Since the 1980s, instructional techniques have been dominated by a compelling argument: “Look at the difficulty many people have when learning in classrooms. Compare that difficulty with the ease with which people learn outside of the classroom. If we switch our classroom procedures to more closely match those of the natural world, learning will be enhanced.”

This argument has swept all before it.

As one example, whole-language learning was introduced when teaching children to read. Using whole-language learning, students are presented with entire words and sentences without the use of soundletter correspondences. The “natural world” rationale is that since children do not need to be taught basic sounds when they learn to listen to their first language, they should not need sound-letter correspondences when they learn to read.

Similarly, when learning a foreign language in secondary and tertiary institutions, we have been urged to use immersion in the foreign language, with little or no emphasis on grammar and vocabulary. In line with the natural world argument, immersion is more natural and works when learning a first language, so it should work when learning a foreign language.

And the argument has also has been used to justify inquiry- and problem-based learning, which have become dominant in most disciplines, especially mathematics and science. Learners are presented with problems to solve without first being shown how to solve those problems and sometimes even without being presented with requisite basic knowledge.

Why? We solve problems in the natural world without learning the basic structures that underlie those problems, so we should be able to learn curriculum-based problem solving just as easily by “natural” immersion in a problem-solving environment.

The arguments seemed plausible. Students need to construct knowledge for themselves, it was argued, as they do in the natural world, rather than having it presented by teachers. A lack of supporting data was treated as a minor detail.

Different kinds of knowledge

Thankfully, something else occurred during this period: we were steadily gaining knowledge of human cognitive architecture - how we learn, think and solve problems.

Based on that knowledge, some of our instructional procedures, as summarised above, began to look increasingly bizarre. Cognitive load theory, which uses our knowledge of human cognition to devise instructional procedures, provides very different recommendations for teachers.

Humans do learn differently inside and outside of class, but that is because of the different categories of knowledge acquired in the two environments.

David Geary provided the key to understanding this difference by assuming a distinction between knowledge that we have evolved over countless generations to instinctively acquire and non-instinctive knowledge that we have not specifically evolved to acquire.

Instinctive knowledge is acquired automatically and unconsciously. Examples include routine social interactions, recognising faces, using general problem-solving strategies such as generalising from one problem solution to another and, of course, learning to listen to and speak our native language.

Most of these tasks are immensely complex: it takes artificial intelligence programmers years to duplicate them. But they are not complex for us. We acquire the necessary knowledge effortlessly without tuition because we have evolved to acquire it.

In contrast, non-instinctive skills are important for cultural reasons. We are able to acquire them but have not specifically evolved to do so. They require conscious effort from learners and explicit tuition from teachers.

Processing information

Almost everything that is taught in educational institutions is non-instinctive, “artificial” and absent from the natural world. We invented education specifically in order to teach non-instinctive skills, such as reading and writing.

Without education, few people learn to read and write, in marked contrast to listening and speaking. For thousands of years after the invention of writing, most people could not read or write. It was only after the advent of modern mass education that literacy became widespread. The reason almost everyone could listen and speak while few could read and write was not because listening and speaking were “taught” appropriately while reading and writing were not. Rather, it was because we have specifically evolved to acquire listening and speaking skills, but not reading and writing skills. They belong to different categories of knowledge.

To function in the modern world, we need a huge store of non-instinctive knowledge, held in long-term memory. The bulk of that information is obtained from other people. We have evolved to obtain information from other people and do so instinctively even when the information is non-instinctive.

Instruction needs to be designed to facilitate the accumulation of information in long-term memory, which is the major goal of education.

While most information stored in long-term memory is obtained from others, sometimes that route is unavailable. When that is the case, we can attempt to generate it ourselves during problem solving.

Problem-solving moves are generated randomly and tested for effectiveness. While it is possible to obtain information during problem solving using these random procedures, it is a secondary option. Wherever possible, we will attempt to obtain information from other people. It is far more effective and efficient than problem solving.

The process of generating non-instinctive knowledge during problem solving is instinctive and so does not need to be taught.

Novel non-instinctive information, obtained either from others or generated during problem-solving to be stored in long-term memory, must first be processed in a capacity- and duration-limited working memory. Working memory can be considered as consciousness. We are conscious only of what is in working memory and unconscious of most of the contents of long-term memory.

When dealing with novel non-instinctive information, working memory can process no more than three to four elements of information and hold that information for no more than about 20 seconds without rehearsal. Beyond these limits, working memory ceases to function effectively.

Instructional recommendations that ignore the limitations of working memory when learners process novel non-instinctive information are likely to be random in their effectiveness.

The limitations of working memory that apply to novel information, disappear when dealing with familiar information already stored in long-term memory.

There are no known capacity or duration limits when working memory deals with stored information transferred from long-term memory.

The transformative effects of education occur through this cognitive architecture. Once information is processed in working memory and stored in long-term memory, it can be transferred back to working memory, allowing us to take action of which otherwise we could only dream. We are transformed.

Cognitive load theory uses the above architecture to generate instructional procedures. Here are a few examples:

1. Problem-solving

Non-instinctive knowledge should be taught explicitly. As discussed, we have evolved to present information to other people and to obtain information from others.

Inquiry-based learning is used widely, but requiring learners to obtain information using an inquiry-based discovery process when they could be explicitly presented the same information is pointless, despite its effectiveness in acquiring instinctive, natural world knowledge.

Learners initially should be shown how to solve specific categories of problems - rather than being asked to work out how to solve the problems themselves. They should practise solving problems themselves only after they have been shown how to solve them. Futile searches for problem solutions during problem-based learning impose a heavy working memory load with no discernible advantage.

2. Foreign languages and subject content

The current trend to teach students subject content using a foreign language is problematic. Learning a foreign language and subject content simultaneously should be avoided. Because we can learn content while instinctively acquiring a native language as children, it does not mean that procedure should be used as a template for secondary or tertiary students.

Learning a native language is instinctive, while learning a foreign language as an adolescent or adult is not instinctive. The working memory load of learning a new language at the same time as learning a content area is likely to overwhelm working memory.

Learning a language and learning content should be kept separate until sufficient levels of foreign language competency are attained to allow limited working memory resources to be devoted to content rather than language.

Learning a foreign language is a redundant activity when attempting to learn content and vice-versa. We should always attempt to eliminate redundant information.

3. Spoken and written information

The relations between two or more sources of non-instinctive information are critical to instructional design.

Presenting the same information in spoken and written forms results in attempts to process both forms in working memory simultaneously, unnecessarily increasing working memory load. We have found repeatedly that more is learned by using only one form of presentation. The addition of any unnecessary visual or auditory information, including background music, has been shown to decrease learning.

If a diagram and text are both needed to explain a concept, neither is redundant and so neither can be eliminated. In this instance, they should be physically integrated so that learners do not have to mentally integrate them.

Place text in appropriate locations on the diagram rather than next to the diagram, or run arrows between relevant text and diagram locations.

Instead of writing “Angle abc = Angle xyz…” underneath a geometry diagram, place the statement in an appropriate location on the diagram or connect the statement to the appropriate angles using arrows. Searching for referents in a diagram rather than having them physically connected wastes limited working memory resources.

Working memory resources can be increased by using both auditory and visual components of working memory by presenting one piece of information orally and the other visually as a diagram on the page.

While learners are looking at the diagram, it may be better to speak rather than writing “Angle abc = Angle xyz…”. Presenting the information through both visual and auditory channels can increase working memory capacity and facilitate learning, assuming both sources of information are essential and not redundant.

4. Permanence vs transience

While using both auditory and visual components of working memory may increase capacity, we need to remember that written information is permanent, while spoken information is transient and immediately disappears unless it is recorded.

Lengthy, complex non-instinctive information may be difficult to understand because what is presented earlier in a lesson may be needed in order to understand what is presented later. But when presented in spoken form, that information is transient and has disappeared.

In written form, learners can repeatedly return to required information, reducing the working memory load.

Complex information needs to be presented in written form or divided into small chunks when it is presented in spoken form.

There are many other instructional procedures that flow from cognitive load theory. All procedures based on the theory are supported by multiple randomised-controlled trials. For those concerned with instructional issues, knowledge of human cognitive architecture and the instructional procedures that flow from that knowledge should no longer be optional extras. They are essential.

John Sweller is emeritus professor of education psychology at the University of New South Wales, Sydney, Australia