The four criteria, the techniques that actually evidence them, and why member checking is not the safe bet it looks like.
Trustworthiness is how qualitative research demonstrates that its findings can be relied on, without borrowing criteria built for measurement. The framework most theses are examined against is Lincoln and Guba’s: credibility, transferability, dependability and confirmability. Each is demonstrated by something you actually did and can evidence — which is why a trustworthiness section that asserts the four terms without showing the work convinces nobody.
Reliability asks whether an instrument gives the same result on repetition. Applied to qualitative research, that expectation does not hold: a second researcher, with a different position and different rapport, would produce a different account, and in interpretive work that is not a defect to be engineered away.
Lincoln and Guba proposed parallel criteria addressing the same underlying concern — can these findings be trusted — on terms appropriate to the work. Credibility parallels internal validity, transferability parallels external validity, dependability parallels reliability, and confirmability parallels objectivity.
It is worth knowing that this framework is not universally accepted. Some argue that constructing criteria parallel to quantitative ones concedes the wrong ground, and alternative frameworks exist — Tracy’s criteria, among others, are used in some fields. Most doctoral examiners will expect the Lincoln and Guba four, so using them is the safe default. Showing you know they are one framework among several, rather than a natural law, is what distinguishes a candidate who has read the methods literature from one who has read a summary of it.
The pattern to follow throughout: name the criterion, name what you did, and point to where the evidence sits.
Do the findings genuinely represent what participants meant? This is the one examiners press hardest.
What evidences it: prolonged engagement with the setting, so you are not reading a single encounter as the whole picture. Triangulation across sources, methods or investigators, where accounts from different directions converge. Peer debriefing — someone outside the study interrogating your developing interpretation. And, most persuasively, negative case analysis: a deliberate search for material that contradicts your developing account, with a report of what you found and what you did about it.
Negative case analysis is the cheapest credibility available and the most often skipped. An analysis where everything agreed invites the suspicion that disagreement was not looked for.
Could a reader judge whether these findings apply in their own setting? Note the framing: transferability is the reader’s judgement, not your claim. Your job is to make it possible.
What evidences it: thick description — enough detail about the setting, participants, context and conditions that a reader can assess the fit. That means describing the organisation, the timing, the local circumstances and the participants’ characteristics in enough depth to be recognisable.
Stating that findings are transferable is the error here. Providing the detail from which someone else can decide is the correct move, and saying so explicitly reads as methodological confidence.
Could the process be followed by someone else? Not repeated for identical results — followed, and understood as coherent.
What evidences it: an audit trail. Dated records of decisions, codebook versions, memos explaining why a code was merged or a theme abandoned, notes on changes to the interview guide and why they were made. Some studies add an external audit, where someone outside examines that trail.
The audit trail has to be built as you go. It cannot be reconstructed at the end, and examiners can tell the difference between a record kept contemporaneously and one assembled afterwards.
Are the interpretations traceable to the data rather than to your preferences?
What evidences it: coded extracts shown in the thesis or appendices so a reader can see the basis for a theme. Memos showing how interpretations developed. And reflexivity — an explicit account of who you were in relation to participants and how that shaped what you were told and what you noticed.
Reflexivity is not a confession of bias. It is a description of position: your professional background, your relationship to the setting, what participants were likely to assume about you, and how that plausibly shaped the material. A short specific account is worth more than a page of general acknowledgement.
Returning findings to participants is widely treated as an automatic strengthener. It is more complicated than that, and a candidate who knows why is in a stronger position than one who simply reports doing it.
The difficulty is that participants are not necessarily well placed to validate an analytic account. Your interpretation may operate at a level they were not working at, may draw on comparison across participants they have no access to, or may say something about the setting that is accurate and unwelcome. Disagreement then does not straightforwardly mean the analysis is wrong.
There is also the practical problem: months later, participants may not recall the interview clearly, and may respond to how they wish to be seen now rather than to the accuracy of the account.
Member checking works better for some purposes than others. Returning a transcript for factual correction is uncontroversial and useful. Returning a summary of an individual’s own account and asking whether it represents what they meant is reasonable. Asking participants to ratify a cross-case theoretical interpretation is where it becomes questionable.
If you do it, say what you returned, to whom, what you asked, what came back, and what you changed as a result. “Findings were member checked” with no detail is a sentence that invites a question rather than closing one.
Most trustworthiness sections fail the same way: four paragraphs, each defining a criterion and asserting it was met. Definitions belong in a textbook, not in your thesis, and assertion is not evidence.
A section that works does three things for each criterion:
The reliable test: could a reader, from this section alone, tell what you actually did? If it would read the same for any qualitative study in your field, it is describing a framework rather than your research.
It is how qualitative research demonstrates that its findings can be relied on, using criteria suited to interpretive work rather than borrowing those built for measurement. The framework most doctoral theses are examined against is Lincoln and Guba’s four criteria: credibility, transferability, dependability and confirmability. Each is demonstrated through specific practices rather than asserted.
Credibility — whether the findings genuinely represent participants’ meanings, shown through prolonged engagement, triangulation, peer debriefing and negative case analysis. Transferability — whether a reader could judge if the findings apply in their setting, made possible by thick description. Dependability — whether the process could be followed, shown through an audit trail. Confirmability — whether interpretations are traceable to data rather than to the researcher, shown through coded extracts, memos and reflexivity.
Not automatically. Participants are well placed to correct factual detail in their own transcript, and reasonably placed to say whether a summary of their own account represents what they meant. They are less well placed to ratify a cross-case theoretical interpretation, which may operate at a level they were not working at or draw on comparisons they cannot see — so disagreement does not straightforwardly mean the analysis is wrong. If you use it, report exactly what you returned, what you asked and what you changed.
A dated record of the decisions you made and why: codebook versions, memos on merging or abandoning codes, changes to the interview guide, and the reasoning behind analytic choices. It is the main evidence for dependability, and it has to be built as you work — reconstructing one at the end is close to impossible and examiners can usually tell.
For qualitative work, trustworthiness, and it is worth saying briefly why: reliability assumes repetition should produce the same result, which does not hold where the researcher is the instrument. Using quantitative vocabulary for qualitative work is a small error that signals a larger confusion. In a mixed-methods thesis you need both, applied to their own strands, and applying either set to the wrong strand is a common and visible mistake.
For each criterion, say specifically what you did, point to where the evidence sits in the thesis or appendices, and be honest about what you could not do. Avoid textbook definitions — the examiner knows them. The test is whether a reader could tell from the section what you actually did: if it would read identically for any qualitative study in your field, it is describing a framework rather than your research.
Trustworthiness sections are quick to check and commonly the weakest part of an otherwise strong chapter. Send what you have written and a PhD in your field will tell you where it asserts rather than evidences.
Leave an email or a WhatsApp number — whichever you prefer — and tick how we should reach you. We reach out within 30 minutes.
