Free and Frank Advice: Studying Meaning with the Decentred Approach and Q-methodology

10/01/2026

I explain why my thesis studies free and frank advice as practitioner meaning, how Q-methodology tests whether that meaning is shared, and how the three-stage design and its analytical decisions worked.

This is the eighth of twenty posts drawn from my doctoral thesis on free and frank advice in Aotearoa. Having shown that neither the statutes nor the scholarship say what the convention means, I set out here the method I used to go to the people who supply that meaning every working day.

If the content of free and frank advice sits outside the statutes and the scholarship, and is supplied every day by the ministers and officials who give and receive advice, then the inquiry has to begin with what those practitioners understand the convention to be, and it cannot begin from a definition supplied by statute, by the literature or by me as the researcher, because each of those would fix in advance the very thing the research asks practitioners to supply.

In this post I explain three parts of the method in the order I built them: first, I set out the interpretive theory behind the inquiry, the decentred approach, and why it suits a convention made meaningful through judgement; second, I explain why interviews alone were not enough, and how Q-methodology allowed me to test whether practitioner meanings form shared patterns; and third, I describe the three-stage design and the decisions I made in turning interviews, sorts and validation material into findings.

Beliefs, traditions and dilemmas

Interpretive political science begins from the proposition that political and administrative action is meaningful action, and that meaning has to be recovered from those who hold and use it, and the version of that theory my thesis draws on is the decentred approach developed for public administration and governance through the work of Bevir and Rhodes (Bevir & Rhodes, 2006; Rhodes, 2017). Its value for this inquiry lies in three connected ideas: actors act on beliefs, the meanings they give to the world and their place in it; those beliefs are carried in traditions, the inherited webs of understanding through which practitioners learn what counts as proper or improper conduct; and traditions are tested and modified through dilemmas, the moments when new ideas, events or experiences press against inherited understandings and require people to reconsider them. In my view an official deciding how directly to state an unwelcome judgement, or whether to press a legal concern against a minister’s preference, is acting on exactly such beliefs, and doing so inside traditions that are being tested.

Those traditions are inherited unevenly in Aotearoa, and the dilemmas that test them are live. Practitioners draw on Westminster and Whitehall, on the statutory public service, on Te Tiriti o Waitangi and the constitutional order in which public administration in Aotearoa sits, and on the managerial reforms that changed the conditions in which advice is produced and heard. They face dilemmas about whether candour requires saying the unwelcome thing now or keeping the relationship for later, whether responsiveness has become accommodation, whether disclosure changes what is put to paper, and whether working knowledge from delivery has reached the advice at all. A decentred inquiry asks how practitioners understand those inheritances and dilemmas and how they construct meaning in response, and it treats subjectivity, in the sense of a practitioner’s standpoint, as the object of study. I came to regard that as the only defensible starting point for research into a convention that statute names and declines to define.

The approach was developed in and for British governance, however, and imported unchanged it could treat the Westminster inheritance as neutral, which in Aotearoa it is not. I therefore extended it through Āta, a Māori methodological practice that shaped how participants were approached, heard and returned to, and I take that extension up in the next post. For the purposes of this one, the point is that the decentred approach tells me what to look for, which is practitioner meaning, and it leaves open whether that meaning is shared.

Q-methodology and subjectivity as configuration

Interviews can recover practitioner meaning in depth, and the open question they leave is whether those meanings remain individual accounts or cohere into shared viewpoints, which matters a great deal for a convention, because a convention that lived only as individual meaning would hardly be a convention at all. Constitutional scholarship has observed that conventions are widely acknowledged while only partly identified (McLean, 2024), and for free and frank advice to persist as a convention, the views practitioners hold of it will very likely need to cluster. A method was therefore needed that could examine both the pattern of meaning and the limits of that pattern.

Q-methodology studies subjectivity as configuration, and it reverses the usual direction of analysis. Where survey designs begin with variables the researcher defines and ask how people vary across them, Q-methodology asks participants to sort a set of statements drawn from the field itself against one another, and then correlates whole sorts with one another to find shared viewpoints (Brown, 1980; Watts & Stenner, 2012). The sort is a forced choice, and that matters for this convention, because free and frank advice is an ordering of priorities, obligations, risks and limits, in which candour is weighed against relationship and responsiveness is distinguished from accommodation. In my view the forced choice is what brings the judgement inside the convention into view.

The analysis is technical, although its purpose throughout is interpretive: the sorts are correlated, the correlation matrix is factor analysed, the factors are rotated and the strongest are retained, and each retained factor is expressed as an idealised sort, its factor array. A factor is then interpreted from the whole configuration, reading the full array, the highest and lowest placed statements, the distinguishing and consensus statements, the factor loadings and the post-sort reflections together, and never from the statements at the ends of the distribution or from the distinguishing statements alone. That configurational discipline keeps my judgement in play while making it answerable to the arrangement participants themselves produced.

Three stages and the decisions in the analysis

The design ran in three stages, each asking something different of participants, and participation in each stage was defined by completed work within it. Stage One drew 32 interview participants, current and former politicians, senior officials, practitioners from the wider state and academics with sustained engagement in practice, whose agreed interview summaries were carried into the analysis; the interviews were open and semi-structured, and I analysed them abductively, moving back and forth between practitioners’ accounts and the vocabulary of beliefs, traditions and dilemmas (Blaikie & Priest, 2019; Braun & Clarke, 2006). Their output was the concourse, the wider field of statements and contestable propositions about what free and frank advice is, why it matters and what turns it on or off, and from that concourse I built a Q-set of 34 statements, which a number of Stage One practitioners reviewed for fidelity to their own language before it was used. Stage Two drew 58 participants, most of them new to the study, who sorted the 34 statements across an eleven-point distribution from minus five to plus five, with columns of 1, 2, 3, 4, 4, 6, 4, 4, 3, 2 and 1 statements, and who each reflected on their sort afterwards. Stage Three drew 88 participants into validation through two wānanga, facilitated workshops and fifteen one-on-one expert interviews, and five people were carried through all three stages as a continuity check. I see the three stages as checking one another, with the interviews keeping the work close to practice, the sorting disciplining my interpretation, and the validation testing whether the result had become too neat.

Turning the sorts into findings required five decisions, which I made by reading statistical output and interpretive clarity together. A chi-square test first confirmed that participants had used the Q-set in a differentiated way. I then chose Spearman rank-order correlation over Pearson, because the completed sorts are ranked and ordinal. Of the two extraction methods I examined, centroid factor analysis and principal component analysis, the second gave a clearer and more legible structure for this study, and Varimax rotation suited a search for distinguishable postures. The number of factors to retain was treated as a matter of judgement disciplined by statistical criteria, weighing eigenvalues, explained variance, defining sorts and distinctiveness against the intelligibility of each array and the reasoning in the post-sort reflections (Brown, 1980), and that weighing produced the six-factor solution I report later in this series. My own view is that retention in Q-methodology is properly a judgement, and that the honest course is to make each step of that judgement visible.

Two further steps kept the factor solution answerable to something beyond its own statistics, and in the first I read the consensus statements, those placed in similar ways by participants who otherwise sorted very differently, with particular care, because they mark what is shared across the postures and later supported the interpretation of practical statecraft as the shared boundary of the convention. In the second, once the factors were interpreted, I tested whether they could be explained by structural or demographic categories such as institutional location, seniority, tenure, elected or appointed status, gender and ethnicity, since a decentred account expects meaning to cut across formal position, and the claim would have needed qualification if the postures had simply followed role. The narrowing all this required is real, because an open field of practitioner talk became 58 completed sorts of 34 statements, and what was left out of the Q-set never reached the sorting. In my view that loss is the price of being able to test whether practitioner meanings are shared, which is the question the convention itself poses.

In the next post I turn to the research relationships behind the method: Āta as methodological practice, my position as an insider-Indigenous researcher, and the ethics of confidentiality in a small senior public service.

References

Bevir, M., & Rhodes, R. A. W. (2006). Decentred theory, change and network governance. In E. Sørensen & J. Torfing (Eds.), Theories of democratic network governance (pp. 1–20). Palgrave Macmillan.

Blaikie, N., & Priest, J. (2019). Designing social research: The logic of anticipation (3rd ed.). Polity Press.

Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101.

Brown, S. R. (1980). Political subjectivity: Applications of Q methodology in political science. Yale University Press.

McLean, J. (2024). Conventional norms and the comparative constitutional law project. Comparative Constitutional Studies, 2(2), 272–291.

Rhodes, R. A. W. (2017). Network governance and the differentiated polity: Selected essays (Vol. 1). Oxford University Press.

Watts, S., & Stenner, P. (2012). Doing Q methodological research: Theory, method and interpretation. SAGE Publications.