The Kirkpatrick Model: What the Four Levels of Training Evaluation Actually Measure

Integrations mentioned in this article
Why the best-known model for evaluating corporate training is still useful, what it is criticized for, and why it should be neither idealized nor written off.
Introduction
The Kirkpatrick model is arguably the best-known system for evaluating corporate training. It appears in presentations, methodological guidelines, training program design, and in every conversation about how to prove the value of L&D to the business.
The four levels — reaction, learning, behavior, and results — look simple and logical. First we find out whether participants liked the training, then we check what they learned, then we look at whether they apply the new knowledge and skills at work, and finally we assess whether the training affected the organization's results.
But behind this clear scheme lie several important questions. Do the four levels really form a sequential ladder? Can we assume that a positive reaction leads to learning, learning leads to a change in behavior, and a change in behavior leads to business results? And how legitimate is it to attribute changes at work specifically to training?
To answer these questions, it helps to go back to Donald Kirkpatrick's original publications, and then to see how the model developed and what criticism it received from researchers.
1. Where the model came from
In 1959–1960, Donald Kirkpatrick published a series of four articles in the Journal of the American Society of Training Directors under the general title Techniques for Evaluating Training Programs.
Each article was devoted to a separate aspect of evaluation:
- November 1959 — Reaction;
- December 1959 — Learning;
- January 1960 — Behavior;
- February 1960 — Results.
In the original publications, Kirkpatrick called them not levels but steps. He proposed four directions for evaluating the effectiveness of training and described practical ways of doing it 1.
The historical context matters. This was not an attempt to create an exhaustive scientific theory of adult learning or a universal model of cause and effect. Kirkpatrick addressed training managers first and foremost, and aimed to give them clear ways of evaluating programs systematically.
At the start of the series he formulated a principle that conveys the practical orientation of the work well: the results of evaluating one program cannot simply be transferred to another, but evaluation methods can be borrowed.
In other words, Kirkpatrick offered not universal success indicators but approaches that organizations could adapt to their own goals.
2. The four levels: what exactly we evaluate
1. Reaction
2. Learning
3. Behavior
4. Results
Level 1. Reaction
The first level answers the question: how did participants perceive the training?
Kirkpatrick defined reaction as how well participants liked the program. It concerns their impressions of and attitude toward the content, the instructor, the format, and the organization of the training.
In practice these are usually feedback questionnaires that participants fill in after a session or a course.
You can find out:
- how well the content matched expectations;
- how clear and useful the presentation was;
- how satisfied participants are with the instructor's work;
- how comfortable the learning format was;
- whether participants consider the knowledge they gained applicable at work.
It is important here not to confuse satisfaction with training with the results of training. Kirkpatrick himself stressed that evaluating reaction does not include measuring what participants learned 1.
A high rating for a course does not mean that participants actually mastered the material. And the other way round: a demanding, difficult program can be useful even if not all participants rate it highly.
So the first level shows the quality of the perceived learning experience, but on its own it does not prove that learning took place.
Level 2. Learning
The second level answers the question: what did participants learn or what can they now do?
This level evaluates the change in knowledge, skills, or attitudes as a result of training.
The following can be used for this:
- testing before and after training;
- practical assignments;
- skill demonstrations;
- case solving;
- checking that certain actions are performed;
- assessing the change in attitudes, if they are the goal of the program.
For example, after a project management course a participant may know the basic principles of planning, be able to draw up a project plan, and understand how to work with risks.
But it is important to distinguish knowing something from being able to apply it in a real work situation. A person may pass a test successfully and still not use the new knowledge at work.
Kirkpatrick recommended comparing results before and after training where possible, and also using control groups and statistical analysis. This shows that even in his original work he treated evaluation not merely as collecting impressions or final testing, but as a task requiring more rigorous methods 1.
Level 3. Behavior
The third level answers the question: has the person's behavior at work changed?
This is one of the most important and at the same time most difficult levels of evaluation.
Training can lead an employee to learn new approaches and demonstrate the required skills in class. However, this does not yet mean that they will start applying them in real work.
In his article on behavior, Kirkpatrick gives a telling example of a manager who learned the principles of human relations well during training but did not apply them after returning to the workplace 1.
This example matters: there is a gap between mastering knowledge and changing behavior.
Kirkpatrick drew attention to the conditions necessary for behavior change. Among them are the person's own wish to improve their work, awareness of their own weaknesses, a supportive work environment, help from an interested and competent manager or mentor, and an opportunity to try new ways of working.
To evaluate behavior, he proposed comparing the situation before and after training, collecting assessments from several sources — the employee, the manager, colleagues, and subordinates — carrying out the evaluation some time after the training and, if possible, using a control group 1.
Thus the third level is no longer just a check of knowledge. It requires observing a person's work and understanding the conditions under which learning is transferred into practice.
Level 4. Results
The fourth level answers the question: what results did the training produce for the organization?
Unlike the previous levels, attention here shifts from the participant to the organization.
The following can be considered results:
- increased productivity;
- improved quality of products or services;
- cost reduction;
- fewer errors and defects;
- lower staff turnover and absenteeism;
- fewer complaints;
- higher customer satisfaction;
- improvement in other indicators related to the goals of the training.
Kirkpatrick gave examples of such indicators already in the original article on results 1.
However, a fundamental methodological problem arises here: even if a business indicator improved after training, this does not yet prove that training was the cause of the change.
Results can be influenced by new technology, process changes, managers' actions, employee motivation, the market situation, and many other factors.
So evaluating the fourth level requires not only recording changes, but also an attempt to understand what role training played in them.
3. Do the four levels really form a sequential ladder?
In professional practice the model is often drawn as a pyramid or a sequential ladder: from participants' reaction to learning, then to behavior change, and finally to business results.
This picture is convenient for explanation, but it can create a wrong idea of the model's own logic.
First, the four levels describe different objects of evaluation. Reaction is participants' impressions, learning is acquired knowledge and skills, behavior is the application of what was learned at work, and results are changes at the organization level.
Second, these indicators are not necessarily linked automatically. High satisfaction does not guarantee learning. Successful mastery of material does not guarantee its application. A change in behavior does not always lead to a noticeable business result — at least within the observation period.
Third, a higher level does not always mean fuller or better information. For example, to improve a training course, detailed participant feedback may be more useful than an attempt to measure distant business results.
It is these assumptions that became the subject of scientific criticism of the model.
4. What researchers say: the main limitations of the model
Criticism of the causal link
In 1989, George Alliger and Elizabeth Janak published the article Kirkpatrick's Levels of Training Criteria: Thirty Years Later 3.
The authors identified three disputed assumptions often associated with the model:
- Each subsequent level provides more information than the previous one.
- There is a causal relationship between the levels.
- The indicators of different levels correlate positively with each other.
The researchers subjected these assumptions to critical analysis. Their work became one of the best-known publications to question the simplified understanding of the model as a sequential causal chain.
This does not mean that the four directions of evaluation are useless. Rather, we cannot automatically assume that a positive result at one level guarantees a positive result at the next.
The model as a classification of outcomes, not a full evaluation theory
In 1996, Elwood Holton published the article The Flawed Four-Level Evaluation Model 4.
His criticism was more fundamental. Holton argued that the four levels are better seen as a classification of training outcomes than as a full evaluation model.
In his view, explaining the effectiveness of training requires taking into account intermediate factors that influence the outcomes. They include:
- motivation to learn;
- individual characteristics of participants;
- attitude to work;
- readiness to apply new knowledge;
- work environment conditions;
- opportunities to transfer learning into practice.
In other words, it is not enough simply to measure four types of outcomes. We need to understand which factors influence their appearance and how they interact.
This is an important distinction: the Kirkpatrick model helps structure evaluation, but by itself it does not explain the whole mechanism through which training leads to change.
Participants' reaction is not as simple an indicator as it seems
In 1997, Alliger, Tannenbaum, Bennett, Traver, and Shotland published a meta-analysis of the relationships among training criteria. It included 34 studies and 115 correlations 5.
The authors proposed to separate participants' reaction in more detail. In particular, they distinguished the affective reaction — whether the training was liked — from the assessment of its utility, that is, how applicable participants consider the content they received.
The results showed that the assessment of utility was linked to learning and to transfer to job performance more strongly than affective satisfaction.
Limitations for practical evaluation
In 2004, Reid Bates published a critical analysis of training evaluation practice in which he examined the Kirkpatrick model in terms of its possibilities and limitations 6.
One important question is whether the model helps not only to determine whether the training was effective, but also to understand how to improve the program itself.
Collecting indicators across the four levels can give a general picture, but it does not always explain why the training worked or did not work. That requires analysis of the causes, the context, and the conditions in which the training took place.
In 2017, Thomas Reio and co-authors returned to the criticism of the model. They examined its influence, alternative approaches, and empirical research. Their conclusion is fairly balanced: the model remains useful and influential, but a full evaluation across all four levels presents significant difficulties in practice 7.
5. What the Kirkpatrick model offers modern L&D
Despite the criticism, the Kirkpatrick model has not lost its practical value. Its strength lies in a simple and clear way of broadening the idea of what training evaluation is.
Without such a framework, organizations often limit themselves to two questions: how many people completed the training and how much they liked the course.
Kirkpatrick proposed looking further — at what a person actually learned, whether their behavior changed, and whether results for the organization appeared.
This is an important shift in the understanding of the role of training. The mere fact of running a program, the number of participants, and positive reviews do not yet say anything about its effectiveness.
However, the model should be applied consciously.
First. You do not have to measure all four levels for every program
The depth of evaluation should match the goals, cost, and significance of the training. For a short introductory course, checking that the material was absorbed may be enough. For a leadership development program aimed at changing management practices, evaluating behavior will be far more important.
Second. Do not treat the levels as an automatic causal chain
Results at one level do not guarantee results at the next. A positive reaction does not prove learning, and acquired knowledge does not guarantee its application at work.
Third. Take the work environment into account
If an employee cannot apply new knowledge because of a lack of authority, time, manager support, or necessary tools, the problem may not lie in the quality of the training.
Fourth. Define in advance what exactly should change
Evaluation should not begin after the course is over. If an organization wants to measure behavior change or business results, the relevant indicators and baseline values must be defined before the training starts.
Fifth. Do not attribute all positive changes to training
If sales grew or the number of errors fell after a program, this does not yet mean that training was the cause. More well-founded conclusions require additional data and a suitable evaluation design.
6. So should you use the Kirkpatrick model?
Yes. But it is important to understand what for.
The Kirkpatrick model remains a convenient framework that helps keep training evaluation from being reduced to participant satisfaction and the fact of completing a course. It sets four directions of attention: reaction, learning, behavior, and results.
At the same time, it is not a universal tool that automatically lets you prove the impact of training on the business. It does not replace analysis of the work environment, the study of the causes of change, and a well-thought-out evaluation design.
Its usefulness lies in asking the right questions. But answers to these questions require different methods, data, and time horizons.
Perhaps this is exactly why the model keeps its significance more than six decades after the first publications. It does not solve all the problems of training evaluation, but it helps us see how much wider the task is than collecting feedback after a course or checking knowledge.
And most importantly, it reminds us that training should be evaluated not only by what happened during the program, but by what changed after it.
Frequently asked questions
Sources
1 Kirkpatrick, D. L. (1959–1960). Techniques for Evaluating Training Programs. Journal of the American Society of Training Directors, 13(11), 3–9; 13(12), 21–26; 14(1), 13–18; 14(2), 28–32. The four original articles: Reaction, Learning, Behavior, and Results. An available reprint of the whole series: Training and Development Journal, June 1979, pp. 78–92. Open the reprint PDF
2 Kirkpatrick, D. L. (1994). Evaluating Training Programs: The Four Levels. San Francisco: Berrett-Koehler Publishers.
3 Alliger, G. M., & Janak, E. A. (1989). Kirkpatrick's Levels of Training Criteria: Thirty Years Later. Personnel Psychology, 42(2), 331–342. https://doi.org/10.1111/j.1744-6570.1989.tb00661.x
4 Holton, E. F. III. (1996). The Flawed Four-Level Evaluation Model. Human Resource Development Quarterly, 7(1), 5–21. https://doi.org/10.1002/hrdq.3920070103
5 Alliger, G. M., Tannenbaum, S. I., Bennett, W., Traver, H., & Shotland, A. (1997). A Meta-Analysis of the Relations Among Training Criteria. Personnel Psychology, 50(2), 341–358. https://doi.org/10.1111/j.1744-6570.1997.tb00911.x
6 Bates, R. (2004). A Critical Analysis of Evaluation Practice: The Kirkpatrick Model and the Principle of Beneficence. Evaluation and Program Planning, 27(3), 341–347. https://doi.org/10.1016/j.evalprogplan.2004.04.011
7 Reio, T. G., Rocco, T. S., Smith, D. H., & Chang, E. (2017). A Critique of Kirkpatrick's Evaluation Model. New Horizons in Adult Education and Human Resource Development, 29(2), 35–53. https://doi.org/10.1002/nha3.20178
Additional reading on transfer of training:
Baldwin, T. T., & Ford, J. K. (1988). Transfer of Training: A Review and Directions for Future Research. Personnel Psychology, 41(1), 63–105.
Grossman, R., & Salas, E. (2011). The Transfer of Training: What Really Matters. International Journal of Training and Development, 15(2), 103–120.

















