Having collected information about society, having analysed it through open coding and constant comparison, and having distilled it into grounded theory, you wish to share your understanding of society with an audience. Writing is central to all social research, but it has a special place in naturalistic inquiry.

A conventional view of writing in social research is that it is simply the reporting of the findings of research, of writing up once the other phases have been completed. This view, connected to positivism, advocates keeping the language of reporting descriptive and factual in order to avoid ambiguities and using a scientific, detached vocabulary to ensure objective reporting. Ideally, the language of mathematics is used, but when this proves to be impractical, it is recommended to use a conceptual apparatus that has been tried in previous studies and that has been connected at the outset with the data in a procedure known as ‘operationalization’.

That is not the approach to writing in naturalistic inquiry. Writing therein is seen as flowing from analysis, as the final stage of the arc of naturalistic inquiry, and as a capstone put on the entire research process. It encompasses coming to terms with the complex relations between description, interpretation, and explanation in a coherent master narrative.

Naturalistic inquiry does subscribe to the ideal of reporting though, but that is not the same as sharing raw data with our readers. Writing entails a process of careful selection of the most meaningful (emically) and most significant (etically) parts of those data and organizing them in a single, coherent theoretical framework. This is not a subjectivist procedure, nor an arbitrary one. The selection is based on a thorough familiarity with the society under study – a familiarity built up over months or even years of fieldwork, and it is guided by the careful and systematic analysis of all the materials. Eventually and necessarily, those events and situations are selected that are the most telling. Still, and ideally, they also represent the rest of the data.