The interest in the post on how not to do qualitative research went way beyond what I expected. Soon after I posted it, however, I was challenged to write its positive version: how to do qualitative research. And so, here it is. This post contains some ‘bad language’.
Writing this post was considerably more difficult than I expected. How do I pitch it? There are many textbooks and other texts which tell you all you could possibly want about qualitative research. I can’t and don’t want to compete with them. This post cannot be a general introduction to qualitative research; to be honest, I really don’t feel equal to the task.
So, what I decided to do is to take the post on how not to do qualitative research and use it as a template for my positive post. I will need to ignore a couple of things, I will need to add a couple of things, but somehow, I thought it might work. Once again, I do not pretend to offer an ultimate guide to qualitative research. Rather, I want to make a few comments on its aspects that, in my opinion (and only in my opinion), researchers could benefit from.
1. Language and thinking. No, you cannot claim that I think something just because I say something. So what’s the positive message? Well, you could say that linguistic patterns you found are unlikely to belong only to your informants. In other words, it is very unlikely that only your informants say things in a particular way.
Let me give you an example. I wrote about men talking about depression and I wrote about a number of patterns in their stories. For example, they never spoke about their depression directly, always distanced themselves from it. And so, they would never say “I am ill.”, they would talk about “the illness”, or that “one is ill”. Now, I interviewed over 30 men and qualitative research does not allow me to talk about anyone else.
However, we also know that people don’t consciously choose the linguistic form. In other words, we don’t normally say to ourselves, mmm, I’d better do passive voice now. And so, if you put these two things together, you can postulate that the patterns you observe are likely to go beyond the corpus of narratives you have. For it is unlikely that a group of 30+ men, from different regions, interviewed in different hospitals, will have developed their own ‘discourse’. But do bear in mind that I said ‘postulate’ and I do refer to the linguistic form. Can more be characteristics of stories be postulated to be like that? I don’t know.
3. Analysis. That’s probably the most difficult thing you can give ‘advice’ on as there isn’t one uniform, homogenous analysis. The kind of analyses I do are very different from the analyses done in other forms of DA, CA, TA, IPA and their multiple versions. I do have one point to make though, though I realise that it will not be applicable in equal measure to the different variations of qualitative research (and that things are complicated).
One of the points I insist on when reviewing papers or examining work is that there is a clear difference between the analytic and interpretative stages of our research. The difference is that analysis should come straight from the methodology. I tend to think that you and I should come up with the same analytic conclusions, if we analyse something. In other words, if, for example, a psychologist writes in their clinical notes (a genuine example):
Support was given.
we will agree that the sentence is in passive voice. We will also agree that the linguistic agent (the giver of support) was removed from the sentence structure. Easy. And if you tell me what you’re looking for, I should be able to come up with a similar answer.
Interpretation is when we might differ. I might think that the psychologist was uncaring, afraid of taking responsibility, you might think that it’s to do with practices of writing clinical notes, someone else might simply say that the psychologist was just uneducated. This is how we interpret our analytical results, sometimes it’s more informed (you can check the writing practices, for instance), sometimes less, but this is what interpretations are for.
Long time ago, I co-wrote an article about the CIA website. It was all going well, until my co-author (a British communist) and I (a child of communism) wanted to make sense of what we found. And we simply could not agree. Our biographies were so different that we simply could not come up with a common account of what we make of the data. And so, we didn’t, we actually wrote that we couldn’t agree. And that’s interpretation for me. It can be more subjective, personal, reflective. And it is precisely because of the subjectiveness of interpretation that I think the two stages of the analytic process should be (explicitly) separated. Yes, sometimes it’s difficult.
3. Analysis again. I am frequently quite surprised how sketchy the accounts of data processing are. Data are collected, themes emerge, coding is done. And I keep wondering what’s behind it. In a Twitter exchange I called it the black box of analysis. What’s inside?
I think ‘we’ would all benefit from being more clear (sometimes considerably more clear) about what exactly we analyse. Of course, this statement applies to various versions of discourse analysis in different ways. So, when you code things, what do you focus on? I once tried to answer that question for myself. This is because I do use coding as informal pre-analytic data-sorting strategy. It’s informal in the sense that I make no claims at all on the basis of my coding, yet, for transparency, I thought it would be a good idea to be able to account for it. And so, it turned out that I tended to focus on lexical material only. This is because I want to cast the net as widely as possible.
So, if someone uses the word ‘suicide’ and something within its semantic field (like ‘kill oneself’ and others) I will code the fragment accordingly. In such a way, I come up with all those fragments in which my informants talk about suicide. But, to stress, this is just sorting the data – making it manageable, if it’s large (in qualitative research terms, of course). I don’t use coding on small corpora.
However, if I wanted to make claims on the basis of such coding, so for example, to claim thematic configurations in my data, I think I would have to be considerably more specific and structured. For admittedly, sentences like
Suicide is something I have thought about.
I once thought about killing myself.
are quite different, even though both refer to suicide. But can you claim, for example, that both are consistent thematically? That is to say, can they both be said to have a ‘theme of suicide’? I’m afraid I’m not certain at all this is the case, and I really would like to see a cogent argument that the theme of the second sentence is suicide, and not self or thinking. If, on the other hand, it can be all three, that raises all sorts of issues.
And here we come to my third piece of impressionistic advice. I would like to know what exactly you do when, for example, you do your coding. Since we, qualitative researchers, tend to analyse text, what is it in the text that you focus on? Even if your analysis is based on intersubjective agreement of researchers, in my view, this agreement should be explicitly about something. And so, I would say: agree not on what you say, but agree first on what you say it about!
4. Transcripts. I need to repeat this. I’m always astonished with the presented extracts, it’s as if people read them.
There are many levels of transcription you can use. From barely noting pauses and repairs, all the way to CA-like transcriptions which note a gust of wind outside if it overlapped with the informant’s sneeze. Use whichever is useful for your research, but use one! And, most importantly, don’t sanitise what your informants said. If your informant uses ‘bad language’ do keep it.
Here is one of the examples I give. Many years ago, I was witness to a conversation between two Polish soldiers. To say that one of them was using ‘bad language’ is an understatement. What he was saying was just about only bad language. Needless to say, I haven’t got a recording (also, it was in Polish), but just to give you some flavour of what he was saying, it could be rendered like this:
And so, fuck, I mean this fucking fuck was just fucking telling me, the fuck, that I should get the fuck a fucking move on…
I’m sure we can all agree that this statement is very very different from its sanitised version:
And so, he was telling me that I should get a move on.
We do lose quite a lot, don’t we? Even though removing all the f-words doesn’t change the contents of what is said, still, the message is very different. And just to be clear, I reject the notion that at an academic conference quoting from such data is somehow embarrassing, offensive etc. Data are data, language is language, beautiful in its variety.
5. Claims/knowledge. To a considerable extent this is what I wrote in a recent post: not everything is ‘a discourse’. I believe ‘we’ use the word ‘discourse’ too often.
Consider what exactly your data let you claim. Instead of talking about discourses, consider talking about practices, rhetoric, strategies. You could also talk about constructions, ways, patterns. For when I see a claim to a ‘discourse of book titles’ (genuine example), I can’t keep wondering whether it is a discourse too far.
6. Context. When I interviewed men in depression, in a number of interviews, my interviewees referred to me as a man. Whether I liked it or not, it became relevant in the interview. It is more than likely that it also had some impact on what and how the interviewees talked during our conversations. And so, I’d risk a statement that who you are might sometimes be more important than who your informants are. Reflecting on it is, I think, important and, incidentally, I am directing these words to myself, too. I don’t think I do enough of it.
It’s important because it’s about co-constructing the data. How? I don’t know. But if the ‘men in depression’ were roughly my age plus/minus 10 years, there is little doubt, I think, we shared many experiences of what it is to be a man, a father, which, sometimes, was made explicit. That helped my interviewing, of course, but it also raised the issue of ‘whose data’ I was collecting.
To be honest, I haven’t got clear and simple ‘advice’ as to what to do. I think it’s complex, but I also think it’s important to engage in such reflection. There aren’t ‘interviews’, there are ‘Dariusz’s interviews’ and that’s very different, I think.
So, here you are. This is my take on how to do qualitative research. It’s by no means comprehensive, it needs considerably more nuance – after all, I’m sure we all have said a number of times that there isn’t any ‘qualitative research’, as if it were this homogeneous blob of research. But I hope we can continue debating.