The Gold Standard of Anthropology | Member Checking, Power Relations and Interpretive Authority

1. Member checking is not one procedure

Back to contents

One researcher sends a transcript and asks whether names and events are correct. Another shares a summary of themes and asks what feels missing. A third invites participants to analyse a case together. All may be called member checking, but they produce different evidence.

Begin by naming the purpose: factual correction, transcript clarification, resonance, generation of new data, or co-interpretation. Without that statement, a later claim that “participants validated the findings” hides what actually occurred.

2. Interpretive authority is shared unevenly

Back to contents

Participants know their experiences, meanings and circumstances. Researchers also make comparisons across interviews, records and theory that no single participant may see. Neither perspective automatically cancels the other.

The practical question is who can challenge which claim, under what conditions and with what consequence. A participant can correct a date, reject a label or offer a different explanation. The researcher must record that response and justify whether the interpretation is revised, bounded or retained.

3. Power shapes who responds and what is safe to say

Back to contents

An invitation is not equal access. Literacy, language, time, disability, digital access and fear of a gatekeeper can determine who returns comments. A group workshop may silence a junior member when a leader is present.

Plan private and accessible response routes, translation where needed, compensation consistent with the study, and a clear statement that disagreement will not affect services or relationships. Report non-response without treating silence as agreement.

4. Match the material to the purpose

Back to contents

For factual checking, a dated excerpt or event timeline may be enough. For interpretive checking, a short plain-language account of the theme, its scope and two contrasting examples is usually more useful than a long academic report.

Do not expose one participant’s identifiable material to another. Remove unnecessary detail, explain that the summary is provisional and say whether feedback will become new research data. If it will, consent and secure handling must cover that use.

5. Keep disagreement as data

Back to contents

A participant may say, “That theme is not how I see myself.” The response can reveal a category imposed by the research team, a change over time, a conflict between public and private accounts, or a legitimate difference in interpretation.

Log the original claim, feedback, context, analytic decision and reason. Options include revising wording, narrowing the claim, adding a counter-theme, distinguishing factual from interpretive disagreement, or retaining both readings.

6. Worked example: feedback changes the boundary

Back to contents

A study summarises twelve interviews as “volunteers stay because the organisation feels like family.” Eight participants endorse the summary, two do not reply, and two say family language describes pressure to accept unpaid overtime.

The result is not 8 votes for truth against 2. The team revises the claim: family language can create belonging and can also legitimise unequal obligations. It adds the two accounts as a boundary condition, reports the two non-responses and notes that managers did not attend the feedback sessions.

7. Separate correction, resonance and endorsement

Back to contents

Correction concerns an identifiable error: a date, role or sequence. Resonance asks whether an interpretation connects with experience. Endorsement asks for approval and is often the least useful standard because people may agree politely, disagree for principled reasons or change their view.

A study can be credible without unanimous endorsement. Credibility rests on a transparent analytic trail, engagement with alternatives and claims proportionate to the evidence.

8. Changed-context transfer check

Back to contents

Move the worked example from a volunteer organisation to interviews with teenagers about school discipline. The need to state purpose and preserve disagreement transfers. The process changes: consent and assent, safeguarding duties, adult authority and peer visibility affect what can be returned and how.

A group session led by a teacher could manufacture agreement. Private, age-appropriate summaries with an independent response route may be safer and more informative.

9. Limits and common misconceptions

Back to contents

Member checking does not prove that an interpretation is true, erase sampling bias or make every participant a co-author. Memory can shift, later identities can conflict with earlier accounts, and some analytic claims operate across cases rather than within one person’s experience.

Do not assume non-response means consent. Do not remove an uncomfortable finding merely because a powerful participant objects. Do not ask vulnerable participants to adjudicate material whose disclosure could expose others.

10. Audit the feedback trail

Back to contents

An audit records who was invited, who responded, in which language and mode, what material they received, what questions were asked, and what changed. It also records access barriers, withdrawals and disagreements that remained unresolved.

Compare the participant-facing material with the published claim. If the final article is broader or more certain than the version participants saw, explain why the inference expanded and what additional evidence supported it.

11. Frequently asked questions

Back to contents

Must every participant be contacted?

No. The sampling and invitation strategy should fit the purpose, while making exclusions and non-response visible.

Can a participant withdraw an interpretation?

They may withdraw their data according to consent terms, but disagreement with a cross-case interpretation is analytically distinct. Follow the agreed ethics process and report the dispute carefully.

Is a group feedback meeting enough?

Sometimes, but group hierarchy can suppress dissent. Offer routes that do not require public disagreement.

12. The gold standard is accountable interpretation

Back to contents

Good member checking creates a documented encounter between analysis and participant response. It does not outsource judgment or stage a ceremonial approval.

Before claiming that findings were checked, ask four questions: what was checked, by whom, under what power conditions, and what changed? If the article answers all four, readers can see both the value and the limits of the process.