Q Methodology: A brief introduction

In preparing this introduction, I have relied heavily on Steven Brown's eight part series on Q methodology. For those who want to learn more about Q methodology, a lively discussion can be found on the q-method list: q-method@kentvm.kent.edu

Although Q methodology sounds complicated, it is really just a systematic way to identify and model the main themes in people's opinions about an issue. It differs from conventional survey research methods in that it treats the various views of an individual as coming together to form a meaningful whole or gestalt, not as a random collection of disparate views.

In a Q study, subjects are asked to consider a set of statements about some topic and to arrange them in some order, such as from "agree strongly" to "disagree strongly." This procedure is called a "Q sort." The resulting Q sorts are then statistically analyzed and interpreted to produce models of several distinctive sets of opinions on an issue. The statistical procedure used to analyze the Q sorts is called factor analysis. What matters in the analysis are correlations between the total responses of individuals, not correlations between responses on isolated statements.

What difference does this make? For a very simple example, say we want to know people's views on government spending, taxation and budget deficits. A conventional public opinion poll might tell us that 66% of the people support reducing the deficit, 66% support lower taxes and 66% support government programs that would require increased spending. On the face of it, this result seems inconsistent until we realize that those 66% are not always the same people. We would have a much clearer picture if we knew that 33% supported lower taxes, lower spending and reduced deficits; another 33% supported increased spending, increased taxation and reduced deficits; and the final 34% supported increased spending, lower taxes and higher deficits. Of course, if things were this simple, we wouldn't need factor analysis to work out the results.

Real issues and real opinions are much more complex than the above example, however. Q methodology provides the tools for handling some of this complexity. Another advantage of Q methodology is that, unlike conventional survey research, it doesn't require large samples of the population to produce a meaningful result. Where an opinion poll might need a thousand responses before it can claim to be representative, a Q sort completed by 40 or 50 individuals can give a good picture of the main trends of opinion. This economy of data collection allows and encourages the researcher to document and investigate much subtler nuances of opinion.


In Q methodology, the sum of what is said about a topic is referred to as a concourse. The first step in a study is to assemble a concourse representing what people have to say about the topic under study. The concourse can be gathered from a variety of sources: interviews, notes from less structured conversations, newspaper articles, scholarly essays -- even eavesdropping. In the proposed study of changes in working time, a concourse would be drawn from the books, articles and policy documents reviewed, from the email discussions, and from media sources and public discussion.

Characteristically, a concourse will contain a large number and variety of statements -- too large to ask volunteer subjects to sort -- so the second step in a Q study is to simplify the task by drawing a sample of statements. This sample is typically drawn by classifying the statements according to a descriptive matrix and then randomly choosing an equal number of statements from each compartment of the matrix.

Once a sample of statements has been drawn, they are ready to be sorted. The procedure here is to print the statements on a pack of randomly numbered cards, one to a card, and to ask subjects to sort them according to some rule, called a "condition of instruction". For example, the instruction could be to rank the cards in order from most agree to most disagree.

Subjects are asked to sort the sample statements into a set number of piles, each of a specified size. For example, the researcher may ask subjects to sort statements into 11 piles, ranging from +5 for most agree to -5 for most disagree. The researcher may specify that the piles, from end to end, resemble a flattened bell curve -- lower on both ends and higher in the middle. The number, size and distribution of piles is mainly for the convenience of administrating and statistically correlating Q sorts. It doesn't have a major effect on the results of the study (other than making them easier to obtain).

The completed Q sorts are statistically analyzed to find correlations and identify factors that are common to the sorts of several individuals. The number of factors identified depends in part on the degree of agreement amongst groups of subjects and in part on how much detail the researcher feels it is useful to analyze. In effect, a Q analysis is a way of summarizing a large number of opinions, there is always a point in a summary where more detail yields diminishing returns.

The final step in a Q study is interpretation of the findings. Factor analysis identifies which statements most distinguish each identified factor. Interpretation is a matter of seeing how the statements that characterize each particular factor fit together. In putting the picture back together, it is useful for the researcher to return to sources, reviewing interview notes from the Q sorts and documents from which the concourse was assembled.

The result of a Q study is a description of the diversity of coherent opinion on a particular topic. It is much more concise than the original concourse, yet is substantially complete in its consideration of important detail. One of the most compelling features of Q methodology is that the analysis can reveal unexpected patterns of opinion. Thus Q methodology is uniquely suited for generating hypotheses.

[Return to TimeWork home page] [Overview of the research process]

Copyright 1995 by knoW Ware Communications