Straussian Grounded Theory: open, axial, and selective coding

Open coding breaks data apart: line-by-line or incident-by-incident, naming concepts, and grouping them into provisional categories with properties and dimensions. A property is a characteristic of a category; a dimension is the range that property can take. “Disclosure” (category) has timing (property) running from immediate to never (dimension).

Axial coding puts data back together: for each major category, the analyst asks what conditions give rise to it, the context in which it occurs, the actions and interactional strategies by which it is handled, and their consequences. This is the coding paradigm, and it is applied category by category, with subcategories linked along those axes.

Selective coding integrates: one central category is chosen, the others are systematically related to it, the relationships are validated against the data, and categories needing further refinement are filled in. The product is a theory with a storyline that can be stated in a paragraph.

The phases are iterative rather than strictly sequential — analysts cycle back — but each has distinct questions, and knowing which phase you are in is what keeps the analysis moving instead of endlessly re-coding.

The procedural canon is Corbin and Strauss’s Grounded theory research: Procedures, canons, and evaluative criteria, which states the method’s rules — constant comparison, theoretical sampling, memoing — and, unusually for qualitative work of its era, explicit evaluation criteria.

The book-length treatment is Corbin and Strauss’s Basics of Qualitative Research (SAGE), now in its fourth edition, in which later editions notably soften the paradigm from requirement to heuristic — a shift worth knowing when a reviewer cites an edition at you.

For the debate about what the procedures do to emergence, Kelle’s “Emergence” vs. “Forcing” of empirical data? is the fairest treatment, arguing the paradigm is defensible precisely as an explicit, inspectable theoretical lens.

• Choose it when you want procedure you can show. Committees, funders, and journals that ask “how exactly was the theory derived?” are answered by the paradigm’s audit trail.
• Choose it for team analysis. Shared procedures make multi-coder grounded theory workable; classic GT’s reliance on one analyst’s emergent judgement scales badly.
• Prefer classic GT if prior structure worries you more than procedure comforts you — especially in genuinely unmapped territory.
• Prefer constructivist GT if reflexivity and co-construction are central to your epistemology — the Straussian texts acknowledge interpretation but the procedures foreground technique.

A two-researcher team studied how ward nurses incorporated a sepsis early-warning system. Open coding of the first eight interviews and four observation sessions produced 61 concepts, consolidating into categories including alarm fatigue, workaround crafting, and credibility testing — this last with properties like who tests (novice vs veteran) and stakes (routine vs deteriorating patient).

Axial coding on credibility testing specified its conditions (false-alarm history, staffing pressure), strategies (shadow-scoring patients manually, consulting the informal ward expert), and consequences (selective compliance — following alerts for unfamiliar patients, overriding them for known ones). Theoretical sampling then deliberately added night-shift nurses, where staffing pressure dimensionalised differently, and two float nurses, for whom “known patients” barely existed.

Selective coding integrated around earning the right to override: a process theory in which nurses grant the system authority provisionally and re-negotiate it after every high-stakes event. The storyline paragraph survived validation against all 26 interviews, with one negative case — a nurse who never overrode — explained within the theory by her float status rather than treated as an exception.

• Paradigm as questionnaire. Filling conditions/strategies/consequences slots for every category, with thin data, produces bureaucratic theory. The paradigm is applied where the data support it.
• Open coding forever. Hundreds of codes and no axial move is the most common stall; the phase discipline exists to prevent it.
• Axial coding without incidents behind the arrows. A diagram is a claim; every link needs compared data.
• Core category by fiat. Choosing the core because it matches the research proposal, rather than because it explains the most variation, defeats the method.
• Reporting the procedures instead of the theory. A methods section that narrates coding phases while the findings present themes (not an integrated theory) signals the integration never happened.

The paradigm is a theory of action embedded in the method — conditions, strategies, consequences presuppose that the phenomenon is a managed process. Phenomena that are not processes (identities, atmospheres, meanings) fit it badly, and forcing them through it is the failure Glaser named.

Procedure also tempts substitution: it is possible to perform every step competently and produce a correct, boring theory of the obvious. The method guarantees traceability, not insight.

And full Straussian analysis is slow — the paradigm applied honestly across a dozen categories with theoretical sampling is a year’s work, which is why so many studies borrow the coding vocabulary while running something faster and thinner.

Classic grounded theory is the emergence-first original; the contrast clarifies what the paradigm adds and costs.

Constructivist grounded theory keeps iterative coding and memoing while rejecting the discovery epistemology.

The constant comparative method is the underlying engine; axial coding is one way of structuring its output.

Directed content analysis is the honest label when the categories come from prior theory rather than from the data.

The Straussian audit trail — codes, categories, properties, dimensions, paradigm links, memos — is a data-management problem as much as an analytic one. AI-assisted coding accelerates the open-coding pass and keeps every code application quote-linked, which is what makes axial claims checkable later; Evidano handles the corpus-scale first pass and the retrieval, so the team’s time goes to the paradigm work and negative-case testing.

Axial and selective integration remain analyst judgement. A tool can show every incident behind “credibility testing”; deciding that the core category is earning the right to override is the researcher’s move.