How do I show rigour in qualitative research?

Quantitative rigour is about whether the measurement was accurate and the inference valid. Qualitative rigour is a different question, because the researcher is the instrument and the analysis is interpretive. Borrowing the quantitative vocabulary produces claims that sound rigorous and do not fit, which examiners notice.

The criteria that do fit are about trustworthiness: could someone else follow your reasoning, are your interpretations grounded in the data, would the account be recognizable to participants, and is the context described well enough for a reader to judge transferability. Each has a corresponding practice.

Those criteria have names you can use in a methods chapter. Korstjens and Moser set out four for novice researchers, a group they define to include Master's students: credibility, transferability, dependability, and confirmability. Credibility asks whether your findings plausibly represent what participants said. Transferability rests on thick description of both experience and context, so a reader can judge relevance to another setting. Dependability is about consistency of the research process, and confirmability is about showing that your interpretations came from the data rather than from your preferences (Korstjens and Moser, 2018, p. 120). The same guidance says internal validity, generalizability, reliability, and objectivity are unsuitable as direct criteria here, which is why the borrowed vocabulary reads badly (Korstjens and Moser, 2018, p. 121).

The audit trail is the single most useful of those practices and the most neglected. Keep a dated record of analytical decisions: why you merged two codes, why a theme was dropped, what changed after the third read. It costs a few minutes a session and it is the evidence that your themes were developed rather than asserted.

Korstjens and Moser list what belongs in that record: decisions and their rationale, team meetings, reflective thoughts, sampling, research materials, the emergence of findings, and data management (Korstjens and Moser, 2018, p. 122). If you work from a coding template, Coates and colleagues add a procedure for the coding stage itself. Every coder rereads every transcript and applies the finalized template, preferably on a fresh copy. Any response that attracts more than one code gets flagged for discussion. An independent researcher who took no part in developing the scheme can apply the template as an additional coder, which those authors treat as a reliability gain (Coates et al., 2021, p. 3).

Disconfirming case analysis is the second. Go looking for the data that does not fit your emerging account and report what you found. A study that names its exceptions and explains them is far more convincing than one where everything agrees.

Sample size is the claim reviewers challenge most often. Vasileiou and colleagues analyzed interview-based qualitative health studies across medicine, psychology, and sociology. Among the studies that justified their sample at all, saturation accounted for 55 percent of justifications and pragmatic constraints such as resources, response rates, and participant availability for about 10 percent. Claims of saturation were never substantiated by the procedures those studies actually reported (Vasileiou et al., 2018, p. 15). Their recommendation is not to run more interviews. It is a study-specific judgment that ties your sample size to your design, your sampling strategy, the richness of the data, and what your analysis requires (Vasileiou et al., 2018, p. 16).

Reporting frameworks are a useful check and a poor substitute for that argument. SRQR runs to 21 items covering title, abstract, introduction, methods, findings, discussion, conflicts of interest, and funding. COREQ runs to 32 items for interview and focus-group studies, covering the research team and reflexivity, study design, analysis, and reporting (Korstjens and Moser, 2018, p. 124). Use one to test whether a reader could reconstruct what you did. A completed checklist will not stop a reviewer applying the wrong standard, and Vasileiou and colleagues found that qualitative samples were still being assessed against an implicit quasi-quantitative one (Vasileiou et al., 2018, p. 16).

Do I need inter-rater reliability?

Only if your design treats coding as measurement. In interpretive traditions, agreement between coders is not the goal and a high figure can indicate a shallow coding frame. What is expected instead is documented discussion between coders about disagreements, and a record of how they were resolved.

Content analysis and other traditions that count coded instances do need reliability statistics, and there the expectation is a specific coefficient reported with its calculation.

If a supervisor or reviewer asks for a reliability figure in an interpretive study, it is worth asking what they want it to demonstrate. Usually the answer is that the analysis was not done by one person alone in a room, and there are better ways to show that.

What does reflexivity look like in practice?

A record, not a paragraph of good intentions. Note your prior expectations before analysis, note where your position shaped access or what participants said, and note the moments you changed your mind. Then report the ones that affected the findings, with the effect stated.

The reflexivity sections that fail are the ones that describe the researcher's identity and never connect it to the analysis. Stating your background is a starting point. Saying what it made easier, what it made harder, and what you did about it is the actual work.

Keep the record while analyzing rather than writing it at the end. Reconstructed reflexivity is a description of how you would like to have thought, and it reads that way.

Member checking is worth considering and is not obligatory. Returning your interpretation to participants can strengthen an account, and in some traditions it makes little sense, since your analytical reading is not one a participant is positioned to confirm. Decide deliberately and say why.