These days, teachers can be bombarded by “evidence”.

The field of cognitive science, in particular, is seen by some as offering a general framework for “how we learn” that can support all teachers to improve their practice.

But to what extent are such principles really “general”? Surely much of the knowledge that we gain in school is subject-specific. So when it comes to applying evidence, how far can we disregard school subjects?

The answer to this is not straightforward and there are several factors we need to consider. Firstly, we can look at the role of prior knowledge. We know the predictive power of prior knowledge for learning is low. A recent meta-analysis, which I flagged up in a previous column, shows that prior knowledge is not always as helpful as we might think when you bring subjects into the mix.

In fact, some prior knowledge can have negative or negligible effects. For example, you might learn to avoid using certain units in mathematics, but then be required to use those same units in science, which can confuse students.

Secondly, we need to consider transfer of learning - the extent to which learning in one subject can be applied in others.

“Far transfer” is the holy grail here - where a pupil learns something about, say, the concept of logic, and then applies this understanding in many different subjects. Far transfer is not easy to come by.

“Near transfer”, in which one subject helps with an associated subject, is still hard but it happens more commonly.

Finally, when looking at research, we need to be aware that some school subjects are simply less popular with researchers than others. It is notable that, even if the effects of an approach seem to be quite robust - as in Agarwal’s recent review of retrieval practice, or the Education Endowment Foundation’s review of applying cognitive science in the classroom - the research into those effects is often scarce in some subject areas, especially non-science subjects.

It might be that something that works in science also works in art, but we nevertheless need to be careful with making generalisations across all subjects - especially given that there are challenges around the best ways to learn in a particular subject.

Consider learning how to play a musical instrument. This may require a completely different approach to learning how to solve an equation in mathematics, which, in turn, might require a different approach to learning about the causes of the Second World War.

The problem is that some commentators aren’t prepared to fully acknowledge this. They want to make claims about the best ways to acquire skills and knowledge overall.

It is not a given that principles of teaching should work because of some general idea of the universal cognitive architecture that underpins how we learn.

What’s more, even if we adopt such frameworks, we can’t escape the fact that the knowledge we already possess will play an important role in any further learning.

What does this mean for evidence-informed teaching? Quite simply, it means the most suitable person to determine how best to apply insights from evidence in combination with best practice in a given subject will always be the subject teacher themselves.

Christian Bokhove is associate professor in maths education at the University of Southampton and a specialist in research methodologies

This article originally appeared in the 20 August 2021 issue under the headline “What’s the best way to learn? That depends”