Cognitive load theory holds that working memory is extremely limited in both capacity and duration, and that learning breaks down when it is overloaded. Developed by John Sweller in the late 1980s out of a study of problem solving, the theory aims to provide guidelines for presenting information in ways that encourage learner activities optimising intellectual performance. Its practical core is a simple design rule: eliminate extraneous load, manage intrinsic load by sequencing, and protect room for germane load.
What Cognitive Load Theory Actually Claims
Sweller's foundational paper on cognitive load during problem solving was published in 1988 in Cognitive Science. What began as an investigation into how learners solve problems became a broader framework for instructional design. Cognitive load theory deals with the types of load in human working memory and how to manipulate instructional designs so that working memory is available for learning.
Two assumptions do most of the work. The first is a limited-capacity working memory with partially independent subcomponents for auditory/verbal and visual information. The second is an effectively unlimited long-term memory holding schemas that vary in their degree of automation. Because the second store is generous and the first is not, instruction succeeds or fails largely on how well it respects the bottleneck between them.
Why Working Memory Is the Bottleneck
Working memory is extremely limited in both capacity and duration, and these limitations will, under some conditions, impede learning. The theory uses the schema as the primary unit of analysis for designing instructional materials, because a schema is what lets a learner treat many separate elements as one familiar chunk. Once a schema is automated in long-term memory, it consumes far less working memory, which is why an expert and a novice can look at the same problem and experience completely different mental effort.
Capacity figures deserve caution. Miller's 1956 paper suggested human short-term memory holds about seven units, plus or minus two, but that historical estimate is a rough guide rather than a current specification. Treat "about seven chunks" as a piece of history, not a design parameter.
Intrinsic, Extraneous and Germane Load Explained
Cognitive load theory distinguishes three types of load: intrinsic, extraneous and germane. Each names a different source of mental effort.
| Load type | What it refers to | Simple example |
|---|---|---|
| Intrinsic | The effort associated with a specific topic | Solving a multi-step equation with several interacting variables |
| Extraneous | The way information or tasks are presented to a learner | A diagram on one page and its labels on another, forcing the reader to hold one in mind while finding the other |
| Germane | The work put into creating a permanent store of knowledge, a schema | Deliberately comparing two worked solutions to extract the underlying rule |
The design rule that follows is to eliminate extraneous load, manage intrinsic load by sequencing, and protect room for germane load. In practice, that usually means cutting presentation clutter first, since it is the easiest to remove and the least useful to the learner.
Do the Three Loads Add Up?
For years the three types were treated as separable quantities that could be summed. The additivity of these load types has been investigated and questioned, and they are now thought to influence each other circularly. A review notes it is still not clear whether intrinsic, extraneous and germane load are independent and can be added into an overall measure of load.
One reformulation treats intrinsic and extraneous load as the two primary additive types and redefines germane load in terms of working memory resources devoted to intrinsic load. This matters for anyone applying the theory: you should not expect to measure three separate loads, subtract the bad ones, and read off a precise number. The categories are more useful as diagnostic questions than as arithmetic.
What the Theory Recommends in Practice
Sweller studied learners solving problems and found they often used means-ends analysis, a strategy that consumes processing capacity that might otherwise go to schema construction. That finding drove the instructional recommendations.
Sweller proposed that instructional designers avoid unnecessary load by using materials such as worked examples and goal-free problems instead of conventional problem solving. Empirical work in the 1990s demonstrated several learning effects, including the completion-problem, modality, split-attention, worked-example and expertise reversal effects. Each describes a recurring mismatch between how material is presented and how working memory actually handles it.
The Expertise Reversal Effect
Support that helps novices, such as worked examples, can start to harm learners as their expertise grows, which researchers call the expertise reversal effect. A fully worked example gives a novice a clear path through the problem. The same example gives a more experienced learner redundant information to reconcile with what they already know. Guidance should therefore fade as competence builds, rather than staying constant across a course.
Where the Theory Is Contested
Cognitive load theory has been criticised for conceptual clarity and methodological approaches. The measurement problem is concrete. Commonly used measures of cognitive load have been found to respond in different ways to extraneous, intrinsic and germane load, so a single instrument may not track what you think it tracks. A 2020 study suggested extraneous cognitive load may consist of various demand components that may need different questionnaires to measure.
This does not make the theory unusable. It means the load categories are best treated as hypotheses about where effort is going, tested against what learners actually do, rather than as settled measurements.
A Practical Checklist for Lessons and Materials
Use these questions when building a lesson, slide deck, course module or help page. They follow the theory's core recommendations without requiring a formal measurement of load.
- Cut extraneous load first. Remove decorative elements, duplicated explanations and competing channels that force learners to split attention.
- Sequence intrinsic load deliberately. Introduce interacting elements in a workable order rather than all at once.
- Start novices with worked examples. Give a complete solution path before asking for independent problem solving.
- Fade guidance as expertise grows. Withdraw scaffolding for learners who no longer need it, in line with the expertise reversal effect.
- Protect room for schema building. Leave enough mental space for learners to consolidate patterns rather than just complete tasks.
Conclusion
Cognitive load theory explains why good instruction is largely a matter of protecting a small, temporary workspace: working memory is limited in capacity and duration, and learning breaks down when it is overloaded. Its design rule is to cut extraneous load, sequence intrinsic load, and leave room for schema building. The theory's categories are contested and hard to measure precisely, but its central claim about the working memory bottleneck remains the most useful lens for diagnosing why a lesson or a help page feels harder than it should. For readers interested in adjacent ideas, attention residue and context switching describe related costs that persist when people move between tasks.