The Gold Standard of Anthropology | Visual Elicitation, Image Ownership and Unequal Legibility

1. An image is a prompt, an artefact and a relationship

Back to contents

In photo elicitation, a photograph may prompt a participant to remember a route, explain a household routine or question the researcher’s assumptions. The spoken account is not a caption waiting inside the picture. It is produced in a particular conversation, with a particular person, at a particular time.

The same image is also an artefact with a maker, a technical history and a route of circulation. It may have been taken by a participant, selected from a family archive, supplied by the research team or found online. Those origins shape what may be inferred and what permissions are required.

Begin the analysis by recording four layers separately: what is visibly present; who made or selected the image; what was said around it; and what the research team later interpreted. This prevents the visual surface from swallowing the social conditions that made it meaningful.

2. Choose who controls the camera and why

Back to contents

Researcher-produced images can standardise prompts across interviews, but they also embed the researcher’s framing before the conversation begins. Participant-produced images can shift attention toward what participants consider important, yet access to devices, time, mobility and confidence may still distribute authorship unevenly.

A participant asked to photograph “a safe place” may choose a doorway, a person or an empty street. The value lies not in deciding which picture most accurately depicts safety, but in asking how the selection connects place, memory and obligation.

State the production rule before fieldwork: who may take images, what may not be photographed, whether existing images can be used, how equipment is provided and what happens when someone prefers words, drawing, mapping or no visual task at all.

3. Consent has several separate objects

Back to contents

Consent to join an interview does not automatically include consent to be photographed, to have someone else’s photograph discussed, to archive the file, to publish it online or to reuse it in teaching. Each action changes the audience and the possible consequences.

Use a permission matrix rather than a single yes-or-no box. A participant might allow analysis inside the research team, permit a cropped image in a closed classroom and refuse public web publication. Another may allow quotation from the interview but not reproduction of the image.

Revisit consent when the proposed use changes. Digital publication makes copying and recirculation difficult to control, so “public” should be explained with concrete examples rather than treated as an obvious category.

4. Ownership, licence, attribution and ethical permission differ

Back to contents

The person who presses the shutter may hold legal rights under the relevant jurisdiction, but the image can also contain people, artworks, private spaces or culturally restricted knowledge. Legal permission from a photographer is therefore not the same as ethical permission from everyone affected.

A research agreement should identify the maker, file custodian, permitted uses, attribution preference, duration of retention, withdrawal limits and any community authority governing circulation. Do not promise deletion from the wider internet after publication if that cannot be delivered.

When law, institutional policy and community expectations differ, follow the most protective workable route and document the decision. If rights remain unclear, describe the image without reproducing it.

5. Unequal legibility changes what an image can expose

Back to contents

An image that looks anonymous to an outsider may be immediately recognisable to neighbours through a doorway, uniform, landscape, tattoo or sequence of objects. Blurring a face does not remove location, relationships or social identity.

Legibility is unequal because viewers possess different contextual knowledge. A family member may recognise a room that a journal editor cannot. A platform may extract metadata or use automated recognition even when a casual reader sees little.

Conduct a layered disclosure review with people who understand the setting. Check faces, bodies, text, reflections, metadata, landmarks, filenames and the narrative beside the image. Record who considered the risk and which audience their judgement represents.

6. Worked example: one market photograph, four decisions

Back to contents

A participant photographs a market stall to explain how informal credit works. The image includes the participant’s handwritten ledger, a neighbouring seller, a school uniform and a distinctive sign. The photographer approves publication because the stall matters to the story.

The research team separates the decisions. The ledger contains financial information; the neighbour has not consented; the uniform could identify a child; and the sign fixes the location. They retain the original in restricted storage, create a fieldnote describing the image and return to discuss alternatives.

The final article uses a participant-approved drawing of the credit relationship and a short quotation about trust. The photograph still generated evidence, but publication was not required for the image to be analytically valuable. The audit trail records why representation changed.

7. Interpret the conversation, not an imagined visual truth

Back to contents

Two participants can give different accounts of the same family photograph because the image activates different relationships and memories. Their disagreement is evidence about position and history, not a defect to be solved by asking which account the picture proves.

Code visual form, elicited talk and contextual information in linked but distinct fields. Note pauses, pointing, sequencing and what the participant declines to explain. Avoid importing symbolic meanings from a generic visual dictionary.

Compare interpretations across cases only at the level the design supports. A repeated colour, object or composition may be meaningful, habitual or produced by the camera task itself. Test the pattern against interviews, observation and negative cases.

8. Changed-context transfer check

Back to contents

Move the market example to a digital ethnography in which participants select screenshots from a private messaging group. The layered logic transfers: a screenshot is a prompt, a record and an act of circulation. The risks change because usernames, timestamps, message order and platform metadata connect several people who did not volunteer for the study.

Permission from the participant who captures the screen may not authorise disclosure of other contributors. Redaction can also destroy the conversational sequence needed for interpretation.

A safer design may use participant-created reconstructions, paraphrased message maps or analysis without reproduction. Transfer the questions about control, audience and legibility; do not assume that “already digital” means already public.

9. Limits and common misconceptions

Back to contents

Visual methods do not automatically give participants more voice. A researcher still defines the task, selects what enters the article and controls many routes of dissemination. Providing cameras can widen participation or create new burdens and exclusions.

An image is not more objective than speech, and a vivid photograph is not stronger evidence merely because readers find it persuasive. Consent forms cannot predict every future context, anonymity cannot be guaranteed by face blurring alone, and copyright ownership does not settle relational harm.

Do not treat refusal to photograph as missing data. Do not publish a vulnerable person’s image to demonstrate authenticity. Do not use stock images as if they document the field site. State what the visual record can and cannot support.

10. Audit the full image journey

Back to contents

Keep an image register that records origin, maker, date, location precision, people or restricted material depicted, consent scope, transformations, storage, copies, planned audiences and final disposition. Link it to the interview and analytic memo without putting sensitive identifiers into public filenames.

Before dissemination, compare the proposed use with the permission matrix and run a contextual recognition check. Confirm captions, credits and alt text do not reveal what the image treatment tried to conceal.

After publication, retain a contact route for concerns and a realistic response policy. Removing a file from the project site may reduce exposure even when copies elsewhere cannot be recovered. Accountability is an ongoing practice, not a completed form.

11. Frequently asked questions

Back to contents

Must every person visible in a photograph sign the research consent form?

Requirements depend on context, jurisdiction and use. The research team should assess identifiability, expectation of privacy, risk and publication scope, and obtain appropriate consent or avoid reproduction when permission is not defensible.

Can an image be analysed but not published?

Yes. Analysis, storage, archiving, teaching and public reproduction are separate uses and can carry different permissions.

Does participant authorship remove the researcher’s responsibility?

No. Participant control can improve the method, but researchers and publishers still make decisions about risk, context, representation and circulation.

12. The gold standard keeps meaning and circulation accountable

Back to contents

A strong visual study makes the image useful without pretending it is self-explanatory. It shows who created the prompt, how interpretation emerged and which audiences were authorised to see the result.

The practical test has five questions: who made or selected the image; who and what can be recognised; what permission covers this exact use; whose interpretation is being reported; and what happens if circulation causes harm?

If those questions are answered in the methods, analysis and archive, visual elicitation becomes a transparent way to learn. If they are hidden, the most compelling image may also be the least defensible evidence.