A nurse learns to recognize a warning pattern in a classroom simulation. Weeks later, the same pattern appears during a crowded shift, without the colored label that marked it in training. The question is no longer whether the nurse can repeat the lesson. It is whether the lesson helps when the room, timing and consequences have changed.
Transfer of learning is the use of earlier learning in a new situation. The new situation can be almost identical to the practice task or different in several important ways. Transfer can help, fail to appear, or even mislead when an old response fits the new situation poorly. The concept matters because much education and training promises value beyond the place where instruction occurred.
What has to move?
A person may carry a remembered fact, a procedure, a way of noticing a problem or a habit of asking a useful question. These are not the same outcome. A student who can solve another equation with different numbers has transferred something, but that is a narrower claim than saying the student can recognize a mathematical structure inside an unfamiliar workplace problem.
Perkins and Salomon describe positive transfer when prior learning improves later performance and negative transfer when it interferes. Their account also distinguishes near from far transfer and emphasizes that resemblance is not a single quantity. Two tasks may look alike but require different reasoning; two situations may look different while sharing a deeper relationship. The analyst needs to name both what was learned and what changed.
Transfer is therefore not simply a test score after training. It is a relationship among the learning experience, a later situation and a defensible comparison. If a participant solves a later problem, we still need to know whether prior learning helped, whether they already knew the method, and whether the task offered enough cues for them to notice the connection.
Near, far and the distance between
Near transfer usually refers to applying learning to a closely related task. A driver who practices one model of vehicle and then drives a similar model may benefit from familiar controls. Far transfer refers to application across more substantial changes, such as using a principle from a laboratory exercise to make a sound decision in a busy institution. The labels are useful only if the dimensions of difference are specified.
Barnett and Ceci proposed a taxonomy because debates over far transfer often compared unlike cases. They distinguished dimensions of what was learned from dimensions of where and when it was tested. A new physical setting, a delay of months, a different social purpose and a change from a written exercise to an action task are separate changes. A study can be far on one dimension and near on another.
This matters for promises made to learners. “The training transfers” can mean anything from recognizing a slightly modified example tomorrow to handling an unforeseen real-world event next year. Those claims require different evidence. An honest course description says which later performances were tested, under what conditions, and where the claim remains an expectation rather than an observed outcome.
Why a familiar solution may not be used
Sometimes a learner has the relevant knowledge but does not retrieve it when needed. The surface of the new problem may fail to remind them of the old one. Gick and Holyoak’s analogical-transfer studies explored this issue by teaching participants stories with a solution structure and then asking them to solve a different-looking problem. The experiments distinguished having encountered a useful analogy from spontaneously noticing it.
When examples are compared and their common structure becomes clearer, later application may improve. But a cue from an experimenter can also make transfer easier than it would be in the wild. If the teacher says “remember yesterday’s lesson,” the learner’s success demonstrates that the knowledge can be used when prompted. It does not yet show that the learner will identify the opportunity independently.
The distinction between possession and access is practical. A team can know a safety procedure and still fail to recognize the moment when it applies. Training that rehearses only a labeled case may build recall under the label. Training that varies cues and asks people to explain why a principle fits can better test recognition across settings. Whether that improvement persists must be measured in the target context.
Mechanisms, not magic
Nokes’s review of knowledge-transfer mechanisms describes routes such as analogy, compilation of practiced procedures and revision of constraints when a new problem challenges an old approach. The route depends on the kind of knowledge and the demands of the later task. A general instruction to “apply what you learned” does not supply a mechanism by itself.
Analogical transfer asks the learner to recognize a shared relation beneath different details. Procedural transfer may depend on repeated action until relevant steps can be executed reliably. Conceptual transfer may involve identifying when a principle is appropriate and when an exception changes the problem. These forms call for different practice and different assessments.
There is no need to choose one route for all learning. A clinician may need a practiced physical action, a conceptual model of a system and the judgment to recognize when the model’s assumptions fail. An educational design should articulate which of those capacities is meant to travel beyond instruction. Otherwise the word transfer hides several unfinished tasks.
Transfer can be preparation for learning again
Traditional tests often ask whether a learner can perform a new task without help. Bransford and Schwartz argued for a wider question: does earlier learning prepare a person to learn from resources in the new situation? Someone may not immediately solve a novel problem, yet may ask better questions, interpret feedback more productively or learn a new tool faster because of prior instruction.
Their proposal is often called preparation for future learning. It does not excuse weak outcomes by replacing them with a vague claim of readiness. It asks for a different assessment design. Provide both groups with a chance to study new information, then measure how they use it. An earlier learning experience may reveal its value through adaptability rather than through unaided first-try accuracy.
This distinction is useful in changing professions. If software, rules or local conditions evolve, no course can contain every future answer. A learner may benefit from knowing how to inspect a new case, identify missing information and revise a first response. That is a transfer claim, but it still needs a specified later task and evidence that the trained group learned or acted better than an appropriate comparison group.
What a credible transfer test looks like
A test should begin by defining the target situation before choosing the training exercise. If the goal is safer clinical decisions, the outcome should reflect decisions under relevant uncertainty, not merely recall of slide content. If the goal is better mathematical problem solving, the test should vary surface features while retaining or deliberately changing the underlying relationship. A designer should explain why each new task counts as the intended transfer distance.
A comparison condition is also important. Improvement from before to after a course may reflect practice with the test, ordinary experience or a change in motivation. A credible study asks what would have happened without that particular training, or with another plausible form of instruction. It measures baseline knowledge and reports how learners who did not complete the intervention were handled.
Time is another dimension. Performance immediately after instruction may reveal temporary accessibility. A delayed test asks whether the useful representation survived. If the target context itself changes between tests, that change should be described rather than hidden under one follow-up score. Transfer is an achievement under conditions, not a property assigned to a course forever.
Beware the broad cognitive-training promise
Some programs imply that practicing a narrow computer task will improve broad intelligence, workplace judgment or everyday attention. Simons and colleagues’ review of brain-training research found that evidence for broad real-world benefits was much weaker than evidence for improvement on practiced or closely related tasks. This does not mean practice never helps. It means the distance between trained task and promised outcome must be crossed with data, not with an appealing story.
The same caution applies to leadership workshops and digital-wellbeing products. Participants may become better at the workshop’s exercises, enjoy the experience and remember its language. Those are legitimate observations. They do not by themselves show that difficult meetings, decisions or relationships improved. The assessment should follow the claim outward to the actual setting.
A program can also produce negative transfer. A rule that works in routine cases may be applied too confidently to a case with a crucial difference. A novice who has learned a shortcut may need explicit practice recognizing its boundary. Better training teaches both when a principle applies and when to stop using it.
Designing practice for a changing context
One design strategy is to vary examples while preserving a common principle. Another is to compare cases and explain why the same method works in one but not another. A third is to make the learner generate the cue that signals relevance, instead of always supplying that cue in the prompt. These strategies follow from research on analogical noticing and transfer mechanisms, but their effectiveness depends on the domain, learner and target task.
Practice should also include feedback about the reasoning, not just the final answer. If a learner chooses an appropriate action for the wrong reason, a changed setting may expose that weakness. If they make a reasonable choice under uncertainty but miss one detail, feedback can direct attention to that detail. Without such information, repeated examples may strengthen a brittle pattern rather than a transferable one.
The Institute’s synthesis is that training for capacity should include a deliberate “changed conditions” test. Change the room, collaborator, time pressure or tool in ways that resemble the target environment, then observe whether the learner can still orient and act. This is a design proposal derived from transfer research, not proof that any specific Institute exercise produces broad life transfer.
Transfer is not the same as retention
Retention asks whether something learned remains available after time passes. Transfer asks whether it is used in a new situation. The two can interact, but they should not be collapsed. A person may remember a rule and fail to see that it applies. Another may remember little verbal detail but use a practiced procedure effectively when the situation resembles training.
Spacing and retrieval practice can support later memory, as the separate Library entries explain. Whether that memory supports transfer depends on what was learned, how the new task is framed and whether the learner recognizes the connection. A delayed test using the same format measures durable performance on that format. A changed-context test asks an additional question.
This distinction protects both teachers and learners from a common error. A high final-exam score need not mean a course failed if real-world transfer is not yet demonstrated. But neither should it be sold as proof of transfer. Different evidence answers different promises, and the institution should choose the evidence that matches the claim it makes.
How to read a transfer claim
Ask first what the training task was and what the later task required. Were the same materials, responses and cues used? If so, success may be valuable near transfer or retention, but not evidence for distant application. Ask next whether the later outcome mattered to the actual goal. A simulated score can be useful if the simulation captures relevant choices; its validity should be explained.
Ask who was studied. Experts and novices may notice different features of a new case. Learners with strong prior knowledge may use a principle that novices cannot yet recognize. A mean effect can conceal that the course helped one group and confused another. Look for baseline measures, subgroup reasoning specified in advance and enough cases to test the promised variation.
Finally, ask whether the design tested spontaneous application or prompted application. Both are interesting. A hint can reveal usable knowledge that would otherwise remain dormant. In the real world, a good prompt embedded in a workflow may itself be an effective design. But the claim should then include the prompt, not describe an unaided capacity that the study did not observe.
What the evidence does not show
Transfer research does not show that every carefully taught lesson will generalize. It does not show that far transfer never happens. It does not justify a universal formula for how much example variation or abstraction every learner needs. It does not turn a nearby practice improvement into proof of broad human-capacity development.
A taxonomy of transfer distances is a way to specify questions, not a scale on which every lesson can be assigned one objective number. A study of analogy in a laboratory does not establish outcomes in a hospital, school or public agency. A review of brain training does not settle every kind of professional learning. Each source supports a bounded inference.
When the room changes
Return to the nurse facing the warning pattern without the classroom label. Training may have prepared recognition, action and further learning; it may also have left one of those jobs undone. The only honest way to know is to look at what happens when the cues and stakes change.
Transfer of learning is the promise that education travels. The research makes that promise more exacting and more useful. It asks us to identify what should travel, where it must arrive, what might prevent recognition, and what evidence would show that a person can still use the learning when the room is no longer the classroom.
Research references
- Perkins, D. N., & Salomon, G. (1992). Transfer of learning. In International Encyclopedia of Education (2nd ed.). Pergamon Press. View source
- Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychological Bulletin, 128(4), 612–637. View source
- Gick, M. L., & Holyoak, K. J. (1983). Schema induction and analogical transfer. Cognitive Psychology, 15(1), 1–38. View source
- Bransford, J. D., & Schwartz, D. L. (1999). Rethinking transfer: A simple proposal with multiple implications. Review of Research in Education, 24, 61–100. View source
- Nokes, T. J. (2009). Mechanisms of knowledge transfer. Thinking & Reasoning, 15(1), 1–36. View source
- Simons, D. J., Boot, W. R., Charness, N., Gathercole, S. E., Chabris, C. F., Hambrick, D. Z., & Stine-Morrow, E. A. L. (2016). Do “brain-training” programs work? Psychological Science in the Public Interest, 17(3), 103–186. View source