Context cardTransitional topic
Cognitive load and split attention: the three types of load, the split-attention effect and what they mean for documentation
What is cognitive load, how does the split-attention effect relate to it, and which design decisions in technical documentation follow from the two?
The short answer
Cognitive load is the demand that processing information places on a reader's limited working memory; cognitive load theory, developed by John Sweller from 1988, distinguishes intrinsic load (the complexity of the content for this reader), extraneous load (load caused by presentation) and, in its classic form, germane load (effort that builds lasting knowledge). The split-attention effect is one specific, avoidable source of extraneous load: when a figure and a text only make sense together but are presented apart, readers spend working memory searching and matching them (Chandler and Sweller, 1991). In documentation this supports integrated figures and text, no needless repetition, worked examples for novices and layered information for experts — always checked with real readers and never at the expense of required safety information.
For: Technical writers, illustrators, e-learning authors and information designers who combine text, graphics and media
Key points
- Cognitive load theory rests on one premise: working memory is narrow for new information, while long-term memory holds schemas that let experts treat many elements as one.
- Intrinsic load comes from how many elements of the content interact for a given reader; extraneous load comes from the way information is presented; germane load, the effort spent on building schemas, is a contested category that Sweller himself recast in 2010.
- Cognitive load is the overarching framework; split attention is one concrete design problem within it — a cause of extraneous load, not a synonym for it.
- Split attention only arises when the separated sources are unintelligible on their own; for a self-explanatory figure, adding the same information as text creates the opposite problem, redundancy.
- Related effects shape the remedies: modality (spoken words with a picture), worked examples, segmenting and sequencing, and expertise reversal, which turns novice aids into a burden for experts.
- For documentation: labels on the figure, the figure at the step, one source per piece of information, worked examples that fade into tasks, and layered depth for different readers.
- No load argument removes required content: repeated safety messages at the point of hazard are not redundancy, and no cognitive principle replaces the risk assessment or makes a document compliant.
The context
What cognitive load is
Readers process new information in working memory, which holds only a few elements at a time and loses them quickly. Long-term memory has no known limit and stores knowledge as schemas: once a reader has learned what a pneumatic cylinder is, the whole idea occupies one element in working memory instead of many. Cognitive load theory, developed by John Sweller from his 1988 study of problem solving, asks how instruction uses this narrow channel. Cognitive load is the total demand that processing a piece of information places on working memory at a given moment.
Three types of load
| Type | Where it comes from | What design can do | Example in documentation |
|---|---|---|---|
| Intrinsic load | The number of elements that interact in the content, relative to what the reader already knows | It cannot be designed away, only managed: sequence from simple to complex, build prior knowledge first, present in segments | A changeover procedure in which the setting of one axis depends on three others |
| Extraneous load | The way information is presented: searching, matching, decoding, clutter, distractions | Reduce it: integrate what belongs together, remove what does not serve the task | A legend on another page, a step that says "see Figure 12" twenty pages later, decorative images in a task |
| Germane load | In the classic model, the effort spent on building and automating schemas | Free capacity for it by cutting extraneous load; prompts such as self-explanation can direct it | A worked example that shows why a parameter is set, not only which value |
The first two are the stable core of the theory. Germane load was introduced by Sweller, Jeroen van Merriënboer and Fred Paas (1998); Sweller (2010) recast it as the working memory resources devoted to dealing with intrinsic load rather than a separate, additive source, and the 2019 review of the theory follows that view. The three types are therefore a useful lens, not three quantities that can be measured and added up. (Intrinsic cognitive load and extraneous cognitive load have their own glossary entries.)
What the split-attention effect is
The split-attention effect is a classic case of avoidable extraneous load. Paul Chandler and John Sweller (1991, 1992) compared instruction in which a diagram and its explanation were presented apart with versions in which the explanations were placed in the diagram. When neither part made sense on its own, the integrated versions led to better learning: readers no longer had to hold one source in mind while searching the other and matching the two. The separation can be spatial (a figure and its text on different pages, a key far from the drawing) or temporal (a narration that comes before or after the animation it explains). Sweller and Chandler (1994) showed that the cost depends on element interactivity: it matters for material with many interacting elements and hardly for simple material.
The relationship, side by side
| Aspect | Cognitive load (general concept) | Split attention (specific effect) |
|---|---|---|
| Level | An overarching framework about how instruction uses limited working memory | One concrete design problem within that framework |
| Cause | Complex content (intrinsic), poor presentation (extraneous), or both at once | Sources that belong together and are unintelligible on their own, presented apart in space or time |
| Effect | Productive when capacity goes to understanding; overload when the total exceeds what working memory can handle | Raises extraneous load and, for novices and complex material, worsens learning and performance |
| Remedy | Sequence and segment content, build on prior knowledge, use worked examples, cut extraneous load | Integrate physically (labels in the figure, text at the part it explains), synchronise narration and picture, or use the modality effect where listening is possible |
| Boundary | Effects depend on the reader's expertise and the element interactivity of the content | Does not apply when one source is self-explanatory; then adding the other creates redundancy |
In short: split attention is one cause of unnecessary cognitive load. Integrating the material removes that cause; it does not reduce the intrinsic difficulty of the content.
Related effects that shape the remedies
- Redundancy. When a figure is understandable on its own, repeating it as text does not help and can hinder; the same holds for identical text shown on screen and spoken at once. Slava Kalyuga, Paul Chandler and John Sweller (2004) found this in multimedia technical instruction. The redundancy principle of multimedia learning makes the same point.
- Modality. Seyed Yaghoub Mousavi, Renae Low and John Sweller (1995) found that a diagram explained in spoken words can be easier to follow than the same diagram with written text, because eyes and ears share the work. Wayne Leahy and John Sweller (2011) showed the limit: long or complex spoken explanations are transient and can overload working memory, so the effect reverses.
- Worked examples. John Sweller and Graham Cooper (1985) found that novices learn more from studying solved problems than from solving problems alone. A worked example can fade into a partly completed and then an independent task — the pattern often summarised in classrooms as "I do, we do, you do".
- Segmenting, sequencing and chunking. Presenting new information in small portions with one goal each, simple before complex, and building on what readers already know manages intrinsic load. In memory research, chunking means grouping elements into larger units through prior knowledge; it is not a fixed number of items per list (see the card on working memory and procedure steps).
- Expertise reversal. Slava Kalyuga and colleagues (2003) reviewed findings that aids which help novices — integrated explanations, worked examples — can burden experts, who must process guidance they no longer need. See the expertise reversal effect.
- Multimedia principles. Richard E. Mayer's cognitive theory of multimedia learning arrives at related principles: spatial contiguity, coherence, signalling and redundancy.
The explainer of the Australian Education Research Organisation (2023) translates these findings into teaching practice: present the information needed for a task in one place and at one time, leave out what does not serve the task, break new material into small chunks, scaffold with worked examples and checklists that are faded out, draw on prior knowledge, check understanding, space practice and keep the environment free of distractions.
What it means for documentation
- Labels on the figure. Callouts, short labels or numbered leaders in the illustration, with a key directly beside it — not a legend on another page. The symbols and safety signs in a figure follow the same logic.
- The figure at the step. In a DITA task the figure belongs to the step it illustrates, not to the end of the task or an appendix; on screen, step and figure stay in view together.
- Same names on the page and on the machine. Readers also split their attention between the document and the product. Labels, HMI texts and the terms in the instructions match exactly, so matching costs no search.
- Narration with the picture. In videos and e-learning, spoken explanation synchronised with the animation; no on-screen text that repeats the narration word for word; short segments the viewer controls. Where readers cannot listen — noise, hearing protection — written text stays the source.
- One source per piece of information. A self-explanatory figure is not described again in full; a table and a paragraph do not carry the same values.
- Worked examples that fade. A filled-in configuration, a sample calculation or an annotated screen first, then the task with fewer hints, then the reference.
- Segments with one goal. One action per step, concept before task, prerequisites stated before the procedure; numbered steps also help readers resume after an interruption.
- Layered depth. Novices and experts need different depths: quick reference and full procedure, or progressive disclosure — in the spirit of minimalism, never by leaving out what the task requires.
- Legibility is part of extraneous load. Type size, contrast and the readability of safety information decide how much effort reading costs.
- Check with readers. A think-aloud protocol, a cognitive walkthrough or a usability test shows whether the design actually reduces search and errors.
Limits
Most studies were done with school and university learners working with instructional material, often with short tests afterwards; documentation is used during work, under time pressure and interruptions. Load is hard to measure — often by self-report — and effect sizes vary. A recent conceptual paper by Andrew Sortwell and colleagues (2026) argues that cognitive load theory, applied as a general framework for school learning, attends to cognitive efficiency and overlooks emotional, social and self-regulatory sides of learning; for documentation this is a reminder that motivation, trust and the situation of use count as well. The findings support the direction of design choices, not fixed rules.
No cognitive argument replaces the risk assessment or legally required safety information. Integrating figures and text never justifies dropping required content from the instructions for use; safety messages that are repeated where the hazard occurs are not redundancy in the sense of the theory; and a well-designed page does not make a product compliant.
Questions readers ask next
- Is split attention the same as cognitive load?
- No. Cognitive load is the general demand on working memory; split attention is one specific cause of unnecessary (extraneous) load, which arises when sources that only make sense together are presented apart.
- Does split attention mean every figure needs text inside it?
- No. The effect applies when figure and text only make sense together. A self-explanatory figure needs no added text, and a text that stands alone needs no figure; adding the other then creates redundancy.
- What is the difference between intrinsic and extraneous load?
- Intrinsic load comes from the content and the reader's prior knowledge, such as how many parts of a system interact. Extraneous load comes from presentation, such as searching for a legend on another page. Design can reduce extraneous load; intrinsic load is managed by sequencing, segmenting and prior knowledge.
- Is germane load still part of the theory?
- In a changed form. It was introduced in 1998 as productive effort spent on building schemas; since Sweller (2010) it is treated as the working memory resources devoted to intrinsic load rather than a separate type that adds up with the other two.
- Should instructions be spoken rather than written, because of the modality effect?
- Only in some media. Spoken explanation with a picture helps in short, simple segments, for example in videos; long or complex spoken text is transient and can overload working memory. At the machine, written instructions remain the reference, and safety information stays in its required form.
- Is cognitive load theory proven for manuals?
- It is well supported for learning from instructional material, and some studies used technical instruction, but direct studies of documentation in use are fewer. Its principles are a strong starting point that still needs testing with real users.
Sources
- John Sweller (1988), Cognitive load during problem solving: Effects on learning — Cognitive Science, 13 October 2026
- Paul Chandler and John Sweller (1991), Cognitive load theory and the format of instruction — Cognition and Instruction, 13 October 2026
- Paul Chandler and John Sweller (1992), The split-attention effect as a factor in the design of instruction — British Journal of Educational Psychology, 13 October 2026
- John Sweller and Paul Chandler (1994), Why some material is difficult to learn — Cognition and Instruction, 13 October 2026
- John Sweller, Jeroen J. G. van Merriënboer and Fred G. W. C. Paas (1998), Cognitive architecture and instructional design — Educational Psychology Review, 13 October 2026
- John Sweller (2010), Element interactivity and intrinsic, extraneous, and germane cognitive load — Educational Psychology Review, 13 October 2026
- John Sweller, Jeroen J. G. van Merriënboer and Fred Paas (2019), Cognitive architecture and instructional design: 20 years later — Educational Psychology Review, 13 October 2026
- Seyed Yaghoub Mousavi, Renae Low and John Sweller (1995), Reducing cognitive load by mixing auditory and visual presentation modes — Journal of Educational Psychology, 13 October 2026
- Wayne Leahy and John Sweller (2011), Cognitive load theory, modality of presentation and the transient information effect — Applied Cognitive Psychology, 13 October 2026
- Slava Kalyuga, Paul Chandler and John Sweller (2004), When redundant on-screen text in multimedia technical instruction can interfere with learning — Human Factors, 13 October 2026
- John Sweller and Graham A. Cooper (1985), The use of worked examples as a substitute for problem solving in learning algebra — Cognition and Instruction, 13 October 2026
- Slava Kalyuga, Paul Ayres, Paul Chandler and John Sweller (2003), The expertise reversal effect — Educational Psychologist, 13 October 2026
- Richard E. Mayer, Multimedia Learning, second edition (2009) — Cambridge University Press, 13 October 2026
- Managing cognitive load optimises learning (explainer, September 2023) — Australian Education Research Organisation, 13 October 2026
- Andrew Sortwell, Evgenia Gkintoni, Jesús Díaz-García, Peter Ellerton, Ricardo Ferraz and Gregory Hine (2026), Beyond cognitive load theory: Why learning needs more than memory management — Brain Sciences 16(1), 109, 13 October 2026
Review log and changes
Every context card is checked against its sources before it is published, and again whenever it changes; the date under the byline is the last review. Corrections (something was wrong) and additions (something was missing) are logged below with date and time (Berlin time). Typos, formatting and link fixes are not listed.
Reviewed
Corrections and additions
Every correction and addition to this card, with date and time (Berlin time).
Addition
Addition: the card now explains cognitive load as the overarching concept and split attention as one avoidable cause of extraneous load, with a table of the three types of load (including the current status of germane load), a side-by-side comparison, the related effects (redundancy, modality and its limit, worked examples, segmenting and chunking, expertise reversal), more documentation practices and a recent critique of the theory in the limits.