Having posted something on the list several months ago asking if anyone
was attempting to use Nvivo to support discourse analysis this is indeed
an interesting conversation!
As a PhD student with a department has an Nvivo licence there was
undoubtedly a financial aspect to my decision, as Anne makes reference
too! That said I do think its important to raise and question to
dominance of "coding" in the software - and I think that this is why
there is a reluctance for many qualitative researchers (not just
discourse analysts) to use CAQDAS - as although coding is part of many
approaches largely it is not in and of itself analysis. That said I am
finding uses for coding and this maybe because my data seems very
different from Thomas's newspapers reports. For example, one of the
things I am looking at is the talk of a few participants over a variety
of circumstances - in one-to-one interviews, in focus group sessions and
in meetings I am attending as a participant observer - being able to
code this via a case node has helped to keep track of all this! I am
also collecting a fair amount of what you could probably call background
data while at the organization - and again Nvivo is a logical place to
dump and then sort through that. Maybe this is just "data management"
and maybe there are other packages that I could have used! In terms of
actually doing discourse analysis in Nvivo - I prefered to write
detailed notes within the transcripts rather than code - although I have
found some coding by topic useful!!
Katrina Pritchard,
Dept of Organizational Psychology
Birkbeck
-----Original Message-----
From: qual-software [mailto:[log in to unmask]] On Behalf Of
Thomas Koenig
Sent: 11 June 2004 01:54
To: [log in to unmask]
Subject: Methodologies and CAQDAS, was: Re: Freely available comparisons
of Atlas vs. NVivo
Ann,
this becomes interesting. Thanks for your thoughful comments, here is my
reply: At 18:33 10/06/2004, you wrote:
>BUT.... theres a lot more to much CAQDAS software , than just coding
>(as you say in several places), but I mean a LOT. ...and anyway in
>defence of codes - they are, as Seidel and Kelle suggested (1995) 'just
>heuristic devices for discovery' ...I love that description, and it
>holds a lot more possiblities than are catered for in your statement
>"Coding in my view has a strong elective affinity to counting. By
>coding you standardize your data to a certain extent, which is not a
>bad thing, but something that can easily be used for quantitative
>analysis" ...well I know I took that out of context slightly
>Thomas...but to many of us its 'damning with very faint praise'
Well, actually, I'm quite happy with the quote, even out of context,
because I personally don't find anything inherently wrong with counting.
In fact, I really like counting and stats, and this is where I see loads
of opportunities for CAQDAS, which have yet to be explored. I sometimes
have the feeling that among some so-called (so-called, because I really
see no point in the qualitative/quantitative distinction) "qualitative"
researchers, there is some underlying assumption that quantitative
research is theoretically unchallenging, even by definition
positivistic. It is not. There is good and bad in both qualitative and
quantitative work, so it is not a disgrace, if you compute some
frequencies, or do some more sophisticated quantitative analysis, if
theory and data allow for such analysis.
There also is absolutely nothing wrong with "standardization." Yes,
social life is complex and cannot be neatly fitted into standardized
categories, but after all *any* social theory must to some extent
standardize, otherwise it's not a theory, but simply a description of
reality. So, even if we cannot *neatly* fit social life into categories,
we still have to fit it, in a way that does both justice to complexities
in real life and the need for parsimony in theory.
OK, after this little defense of counting and coding, I would also like
to make the point that coding is not always the way forward.
To give you a simple example from my own work: I am currently looking,
how newspapers in different countries framed the "Berlusconi-Schulz"
incident (inter alia using CAQDAS (MAXqda) for coding btw). Last July,
Italy's prime minister Berlusconi called the German Social Democratic
MEP Martin Schulz a "kapo", an auxilliary concentration camp guard.
Papers framed that incident in all sorts of ways, but - from my
preliminary skimming -- there appear to exist a number of things they do
*not* say. For instance,
- The papers do not mention that the *main* insult of Berlusconi was
(probably inadvertendly) against the victims of the holocaust,
*because*, however nasty Schulz may be, he certainly did not instill the
same fear in his adversaries as did guards in the concentration camp
inmates. A few papers say *that* Berlusconi made an insult to holocaust
victims, but they fail to elaboate, why that is the case.
- They also do not elaborate on the fact that /kapos/ were prisoners
themselves and not even necessarily ethnic Germans.
How would I code something that is precisely *not* to be found in the
data (or so I hope, because, otherwise I will heve to adjust my theory)?
For sure, it would make no sense to code each and every story with
"concept of kapo has not been elaboated." In fact, why should I even
make such a code, after all, there are hundreds of other things that go
unreported, so why would I expect that the papers should explain, what a
kapo is?
Of course, this is a pretty straightford case, where coding does not
help me at all, but theory does. For my purposes, this is a mere aside,
most of the remainder of my work is done with coding -- and counting.
But a more sophisticated theoretical analysis, might not require any
coding at all. Kracauer (1952) makes this point more eloquently than I
can do.
> >.Which *other* than methodological (and therfore implicitly
> >theoretical)
>aspects would you suggest in the choice, if and what CAQDAS to use? I
>can only think of financial aspects here, but I don't think that's what
>you had in mind.
[...]
>And there are *other*
>considerations' Thomas - often researchers don't have a specific
>methodology - yes really!
That's what I fear. Call me old-fashioned (or maybe
anti-post-Feyerabend?), but I still think that methodology is the most
important thing for empirical studies, well after theory, that is.
>They only want a data management tool.
If you really "only" want a data management tool, I still hope you
already have an idea *how* to manage your data. Sometimes, for ordering
data, it might be much easier to use a non-CAQDAS management tool. Even
the Windows Explorer might do, or more sophisticated replacement like
Cardfile. Or a spreadsheet program. Or a database program. It all
despends on your type of data and your approach. For much of
International Comparative work, I would avoid CAQDAS, because even with
recent advances, they don't swallow as many file formats as does Windows
Explorer or Cardfile, and for most purposes, a spreadsheet would do the
job much more efficiently than any CAQDAS. If I want to compare, say
media systems across countries, I might get information from all sorts
of different sources and would store these information in a database.
That is much more efficient than using even the most versatile CAQDAS. I
can even take the relevant information, say, number of local/national
papers, readership figures, degree of state involvement in the media,
etc., from hardcopies and type them into my database. For further
analysis, fs/QCA would beat any CAQDAS hands down.
Ann, your post was quite thought provoking, and so I came up with yet
another problem (not a shortcoming, but a problem) of methodological
guidance through CAQDAS: They tend to guide you towards positivist
theorizing, and by positivist I mean positivist in the original sense,
that is inductive theorizing a la Vienna Circle (Carnap et al.). Just
look at the "in vivo" coding function most CAQDAS offer: If that's not
an ingenious way of inductionism, what is? Of course, sometimes /in
vivo/ coding is a useful thing, but you might get addicted to it, if you
are not careful. It's easy to start out with /in vivo/ coding and then
gradually work yourself up to higher, more general codes, espicially, if
you do not have a clear-cut theory, before you examine your data. In the
end, you might end up with some thoroughly positivistic theory and one
that is not even necessarily supported by the power of numbers. All you
end up with is a summary of reality. A neat one, to be sure, but not
really a theory, even less a critical theory and even less Critical
Theory.
Several social theorists have emphasized that positivistic theorizing is
not only inherently conservative, but when it comes to discourse data,
it also precludes effective analysis, which requires you to distance
divorce your theoretical concepts from everyday life concepts (Bourdieu
called that "breaking with social categories").
I am aware that there are scholars, who think differently. Particularly,
when I look at some of the stuff that is written in the Denzin/Lincoln
reader for Qualiative Research, I see people, who seem to advocate a
theorizing that more or less "arises" from the data (I am definitely
oversimplifying here and will expand on that strawman in the future).
But that's precisely my point: They would not feel "guilty", when coding
/in vivo/, but given my different theoretical outlook, I do (still, I
press that button time and again just for convenience).
Again, nobody is forced to code /in vivo/, but chances are, if the
concept is available, you might use it, even if you don't have any
theoretical justification for it. Just like it is standard practice to
compute correlation matrices in SPSS, despite the fact that is hugely
problematic, too.
Thomas
__________________________
Kracauer, Siegfried (1952): "The challenge of qualitative content
analysis," Public Opinion Quarterly 16: 631-642.
|