Transfer of training: why skills fade after the course, and the cycle that fixes it

What transfer of training is

It’s the classic L&D paradox: a team leaves a workshop inspired, and a month later they are back to their old habits. The course was fine. What failed is transfer of training, the step from the training room to the job.

Timothy Baldwin and Kevin Ford, whose 1988 review still frames most research on the topic, describe positive transfer as “the degree to which trainees effectively apply the knowledge, skills, and attitudes gained in a training context to the job.” For transfer to have occurred, learned behavior must be generalized to the job and maintained over time on the job1.

• Generalization: using the skill in situations that differ from training.
• Maintenance: still using it months later, not only in the week after the course.

“Transfer of training” and “training transfer” mean the same thing; researchers use both, and so does this article.

Why training doesn’t transfer to the job

Baldwin and Ford sorted what drives transfer into three groups: the trainee, the training design and the work environment. A later meta-analysis by Brian Blume and colleagues linked transfer to factors in all three2:

• The learner: cognitive ability, conscientiousness and motivation to use what was learned.
• The design: a learner-centered approach, where people apply the skill through hands-on tasks, role-play, discussion, feedback and reflection.
• The environment: support from the supervisor, peers and the organization, and a positive transfer climate.

Most programs invest in the course and hand people back to a work environment nobody prepared. They stop where real learning begins: the first time a person uses the skill with a real employee or customer, with no facilitator in the room.

What is the theory of transfer of training?

There is no single theory. Siriporn Yamnill and Gary McLean grouped HRD theories around three questions: why people want to change after training, what in the design helps them transfer and what environment supports it5. Two ideas matter most for practice.

Identical elements. In experiments published in 1901, Edward Thorndike and Robert Woodworth looked for one kind of learning improving another and generally failed to find it. Thorndike concluded that transfer depends on “identical elements” shared by the two performances6,7. The more training has in common with the real task, the more of it carries over.

General principles. Transfer also improves when people learn the principle behind a behavior. A National Academies review concludes that the most effective transfer may come from a balance of specific examples and general principles8.

Behavior modeling research agrees: rules behind the model’s behavior worked better than instructions about exactly what to say9. So the scenario should look like the real conversation, and the learner should leave with a rule, such as “name the behavior, then ask,” not a line to recite.

What are the four types of transfer?

Lists of the four types differ. The most common pairs are positive versus negative and near versus far2,7. Some sources add maintenance versus generalization, or typical versus maximum transfer: what people do on an ordinary day versus what they can do when asked to show it2.

Gavriel Salomon and David Perkins also described two roads to transfer10. The low road is reflexive: a well-practiced routine is triggered by a situation like the one where it was learned, and it builds through much and diverse practice. The high road is mindful: deliberately abstracting a principle and searching for where else it applies7. Practice needs both: repetition and reflection on the rule.

The skill-building cycle: practice, reflection, new model, try again

The skill-building cycle is how we design practice at The Reflection. It is similar to Kolb’s experiential learning cycle, from concrete experience to active experimentation11, and it follows the steps of behavior modeling training: describe the skill, show a model, practice, give feedback and take steps to carry it to the job9.

01Stage 1: Practice: show the current behavior

Start with an attempt, before any teaching: it shows the starting point.

02Stage 2: Reflection: find the gap

Compare the attempt with the goal, criterion by criterion.

03Stage 3: New model: show the better way

Introduce the better way and the rule behind it. Show more than one model, including what not to do: behavior modeling research recommends showing both desired and incorrect behaviors9.

04Stage 4: Try again: use the new model

Run the same situation again, then a harder version. This stage turns an insight into behavior.

An illustrative run: Celia, a new team lead at a home insurer, practices corrective feedback with Omar · AI senior claims handler, who left a claim file without notes.

The full situation and criteria are in a feedback scenario with Omar, an AI claims handler.

An honest caveat: in the meta-analysis of behavior modeling training, the effect on knowledge and skills was large, and the effect on job behavior modest9. One good run does not guarantee a changed conversation at work, so the cycle repeats.

Why “try again” is the stage training skips

A typical workshop has time for one or two rounds of practice per person, and a coach for every repetition doesn’t scale. So the fourth stage is left “for the job,” where nobody plays the other side or gives feedback.

Yet low-road transfer builds through much and diverse practice7. One run in a workshop is neither. Three ways to repeat without a coach on every attempt:

1. The same scenario, soon. Run it again within a few days, against the same criteria.
2. A tougher version. A character who pushes back harder, hides the real cause longer or escalates. In our feedback scenario, Omar comes in three versions: open, defensive and escalating.
3. Again after a few weeks. A later run shows maintenance: whether the skill survived without the course around it.

For a worked schedule, see a four-week practice rhythm for sales objections; it works for any conversation skill.

How to make training stick: a checklist for L&D

The checklist follows the research above, and one idea runs through it: frequency of practice matters more than the volume of content.

Before the course

• Pick two or three conversations that matter on the job, and build the program around them.
• Train the managers too: transfer improved when the learners’ supervisors also received the training9.
• Write the criteria for a good conversation as behaviors you can point to.

During the course

• Start with practice, not a lecture. The first attempt is the baseline.
• Show a strong and a weak model, and the rule behind both.
• Let participants bring their own situations: transfer improved when trainees generated the practice scenarios9.
• Have each person set specific goals for using the skill on the job.

After the course

• Repeat the same scenarios, then harder versions.
• Give managers a role: they see the results and discuss them with their team.
• Re-measure on the same criteria.

Closing the cycle with AI roleplay

Every repetition needs someone to play the other side and someone to give feedback. An AI character and an AI coach do both, as often as people need. That is what communication skills training where employees rehearse with an AI character is built for. On The Reflection:

• Practice without an audience. Employees rehearse by voice, with a video avatar or in a text chat, in 50+ languages.
• Scenarios from real work. Your L&D team builds scenarios on any topic, with no developers.
• Feedback after every session, against your criteria. In the feedback scenario, those include “Explains the impact” and “Responds to what he shares.” People rewatch their recording on a timeline, with comments from the AI coach and your experts pinned to the exact second. Second-by-second video review gets people to the target skill level 40% faster than summary-only feedback.
• Managers in the loop. Each manager sees their own team’s sessions. Internal experts can score the same sessions: AI and expert scores correlate at r = 0.80, re-measured quarterly.
• Results in your LMS through xAPI and SCORM, on the Enterprise plan.

How to measure training transfer

Most transfer research relies on self-report surveys: people rate how often they use a skill, or how much it improved their work2. That shows what people think they do, not what they say.

A conversation-based alternative uses the same scenario and criteria at four points:

1. Before training: a baseline.
2. Right after training: learning, not yet transfer.
3. A few weeks later, the same scenario: maintenance.
4. A harder version or a related scenario: generalization.

Then compare with what managers see: the scenario shows what someone can do when asked, the manager sees what they do when nobody asks.

For a whole team, measure conversation skills before and after training. Then estimate the payback of a training program.

Questions L&D teams ask about training transfer

Is it transfer of training or training transfer?

Both; the terms mean the same thing. “Transfer of learning” is the broader term from education and psychology.

What percentage of training transfers to the job?

There is no reliable single number. The “10%” is an estimate quoted in a 1982 article, which researchers later called “the 10% delusion.” Measure transfer on your own conversations: the same scenario and criteria, before and after.

Who is responsible for transfer of training?

The participant, the people who design the program and the work environment, above all the participant’s manager. Georgenson’s answer in 1982 was a partnership: for training to pay off, trainers should work with the managers of course participants.