Written and maintained by CASRAI Editorial Board
Last updated
A researcher preparing to enter the field for a qualitative study faces a real design choice that rarely gets named directly: how much of a framework to bring in the door. Bring nothing, and early fieldwork can be unfocused — everything looks potentially relevant, and it is hard to know what to write down. Bring a fully specified hypothesis, and the opposite risk appears: the framework tells you in advance what you are looking for, which makes it easy to see confirming instances and easy to miss the things the hypothesis never anticipated. Herbert Blumer’s sensitising concept, introduced in his 1954 paper “What is Wrong with Social Theory?”, is the answer sociology and later the broader qualitative-methods tradition converged on for that problem. It names a real third option, not a compromise between the other two but a distinct methodological device with its own logic.
Blumer’s Distinction: Definitive Concepts vs Sensitising Concepts
Blumer drew the distinction against what he called a definitive concept: a concept given a fixed, operational definition in advance — a set of benchmarks or criteria that let a researcher identify a clear-cut instance of the class it names. A definitive concept is what a hypothesis needs to be testable: if “team cohesion” is defined as “a score of 4 or higher on this validated instrument,” you can measure it, compare groups on it, and accept or reject a prediction about it. That precision is the whole point of a definitive concept, and it is exactly right for confirmatory, hypothesis-testing designs.
A sensitising concept, in Blumer’s own words, “gives the user a general sense of reference and guidance in approaching empirical instances… merely suggest[ing] directions along which to look.” It does not tell the researcher what a “cohesive team” looks like in advance. It tells the researcher that team cohesion is a dimension worth watching for, and leaves the actual content of what counts as cohesion — and what produces or erodes it — to be discovered in the specific setting under study. The concept orients attention; it does not pre-specify findings.
This is not vagueness for its own sake. A sensitising concept still has to be precise enough to be useful — “entering with orienting ideas” does not mean entering with no ideas. The skill is choosing a concept abstract enough to survive contact with a setting the researcher does not yet fully understand, while concrete enough to actually shape what gets noticed, written down, and asked about.
Three Ways to Enter the Field, Compared
It helps to lay all three postures side by side, because “no framework” and “sensitising concept” are easy to conflate from the outside — both look, superficially, like the researcher “just went and looked.”
| Entry posture | What’s fixed before data collection | What’s discovered in the field | Typical design |
|---|---|---|---|
| No framework | Nothing — not even a topic of attention | Everything, with no way to prioritise what matters | Rare in practice; usually a mischaracterisation of exploratory work, which still has some orienting interest |
| Sensitising concept | A named dimension worth watching (e.g. “team cohesion,” “legitimacy,” “boundary work”) | What that dimension actually consists of, how it operates, and whether it matters, in this setting | Grounded theory, ethnography, most inductive and abductive qualitative designs |
| Hypothesis | A specific, operationalised prediction about a relationship between defined variables | Whether the prediction holds, within the terms it was stated in | Confirmatory quantitative designs; deductive qualitative designs using a pre-built codebook |
The practical difference shows up in what the researcher’s field notes and interview guide look like on day one. A hypothesis-driven researcher walks in with an instrument or a structured protocol built to detect a specific, already-named pattern. A sensitising-concept researcher walks in with a short list of dimensions to stay alert to, an open-ended interview guide, and no pre-built answer key.
Worked Example: The Same Topic, Two Different Framings
The scenario below is an illustrative composite constructed for this guide, not a real study, ward, or institution — it exists to make the sensitising-concept/hypothesis contrast concrete on identical subject matter. The numbers in the hypothesis-test version are computed from a seeded, reproducible simulation (Mulberry32 pseudo-random generator, seed 20260826, Box-Muller normal draws, Welch’s two-sample t-test), not invented to look plausible; anyone can regenerate the same values from the same seed.
Suppose a researcher is interested in team cohesion on a hospital ward. Framed as a hypothesis, the design has to commit to an operational claim before data collection starts — for example: “Nurses with more than five years of tenure will report higher team-cohesion scores than nurses with less than two years of tenure, measured on a validated 1–7 cohesion instrument.” That is testable, and testing it is exactly what a simulated dataset built to that specification does:
| Group | n | Mean score (1–7) | SD |
|---|---|---|---|
| Early-career (<2 yrs tenure) | 12 | 4.9 | 0.6 |
| Veteran (>5 yrs tenure) | 12 | 5.2 | 0.6 |
Welch’s t-test on the simulated scores: t = −1.52, df = 21.9, two-sided p = 0.144 — not statistically significant at conventional thresholds. That is a complete, honest answer to the question the hypothesis asked. It is also the only question the design can answer. The instrument never asked about shift-handoff rituals, workstation geography, informal debrief norms, or how a bad outcome the week before gets talked about on the unit — because none of those were part of the operational definition of “cohesion” the hypothesis committed to in advance. A null result here tells the researcher tenure isn’t the driver; it says nothing about what is.
Framed instead as a sensitising concept — “team cohesion” as a dimension to watch, undefined in advance — the same researcher enters with open-ended field observation and interviewing, watching for whatever the ward’s own members treat as evidence of a team working well or badly together. Nothing rules out tenure mattering, but nothing requires it to be the axis either. In an ethnographic or grounded-theory design run this way, it is entirely plausible for the researcher to notice, and later code and write up, that cohesion talk on the unit clusters around who is present at shift handoff and how disagreements from a difficult case get aired afterward — a finding the tenure hypothesis was never positioned to produce, because it was never asked.
The two are not competing answers to the same question. They are different instruments pointed at the same general topic, and the choice of which to use is a choice about what kind of question the study is actually trying to answer at this stage of the research.
Declaring a Sensitising Concept Without Pre-Empting the Analysis
A proposal, protocol, or IRB submission built around a sensitising concept still has to say something concrete — reviewers reasonably want to know what the study is oriented toward, not just that it is “open” or “exploratory.” A few conventions handle this without smuggling a hidden hypothesis in:
- Name the concept and its source explicitly. “This study uses [concept] as a sensitising concept in Blumer’s (1954) sense” tells a reviewer familiar with qualitative methods exactly what kind of methodological commitment is and isn’t being made.
- State what the concept is not. Explicitly note that the concept is not being operationalised into fixed indicators before data collection, and that its content will be developed from the data — this pre-empts the common reviewer reflex of asking for a measurement instrument.
- Write the interview guide and observation protocol around the concept, not a definition of it. Questions should invite participants to describe, in their own terms, whatever they treat as relevant to the concept, rather than asking participants to rate or confirm a researcher-supplied definition.
- Distinguish it from a research question. The research question can still be sharply stated (“how do ward staff describe and account for team cohesion?”) — what stays open is the content of the answer, not the topic of the study.
- Revisit and refine, don’t finalise, as data comes in. Many grounded-theory and constructivist designs treat the sensitising concept as a starting point that gets progressively sharpened into more specific codes and categories through the analysis itself — that refinement is the expected trajectory, not a deviation from the design.
How Sensitising Concepts Relate to Other Qualitative Tools
A sensitising concept is a starting orientation, not a full analytic method, and it typically feeds into one of several established approaches rather than standing alone:
- Grounded theory uses sensitising concepts most explicitly and most carefully — Glaser and Strauss’s original formulation and later variants treat prior theoretical vocabulary (including sensitising concepts) as something to hold loosely so it orients without dictating the categories that eventually emerge from coding. See the full walkthrough in the grounded theory guide.
- Thematic analysis can use a sensitising concept the same way at the familiarisation stage, before initial coding begins — see the Braun and Clarke thematic analysis guide for where that sits in the six-phase process.
- Deductive, codebook-based coding is the approach a sensitising concept is deliberately not: a codebook fixes categories before coding starts, which is appropriate once a study has moved from exploration to confirmation, or when replicating a prior framework, but is a different design decision than sensitising-concept work.
- Symbolic interactionism, the broader theoretical tradition Blumer helped found, is the intellectual home of the sensitising-concept idea — concepts like “self,” “role-taking,” and “definition of the situation” are treated as sensitising rather than definitive throughout that tradition.
- Inductive and abductive reasoning are the logical modes sensitising-concept work depends on: the concept supplies a starting direction, but the actual content of the finding is built up from the data rather than deduced from a prior definition.
- Reflexivity matters more, not less, with a sensitising concept in play — because the researcher’s prior familiarity with the concept can still shape what gets noticed even without a formal operational definition. A reflexivity and positionality statement is the standard place to name that influence explicitly.
Common Mistakes
- Treating the sensitising concept as a hidden hypothesis. If a “sensitising concept” section of a methods write-up already states what the concept will turn out to mean, it has quietly become a hypothesis with an extra label. The test: could the fieldwork plausibly turn up a finding that contradicts the researcher’s current framing of the concept? If not, it isn’t functioning as a sensitising concept anymore.
- Choosing a concept too vague to orient anything. “Culture” or “experience” on their own are often too broad to shape what gets noticed in the field. Effective sensitising concepts are specific enough to point somewhere — “boundary work,” “moral career,” “articulation work” — while still leaving their content open.
- Never revisiting the concept. A sensitising concept that is named at the start of a study and never mentioned again in the analysis wasn’t doing methodological work — it was decoration. The concept should show up again in how findings are organised, refined, or explicitly complicated.
- Using the term as a synonym for “theme.” A sensitising concept precedes data collection and orients it; a theme is a product of coding, produced after data collection. Conflating the two blurs a design decision with an analytic output.
Frequently Asked Questions
Is “sensitising concept” the same as “sensitizing concept”?
Yes — it is the same term, spelled with British (“sensitising”) or American (“sensitizing”) conventions. Blumer’s original 1954 paper uses the American spelling; both forms appear throughout the methods literature interchangeably.
Who coined the term “sensitising concept”?
Herbert Blumer, in “What is Wrong with Social Theory?” (American Sociological Review, 1954), where he contrasted it with what he called a definitive concept.
Can a study use more than one sensitising concept?
Yes, and many do — a short set of two or three related sensitising concepts is common, especially in grounded theory and ethnographic work, as long as each one is genuinely orienting the fieldwork rather than functioning as an unstated hypothesis.
Does a sensitising concept need to be cited to a specific theory?
Not necessarily, but naming the theoretical tradition a concept is drawn from (symbolic interactionism, institutional theory, practice theory, and so on) helps a reader and a reviewer understand what kind of orienting work the concept is expected to do, and makes the concept’s boundaries clearer.
How is a sensitising concept different from a conceptual framework?
A conceptual framework typically specifies relationships among several concepts in advance, closer to a loosely stated hypothesis. A single sensitising concept makes a narrower commitment: it names a dimension worth attending to, without specifying how it relates to anything else.
