could be more confident that the person I eventually found to transcribe the interview would not know the respondent.

As detailed in the extract above, once I had received the completed transcriptions and realised the extent to which I still needed to work on them, that was a very salutory lesson. Firstly, about the utility of doing the transcription myself as the researcher (I was able to immerse myself back in the data, pick up on things I hadn’t realised at the time, remind myself of things I had forgotten, realise how often as the researcher I interrupted or mumbled over the top of the speaker, etc). However, in my case I have the advantage and luxury of a fast touch-typing speed, which means that I can be considerably faster at the job than many, for whom not only would this be a dull task but also really, really slow. Secondly, it really opened my eyes to the political and values-laden possibilities inherent in transcription (which, as a healthcare worker who has worked with interpreters, I would say is pretty similar to the issues around interpreting and translation). Now that I have started my own transcription business, I am often surprised at how often I have to make quite active decisions about what I’m hearing – the process is so much more than just putting words down on a page. Even seemingly simple things like punctuation require me sometimes to stop and think, and actively decide what to do – we don’t realise until we listen to a recorded conversation that we don’t talk the way we write. Is that pause indicative of a finished sentence, or would a comma or ellipsis be more appropriate? If I put a full stop here, does that even change the meaning of what the respondent was saying? (it isn’t usually that critical, but theoretically could be). How do I render the conversation on the page when people are talking over each other? (a huge issue when transcribing focus groups, but not unheard of in 1:1 interviews too).

I like to think that, as someone who has undertaken qualitative research at a pretty high level, I can make informed and intuitive decisions about what I am hearing, and do my best to get those down on the page as accurately as I can. In those cases where I’m really not sure (beyond the unclear words which I will just mark and timestamp and expect the researcher to check themselves), I will always get back to the researcher if necessary and say ‘listen to this bit again – this is what I think they were saying, but I’m not 100% sure’. I remember one interview where the interviewer had laughed at something the respondent had said, which was a totally natural part of the conversation (I would have laughed too), but it then obscured a couple of words, and so I couldn’t actually tell if they were saying that they had done something or they hadn’t. Either way would have been making a really important (but opposing) point, and I simply didn’t feel that I could make a decision one way or another, and had to get it back to the person who was one half of that conversation to make that call.

An issue which wasn’t mentioned in Julie’s post (although I gather that subsequently it has been raised) is the responsibility of the researcher towards the transcriber. She talked extensively in her post about the ethical dilemmas of informing interview participants about the possibility of third party transcription, and I will direct you back to her post as I agree fully with what she says. However, there is also of course the issue, particularly in sensitive research, of the impact of listening to the interview on the transcriber. I am always grateful when a researcher sends me a file and includes in their email a note to the effect that this interview was a bit full-on, we discuss XY&Z, so you might want to brace yourself. Ultimately of course it is my responsibility to manage my feelings, and to speak out if an interview topic is more than I can handle (not that that has ever happened!) – and it has more often been the case that it’s been the unexpected, seemingly simple topics which have struck a nerve in me as I listen, and I’ve had to pace myself to manage the process well (eg by working on a different interview for a bit and then coming back to the difficult interview later). But I do think that that is an important issue for researchers to bear in mind – to what extent do they have a responsibility towards the transcriber to ensure that their wellbeing is considered too?

Although this isn’t primarily meant to be a pitch promoting me and my service, obviously I do have a business and am happy to discuss working with researchers and seeing how I can help support their research. I can’t finish this post though without acknowledging the ethical complications of funding for services like mine. The fact is, although I believe my rate for transcription is pretty reasonable (£15 per hour of my time, and I would expect to take between 3-4 hours per hour of recording for a 1:1 interview where the sound quality is reasonable; longer for focus groups and recordings with lots of background noise, etc) and compares well with agencies and other services providing transcription by an actual person rather than a bot, I am well aware that my service is probably out of reach of most postgraduate researchers, unless they happen to be working on a project which already has specific additional funding allocated for transcription. In those cases it is completely understandable that student researchers turn to the free or low-cost automated transcription services available online, and these might be a useful first port of call for them. However, it does need to be said that in those cases the researcher will still need to go through and listen and check and edit – as I understand it, many of those services are not great when it comes to things like punctuation, differentiating between speakers, or coping with people talking over each other – so it may end up being almost as time-consuming as doing the transcription themself. I’d also like to reiterate Julie’s point about transcription agencies, which whilst they may be charging less (so represent good value for the public purse), are still using people who will probably end up being paid considerably less for the same service (it is one of the reasons I chose to take my chance and go it alone). Whilst my transcription business is only part of what I do, and therefore I am under less pressure to bring in the money than others, it does need to be borne in mind that that the £15 per hour I charge is gross – I have to do a yearly tax return, and pay both tax and NI from that. That £45-60 I get for transcribing an hour-long interview might seem a large amount of money to the person paying it, but for me it represents half a day’s work, of which, once I’ve paid my costs (insurance, taxes, NI, etc) I will get to see not much more than £30. £15 per hour might sound a lot, but believe me I am not making my fortune this way! What I am doing is getting the satisfaction of a job well done, of being able to work for myself (and all the flexibility that that provides), of being able to support good research with a good-quality service. Just don’t think that what I’m doing is mind-numbing and basic – I hope this post has gone a way to show that there is so much more to transcription than getting a word from sound file to page.

References:

Lapadat, JC and Lindsay, AC (1999), “Transcription in Research and Practice: From Standardization of Technique to Interpretive Positionings”, Qualitative Inquiry, Vol. 5, No. 1, pp.64-86.

Silverman, D (2001), Interpreting Qualitative Data: Methods for Analysing Talk, Text and Interaction (2nd edition), London: Sage Publications Ltd.

Tilley, SA (2003) “”Challenging” Research Practices: Turning a Critical Lens on the Work of Transcription”, Qualitative Inquiry, Vol. 9, No. 5, pp.750-773.