Glaser and Strauss said to let theory emerge from data, not from prior frameworks. IS researchers often invoke grounded theory while doing something considerably different.
The basic idea behind grounded theory is almost perversely simple. Instead of starting with a theory and collecting data to test it, you start with data and let the theory emerge. You go into the field with a question, not a hypothesis. You collect data, you analyze it, and the analysis tells you what to collect next. You keep going until nothing new is appearing. The theory that comes out at the end is grounded in the data that produced it, hence the name.
Barney Glaser and Anselm Strauss developed this approach in 1967, partly as a reaction against what they saw as the dominance of verification-focused social science. The field, in their view, was spending too much energy testing theories that were often not very good fits for the phenomena they were applied to, and not enough energy generating theories that actually explained what was happening. Grounded theory was their answer: a systematic, rigorous method for inductive theory generation.
The core techniques are open coding, axial coding, and selective coding, and they are most associated with the Strauss and Corbin formulation that came later. Open coding means going through your data and labeling concepts that appear. You are not yet committed to any theoretical framework. You are naming what you see. Axial coding means organizing those open codes into categories and identifying relationships between them. You are starting to build the structure of an explanation. Selective coding means identifying the core category, the central phenomenon that your theory is about, and integrating the other categories around it. By the end of selective coding, you should have a theoretical argument, not just a collection of themes.
The technique that distinguishes grounded theory most sharply from other qualitative methods is theoretical sampling. In a traditional study, you decide your sample before you collect data. In grounded theory, you decide where to collect data next based on what your analysis so far is telling you. If your emerging theory suggests that the phenomenon looks different in small firms than in large ones, you go collect data in small firms. If your analysis suggests that a particular mechanism is absent in early adopters, you seek out early adopters. The sample is not fixed at the start. It is built through the analysis, and it ends when theoretical saturation is reached, when additional data stops generating new categories or refining existing ones.
Theoretical saturation is one of those concepts that sounds clear and is extremely difficult to demonstrate in practice. How do you know you have reached it? The honest answer is that saturation is a judgment call, and that judgment is shaped by the researcher's theoretical commitments, her time constraints, and her patience for continuing to collect. Critics have pointed out that researchers routinely claim saturation earlier than it is actually warranted. The rigor of the claim depends heavily on how honest you are being with yourself about whether additional data would genuinely change your theory.
In IS, grounded theory has been used most productively in contexts where the existing theory is weak or genuinely does not fit the phenomenon being studied. Early research on social media adoption in organizational contexts, novel uses of technology in settings where prior theory predicted one thing and people were doing another, the emergence of new work practices around technologies that had no clear precedent, these are all areas where starting with an existing theoretical framework would have constrained what you could see. Grounded theory creates the space to be surprised by your data. And IS has produced genuine surprises this way.
The problem I keep running into when I read IS papers that claim to use grounded theory is that they are doing something different. The most common pattern is what Roy Suddaby called out explicitly in a 2006 paper: researchers use grounded theory as a label for what is really just inductive thematic analysis. They collect interview data, they code it, they produce themes, and they call that grounded theory. But they entered the field with conceptual frameworks already in place, they did not do theoretical sampling, they did not reach theoretical saturation by any rigorous definition, and the output is a description of themes rather than a theoretical argument about mechanisms and relationships. That is not nothing, but it is not grounded theory in the sense Glaser and Strauss intended.
The tension between the two founders themselves is also worth knowing. Glaser and Strauss split over exactly how structured the method should be. Strauss, and later Strauss and Corbin, developed the more procedural version with specific coding families and axial coding around conditions, context, actions, and consequences. Glaser found this overly constraining and argued it forced the data into a predetermined structure rather than letting theory genuinely emerge. The debate is not resolved. Glaserian grounded theory and Straussian grounded theory are genuinely different methods that produce different kinds of outputs, and researchers who invoke one or the other without knowing the distinction have usually not thought through what kind of inductive work they are actually doing.
Kathy Charmaz's constructivist grounded theory added another dimension that I find important for IS research specifically. She argued that neither Glaser's nor Strauss and Corbin's version adequately accounted for the role of the researcher in constructing the theory. The researcher is not a neutral channel through which data flows into theory. She is an active participant in the interpretive process. The theory that emerges is co-constructed by the researcher and the data, shaped by the researcher's positionality, theoretical training, and the relationship she builds with participants. Constructivist grounded theory accepts this rather than pretending it does not happen. It asks researchers to be reflexive about their own role in the construction of the theory, and to be transparent about how that role shaped what emerged.
For IS, the constructivist version matters because IS researchers are rarely theoretically neutral when they enter the field. We know TAM. We know structuration theory. We know institutional theory. We cannot unknow it. The question is whether we treat that prior knowledge as a lens that sensitizes us to what to look for, or as a framework we are sorting data into. The former is legitimate theoretical sensitization. The latter is deductive coding with a grounded theory label on it.
The constant comparative method, which runs throughout all versions of grounded theory, is the part I find most useful regardless of which tradition you follow. You constantly compare new data against existing data, new codes against existing codes, emerging categories against each other. The comparison is what produces the theoretical precision. You notice when two interviews describe what looks like the same phenomenon but from different angles, and that noticing forces you to refine your category. You notice when a new case contradicts your emerging pattern, and that forces you to either revise the theory or explain the exception. The comparison does not stop until writing is done, and sometimes not even then.
What IS sometimes treats grounded theory as is a license to do inductive work without specifying what the rules of that induction are. That is the opposite of what Glaser and Strauss intended. They developed grounded theory because they thought induction needed to be as rigorous and systematic as hypothesis testing, not because they thought leaving structure behind was a virtue. The method has rules. Applying the label without the rules is what produces the gap between what IS says it is doing and what it is actually doing.
About the author
Share
More notes
Related notes