What Is Cognitive Load? How It Affects Learning

What is cognitive load?

Working memory is the small mental workspace that holds information for a few seconds while you use it. We explain what it is, and how little it holds, in What is working memory? Cognitive load is how much of that workspace a task demands at a given moment.

Cognitive load theory is a theory of instructional design built on that limit. In its authors' own summary, all new information is first processed by a working memory that is limited in capacity and duration, and then stored in an unlimited long-term memory; once information is stored there, the capacity and duration limits of working memory disappear when you use it ^[1]. They argue that most earlier advice on designing instruction proceeded as though working memory and long-term memory did not exist ^[1].

The idea began with problem solving. In a 1988 paper, Sweller started from evidence that experts differ from novices mainly in their stored knowledge of problem types, called schemas. He argued that the usual way of solving unfamiliar problems, searching for moves that reduce the difference between where they are and the goal (means-ends analysis), takes up so much processing capacity that little is left for acquiring those schemas; he supported the argument with a computer model and experiments ^[6].

Cognitive load theory is a model of how learning works, so treat it as background. What matters in practice is whether the teaching methods it led to work, which the studies below tested.

Intrinsic, extraneous and germane load

The theory originally described three kinds of load ^[1].

Intrinsic load comes from the material itself: how many elements you must handle at once because they depend on each other, which researchers call element interactivity. It also depends on what you already know ^[1]. The authors' example is the written word "characteristics": for a fluent reader it is a single element pulled from long-term memory, but for someone learning to read English it is many interacting letters that can overwhelm working memory ^[1]. Intrinsic load can only be changed by changing what is to be learned or by the learner gaining expertise ^[1].

Extraneous load comes from how information is presented and what the learner is asked to do, not from the material itself. Unlike intrinsic load, it can be changed by changing the instruction, and most of the theory's instructional effects are explained as reductions in extraneous load ^[1].

Germane load was first described as a third, helpful kind of load devoted to learning. Its authors have since redefined it: it is no longer counted as a load of its own, but as the working-memory resources a learner devotes to the intrinsic part of the task rather than to extraneous demands ^[1]. A 2011 critique argued that, as traditionally used, germane load could not be distinguished from intrinsic load and had been added on theoretical rather than empirical grounds ^[7].

How it differs from working memory

Working memory is the workspace; cognitive load is how much of it a task takes up. The two are linked by knowledge: what counts as one element depends on what you have already learned, so the same material imposes less load on someone who knows more ^[1]. Chunking, recoding small units into larger familiar ones, shows the link. In four laboratory experiments, familiar chunks were recalled better and also improved recall of other information held at the same time, which the authors took as evidence that a chunk reduces the load on working memory ^[8]. The chunking method page explains how to practise it.

Three well-studied effects

Worked examples

A worked example shows a problem with its full solution, step by step, for the learner to study. In a series of five experiments on algebra in 1985, worked examples took much less time to study than the equivalent problems took to solve, and learners who had studied them then solved similar problems faster and with fewer errors ^[9]. The benefits applied only to problems with the same structure as the examples ^[9].

A 2023 meta-analysis of 55 studies in mathematics, from primary school to university, found that worked examples improved performance by a medium amount (g = 0.48) ^[2]. Examples showing correct solutions helped more than examples showing errors, alone or mixed with correct ones ^[2].

The same review contains a conflicting result. Adding prompts that asked learners to explain the examples to themselves produced a negative effect compared with worked examples without prompts ^[2]. Yet a separate meta-analysis of 69 effect sizes found that self-explanation prompts improved learning overall (g = 0.55), across tasks that included studying worked problems ^[10]. The two reviews pooled different studies and comparisons, so whether prompts add anything to a good worked example remains open. The self-explanation method page describes the technique.

Worked examples are not meant to be used forever. The theory's authors describe a sequence that starts with worked examples, moves to partly completed problems, and ends with problems solved from scratch ^[1].

Split attention

Split attention happens when you must mentally combine two separate sources that only make sense together, such as a geometry diagram and solution steps printed elsewhere on the page ^[1]. In six experiments with electrical engineering and biology materials, instructions that built the text into the diagram worked better than conventional layouts over a three-month industrial training course, where the material made no sense without combining the two ^[11]. When the text and diagram could each be understood alone, integrating them made no difference, and adding explanatory material that was not needed could harm learning even when it was integrated ^[11].

Two meta-analyses support the main finding. One, of 50 studies, found that placing related information close together in space or time improved learning, especially for complex material ^[12]. Another, of 58 comparisons with 2,426 participants, found an average benefit of g = 0.63 for integrated designs ^[3]. Linking words and pictures in this way is part of dual coding.

Expertise reversal

The same support does not suit every learner. A 2003 review described what its authors called the expertise reversal effect: techniques that work well for inexperienced learners can lose their benefit, or even become harmful, for more experienced ones ^[13].

In a 2001 study, inexperienced mechanical trade apprentices learned more from studying worked examples than from solving problems, and rated the mental load as lower ^[14]. As the apprentices gained experience, the worked examples became redundant and solving problems became the better way to learn ^[14].

A 2025 meta-analysis pooled 176 effect sizes from 60 experiments with 5,924 participants ^[4]. Learners with little prior knowledge learned more from instruction that gave a lot of help (d = 0.505), while learners with more prior knowledge learned more from instruction that gave less (d = −0.428) ^[4]. The effect was not symmetrical: helping novices mattered more than holding help back from experts, and the evidence was less clear for younger students and for the humanities and language learning ^[4].

Why it matters for learning

If you are teaching yourself something new, the research suggests a few habits:

• Start with examples, then fade them. Study a few worked solutions before solving problems on your own, then move to partly worked problems and finally to full ones ^[1].
• Keep related information together. Put labels on the diagram rather than in a key, and keep an explanation next to the figure it describes ^[3].
• Cut what you do not need. Extra explanation of something that is already clear can hinder learning ^[11].
• Change the method as you improve. Once a topic is familiar, solving problems yourself becomes more useful than studying more examples ^[14][4].

For the study techniques that help you remember what you have learned, see How to improve memory for studying.

How cognitive load is measured, and the main criticisms

Cognitive load is hard to observe directly, and this is where the theory draws most criticism.

Load is often inferred, not measured. In many studies, load is not measured at all: it is inferred from learners' test results ^[5]. The theory's authors acknowledge that many of the experiments behind its main effects did not try to measure load directly; the effects were taken as support for the theory when the predicted learning outcomes appeared ^[1].

The usual measure is one question. The most widely used measure asks learners to rate their mental effort on a single 9-point scale, introduced in 1992 ^[5]. Researchers use it in many forms, with different labels, numbers of points and timing ^[5]. In studies that compared the two approaches, a single rating given once after learning was higher than the average of ratings given after each task ^[1].

What the ratings cannot show. In a 2010 critique, Ton de Jong pointed out that these measures are relative: they compare conditions but cannot show whether working memory was actually overloaded, because no one knows where that threshold lies ^[5]. A single overall rating cannot separate intrinsic from extraneous load, and the common measures are not sensitive to changes over time ^[5].

Explaining results after the fact. Because load can be "good" (germane) or "bad" (extraneous), de Jong argued, almost any result can be fitted to the theory afterwards, which makes it hard to falsify ^[5]. Other critics have said that element interactivity, the theory's concept of complexity, is not defined in a quantitative, measurable way ^[15].

Short, artificial lessons. A substantial share of studies give participants only a few minutes with the material, from about 2.5 to 6 minutes in the examples de Jong cites, often psychology students learning a topic chosen by the researcher ^[5]. Real learners can usually pace themselves, go back and take notes, which many experiments do not allow; de Jong noted that one effect, the advantage of narrating text instead of printing it next to a picture, seemed to hold mainly when presentations were short and paced by the computer ^[5].

Newer questionnaires. Researchers have tried to measure the types of load separately. A 10-item questionnaire developed with statistics students produced the three intended components in 56 PhD students, and the structure was confirmed in three further groups of bachelor students (171, 136 and 148) ^[16]. A 2022 meta-analysis of four such questionnaires found their internal consistency satisfactory; the correlations between the load types partly contradicted what the theory assumes, while their correlations with learning outcomes fitted it ^[17]. The theory's authors maintain that the single rating scale has good psychometric properties and has compared well with physiological measures, while agreeing that it cannot easily separate the types of load ^[1].

What the research says

Frequently asked questions

Is cognitive load a bad thing? Not in itself. Intrinsic load is part of learning anything complex. The theory's advice is to cut extraneous load, so that more of your limited working memory goes to the material itself ^[1].

Should I always learn from worked examples? For a new topic, worked examples helped on average in mathematics ^[2]. As you gain experience, solving problems yourself becomes the better way to learn ^[14][4].

How can I tell if I'm overloaded? There is no reliable measure, even for researchers: rating scales compare conditions, but nobody knows where the overload threshold lies ^[5]. In everyday study you can do what many experiments did not allow: slow down, go back, break the material into smaller parts and take notes ^[5].

References

1. Sweller, J., van Merriënboer, J. J. G., & Paas, F. (2019). Cognitive Architecture and Instructional Design: 20 Years Later. Educational Psychology Review, 31(2), 261–292. https://doi.org/10.1007/s10648-019-09465-5
2. Barbieri, C. A., Miller-Cotto, D., Clerjuste, S. N., & Chawla, K. (2023). A Meta-analysis of the Worked Examples Effect on Mathematics Performance. Educational Psychology Review, 35(1), 11. https://doi.org/10.1007/s10648-023-09745-1
3. Schroeder, N. L., & Cenkci, A. T. (2018). Spatial Contiguity and Spatial Split-Attention Effects in Multimedia Learning Environments: a Meta-Analysis. Educational Psychology Review, 30(3), 679–701. https://doi.org/10.1007/s10648-018-9435-9
4. Tetzlaff, L., Simonsmeier, B., Peters, T., & Brod, G. (2025). A cornerstone of adaptivity – A meta-analysis of the expertise reversal effect. Learning and Instruction, 98, 102142. https://doi.org/10.1016/j.learninstruc.2025.102142
5. de Jong, T. (2010). Cognitive load theory, educational research, and instructional design: some food for thought. Instructional Science, 38(2), 105–134. https://doi.org/10.1007/s11251-009-9110-0
6. Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285. https://doi.org/10.1207/s15516709cog1202_4
7. Kalyuga, S. (2011). Cognitive Load Theory: How Many Types of Load Does It Really Need? Educational Psychology Review, 23(1), 1–19. https://doi.org/10.1007/s10648-010-9150-7
8. Thalmann, M., Souza, A. S., & Oberauer, K. (2019). How does chunking help working memory? Journal of Experimental Psychology: Learning, Memory, and Cognition, 45(1), 37–55. https://doi.org/10.1037/xlm0000578
9. Sweller, J., & Cooper, G. A. (1985). The Use of Worked Examples as a Substitute for Problem Solving in Learning Algebra. Cognition and Instruction, 2(1), 59–89. https://doi.org/10.1207/s1532690xci0201_3
10. Bisra, K., Liu, Q., Nesbit, J. C., Salimi, F., & Winne, P. H. (2018). Inducing self-explanation: A meta-analysis. Educational Psychology Review, 30(3), 703–725. https://doi.org/10.1007/s10648-018-9434-x
11. Chandler, P., & Sweller, J. (1991). Cognitive Load Theory and the Format of Instruction. Cognition and Instruction, 8(4), 293–332. https://doi.org/10.1207/s1532690xci0804_2
12. Ginns, P. (2006). Integrating information: A meta-analysis of the spatial contiguity and temporal contiguity effects. Learning and Instruction, 16(6), 511–525. https://doi.org/10.1016/j.learninstruc.2006.10.001
13. Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The Expertise Reversal Effect. Educational Psychologist, 38(1), 23–31. https://doi.org/10.1207/S15326985EP3801_4
14. Kalyuga, S., Chandler, P., Tuovinen, J., & Sweller, J. (2001). When problem solving is superior to studying worked examples. Journal of Educational Psychology, 93(3), 579–588. https://doi.org/10.1037/0022-0663.93.3.579
15. Karpicke, J. D., & Aue, W. R. (2015). The Testing Effect Is Alive and Well with Complex Materials. Educational Psychology Review, 27(2), 317–326. https://doi.org/10.1007/s10648-015-9309-3
16. Leppink, J., Paas, F., Van der Vleuten, C. P. M., Van Gog, T., & Van Merriënboer, J. J. G. (2013). Development of an instrument for measuring different types of cognitive load. Behavior Research Methods, 45(4), 1058–1072. https://doi.org/10.3758/s13428-013-0334-1
17. Krieglstein, F., Beege, M., Rey, G. D., Ginns, P., Krell, M., & Schneider, S. (2022). A Systematic Meta-analysis of the Reliability and Validity of Subjective Cognitive Load Questionnaires in Experimental Multimedia Learning Research. Educational Psychology Review, 34(4), 2485–2541. https://doi.org/10.1007/s10648-022-09683-4
18. Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students' learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. https://doi.org/10.1177/1529100612453266
19. Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. https://doi.org/10.1037/h0043158