You research politics and are interested in political communication? Then chances are your most common source material is the text, and rightly so. Much of politics is expressed through texts, and closely examining both written and spoken language can provide useful insights into the political position of actors or the rhetoric that informs an argument.
Yet exploring politics through texts is by no means easy. Conducting a professional analysis requires not only access to the right texts, but also time and possibly other resources. This can lead even experienced academics to cut corners. One understandable temptation might be to limit oneself to reading political texts, to summarize them for others, and to then offer a personal interpretation. This may be a legitimate way to produce a review of the relevant sources, but is it an actual analysis? I believe that an analysis has to be more systematic than this. More importantly, I think that good, confident research should be based on evidence and should be transparent, so that others can check the evidence.
In a previous post, I gave an introduction to the theory of discourse. In this post, I turn to the practical question of how to set up a discourse analysis. I will first discuss what is a good research question for a discourse analysis, and will then go over some of the fundamental issues you might want to settle before starting your project. Along the way, I will introduce you to a couple of analytical concepts and approaches that I find helpful.
Finding the right research question
Like with any type of research, a good discourse analysis starts with the right question. Before you pick your sources, and decide what tools to use on them, ask yourself: what is my concern? What motivates you to conduct this particular research, and what do you hope to achieve? For instance, are you trying to find out what the position of a particular government is on a specific topic? Do you want to explore how specific political actors made sense of a crisis event? Maybe you want to know what arguments informed a major political decision, or you want to see how a politician manipulated a debate. All of these are legitimate concerns that could stand at the beginning of a detailed discourse analysis.
Try to turn your concern into a question and then narrow the scope of that question enough to make it “operational”. This will allow you to find sources that promise to address your concern. A poor research question would be how South Koreans think about North Korea. This question is too broad, not to mention that it assumes we can find out what a large group of people actually think. A good research question, on the other hand, would be what position the South Korean President Park Geun-hye voiced towards North Korea during the 2013 crisis between the two states. This question has everything a good research project needs: relevance, a clear topic, potential sources, and a clear time-frame.
In some cases, you may have simply come across a text that you liked, and that you feel deserves to be analysed. This is perfectly acceptable, but make sure you ask yourself what about the text attracted your attention. You likely had an implicit concern, and it is this concern that you can now turn into an explicit research question. Did a speech have a strong emotional effect on you? Then why not explore how the text elicited that emotion. Did the broad public reaction to a political debate intrigue you? Then how about trying to trace which arguments from the original debate made their way into journalistic texts and social media outlets, and in what form.
Without such a concrete agenda, you may still enjoy analysing the text that sparked your interest, but your hard work may not be able to carry a larger project, like a graduate or even post-graduate thesis. Pick your materials based on your questions, not the other way around.
Five points you should address before getting started
To narrow down your project and decide what sources and tools will be appropriate in your case, you may want to clarify the following five points:
1) What topic will you explore?
For instance, you could look at statements on national security, or statements on nuclear energy, or statements on health. The discourse analyst Siegfried Jäger (2004: 160-168) calls such general themes discourse strands. The idea is that such strands are intertwined, and that it can be helpful to explore not only what statements people make within one strand (Jäger calls such statements discourse fragments), but also to explore how one strand relates to others. Think of the disaster that struck Japan in 2011: what statements did various political actors in Japan make on nuclear energy before and after the event? How did these statements draw from assumptions, beliefs, and arguments that have their roots in the discourse strand on national security, on health, or on any number of other such themes? In short: define what topic you will analyse, and note down which various discourse strands you think might be important in that regards. These notes can then become your first coding categories: the analytical attributes you assign to different units of text, such as paragraphs, sentences, or even words, to later explore their distribution across the text, and their relation to one another.
2) What is your time frame?
There are generally two types of discourse analysis: the first focuses on a specific moment in time, and is called a synchronous analysis (Jäger 2004: 171). Discourse stretches out through time. Think of discourse for a moment as a bundle of intertwined wires, each with a different colour, that cross and twist as they stretch forward. These individual wires are the discourse strands, and the wire bundle is the discourse in its entirety. What a synchronous analysis does is dissect the bundle of wires at one spot and look at the incision: where is a specific wire located at that point? Does it touch other wires? The section where you slice into the discourse can be a major event that generates discussion. Such a discursive event could be an earthquake, or a terror attack, or an election.
The second kind of analysis looks at different sections of the wire-bundle and compares them. This is called a diachronic analysis, and is the sort of approach that the famous discourse theorist Michel Foucault favoured: he looked, for instance, at how the institutional setting of the hospital and the role of medical practitioners changed over time, and how these changes were a reflection of (and in turn an influence on) the kinds of “truths” that people held about health at different points in time (Foucault 2005/1989: 55-60).
3) What part of society is your analysis looking at?
Are you mainly interested in the discussions of politicians? Do you want to know how a political discourse plays out among academics, or in the press, or in private setting of people’s homes? Each of these different spheres, or discourse layers (Jäger 2004: 163), is interesting in its own right, but they also influence each other. Tracing how ideas travel between these layers through communication can be a rewarding analysis.
Think about the way in which an argument has influenced debates in different parts of society, for instance Samuel Huntington’s famous claim that the conflicts of the future will come from a clash of civilizations (Huntington 1996). How did this claim make its way into journalistic texts? When did politicians start positioning themselves and their political statements in relation to Huntington’s theories? What kind of statements did this lead to? These are the kinds of questions that a multi-layer discourse analysis would ask.
4) What medium and what language will you be working with?
A major decision on your part will be where you will look for discursive statements. You will need to be clear about the kinds of sources that promise to help you answer your question. A good analysis should explain what texts you used, where they came from, and why you chose them. You might decide, for instance, that your question is best answered by analysing Chinese newspapers, or speeches by Japanese politicians, or interviews with Korean activists.
If the source only exists in a verbal format, this does not need to stop you. You can transcribe such sources and turn them into written data, and can even add special markers to show intonations and emphases. Chilton (2004: 206) has provided a very useful list of such annotations.
Another important issue is whether your source material is already available digitally, or if you can somehow digitize your texts. It is possible to do a discourse analysis with paper-based sources, but digital texts allow you far more analytical options. Think only about how hard it is to do a text analysis without a search function. My advice would be to get your hands on digital versions of your sources, or to generate these yourself by typing up speeches, transcribing interviews, or scanning news articles and running them through software that supports Optical Character Recognition (OCR).
5) Will you need to work qualitatively or quantitatively, or maybe both?
If you are looking at one source, for instance a speech, a journalistic article, or a constitution, then your main concern will be with the kinds of discursive statements that this text makes, and the manner in which it makes them. In other words, you will likely be exploring qualitative aspects of the discourse. On the other hand, if your subject matter generates a large amount of text, for instance thousands or tens of thousands of words, then you will have a hard time deciding which statements to analyse in detail. In such a case, it makes sense to look at the numbers first, for instance by exploring which key words appear most commonly across the different texts.
Word distribution can be an interesting research project in its own right, particularly if you can highlight major emphases in a text. Just take a look at the two very different word clouds below that I created as part of my own research on Digital Nationalism in China, using a handy visualization tool called Tagxedo. One cloud represents the most common words in the Chinese government’s white paper on the relevance of the internet. The other is a similar representation of key words in a speech by Hillary Clinton on the same topic. You can immediately see how different the two political positions are, and where the differences lie.
I personally use such quantitative analyses as a starting point: they can reveal regularities or irregularities across vast amounts of data, and can highlight which specific parts of the text corpus might then lend themselves to a detailed qualitative analysis. In the example, it might now be interesting to check which key words appear in close proximity with one another. It might also be worth analysing what exactly the Chinese government has to say on internet security, or how the word government is used in the text, or how the two different texts use the word people.
Thanks to Florian Schnider (Politics East Asia) for writing the above.