Ethnographic Interview Transcription: Field Notes to Text

You spent six months in a fishing village on the Senegalese coast. Your field recorder holds 47 hours of conversations, plus another 23 hours of formal interviews. Half the audio is in Wolof, a quarter in French, and the rest is a mix of both. Back at your desk, the project of turning that audio into a dissertation feels overwhelming. This guide walks through how working ethnographers actually handle the transcription stage of their fieldwork without losing the contextual richness that makes ethnographic research distinctive.

What Makes Ethnographic Transcription Different

Ethnographic interviews are not the same as structured research interviews. They happen during walks, over meals, in workshops, sometimes mid-task. The audio reflects that.

Three differences shape the transcription workflow.

Setting matters. A 20-minute conversation that took place while a participant was cleaning fish at the dock cannot be analyzed without that context. The transcript needs to capture or reference the setting.

Speech is informal. Participants do not give clean monologues. They interrupt themselves, switch languages, gesture in ways the recorder cannot capture. The transcript inevitably loses information that field notes need to restore.

Relationships shape the conversation. Ethnographers build relationships over months. The data is partly the conversation, partly what the conversation reveals about the relationship. The transcript captures only the words.

A defensible ethnographic transcription workflow accepts these limitations and supplements transcripts with field notes that restore context.

The Field-to-Transcript Pipeline

The seven-step workflow below covers a typical 6 to 18 month fieldwork project with 30 to 80 hours of audio.

Step 1: Record With Field Notes in Parallel

Every recording should have an accompanying brief field note. Date, location, participants, setting, what was happening before the recording started, what triggered the conversation. Two or three sentences.

These notes are what let you reconstruct context months later when you transcribe. Without them, a recording from June feels like a stranger's conversation by November.

Step 2: Back Up Audio Daily

Field environments are hostile to electronics. Heat, humidity, dust, drops, theft. Audio that exists only on one device will eventually be lost. The standard practice is daily backup to a second device, plus weekly backup to encrypted cloud storage when connectivity allows.

For projects in regions with intermittent internet, encrypted external drives carried in a separate bag from your recorder are essential.

Step 3: Triage Recordings

Not every recording needs full transcription. A 90-minute conversation that turned out to be about football scores deserves a short summary in your field notes, not a full transcript.

The triage decision matters because transcription, even with AI, costs review time. Spending review time on 60 hours of audio when only 20 hours actually feed the analysis is wasted effort.

A practical triage rubric:

• Full transcription: Substantive content addressing your research questions.
• Summary transcription: Important context but secondary to your questions.
• Note-only: Background, social, or off-topic content.

Mark these categories in your file system so the transcription stage is targeted.

Step 4: Transcribe in Original Language

Ethnographic data often involves multiple languages, code-switching, and dialects that academic translators may not handle well. The defensible approach is to transcribe in the original language first.

For audio in Wolof, Swahili, or Hausa, the AI pipelines have improved dramatically. Accuracy varies by speaker, recording quality, and the specific variety of the language. Always do a careful review.

For French, Portuguese, or Spanish fieldwork, AI handles regional accents well. The Arabic transcription pipeline handles Modern Standard Arabic and some major dialects, but performance on minority dialects is uneven.

Step 5: Annotate, Do Not Just Transcribe

After the AI produces a transcript, your review should add context, not just fix errors.

Useful annotations:

• Code-switching markers: [switches to French] or [switches to Wolof]
• Setting notes: [interrupted by colleague entering]
• Non-verbal: [laughs], [gestures toward the boat]
• Untranslated terms: nguël gi (the older brother) with a brief gloss

These annotations are what let your transcript serve as analytic data rather than just text. They also become the basis for the thick description that distinguishes strong ethnographic writing from weak ethnographic writing.

Step 6: Connect Transcripts to Field Notes

Each transcript should reference its accompanying field notes. The simplest method is consistent file naming. 2025-09-14_dock-conversation_M.Sow.txt for the transcript, 2025-09-14_field-note.md for the field note.

In your analysis software, create a project structure that lets you view both side by side. Most QDA software supports this through linked documents.

Step 7: Selective Translation

For ethnographic data, translation happens during the writing stage, not the transcription stage. You translate the specific quotes you intend to use in your write-up, not the full transcript.

This approach has two benefits. First, you preserve the original language for your own analytic memory. Second, you make translation choices visible. Footnotes can explain difficult translations, which strengthens the analytic claim.

The research interview template at CATT helps you identify candidate quotes during the review stage. You then make targeted translation decisions for the ones that matter most.

Detail Conventions for Ethnographic Transcripts

Different ethnographic traditions favor different transcription conventions.

Sociolinguistic ethnography often uses Jefferson notation or GAT, capturing turn-taking, overlap, and pause length with detailed symbols. This level of detail matters when interaction patterns are the analytic focus.

Cultural anthropology typically uses cleaner transcripts with annotations for context and code-switching. The conversation is data, but the analytic focus is usually content rather than interactional patterns.

Critical ethnography sits between the two. Detailed enough to support discourse analysis when needed, but readable enough to feed thematic and narrative analysis.

Pick a convention before you start transcribing and apply it consistently. Mixed conventions across a project make analysis harder, not richer.

Handling Sensitive Field Data

Ethnographic research often involves sensitive material. Marginalized communities, illicit economies, politically risky environments. The data layer of transcription needs to match the ethics of the research.

Three practices matter.

Pseudonyms in transcripts. Replace real names with pseudonyms during transcript review, before sharing transcripts with anyone outside your immediate team. Maintain a separate, encrypted key file that maps pseudonyms to real names.

Local-only transcription for high-risk material. When the risk profile is high enough, run transcription on a local machine rather than a cloud service. Open-source Whisper running on your laptop is the standard tool for this.

Storage hygiene. Audio files of sensitive material should not sit indefinitely in cloud services. Download, transcribe, delete from cloud, retain locally with encryption.

The ethics of interview transcription guide covers the broader ethical layer that applies to all interview research, with specific attention to anonymous and sensitive sources.

Time and Cost Budget

A typical ethnographic dissertation includes 30 to 80 hours of audio. Not all of it gets transcribed in full. After triage, expect 15 to 40 hours of full transcription work.

AI transcription time: 30 to 90 minutes processing.

Review and annotation time: 30 to 60 minutes per hour of audio. For 25 hours of triaged audio, 12 to 25 hours of review.

Targeted translation: 1 to 3 hours per dissertation chapter for the quote-by-quote translation work.

Total transcription effort: 15 to 30 hours of focused work for a dissertation-length project.

At a monthly cost of $9.99 through the CATT unlimited plan, the financial cost of AI transcription is trivial compared to the time savings against manual transcription, which would consume 100 to 200 hours of typing for the same audio total.

Working Across Languages in NVivo or MAXQDA

Multi-language ethnographic projects break some assumptions in standard QDA software. NVivo and MAXQDA can both import non-English transcripts, but coding and theme-building across languages is still rough.

The pragmatic approach is to maintain two parallel coding systems. Code in the original language for analytic depth, code translated quotes in English for committee-facing write-ups. The NVivo vs AI transcription comparison covers the broader workflow for combining AI tools with QDA software.

Common Mistakes That Show Up in Ethnographic Writing

Three patterns weaken otherwise strong ethnographic work.

Decontextualized quotes. A vivid quote without setting and relationship context becomes generic. The setting note and the field note relationship are what make the quote ethnographically rich.

Over-translated transcripts. Translating away the participant's voice strips out exactly what ethnographic research is supposed to surface. Keep some original-language phrases visible in the final write-up.

Erased researcher presence. The ethnographer's question shapes the participant's answer. Transcripts that hide the researcher's prompts produce analyses that hide their own positionality. Include the researcher's turns in the transcript.

The discipline of ethnographic transcription is in keeping the context attached to the words. Strip the context and the words mean less than they did in the field. Keep the context and the analytic possibilities open up. The tooling is the easy part. The intellectual discipline is the work.