Classical Ethnography: long-term fieldwork as method

Presence over time. Long enough for the setting to resume normal life around the researcher, for events to recur, and for early interpretations to be broken by later ones. In classical anthropological practice this meant a year or more; organisational ethnographies compress it, but the logic — repeated exposure across cycles — is non-negotiable.

Participation as instrument. The researcher’s own learning curve is data: what one must learn to pass in a setting is a map of the setting’s tacit rules.

Thick description. Following Geertz’s classic formulation in The Interpretation of Cultures (Basic Books), behaviour is recorded with the meaning structures that make it intelligible — a wink is not an eyelid contraction. The written product interprets, not just inventories.

Holism with a focus. Ethnography attends to how parts of a setting connect — work, talk, space, hierarchy — while modern studies define a focused problem within that holism rather than promising a total portrait.

The standard methodological text is Hammersley and Atkinson’s Ethnography: Principles in Practice (Routledge) — the fullest treatment of access, field relations, note-taking, and the ethics of presence.

For a compact, applied orientation, Reeves, Kuper and Hodges’ Qualitative research methodologies: ethnography in the BMJ series is the reference most health and organisational researchers start from.

On the observer’s stance, Gold’s classic typology — complete participant, participant-as-observer, observer-as-participant, complete observer — remains the working vocabulary for describing and defending where you stood, and every methods section in this tradition should be able to place itself on it.

• When the say–do gap is the question. Compliance cultures, safety practices, service delivery “as implemented” — anywhere official accounts are known to diverge from practice.
• When the knowledge is tacit. Skills and norms people cannot articulate because they no longer notice them are recoverable only by watching and learning them.
• When the setting is unmapped. Before designing surveys or interventions for a world nobody has described, describe the world.
• Not when time is short and questions are narrow. A three-week window with a fixed question is a case study or rapid ethnography, and calling it classical ethnography over-claims.
• Not when presence would distort or endanger. Some settings cannot absorb an observer ethically; documents and interviews must serve instead.

An ethnographer spent nine months, two to three days a week, in a hospital pharmacy studying medication-safety practice. The first weeks of wide-angle notes recorded an official world of barcode scans and double checks — and a mounting list of small anomalies: trolleys parked at odd angles, labels folded a particular way, a drawer everyone used and nobody mentioned.

Focused observation followed the anomalies. The folded labels marked doses a senior technician had already eyeballed — an unofficial verification layer; the drawer held pre-assembled “ward emergency” packs that violated policy and had prevented, by the staff’s accounting, at least two crisis delays that quarter. Ethnographic interviews in month six, armed with these specifics, produced candid explanations no early interview had offered: the workarounds were a maintained, taught system, invisible to audit because audits sampled documents, not drawers.

The final account described a two-layer safety culture — the audited system and the artisanal one — and located risk precisely where the layers met (new staff taught the official system only). The hospital revised its induction rather than punishing the drawer, which is what an account of meanings, rather than a list of violations, made possible.

• Time and exposure declared. How long, how often, at what hours — a study of weekday day-shifts should say so and bound its claims accordingly.
• Concrete detail carrying the argument. Claims about culture cashed out in observed incidents, not adjectives.
• The say–do comparison used. Interviews and observation confronting each other, with divergences analysed rather than smoothed.
• Reflexivity with consequences. Not a positionality paragraph, but an account of what the researcher’s presence and standpoint did to the data.
• Negative evidence reported. The events that did not fit the emerging pattern, and what was done about them.

Ethnography trades breadth for depth on terms no other method accepts: one setting, one period, one researcher’s eyes. Its findings transfer by mechanism and analogy, not by statistical generalisation, and the write-up must claim accordingly.

The instrument is a person, so the data bear the person’s limits — where they could stand, who trusted them, what they were equipped to notice. Two ethnographers in one site will not write the same book; the discipline’s answer is transparency and evidence, not interchangeability.

And the costs are real: months of access, ethical weight in studying people who came to trust you, and analytic drowning in hundreds of pages of notes. The method is unmatched for what it does and unsuitable for most questions — the misuse is almost always compression into timeframes where only its vocabulary survives.

Participant observation is ethnography’s core technique, usable in bounded doses inside other designs.

Netnography adapts the ethnographic stance to online communities, with its own access and ethics problems.

Situational analysis offers a mapping-based analytic apparatus that pairs well with ethnographic data.

Discourse analysis shares the interest in situated meaning but works on recorded language rather than lived presence.

Nine months of field notes, interview transcripts, and photographed documents is a corpus problem wearing a notebook’s disguise. Coding and retrieving across it — every incident touching “verification”, every note mentioning the drawer, say–do comparisons between interview claims and observed events — is where Evidano’s quote-linked coding and cross-document search repay the field time, and interview audio transcribes straight into the same workspace.

What no tool provides is the presence: the access, the noticing, and the learned competence that generate the notes in the first place. Software makes the corpus answerable; the fieldwork makes it worth asking.