Models of learning transfer

“An individual understands a concept, skill, theory, or domain of knowledge to the extent that he or she can apply it appropriately in a new situation.”
Howard Gardner[1]

There are over 20 learning transfer models that have been proposed in what has become a field of study within learning theory. Many of these transfer models are based on the work done by Edward Thorndike[2] over 100 years ago. In a seminal article, Thorndike and Woodworth (1901) proposed their common-elements theory, according to which transfer is a function of the extent to which two domains share common features. Their theory of transfer states that the extent to which information learned in one situation will transfer to another situation is determined by the similarity between the two situations. The more similar the situations are, the greater the amount of information that will transfer. Similarly, if the situations have nothing in common, information learned in one situation will not be of any value in the other situation.

Near and far transfer

In effect, the theory predicts that while near transfer takes place often, far transfer is much less common. This leads to the Near/Far model, which is probably the most commonly known, partly because it is useful when considering learning design, and partly because it is still widely discussed in the literature despite being with us for more than a hundred years.

In many respects, transfer is the most critical concept in teaching. However effective instruction might otherwise be, if a learned behaviour or skill does not transfer to relevant functional application contexts and/or is not maintained over time, then the instruction has failed. Skill decay is a major problem in training. Winfred Arthur[3] and his colleagues did a metanalysis of skill decay studies and reported that the day after training, trainees exhibit little to no skill decay, but one year after training, trainees have lost over 90% of what they learned. There is greater decay for cognitive versus physical tasks, but, most importantly, Arthur et al. also reported that overall retention decreases dramatically with longer periods without practice or use of the skill. Understanding the nuances of transfer gives us a head start in designing our training to achieve it.

In the Near/Far model, near transfer occurs when the training context and trained behaviour are almost identical to the application context and application behaviour. A common example used is that of tying shoelaces. Once we have learned to tie a shoelace, it is highly likely that the skill generalises to tying all shoelaces, regardless of the length or colour or thickness of the lace or the design of the shoe. Near transfer involves the study of a problem or task and then practising it to a high level of automation. When a nearly similar problem or task is encountered, it is automatically solved or accomplished with little or no conscious thought.

Shoelaces and taps

This ability to generalise a skill to solve nearly similar problems has been, and still is, crucial to our species for survival. It is hardwired into us. If we didn’t have it, we would have to ‘learn’ all over again how to tie a shoelace every time we bought a new pair of shoes. We would have to ‘learn’ to use a tap every time we were confronted with a tap with a slightly different design. We would have to ‘learn’ to open a door every time we encountered a different design of door. And this brings us to the idea that sometimes the application context and application behaviour is sufficiently different for near transfer not to work. Imagine every tap you have seen and used has a knob to turn, and then you encounter for the first time a tap with a push button. What do you do?

Imagine every door has a lever handle or a knob to turn, and you encounter a door with neither. Instead, it has a push button on the wall beside the door. What do you do? We must learn to expand our skill to include a wider range of contexts, and in doing so, create some higher level ideas in our mind that will enable us to solve problems that are even further away from what we first learned. Now we are transferring concepts to guide problem solution rather than directly applying automated routines. This is called far transfer. Far transfer tasks involve skills and knowledge being applied in situations that change, where the application of the skill is executed differently depending on the situation. In far transfer, the learner adapts their actions, based on their judgement of the situation.

It is important to realise that near and far transfer occur on a continuum and the transfer is either nearer or farther away from the training context and behaviour. It is also important to realise that people vary considerably in their abilities to see, feel or sense similarities between different problem situations. In any problem-solving situation, some people seem to be innately much better at far transfer than are others. Or rather, they see the similarities more easily, and thus for them it is nearer transfer, and comes more easily.

Low road and high road transfer model

The difficulty with the theory of near and far transfer is that it does not provide a foundation or a plan for helping a person to become better at far transfer when dealing with novel and complex problems. And it does not tell us how to teach to increase far transfer. Then, in 1992, David Perkins and Gavriel Salomon published their Low Road and High Road Transfer of Learning Theory on the mechanisms of transfer.

Low road transfer happens when stimulus conditions in the transfer context are sufficiently like those in a prior context of learning to trigger well-developed semi-automatic responses. A relatively reflexive process, low road transfer figures most often in near transfer. High road transfer, in contrast, depends on mindful abstraction from the context of learning or application and a deliberate search for connections. What is the general pattern? What is needed? What principles might apply? What is known that might help? Such transfer is not, in general, reflexive or automatic. It demands time for exploration and the investment of mental effort.

The ideas encompassed in the concept of low/high road transfer start to show us how we need to teach to achieve transfer, particularly the more difficult far transfer. We all make use of strategies as we attempt to solve problems and accomplish tasks. The research literature on problem solving indicates that most people have a relatively limited repertoire of problem-solving strategies and that it is helpful to increase that repertoire. Teaching techniques for high road transfer of learning can help trainees to increase their repertoire. However, there is more to problem solving than a toolbox of strategies. Improving problem solving in a specific domain requires increasing knowledge that is specific to that domain. The learner must consciously practise utilising those new strategies in conjunction with their domain knowledge with a large range of problems and over a long period of time. Far transfer requires effort and practice, and the farther the transfer, the greater the effort required.

In today’s world, employees need to solve problems and make decisions in an ever changing and complex environment. Far transfer focuses on trainees learning general concepts that may be applied in a wider set of contexts than those represented in the training setting. This is not to say that you should never train for near transfer. For example, there is little room for error when training someone to safely operate a power tool. Here, a near transfer training approach is more appropriate, so trainees can replicate the training behaviour as closely as possible.

Some examples of transfer

We can sit here nodding and say that these transfer theories kind of make sense, but then what? How does an understanding of these theories affect how we should design and train? If we accept that near transfer is easier to accomplish and more likely to happen, then where we can, we should train so that the required transfer to get the desired performance improvement is as ‘near’ as possible. In other words, we should design training that simulates or mimics the working environment and then teach procedures. Think of this as teaching ‘how’ to do something and developing routine expertise. Stop for a moment and remember when you learned ‘how’ to do things, such as drive a car, ride a bike, handle a power tool, swim or use a computer. Remember the process you went through, and where you did it, so that these skills are now largely reflexive responses to any situation that requires them. As you practised your new skill, repeated the new behaviours, and refined your learning, you became consciously competent and eventually unconsciously competent.

Although few of us have done this, imagine the process that an airline pilot goes through to learn their skills and how their training programmes make extensive use of tools such as flight simulators. You might have encountered learning in a simulated environment if you have practised various tasks using a software system in a test environment or even doing something as simple as practising a mail merge task with test data before doing it with real data. The skills gained from practice opportunities during a training programme are more likely to transfer well into the work environment when they fit without reworking or modification. Building this training and practice environment to mimic the real world is obvious in principle, but often programme designers and trainers don’t have a sufficiently profound understanding of the environment where the learner will use their new skills. It’s therefore necessary to collaborate with both the organisation and the participants to build relevant practice into the programme. Too often, this research into the learners’ work environment is not done, and the resulting learning transfer is poor because the classroom instruction is too ‘far’ away from the learners’ working reality.

Star Trek holodeck

Another problem occurs when a training course, once developed, is rolled out to users who are in a different work environment, perhaps in a different country or factory, or when the course is continued although changes in the work environment have made it much less relevant and therefore less effective. There are many things you can do in the classroom to create real-life scenarios. You can set the ‘classroom’ up to be the same as a house to practise installing wiring or plumbing; you can use role-play and actors, or you can immerse trainees in solving realistic problems that they will face back in the workplace. You can introduce digital solutions, such as videos, with branching right through to simulators and the virtual reality tools that do an amazing job of recreating the real working environment. In the training centre of a big furniture retailer I visited, they had built part of a house that included stairs with an awkward corner, and delivery trainees would practise carrying a new bed up the stairs without damaging the wallpaper. The ultimate practice environment, of course, would be the Star Trek holodeck, where the learner would be in a totally safe yet totally real scenario, in which they could practise, experiment, and learn. But it’s going to take a while before Star Trek technology is available to L&D; indeed, most L&D budgets cannot even stretch to many of today’s simulation tools, although virtual reality programmes are looking promising.

Wait a minute! Haven’t we already got something close to holodeck technology? It’s called real-life! Within the real-life context of the workflow, we can set up scenarios and, with appropriate risk management, we can get the learner to practise, experiment and learn, just where they will be required to use their skills for real. Unless we consider the inevitable complexities of the real world, the instruction we design is likely to be far too simplistic for transfer to occur to the degree needed. Sometimes we trainers forget that we can ‘extend’ our programme well beyond the classroom or formal event and into real life, providing varied practice in a range of contexts. How can you use the real-life ‘holodeck’ that is available to you? On-the-job training has grown in popularity because of the frequency with which out-of-context training fails to transfer to on-the-job performance. This is another way to reduce transfer distance because we are using the holodeck of on-the-job context. Within that context, as trainees begin to master a skill, the training and practice conditions should be increasingly difficult; there should be less and less trainer support, and practice conditions should increasingly resemble transfer conditions. Practice opportunities should require trainees to engage in the same cognitive processes they will need to use when they graduate from practice to work.

As you look at practice scenarios, you will notice that most require trainees to use judgment in addition to, or instead of, a reflexive response. We are running into the need for high road transfer, where we need to teach more than just ‘how’. We need to also teach ‘why’. We need to teach adaptive expertise and creativity. You could think of this as another higher layer of learning, one that is required above the mechanistic and reflexive response. In reality, the two are on a continuum and cannot be readily separated. Given your desired outcomes for the programme, where on this continuum do your transfer needs fall?

Spoon feeding vs problem solving

One of the things that happens in real life is that we encounter problems that are in some way new to us. So, one thing you can do in your quest to mimic reality is to give people problems to solve before you have trained them how to solve those problems. Give them access to the resources they would have in the real world; in other words, make it real. Provided you ensure that there is sufficient corrective feedback in place, you will find that the learning is often better retained and transfers better to the real world. In addition, as people tackle varying problem scenarios, they will start to develop their own approach and process for solving problems. In many ways, this is what lies at the heart of high road transfer. Solving a problem, with or without support, is always going to be a better training solution than spoon-feeding people information and theory that they then must translate into, and somehow use, in a problem situation they later encounter at work. EM Forster said it best, “Spoon feeding in the long run teaches us nothing but the shape of the spoon”. There is an old saying that in theory, theory works in practice, but in practice, theory never quite works as theory predicts. It reminds me that the military have a saying that no battle plan survives first contact with the enemy.

Some people might say that putting people into a problem scenario without first doing some spoon-feeding will mean that they make mistakes, and it is the mistakes they will remember. This would be the case if you didn’t give them sufficient and supportive corrective feedback and ensure they go through the scenario again and again to practise success. Some difficulty, as opposed to making it too easy to get to a solution, will enhance learning and retention and may also improve motivation to learn in order to solve the problem. Giving people problems to solve may also bring up an interesting situation, in which the trainee is confident at first glance they can solve the problem, but then finds that they can’t. This can happen because they are relying on incorrect information, perhaps because it’s out of date or is incomplete in a way that is now a critical barrier to their success.

Unlearn to learn

An interesting aspect of transfer is that it can be positive or negative. In positive transfer, previous learning facilitates performance and the transfer task, whereas in negative transfer the opposite is the case: previous learning interferes with the transfer task. We are all aware that sometimes we must unlearn things so we can learn new things. In the words of Chester Barnard, “It is what we think we know that keeps us from learning.” Think of the difficulties we have when we first drive in a foreign country on the opposite side of the road to the one we are used to. It is also possible for the method of instruction to create negative transfer or barriers to transfer. This happens when trainers fail to realise the importance of task variation within the classroom. Practising with a single or very limited range of scenarios can encourage a singular response to other scenarios where that response may be inappropriate. Practising on a wider variety of tasks helps the learners to become accustomed to using their newly acquired knowledge and skills in different situations and helps them learn that they will encounter different but similar situations, thus encouraging transfer of learning to the job. In effect, encourage transfer of learning from one practise scenario to another as much as you can, both in the classroom and with activities following the training, so the act of learning transfer itself is learned.

[1] From Howard Gardner’s 1999 book, The Disciplined Mind: What All Students Should Understand

[2] Edward Thorndike

[3] Arthur, Bennett, Stanush, and McNelly reported in Human Performance, (11, 57–101, 1998) on ‘Factors that influence skill decay and retention: A quantitative review and analysis’

Further reading How to reboot training for tangible business impact

Facilitate on-the-job skills application and allow learners to build new habits Learning transfer platform

Learning Transfer at Work