Paper summary - Thematic Networks: An Analytic Tool For Qualitative Research, by Jennifer Attride-Stirling

What I read

In this article, Jennifer Attride-Stirling presents a method for creating thematic networks to summarize themes in research data. She then illustrates how this method works through an example.

The author begins by showing the increased popularity of qualitative research as a respected way of gathering data about human behavior. But, there are few techniques or articles about what to do with the data that is generated through the research. She describes that there must be a rigorous way to structure that data, but most authors have either skipped the process, or failed to indicate their process with their written findings; this makes it difficult to judge the efficacy of their work, and also, to leverage and learn from their techniques. She indicates that her goal with this paper is to show a method suitable for closing the gap between data collection and analysis.

In the first section of the paper, Attride-Stirling references the base of her technique in Argumentation Theory. She describes how Toulmin created a way of thinking about argument, one that focuses primarily on gathering data in order to produce premises, and then use those premises to make a claim. Attride-Stirling indicates a parallel of this three-step framework to grounded theory, which substitutes concepts, categories, and propositions.

Next, Attride-Stirling describes thematic networks analysis, the method that is introduced in the remainder of the paper. She indicates that the core of the technique is to create “web-like networks as an organizing principle and a presentational means.” It’s a methodical way of moving from premises to themes, and then to “super-ordinate” themes which are more metaphorical, or global. The artifact output of this is a web diagram; the intellectual output is a basis for further analysis that is manageable, as compared to trying to analyze all of the data produced in qualitative research.

The three “levels” of theming are described. A Basic Theme is a simple “characteristic of the data, and on their own they say very little about the text or group of texts as a whole.” These basic themes then equate to an Organizing Theme, which creates clusters of similar issues. They are “more abstract and more revealing of what is going on in the texts.” In turn, these taken as a whole lead to Global Themes, which are a claim: a conclusion. These are an interpretation in the whole.

Next, the author explains how to perform this method in a step by step set of instructions.

First, researchers code the material. They “Devise a coding framework… it tends to be done on the basis of the theoretical interests guiding the research questions.” It is based on the researcher’s objectives, and it may be things already established, or from the data itself.

Next, the transcribed data is coded according to the coding framework.

After coding, themes are identified, by extracting the “salient, common, or significant themes.” These are refined, based on interpretation.

The themes then lead to the construction of the networks themselves. A goal is to produce between 5 and 14 groupings. These then lead to Basic Themes, which through rearranging, lead to Organizing Themes. These then lead to Global Themes.

The various elements are presented as an illustration, and the entire network diagram is refined.

The researchers then describe and explore the network. This produces a narrative that often leverages the actual quotes from the data, and elaborates the analysis for a reader.

Next, the network as a whole can be described in a summary in a succinct manner. And finally, the patterns evident in the network can be interpreted in the context of an original hypothesis or research question

After explaining the method in an instructional manner, Attride-Stirling then shows how it is used through a real case study, one focused on nature in culture. She walks through each of the steps above, showing how the various techniques were applied.

She then summarizes the process through a table. The first column shows the creation of the initial codes, based on the researcher’s interests. This leads to the issues that were discussed by the researchers from the raw research data, which then lead to the identification of the Basic Themes. This leads to Organizing Themes, and then to Global Themes.

The output of all of this is a simple thematic network diagram, showing the global element in the middle, and nodes emerging related to the smaller supporting points. Attride-Stirling shows how the diagram, and the process of organization, is then discussed by the research team, and includes partial transcripts of the discussion and how it emerged from the diagram, and added clarity to the material presented.

An example is shown of the next step in the process, summarizing the network, and this then leads to the reflective discussion that supports, rejects, or expands upon the initial hypothesis and claim.

Attride-Stirling concludes by acknowledging, and positively emphasizing, the subjective nature of the research and the analysis itself; the power of all qualitative research is in exploring and explaining, instead of proving. The method described is one of many techniques for moving from data, through the exploration, and to an explanation.

What I learned and what I think

This is a paper describing the synthesis process I’ve used at frog, Blackboard, Modernist and Narrative; it’s one that has become fairly standard for UXR teams, albeit used in those contexts often in a vapid or less rigorous way (due to time constraints, or maybe to poor training, or both.) I know this process works, and I’ve found many ways to refine it operationally to be more effective and powerful for me and my teams. I like that it is acknowledged as a legitimate way of working through academic qualitative research, and that a method I’ve found to be tacit is “okay” to be used in data that I gather throughout my own research in this PhD-getting process.

I see value in the small diagram that is produced in the end, as it is a quick way to communicate a large idea. But as I’ve seen those diagrams in published academic research, without a slide build or talk track that might come in person, they almost always appear to be really thin. It makes sense that they don’t do justice to the research, because the goal is to smush thousands of data points into something small and tidy—it’s overly reductive. In one way, the paper is describing how to make a concept map, but producing the map from the data that’s been gathered. That’s cool; not particularly unique, but maybe that doesn’t matter, because it seems to be important for upwards of 9000 other writers who cited this paper!

One main piece of value of the paper is in showing the power of interpretation and sensemaking, by moving “up and down” in the data, between individual points to groups to points several times. Like the diagramming, this has been a part of a user-centered design process for as long as there has been contextual research and synthesis, so it rings true to me and again, seems to ring true to many other qualitative researchers.

The process also seems to lead naturally to the format of a research paper. For example, quotes are used to create an observational (“objective-ish”) narrative; the discussion is, ultimately, the “discussion” section of the paper; and the summary and interpretation seem to map nicely to a summary, and to an abstract. The creation of the narrative, interspersing quotes with observations, literally is my storytelling technique for clients, except we use images to tell the story (in my opinion, more effectively.)

Something doesn’t sit well with me about where the initial codes come from. The method appears to recommend that these codes are externally provided: that they come from the researchers and their interests. It seems inevitable, then, that the initial frame, after analysis, will lead to… itself. The example used does this, precisely; the core of the diagram is, basically, the codes, transformed into assertions. This is definitely not how we work in design research, where the data itself leads to the initial codes, that then shape the rest of this process. This seems to be because the entirety of qualitative research, as presented here, is that “…one of the principal reasons for using this method is, precisely, to bring to light the meaning, richness and magnitude of the subjective experience of social life… The value of qualitative research lies in its exploratory and explanatory power.” This is only a part of how we use design research; we also use it to identify the subjective experiences people have (and often with the human-made life, not just the social life, although it could be argued that those are the same.) We are trying to identify unmet needs or unfulfilled desires so that we can meet them, because our output is not knowledge for knowledge’s sake, but product or service creation.

I don’t know if this is an accurate read on what is one person’s assertion about the role of qualitative research in academic exploration, but if it’s true that the research is intended to better explore an idea, then it makes sense to start coding with the parts of the idea you want to explore. But if the intention is to uncover new ideas, then the top-down coding doesn’t work at all.

I like the tutorial-nature of this, because it’s how I write and it seems to be an extremely useful style for students. I didn’t expect this to be something that a community of researchers would appreciate so much on a published level. It’s pretty casual, even colloquial (“…but the fun doesn’t really begin until…”) Maybe that’s fine, which is cool with me.

Tomorrow, I’m going to shift to a paper Carl recommended, Designers’ professional identity: personal attributes and design skills. It’s fun bouncing around like this.