Grounded theory is a qualitative research methodology in which a theory is built directly from data rather than tested against it. Instead of starting with a hypothesis and looking for confirmation, the researcher collects data, analyzes it, and lets explanatory concepts emerge, returning to gather more data to develop and refine those concepts until a coherent theory grounded in the evidence takes shape. It was developed by Glaser and Strauss in 1967 and remains the standard approach when the goal is to generate theory about a process for which little existing explanation fits.
The defining commitment is that the theory comes from the data. Grounded theory is not a way to organize interviews around themes you already expected; it is a disciplined process for discovering an explanation you did not have at the start. That ambition makes it powerful for under-theorized topics and demanding to execute well.
What makes grounded theory distinctive
Several interlocking features separate grounded theory from other qualitative methods.
Constant comparison runs throughout: every new piece of data is compared with existing codes and concepts, sharpening categories and revealing their properties. Theoretical sampling means later data collection is guided by the emerging theory, so you deliberately seek the participants or situations that will test and extend your developing concepts, rather than fixing the entire sample in advance. Memo writing captures the analyst's developing thinking, building the bridge from codes to theory. And data collection and analysis happen iteratively, in cycles, rather than as separate phases.
This iterative, theory-driven sampling is the clearest practical difference from a method such as thematic analysis, where data are usually collected first and analyzed afterward. In grounded theory the analysis shapes what you collect next.
The coding stages
Grounded theory analysis moves through coding stages that progressively build abstraction. Open coding breaks the data into discrete concepts, labeling what is happening line by line or incident by incident. Axial coding reassembles those concepts, identifying relationships among categories and their properties. Selective coding integrates everything around a central category that ties the theory together. Terminology varies across the main traditions, but the movement from concrete data to an integrated explanation is common to all.
The endpoint is theoretical saturation, the point at which gathering more data yields no new properties or insights about the categories. Saturation, not a predetermined sample size, governs when data collection stops, which is one reason grounded theory cannot specify its final sample in advance the way a survey can.