“Transfer” is sometimes seen as the holy grail of education. It occurs when people apply information, strategies and skills they have learned in one domain to a new situation or context.

It is not sensible to deny that transfer will ever happen as some learning situations are more or less similar to others. For example, when I step into a new car, the knowledge and skill I have of driving a different car can be used in that new car. This is called “near transfer”.

What people tend to mean when they say that transfer seldom happens is that “far transfer” seldom happens: this is a situation in which the knowledge and skills of a new domain have less to do with previously learned knowledge. So, for example, training mental shape-rotation skills might help with better performance on a spatial test, but does it also help with other mathematical topics or even other subjects?

Such far transfer seems elusive. As long ago as 1901, Edward Thorndike and Robert Woodworth showed that transfer depends on the similarity of the situations or domains. More recently, Sala and colleagues (2019) summarised research on transfer in different types of cognitive training. In working-memory training, there was near transfer, but far-transfer effects were small or null, regardless of the type of population and cognitive-training programme.

So, if there isn’t a lot of evidence for far transfer, should we just let it go?

I think that would be a mistake. Integrating new knowledge into our existing schemata is not always straightforward. Researchers such as Chi (Chi and Brem, 2009) and Ohlsson respectively have proposed general abstract categories of knowledge. For Ohlsson, the brain forms general knowledge through comparison of similarities between the existing body of knowledge and new knowledge. For Chi, the recognition of differences between knowledge enables us to create a general body of knowledge. In short, we integrate new information through recognising similarities and differences.

Meanwhile, Taatgen (2013) tabulates several studies showing transfer in dissimilar tasks.

Even when it comes to something recently very popular, such as the “testing effect”, there are strategies where transfer of learning is greater with more general strategies than very domain-specific strategies like using worked examples (Pan and Rickard, 2018).

The lesson here is that we need to be careful in assuming that knowledge and skills will easily transfer to other domains, but rather than decrying skills as totally domain-specific and non-transferable, we could look at opportunities to maximise transfer by looking at commonalities between domains.

This article originally appeared in the 19 June 2020 issue under the headline “Is transfer of learning an impossible dream?”