Sunday, May 5, 2013

Finding a research question


Students often ask me about finding a research question, since I require one for theses of all sorts, and this blog post is an attempt to provide an answer.

Many students start with a topic that they would like to research. This is natural, but in some ways secondary to the process of scholarly writing. I recommend that students start with: 1) a method based on some well-established discipline, and 2) a source of data. Let me explain.


A method is the tool you use for research. If you were a painter, you might choose a variety of scenes to paint, but the brush and oils would be an essential part of how you approached it. As scholars we build up skills using certain discipline-based methods. In effect, we learn how to paint with a particular set of intellectual tools. If we ignore the tools or never learn how to use them, the chances of painting a satisfying scene are significantly less.

In graduate school I settled on a set of ethnographic tools, which I have used and reused over the decades. This does not mean that my approach has been unchanging, but it has had a consistency based on long practice. I am not a trained ethnographer in the German sense of having a degree in it or even having taken classes, but cultural anthropology was part of the atmosphere of my graduate school environment, and I just keep reading and practicing it in my dissertation and beyond. Having a method  means absorbing a way of thinking. This is essential to formulating a research question.


Access to data is the equivalent to providing a painter with a scene to paint. If the painter has no one who will sit for a portrait, it becomes much harder to paint a portrait. People try sometimes with varying success, but without sufficient experience with real subjects, it is hard for a painter to create a portrait in the abstract. 

Some students want to rely entirely on existing published results and to comment on them. This is more like copying a painting than creating a new one. It can be a reasonable approach if they can do a new analysis, but for a beginner  merely to comment on other people's work without actually analysing the data anew risks superficial results. 

Data are hard to get. Many desirable sources are closed to the public, and many public sources are overworked or unreliable. Data do not have to be perfect to be used in a scholarly study, but they do need to be available and the author does need to understand and be able to explain their imperfections.

The question

Once the method and a source of data are clear, the student can then reasonably begin to formulate the research question. It needs a grounding in the scholarly discourse in the field to explain why that particular question is interesting. Many students want a completely new question so that they can do something original, but wise students often take a well-researched existing research question and approach it with new data or a new method. The advantage of an existing research question is that its importance is already clear. 

The best research questions for a thesis are ones with a straightforward answer. I generally recommend a yes/no question, or one that has a quantitative answer, or one that is a choice among reasonable alternatives. These are not the only possible research questions, but questions involving complex issues about "why" or even "how" tend to be beyond the scope and experience of even the cleverest doctoral students. The virtue of  a yes/no type question is that the student can make a clear choice. A thesis with a vague answer is not a contribution to knowledge, while even a very narrowly stated and highly qualified yes/no answer can be a reasonable step forward.

Choosing a research question is hard, but it is probably the most important step in writing a thesis. The topic matters only in so far as data are available and the research method can reasonably apply. Topics are temporary and can change with the seasons. Good research questions grow ultimately out of the intersection of scholarly methods and quality data. 

1 comment:

  1. See also