[Linganth] qualitative coding software

Wed Dec 12 20:04:25 UTC 2018

Hi Robin!, et al.,

I run the Qualitative Data Analysis Lab at Indiana University (http://www.iu.edu/~quallab.html)<http://www.iu.edu/~quallab/resources.html>, which emphasizes ling anth equipment by virtue of my own research, although I try to listen to what my colleagues in sociology, ethnomusicology, etc. need too. We run Atlas.ti, NVivo, and MAXQDA, as well as free options like ELAN. Dedoose is also worth checking out if you haven't yet. Recently I got a small grant to train a grad student and develop a university-internal seminar directly comparing QDA packages with a common data set, basically to answer your question because it is asked all the time by our lab users. I'll see if the two speakers are OK with sharing their materials with this listserv. But meanwhile, here is what the three of us--from qualitative sociology, sociocultural anth, and ling anth--concluded based on our prep for that workshop:

What all robust QDA software packages have in common is that they have steep learning curves. I think Alex has already alluded to this. It is largely due to their terminological differences: each package has its own meta-language for coding and querying data. E.g., what's called a "case" in one is a "code" in another, words like "attribute" and "node" are entirely program-specific, and even the words "code" and "query" do not denote the same functions across packages. It is also due to the fact that they are huge, powerful programs at this point in their development, with capabilities well beyond what any one individual researcher probably needs for any one individual project. Learning one is a time commitment, but it pays off over the years, if not months.

The other thing that they share--and here I'm talking specifically about Atlas.ti, NVivo, and MAXQDA--is their basic capabilities. Really. They will all let you associate some codes of your own making with textual, audio, and visual material, and they will all let you query that basic metadata in ways robust enough to reach well beyond what any human brain can simultaneously hold in it. They all handle the same basic file types. They all have a range of data export options. They are all currently developing their capacities for integrating social media data. There used to be much more difference between the competing QDA packages, even a few years ago when I founded the lab and a grad student assistant and I selected these 3 from among the wider field, but they have been converging in their basic capabilities and even in their visual layout.

What are their particular strengths and weaknesses? They were developed on different starting platforms, so even as they have converged on the surface over the years, they have weird deep-level differences. The way they store your primary data and metadata is completely different, for instance (inside vs. outside the program, in temp files on your hard drive, etc.). We concluded that NVivo is better for sharing data in a team. Atlas.ti is better at moving between computers and is more flexible in some other ways too, though their most recent release has some hiccups. MAXQDA was developed for Mac and remains the strongest option for Mac users. NVivo continues to have some problems if you move between Mac and PC, but their Mac option has improved dramatically with the latest release. Personally, I have been using NVivo for PC  since 2008 and still prefer it, but partly that's because I know it by now. I used to prefer NVivo for its transcription feature, and for the fact that it handled non-Latin scripts and IPA when the other packages couldn't. It also used to be better with PDFs, of which I had a ton in my media-rich research. At this point, however, Atlas.ti and MAXQDA can do all of those things. MAXQDA in particular is stronger on data visualization options, and the grad student assistant who just trained on all three programs found this one by far the most intuitive. That's not nothing, especially if you'd like a research assistant to get up and running quickly. For transcription, the crucial question is whether you have external transcripts that you would like to link to your audiovisual material inside the program or would prefer to have a single platform in which you do all of your transcribing and coding simultaneously. Atlas.ti is pretty terrible for in-house transcription, compared to the other options; most researchers who use Atlas.ti (and there are a lot of them at IU, because they have better student licenses and it is popular with sociologists) seem to transcribe in Word or ExpressScribe Pro and then dump their files into the QDA program. The better onboard transcription features of MAXQDA are compelling--this is currently the best program for transcribing inside of the program--and if I were starting out with QDA software right now, I think I'd go with MAXQDA. I still think NVivo wins when it comes to coding directly inside audio and video files though, because it directly links those codes in the recordings to the same spots in your transcript automatically. Moreover, you can code specific areas of photos and videos, which would be useful for coding gestures--although, come to think of it, I have never tried doing that in Atlas.ti or MAXQDA, so that might not be peculiar to NVivo. ELAN is still far and away the best option for the money! And it's best for subtitling. But, it requires a little more patience and tech-savvy, and it is more narrowly suited to ling data, when compared to full QDA packages.

So that we don't reinvent the wheel here, the U of Surrey has put together this great resource comparing current QDA packages in general--including several I haven't mentioned:

https://www.surrey.ac.uk/computer-assisted-qualitative-data-analysis

Upshot: Because they all have steep learning curves, I would strongly recommend trying out multiple options, if possible, to see what's most intuitive to you and will meet your own project needs. Unfortunately the trial versions are rarely as robust as the full versions--but you can at least get a sense of which layout and which of the basic features are more intuitive and useful to you. Specifically for coding gestures or stretches of talk in video-recorded interaction, I would recommend NVivo, versions 9-11, or ELAN+, and would also give MAXQDA a close look.

Take care,

Kate

Kathryn E. Graber
Assistant Professor
Department of Anthropology &
Department of Central Eurasian Studies
Indiana University
Frances Morgan Swain Student Building 130
701 E. Kirkwood Avenue
Bloomington, IN 47405-7100, USA
+1 812.856.3777
graberk at indiana.edu
________________________________
From: Linganth <linganth-bounces at listserv.linguistlist.org> on behalf of Janina Fenigsen <jfenigsen at gmail.com>
Sent: Wednesday, December 12, 2018 13:38
To: Riner, Robin
Cc: LINGANTH
Subject: Re: [Linganth] qualitative coding software

It would be good to put it all up on the SLA website, I think

On Wed, Dec 12, 2018 at 10:46 AM Conley Riner, Robin <conleyr at marshall.edu<mailto:conleyr at marshall.edu>> wrote:

Hello all,

Forgive me if this has been answered on this listserv in the past, but do people have recommendations for qualitative coding software that works well with linguistic data, especially video recorded interaction?

Thanks so much,

Robin

Robin Conley Riner, PhD
________________________________
Associate Professor of Anthropology
Department of Sociology & Anthropology
Marshall University
One John Marshall Drive
Huntington, WV 25755-2678
Phone: (304) 696-2788

_______________________________________________
Linganth mailing list
Linganth at listserv.linguistlist.org<mailto:Linganth at listserv.linguistlist.org>
http://listserv.linguistlist.org/mailman/listinfo/linganth
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/linganth/attachments/20181212/e60d1659/attachment.htm>