Cognitive Load Theory in Instructional Design: What Fits in a Lesson

“How long should each lesson be?” is one of the first questions people ask when they plan an online course, and the internet is full of confident answers: six minutes, ten minutes, one idea per video. None of those numbers comes from cognitive load theory, the best-supported account we have of why some lessons overwhelm learners and others don’t.

What the theory offers instead is a way to judge it. The limit on a lesson isn’t time; it’s how many new, interacting pieces of information a learner has to hold in mind at once, and how much of their attention your design wastes. This entry explains what the research says, how to apply it when you cut a course into lessons, and what the main course platforms let you do about it.

What the research says

Cognitive load theory was developed by John Sweller and colleagues. Its most recent full restatement, written twenty years after their landmark 1998 paper, is the place to start (Sweller, van Merriënboer & Paas, 2019).

A small, forgetful workspace for anything new

The theory rests on one idea about human memory. All new information is processed first in working memory (the mental workspace where you hold and manipulate what you’re attending to right now), and working memory is limited in both how much it can hold and how long it holds it. Those limits apply to novel information. Once something is stored in long-term memory and organised into what psychologists call schemas (connected bundles of knowledge that can be handled as a single unit), the limits largely fall away. Sweller and colleagues put it plainly: once information is in long-term memory, the capacity and duration limits of working memory “disappear” (Sweller et al., 2019).

That explains something every course creator has noticed. An expert can watch a dense twenty-minute walkthrough and follow it easily; a beginner watching the same video is lost by minute four. The video didn’t change. The amount of it that counts as new did.

Two kinds of load you can work with

The theory separates the load a lesson places on working memory into two basic kinds.

• Intrinsic load comes from the complexity of the material itself, for this learner. The key idea is element interactivity: how many pieces of information must be understood together. A list of keyboard shortcuts has low element interactivity, because each one can be learned on its own. Reading a balance sheet has high element interactivity, because no line makes sense without the others.
• Extraneous load comes from how the material is presented and what the instruction asks learners to do. It isn’t caused by the subject, so a designer can reduce it.

The distinction matters because the two respond to different fixes.

In practice, that leaves a course designer with three levers. You can reduce extraneous load through better presentation; you can manage intrinsic load by changing what a single lesson asks learners to take on at once; and you can build learners’ expertise, so that the same material carries less load next time.

What happened to germane load

Older explanations of the theory, including many of the pages that rank for it, describe a third type, germane load: the effort that goes into learning itself. The 2019 restatement revises this. The authors now treat germane load not as a third quantity that adds to the total, but as a redistribution of working memory resources towards the material that has to be learned, and they describe this as a departure from the 1998 paper (Sweller et al., 2019). For a course designer, the practical message is the same: reduce wasted effort, so more of a learner’s limited attention goes to the material that matters.

The effects with the strongest support

Cognitive load theory has generated a long list of named instructional effects. Four matter most for online lessons.

Split attention. When learners must mentally combine two sources of information that only make sense together, such as a diagram and its explanation in separate places, they spend working memory on the combining. A meta-analysis by Schroeder and Cenkci found a medium-sized advantage for integrated layouts over separated ones (g = 0.63, across 58 comparisons involving 2,426 learners), with a larger benefit for static illustrations than for interactive ones (Schroeder & Cenkci, 2018, as summarised by the Clearing House Unterricht at TU Munich).

Redundancy. Presenting the same information twice in different forms, for example narration that reads aloud the text on screen, can hurt learning rather than help it, because learners have to process both. In the early experiments Sweller and colleagues describe, presenting the diagram alone was better than presenting the diagram and the same information in text (Sweller et al., 2019).

Transient information. Spoken explanations and animations disappear as soon as they’re shown, so learners must hold them in working memory while the next part arrives. The 2019 review names self-pacing and segmentation as ways to compensate (Sweller et al., 2019). This is the effect most relevant to video-based courses.

Segmenting. This is the closest the research comes to answering “how much belongs in one lesson?” A meta-analysis of 56 investigations, including 88 comparisons, found a reliable segmenting effect, with small to medium benefits for both retention and transfer. Segmentation also reduced overall cognitive load and increased learning time, and the effects held when the system, rather than the learner, set the breaks (Rey et al., 2019).

One finding in that meta-analysis is worth noting because it cuts against a simple reading of the theory: learners with high prior knowledge benefited more from segmenting than those with little or none, at least on retention. Segmenting, it seems, isn’t only a crutch for beginners.

When help stops helping

The final piece is the expertise reversal effect. Techniques that reduce load for beginners, such as fully worked examples or tightly integrated explanations, can lose their effectiveness and even have negative consequences for more experienced learners (Kalyuga, Ayres, Chandler & Sweller, 2003). For an online course with a mixed audience, that’s a design problem, not a footnote.

How strong is the evidence?

Strong, for the core effects. Split attention and segmenting have each been pooled in the meta-analyses above, and the 2019 review gathers decades of experiments behind the others. Two limits are worth keeping in mind. Effects found with one kind of material and learner don’t always carry over to another, as the expertise reversal effect shows, so check who was studied before relying on a single experiment. And the theory explains why a lesson overloads learners; it doesn’t hand you a number of minutes, slides or ideas per lesson. Anyone who quotes one as “what cognitive load theory says” has gone beyond the evidence.

How to apply it in a course

The question to ask of every lesson is not “how long is it?” but “how many new things does the learner have to hold together here, and what am I adding that they don’t need?” Here’s how that plays out when you plan a module.

We’ll use one example throughout: a course teaching small-business owners to read their own profit and loss statement. It’s an invented example, but it’s typical of the material independent creators teach: a handful of ideas that only make sense in relation to each other.

1. List the new elements in each lesson

Draft the module as a single long lesson, then list every idea a beginner hasn’t met before. For the profit and loss module that might be: revenue, cost of goods sold, gross profit, operating expenses, net profit, and how each line is calculated from the one above.

Mark which elements can be learned on their own and which only make sense together. “Revenue” can be taught alone. “Gross profit” can’t be understood without revenue and cost of goods sold. The cluster that must be held together is your minimum lesson: you can’t split it further without breaking the idea.

2. Cut at the seams, not the clock

Make each lesson one coherent cluster, and stop. Our module might become three lessons:

1. What revenue and cost of goods sold are, with examples from the learner’s kind of business.
2. Gross profit: how the first two combine, worked through on a real statement.
3. Operating expenses and net profit, then reading a whole statement top to bottom.

Some will run four minutes and some twelve. That’s fine. The segmenting research supports meaningful breaks; a cut in the middle of an explanation just to hit a target length adds load rather than removing it.

3. Pre-teach the parts that can stand alone

If one lesson is still too dense, move the low-interactivity material out of it. Terms, definitions and the names of things can be taught and practised first, in a short lesson or a glossary with a quick quiz. When learners arrive at the complex lesson already knowing what “cost of goods sold” means, that term no longer competes for working memory.

4. Strip out extraneous load

Go through each lesson and remove effort that doesn’t serve learning:

• Put explanations next to what they explain. Label the parts of the statement directly on the image, or narrate over it, rather than showing the statement on one slide and the explanation on the next.
• Don’t read your slides aloud. If you’re narrating, keep on-screen text to labels and key terms. If learners need the full text, offer it as a separate transcript.
• Cut the decoration. Background music, animated transitions and stock images that don’t carry information all compete for attention.

5. Hand control to the learner

Video is transient, which makes it the format most likely to overload. Give learners ways to take it at their own pace: chapter points or short separate videos, a clear pause after each step, and the key frame or worked example available as a still image or PDF they can look at while they practise.

6. Fade the support as learners progress

Early lessons can lean on fully worked examples: here is a statement, here is each step. Later lessons should give partially worked examples, then problems learners solve alone. For a mixed audience, open each module with a short diagnostic quiz, and tell learners who pass it they can skip to the practice.

This is design work, not production work. It belongs in the design phase of the ADDIE model, before you record anything, and it’s one of the checks in our online course design checklist.

How the platforms support it

Every course platform lets you break a course into modules and lessons, so segmenting itself is never the problem. The differences are in how much control learners get over pace, and whether you can put quick checks between segments. We haven’t yet published our reviews of Thinkific, LearnWorlds, Teachable and Kajabi, and plan features change often, so check these three things in your platform’s current documentation, on the plan you’re paying for:

• Quizzes between lessons. Can you place a short quiz anywhere in a module, not just at the end?
• Gated progress. Can you require a lesson or quiz to be completed before the next one opens?
• Pause points inside video. Can you add questions or stops inside a video, rather than splitting it into separate files?

Whatever the platform, check two things in your own course: that learners can pause, rewind and scrub video freely, and that you can see where in a lesson they stop watching. A lesson where many learners drop off at the same point is often a lesson that asked them to hold too much at once.

Further reading

• Sweller, van Merriënboer & Paas (2019), “Cognitive architecture and instructional design: 20 years later”. The authors’ own restatement of the theory, including the revised view of germane load and the newer effects. Long, but the most authoritative single source. Open access.
• Centre for Education Statistics and Evaluation (2017), “Cognitive load theory: Research that teachers really need to understand”. A short, clear literature review from the NSW Department of Education, written for teachers. The best plain-English introduction we know of.
• Rey et al. (2019), “A meta-analysis of the segmenting effect”. The evidence behind breaking lessons into learner-paced pieces, including what moderates the effect.
• Kalyuga, Ayres, Chandler & Sweller (2003), “The expertise reversal effect”. Why a course built for beginners can hold back learners who already know the basics.
• Schroeder & Cenkci (2018), “Spatial contiguity and spatial split-attention effects in multimedia learning environments: a meta-analysis”. The case for putting explanations next to what they explain, with the conditions where it matters most.

For the rest of the principles behind good course design, browse our research library.