Critical Incident Technique
- Introduced by
- John C. Flanagan
- Year
- 1954
- Type
- Qualitative behavior-elicitation and classification procedure
- Family
- Organizational behavior / job and task analysis
What does it do?
The critical incident technique (CIT) is a qualitative procedure for studying human behavior by collecting and classifying detailed accounts of specific incidents in which behavior was especially effective or ineffective in achieving an aim.
John Flanagan introduced it in his landmark 1954 Psychological Bulletin article, drawing on his work selecting and classifying aircrew in World War II, where vague trait descriptions had proved useless and concrete behavioral accounts proved decisive. Rather than asking people for opinions or generalities, CIT asks observers to recount what actually happened, what the person did, and why it mattered, then builds a framework of behavioral requirements inductively from those accounts. The technique gave applied psychology a rigorous, replicable way to derive job requirements, performance criteria, and training content from real behavior. It remains a foundational method underlying job analysis, behaviorally anchored rating scales, and competency modeling. Its hallmark is grounding abstract requirements in observable, situated action.
The intuition behind the method
If you ask people what makes a good nurse, a good pilot, or a good manager, you get adjectives: caring, decisive, organized. These are nearly useless for selecting or training anyone, because everyone interprets them differently and no one can act on them. Flanagan's insight was to refuse the abstraction and demand the story instead. Tell me about a specific time you saw a nurse do something that made a real difference, good or bad. What was the situation, what exactly did they do, and what happened as a result? A single such story is just an anecdote, but hundreds of them, gathered from people who were actually there, begin to cluster. The same kinds of effective and ineffective behaviors recur, and from those clusters you can build a precise, behavior-based picture of what the job really requires. The cleverness is methodological humility: instead of imposing categories from theory, you let the categories emerge from the incidents, so the resulting framework describes the job as it is actually performed rather than as someone imagines it.
History
Flanagan developed the critical incident technique from the Aviation Psychology Program of the U.S. Army Air Forces during World War II, where the task of selecting and classifying aircrew exposed the futility of trait-based job descriptions and the value of concrete behavioral records of why missions and training succeeded or failed. He synthesized these wartime studies, and earlier work on activity analysis at the University of Pittsburgh, into the comprehensive 1954 Psychological Bulletin article that named the method and laid out its principles. CIT spread rapidly through industrial and organizational psychology as a foundation for job analysis and performance measurement, and it directly enabled later developments such as Smith and Kendall's behaviorally anchored rating scales. Over the following decades it was adopted well beyond its origins, in nursing, education, marketing, and human-computer interaction, becoming one of the most widely used qualitative methods in applied research.
When to use it
Use the critical incident technique when you need to understand what distinguishes effective from ineffective behavior in a role or activity and you want that understanding grounded in real, observable action rather than opinion or theory. It is well suited to job and task analysis, deriving performance criteria, designing behaviorally anchored rating scales, building competency models, and studying service encounters or safety events where concrete incidents are memorable and consequential. It is less appropriate when the phenomenon does not break naturally into discrete incidents, when knowledgeable observers are unavailable or incidents cannot be recalled accurately, or when the goal is to estimate prevalence or test a causal hypothesis, since CIT describes and classifies behavior rather than quantifying or experimentally testing it.
Limitations and considerations
- Relies on recall of incidents, which is subject to memory bias, salience effects, and selective reporting.
- The categorization step requires judgment and is the most subjective and least replicable part of the procedure.
- It captures extreme, critical incidents and may neglect routine but important everyday behavior.
- It describes and classifies behavior but does not by itself establish prevalence, weighting, or causal effects.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Grounds abstract requirements in concrete, observable behavior drawn from real incidents rather than opinions or traits.
- Produces categories inductively from the data, yielding a framework that reflects the activity as actually performed.
- Is flexible and economical, adaptable across jobs, services, and settings via interviews, questionnaires, or observation.
- Feeds directly into practical tools such as performance criteria, training content, and behaviorally anchored rating scales.
Applications
- Conducting job and task analysis by identifying the behaviors that distinguish effective from ineffective performance.
- Deriving performance criteria and the behavioral anchors used in behaviorally anchored rating scales.
- Building competency models and selection standards grounded in observed effective behavior.
- Analyzing service encounters, safety events, and customer experiences from concrete incident accounts.
- Designing training content by cataloguing the specific behaviors that lead to success or failure.
How it works
How it works
The technique begins by stating the general aim of the activity under study and defining what makes an incident critical relative to that aim. Flanagan insists that the purpose be made explicit, because an incident counts as critical only if it makes a significant positive or negative difference to the stated objective. This step sets the boundary of relevance: without a clear aim, observers cannot judge which incidents matter, and the eventual classification has no anchor. Defining the aim and the criteria for criticality up front is what keeps the open-ended collection of incidents focused and interpretable.
Formula
\text{general aim } A;\quad \text{incident critical} \iff \text{it materially affects } A \text{ (effective or ineffective)}Next, firsthand accounts of specific incidents are collected from people in a position to observe them, through interviews, questionnaires, or recorded observations. Each account must be concrete: the circumstances leading up to the incident, exactly what the person did, and the consequences. Flanagan stresses that the incidents be recent, witnessed directly, and described in enough detail to leave little to inference. Gathering a large number of such specific, behavioral accounts, rather than opinions or summaries, is the empirical heart of the method, since the validity of everything that follows depends on the incidents being real and precisely reported.
Formula
I = \{(\text{context}, \text{action}, \text{outcome})_k\},\quad k=1,\dots,N \text{ firsthand, specific}From each incident the analyst extracts the observable behavior, separating what the person actually did from the observer's interpretations, judgments, or inferred motives. The point is to retain the action and its situated consequence while stripping away evaluative language that cannot be verified. Flanagan's procedure treats the behavior as the unit of analysis, so an incident describing that someone 'was unprofessional' must be pushed back to the specific actions that prompted that label. This disciplined reduction to observable behavior is what gives CIT its objectivity and makes the later categories defensible.
Formula
b_k = \text{observable behavior in incident } k,\quad \text{strip inference, retain action}The extracted behaviors are then sorted into a classification system developed inductively and iteratively. Rather than imposing a preexisting taxonomy, the analyst groups similar behaviors, names emerging categories, and repeatedly revises the scheme as new incidents are coded, much like later grounded-theory coding. Flanagan acknowledged that this categorization step requires judgment and is the most subjective part of the procedure, and he recommended checking it through independent classification by multiple analysts. The output is a framework of behavioral categories that summarizes the many incidents into a structured set of requirements or dimensions.
Formula
C = \{c_1,\dots,c_m\},\ \ b_k \mapsto c_j,\quad \text{inductive, iterative classification}Finally, the analyst judges when enough incidents have been collected, typically when additional incidents stop adding new categories, a saturation criterion, and interprets the resulting framework for its intended use. The categories can be turned into job requirements, performance criteria, training content, selection standards, or the behavioral anchors of rating scales. Flanagan presented CIT as a flexible set of principles rather than a rigid recipe, and this final step adapts the classified incidents to the practical decision at hand, completing the path from raw behavioral stories to actionable, evidence-based conclusions about effective performance.
Formula
\text{saturation: } \Delta C \to 0 \text{ as } N \uparrow;\quad C \Rightarrow \text{requirements, scales, training}Common pitfalls
- Accepting vague or evaluative accounts ('handled it well') instead of pressing for the specific observable behavior.
- Imposing a preexisting taxonomy on the incidents rather than letting categories emerge inductively.
- Stopping collection before category saturation, leaving the behavioral framework incomplete.
- Treating the frequency of incident types as a measure of behavioral prevalence, which the salience-driven sampling does not support.
Frequently asked questions
What exactly makes an incident 'critical'?
An incident is critical when it makes a clear, significant difference, positive or negative, to the stated aim of the activity. Flanagan's definition has two parts: the incident must be a specific, observable piece of human activity complete enough to permit inferences about the person, and it must matter to the objective, either contributing to it or detracting from it. A routine action that has little bearing on success or failure is not critical. This is why defining the general aim first is essential, because criticality is judged relative to that aim, and it is what keeps the method focused on behavior that actually distinguishes effective from ineffective performance.
How many incidents do I need to collect?
Flanagan tied the number to category saturation rather than a fixed count: you keep collecting until additional incidents stop producing new behavioral categories. The number needed therefore depends on the complexity of the activity. Simple, narrow activities may saturate after a few dozen incidents, while complex jobs can require many hundreds or even thousands. A practical check is to monitor how many new categories each batch of incidents adds; when that approaches zero, the behavioral framework is reasonably complete. Reporting how saturation was judged is an important part of demonstrating the rigor of a CIT study.
Is the critical incident technique qualitative or quantitative?
It is fundamentally a qualitative method, since its core activity is eliciting and inductively classifying rich, behavioral accounts of incidents. However, it has a quantitative side: once incidents are categorized, researchers can count how many fall into each category and use those frequencies to indicate which behaviors recur. This counting should be interpreted cautiously, because incidents are sampled by salience and memorability rather than by representativeness, so category frequencies reflect what observers found striking rather than the true base rate of behaviors. Used carefully, CIT bridges qualitative depth and a modest, descriptive quantification of behavioral categories.
Related methods
Sources
Flanagan, J. C. (1954). The critical incident technique. Psychological Bulletin, 51(4), 327-358.
Open DOI referenceCopied.