Narrative Analysis: reading the story, not just the content

The claim underneath narrative analysis is that people do not merely report experience, they organise it — and the organising is itself informative. Where a person chooses to begin a story about becoming ill, what they treat as the turning point, whether they cast themselves as agent or as something acted upon, and whether the story is given an ending at all are choices that carry meaning the content alone does not.

Two people can report the same events and tell entirely different stories about them. Thematic coding will find them similar; narrative analysis will find the difference, and the difference is usually what the research question was actually about.

The corollary is that narrative analysis works on whole accounts, and interviews for it are conducted differently: fewer questions, longer turns, and a deliberate willingness to let someone tell it their way rather than steering them through a topic guide.

The structural lineage begins with William Labov and Joshua Waletzky's analysis of oral narratives of personal experience, reprinted with commentary as Narrative Analysis: Oral Versions of Personal Experience. They identified a recurring six-part structure — abstract, orientation, complicating action, evaluation, resolution, coda — and argued that the evaluation clauses, where the teller says why the story is worth telling, are the analytic centre.

The model has been debated ever since, and the debate is worth reading: Jerome Bruner's Labov and Waletzky Thirty Years On is the most incisive of the retrospectives, and Struggling Beyond Labov and Waletzky sets out the case that the model fits Western, canonical, single-event stories better than anything else.

The interpretive lineage runs through Catherine Kohler Riessman, whose Narrative Analysis distinguishes the analytic approaches now standard in the field. In the adjacent tradition of narrative inquiry, Connelly and Clandinin's Stories of Experience and Narrative Inquiry treats narrative as both the phenomenon studied and the method of studying it.

In a study of people diagnosed with type 2 diabetes in the previous year, two participants described what a thematic coding would treat as the same content: symptoms, a GP appointment, a diagnosis, a change of diet, and mixed feelings.

Participant A began the story two years before the diagnosis, with a period of stress at work. The complicating action was the appointment; the evaluation clauses clustered around the moment the GP used the word "diabetic" — "and that was it, that was what I was now". The resolution was a management routine described in the passive: "the medication was started", "an appointment was made for me". A regressive narrative in which the teller has agency before diagnosis and very little after it.

Participant B began on the morning of the appointment. There was no orientation section to speak of. The complicating action was not the diagnosis but a subsequent argument with a nurse about a target; the evaluation clauses were all about that exchange — "I said, that is not going to work for me, and I meant it". The resolution was an agreement he had negotiated. A quest narrative in which the diagnosis is the setup and the real story is about being taken seriously.

Both would code as "diagnosis experience", "medication", "dietary change" and "relationship with clinicians". The analytic finding — that the same clinical event was made into a story of loss of agency in one case and a story of asserting it in the other — is only available if the accounts are kept whole. The study's recommendation followed from the second reading: the consultation where the target is negotiated matters more to how people narrate their condition than the consultation where the diagnosis is given.

• Coding to themes and calling it narrative analysis. The commonest failure by a wide margin. If the stories were fragmented, the method was thematic analysis.
• Sanitised transcripts. Removing hesitations and false starts removes the evidence for most structural claims.
• Forcing the Labovian model. It fits some stories and not others, and the misfits are informative.
• Extracts too short. A structural claim needs enough text for the reader to check it.
• Treating the story as a factual report. Narrative analysis is about how the account is constructed, not whether it is accurate — and confusing the two produces claims the method cannot support.
• Over-steering the interview. A topic guide with fifteen questions will not produce narratives.
• Ignoring the audience. Accounts are produced for someone. Who the interviewer was, and what the participant took the interview to be for, belongs in the analysis.

A strong narrative analysis shows: stories preserved as wholes; transcripts detailed enough to support structural claims; extracts long enough to check; explicit attention to form, positioning and agency rather than content alone; and reflexivity about the interview as an occasion that shaped the telling.

A useful test is to ask whether the findings could have been produced by thematic analysis. If they could, the method has not earned its cost.

Narrative analysis handles small numbers. Six to twelve accounts is typical, and comparing narrative shapes across sixty is not practical — so it cannot answer questions about distribution or prevalence.

It also requires participants who narrate. Some people answer questions rather than tell stories, some cultural settings organise accounts in ways the available models do not describe, and interviews conducted under time pressure rarely produce narrative at all.

And the reading is interpretive. Calling something the turning point of a story is a claim, not an observation, and the method's defence — long extracts and visible reasoning — is transparency rather than verification.

Narrative inquiry is the adjacent tradition that treats narrative as the phenomenon and the method together, usually over longer relationships with fewer participants.

Discourse analysis shares the attention to language but asks what the talk is doing rather than how the story is shaped.

Conversation analysis works at a finer grain still, on the sequential organisation of interaction.

IPA also stays close to individual accounts, but is after the meaning of an experience rather than the construction of a story.

Narrative analysis is the method most at risk from a tool that makes coding easy, because the easy thing is exactly the wrong thing. A platform that encourages fragmenting into themes will quietly turn a narrative study into a thematic one.

Used carefully it still helps: marking narrative boundaries within long transcripts, holding structural annotations against whole segments rather than short passages, and retrieving complete stories side by side for comparison. Evidano supports narrative analysis as a named methodology and works over whole accounts rather than fragments. Judgements about where a story begins, what its turning point is and what its form means are the analyst's, and they are the analysis.