The trouble with transfer: How can we make learning more flexible?

I define learning as the long-term retention of knowledge and skills and the ability to transfer between contexts. The retention bit is fairly straightforward and uncontroversial: if you can’t remember something tomorrow, can you really be said to have learned it? As Kirschner, Sweller & Clark put it, “If nothing has changed in long-term memory, nothing has been learned.”

Transfer though is a bit trickier. In essence it’s the quality of flexibility; can what you know in one context be applied in another? As Daniel Willingham says, “Knowledge is flexible when it can be accessed out of the context in which it was learned and applied in new contexts.”

Memory is context dependent. Because transfer from one context to another is affected by the context of original learning, students can learn in one context, yet fail to transfer to other contexts. One study found that street children could perform complex mathematics when making sales in the street but were unable to answer equivalent problems presented in a school context. In another study, subjects did very well at making supermarket best-buy calculations despite doing poorly on similar paper-and-pencil mathematics problems. What we remember depends, to a greater or lesser extent, on the context in which we both learn and retrieve information as these studies by Weingartner & Smith indicate. 

Back in 1901, Thorndike & Woodworth developed their Principle of Identical Elements which states that transfer depends on the level of similarity between training and performance environments. We haven’t really moved on much since then. So-called ‘far transfer’ between different subject domains – the idea that you could learn the skill of analysis in history and then apply it physics – is much more difficult than is often supposed.  Experts (those who know a lot about a particular subject domain) find it far easier to transfer their knowledge to new contexts than novices (those who know little within a subject domain). This is not to claim that an expert physicist is more able than a physics novice to apply what they have learned in physics to history, just that they will find it more straightforward to transfer their knowledge to other areas of physics. As we’ll see in the next chapter, expertise is highly specific and experts and novices think in qualitatively different ways.

Susan Barnett and Stephen Ceci’s research has laid the building blocks for us to begin to understand the processes involved in transferring material learned from one context to another. In their 2002 paper When and Where Do We Apply What We Learn? A Taxonomy for Far Transfer they identified several factors that affect transfer.

The first three are concerned with the type of content being transferred: whether it is a learned skill, a change in performance or a feat of memory.

Content: what is transferred?

The other six are connected with the context in which the transfer takes place: the knowledge domain (whether learning can be transferred between the domains of say, history and biology), the physical context (from one place to another), temporal contexts (from one time to another), the functional context (between the academic and non-academic settings) and social contexts (between group situations and individual situations) and the modality (whether material can be transferred between speech and writing). In school, students may be able to transfer what they have learned from a history classroom to the examination hall but be unable to transfer the skill of analysing historical source material to analysing literary texts.

Context: when and where transferred to and from

Barnett and Ceci conclude that instances of far transfer are rare, but under the right conditions can happen, and may even turn out to be predictable. However, it is probably more productive for teachers to consider the possibilities of ‘near transfer’. Transfer from one problem to another within a course, from one year in school to another, from the classroom to the examination hall, between school and home, and from school to workplace.

As we’ve seen, our ability to retrieve information is heavily context dependent – we link what we know to related subject matter, times, places, people and feelings. These contextual links provide cues or prompts which helps us to retrieve what we need when we need it. The trouble is, when we learn a thing in one context we rely on environmental cues in order to recall it, when we change the context the absence of those cues can cause us to be unable to retrieve what may have been secure in another location.

Some kinds of near transfer appear quite straightforward. When subjects are asked questions which are closely related to material they have previously exposed to, they are able to transfer what they’ve learned, (Wooldridge et al, 2014) but transfer does not seem to occur when subjects are asked questions on related material taken from the same section of a textbook. The finding seems to be that students have little difficulty in transferring learning when the exact information on which they are quizzed is used in a new situation, but quizzing on related information doesn’t help students to transfer information, even between closely related topics. Unfortunately, what we learn does not spontaneously or automatically generalise to new contexts and so teachers need to facilitate this process (Butler, 2010).

When students were prompted that they would need to transfer related information to novel situations they were far more likely to successfully apply what they had learned in a new context. It seems that prompting students may be a necessary condition for transfer to occur .

While we cannot, of course, predict every conceivable context in which students will need to apply the knowledge and skills they teach, we can prepare students by explicitly informing them that the material being studied will need to be applied to a new context and then provide practice opportunities where they have to apply prior knowledge to novel situations. With luck, this will help students to recognise situations where they can apply what they know when they come up in the future.

We also need to recognise that the transfer of knowledge or skills to a novel problem requires both knowledge of the problem’s context and a deep understanding of the problem’s underlying structure. Experts and novices think in qualitatively different ways and what may be obvious to an expert may make little sense to a novice. Seeing underlying structures can be facilitated by using concrete examples. For instance, it’s hard to understand the structure of this problem because it’s so abstract:

The Wason Card Test

But, when the problem is presented differently, with a more familiar context, it’s much easier to understand:

Cosmides & Tooby (2005)

The lesson for teachers is that the abstract example which makes perfect sense to us, may be meaningless for our students. We can, however, help students to see the underlying structures by presenting them with a concrete example that shares similar features.

Transfer between contexts is hard and, as I’ve explained before, students might not necessarily be able to transfer from the context of one seat to another seat in the same classroom! It’s worth knowing that we can weaken students’ dependency on context by varying the conditions in which ideas are encoded and retrieved. We can also facilitate transfer by explicitly telling students that they will need to retrieve what they’re studying in a new context. Using concrete examples will also help. But probably the most important thing for teachers to know is that the more you know, the easier it is to transfer what you know to new contexts.