Why cognitive load may be less important than you think

Cognitive load theory (CLT) is “the single most important thing for teachers to know”, according to Emeritus Professor Dylan Wiliam - and this is a message that proponents of CLT love to flaunt.

Coined by John Sweller in 1988, CLT states that the mental effort required for learning tasks affects the ability of short-term memory to process new information and of long-term memory to store information.

This assumption is based on much older research by George A Miller, which found that working memory cannot process too much information at once. There was also a 2010 follow-up study by Nelson Cowan.

CLT has since become a core component of the “evidence-informed” approaches to teaching and learning that have exploded in recent years, and that increasingly emphasise the importance of memory.

But when we talk about “CLT” in education, what exactly do we mean? It’s important for teachers to understand that the theory has expanded considerably over the years.

In the 1980s, John Sweller laid the foundations of CLT from experiments before publishing the 1998 paper (with colleagues Van Merriёnboer and Paas) in which he set out how his theory. This changed our understanding of cognitive architecture and what the implications were for instruction.

This model suggests there are three types of cognitive load: intrinsic (inherent to the learning task), extrinsic (irrelevant) and so-called germane (useful) load. Some teachers and researchers still refer to this version.

But Sweller took his ideas further and in the same year that he published a popular article, written with colleagues Paul Kirschner and Richard Clark, he proposed a fundamental change to the existing theory.

Evolutionary explanations

In the earlier version of his theory, Sweller had been struggling with a question: how is it possible that children can learn something without explicit instruction?

In 2006, with the help of his wife, biologist Susan Sweller, he hit upon what he thought was the answer, in the work of developmental psychologist David Geary.

Geary proposed that there are two types of knowledge: biologically primary (which we store without conscious processing) and biologically secondary (for which we have to actively learn).

Today these evolutionary principles are presented as the “foundation” of CLT.

However, it is not always clear whether a learning task requires biologically primary or secondary knowledge. If something turns out to be easy to learn, it is often claimed to be biologically primary knowledge. The reasoning is circular: biologically primary knowledge is easy to learn, so what you learn easily must therefore be primary knowledge.

Several publications have argued that such knowledge can’t really be taught, because the researchers were not able to find evidence that it was possible (though some would say they didn’t look very hard).

Furthermore, evolutionary explanations are always difficult to verify because they rely on speculation about thousands of years of development and the multiple possible reasons why we evolved in a certain way.

Measurement challenges

And the complications don’t stop there. When talking about CLT, we also need to consider exactly how we measure the magnitude of a student’s mental load. What is too much? What is just right?

To assess this, we rely on teachers’ expertise and experience. But in a serious theory, concepts must be clearly defined and we must also have a clear way of measuring those concepts.

For many years, the most popular way of measuring load in CLT has been self-reporting, using a nine-point‑scale developed by Paas, which asks students to report the mental effort they are investing into a particular learning task. The scale is said to be easy to use and sensitive to changes in load, but it has also received plenty of criticism.

For example, students may not know how much strain they are actually experiencing. We also don’t know whether the perceived load should be high or low for the best performance, or somewhere in the middle.

Finally, we need to think about the point at which we take the measurement. Do we measure cognitive load during an activity or after it?

Some say that none of this matters: as long as CLT makes the right predictions, it doesn’t matter how you measure it. But that makes it tricky to use the theory to offer concrete recommendations to teachers.

Recent developments have included eye tracking and brain scans, but it is disappointing that after almost 30 years, load measurement is still a challenge.

A web of effects

As if CLT wasn’t complex enough by itself, over the decades all kinds of effects have been added to it.

For example, the “split attention” effect suggests that students will learn less effectively if they are asked to refer to different sources (text and image, for example) and that learning can be maximised if these sources are combined.

Meanwhile, the “self-management” effect suggests that students can be taught strategies to reduce extrinsic load themselves.

Effects can even influence and contradict each other. For example, the “modality effect” emphasises that it is best to combine visual and audio sources, while the “transient information” effect could be used to argue against the use of audio.

A 2019 review even referred to what it called “compound effects” that alter existing effects, like those mentioned above.

The web of potential effects allows CLT to go in many directions, meaning the theory can always be right in some way. And although Sweller presents this as a strength, there is a strong risk of the theory just expanding endlessly, to assimilate any findings that contradict it.

What does this mean for the future of CLT?

So, what does all of this mean for schools?

In general, it is useful for teachers to consider the potential cognitive load imposed on their students by learning tasks.

However, this should only ever be considered a “rule of thumb” because the theory does not give precise pointers about the best way to teach a topic.

What’s more, if a teacher reads about an effect related to CLT, they should take the time to consider how this might interact with other effects.

From a scientific point of view, it remains to be seen whether the integration of Geary’s research makes the theory more robust. The version from 1998 seemed more focused on recommendations for teaching and learning, with a central role for building schemas and productive load.

Perhaps most importantly, further additions and effects, as well as the method of measurement, should be strengthened before the theory continues to expand. Otherwise it will simply become like the Very Hungry Caterpillar, growing ever larger by swallowing up everything in its path.

In the meantime, teachers should just do what they’ve been doing for decades: provide their best teaching.

Christian Bokhove is professor in mathematics education at the Southampton Education School, University of Southampton, UK