Constructivist Grounded Theory: Charmaz’s approach

The label is not decoration. Four things follow from it.

The literature is read early. Classic grounded theory advised delaying the literature review to avoid contaminating discovery. If theory is constructed rather than found, the researcher's existing knowledge is unavoidable and better made explicit than pretended away.

The interview is treated as a joint production. Data is co-constructed in the encounter, so the researcher's questions, presence and relationship to the participant are part of the analysis rather than noise around it.

Coding paradigms are refused. Charmaz declines Strauss and Corbin's axial coding template — conditions, context, strategies, consequences — on the grounds that it imposes an external structure on data it should be drawn from.

Participants' meanings are the analytic priority. The theory should stay recognisable to the people it is about, and abstraction that loses their world has abstracted too far.

Charmaz sets out the lineage and the argument in The Genesis, Grounds, and Growth of Constructivist Grounded Theory.

Glaser and Strauss's original 1967 statement split within a generation. Glaserian grounded theory kept the discovery language and the emergence of a core category. Straussian grounded theory added structured procedures, including the coding paradigm. Charmaz, a student of both Glaser and Strauss, produced the third position — her practical treatment is Constructing Grounded Theory Analyses, and an earlier statement appears as Constructivist-Grounded Theory.

A useful way to hold the difference: all three do constant comparison and theoretical sampling. They disagree about where the theory comes from. Glaser says it emerges from the data; Strauss says it is built with procedures; Charmaz says it is constructed by a situated researcher who should say where they were standing.

A recorded discussion of the approach in her own terms is available as Kathy Charmaz Discusses Constructivist Grounded Theory.

A study of people managing a stigmatised long-term condition at work began with 12 interviews and coded line by line.

A noun-phrase coding of one passage — a participant describing how he schedules medical appointments — would have produced "appointment scheduling", "workplace disclosure", "sick leave". Topics. They would have grouped into a theme about workplace arrangements and gone nowhere.

Gerund coding of the same passage produced "booking around meetings nobody notices", "storing up minor symptoms for a single absence", "letting colleagues assume a different explanation". Processes. Comparing these against codes from other transcripts — "rehearsing an answer", "choosing who gets the real version" — produced a focused code, managing the account, and eventually a category about the continuous, uncompensated labour of maintaining a plausible version of oneself at work.

Theoretical sampling followed from the category rather than from a recruitment plan. The category predicted that people whose condition had become visible would have a different problem — no account left to manage — so the study recruited four such participants. They described something the first group had not: not relief, but the loss of a skill they had spent years developing and a resulting loss of control. That contrast is what turned a category into a theory, and no amount of further coding of the original twelve would have produced it.

• No theoretical sampling. A fixed sample recruited up front makes the method impossible. This is the most common defect in published "grounded theory".
• Few or no memos. The theory is built in memos. Without them the output is a coding scheme.
• Noun-phrase codes. Topics do not aggregate into process theories.
• Claiming saturation without theoretical sampling. The claim is about categories developed through targeted data collection, not about having run out of interviews.
• Importing the coding paradigm. Axial coding in Strauss and Corbin's form belongs to a different variant and contradicts the constructivist argument.
• Producing themes instead of a theory. A grounded theory should explain a process or account for variation, not describe categories.
• Omitting the researcher. Constructivist grounded theory requires saying who constructed the theory and from where.

Charmaz proposes four criteria, and they are more useful than generic qualitative checklists. Credibility: is there enough data, and are the claims evidenced? Originality: does the analysis offer new insight or refine an existing concept? Resonance: would it make sense to participants and illuminate their experience? Usefulness: does it contribute something a reader can work with?

Practically, a reviewer should look for evidence of theoretical sampling in the methods section, memos referenced in the analysis, gerund-form codes, and a result that explains something rather than categorising it.

Theoretical sampling requires ethical approval flexible enough to permit it and a timeline that allows collection and analysis to interleave. Many funded and student projects have neither, which is why so much work is labelled grounded theory without doing the thing that makes it grounded theory.

It is also slow and hard to plan. You cannot state the final sample size in advance, which sits badly with committees that require one.

And the constructivist position, while coherent, means two researchers can construct different theories from the same field. Charmaz accepts this; reviewers expecting convergence often do not.

Classic (Glaserian) grounded theory keeps discovery language and the emergent core category.

Straussian grounded theory supplies more procedural structure, including the coding paradigm this approach rejects.

Situational analysis is Clarke's extension, replacing the search for a core process with situational mapping.

Reflexive thematic analysis shares the interpretivist stance without the theory-building ambition, and is often the honest alternative for a project that cannot do theoretical sampling.

Line-by-line gerund coding generates a very large number of codes, and constant comparison means constantly retrieving and re-comparing them. That retrieval is exactly what a qualitative platform does well.

Memos matter as much as codes here, and keeping them attached to the codes and extracts that prompted them — rather than in a separate document — is what makes memo sorting possible at the end. Evidano supports constructivist grounded theory as a named methodology, keeping memos linked to the material that generated them. Theoretical sampling is a fieldwork decision no tool can make, and it is the step that determines whether the study produces a theory at all.