In short
Cognitive load is the amount of working-memory effort a task demands. In interface terms it is the sum of what the user must remember, work out and decide to get something done, and working memory holds roughly four items at once. Intrinsic load comes from the task itself and cannot be removed; extraneous load comes from how the interface presents it and is the designer's responsibility; germane load is the useful effort of learning. Reduce load by showing rather than requiring recall, chunking, setting defaults, removing decisions, keeping consistency, and revealing complexity progressively.
The concept
Cognitive load theory comes from educational psychology, where John Sweller and colleagues studied why some instruction works and some overwhelms. The core finding is that working memory is small, holding around four chunks of information at a time, and that when a task demands more, performance collapses: people forget, err, give up. The same limit governs anyone using software, and interface design is largely the practice of keeping demand under it.
The theory splits load into three kinds, and the split is useful because only one of them is the designer's fault.
Intrinsic load is inherent in the task. Configuring a complex product, comparing four insurance policies, or filing a tax return is hard because the thing itself is hard. Design can structure it, not remove it.
Extraneous load is created by the presentation: cluttered screens, inconsistent controls, information the user must remember from an earlier step, choices that could have been defaults, jargon, motion, and every element that competes for attention without helping the task. This is the load design should eliminate, and most interfaces carry a great deal of it.
Germane load is the productive effort of understanding, forming a mental model of how the product works. Good design leaves room for it by clearing out the extraneous kind.
Where interfaces waste it
Requiring recall. Asking for a reference number from a previous email, a setting on another screen, a code that was shown once. Every recalled item occupies a slot.
Too many simultaneous options. Twenty filters, fifteen toolbar buttons, six pricing plans. Hick's law describes the time cost; the memory cost is that the user has to hold the options while comparing them.
Inconsistency. A control that behaves differently in two places forces the user to remember which is which. Consistency lets one learned rule cover many screens.
Unfamiliar language. Internal terms, abbreviations, invented labels. Each one is a translation the user performs on every read.
Split attention. Instructions on one part of the screen for a form on another; a diagram whose labels are in a separate legend; an error at the top for a field at the bottom. The user shuttles between them, holding one while finding the other.
Noise. Decorative motion, competing colours, promotional banners, anything that draws attention away from the task and must be actively ignored, which costs effort too.
Long forms without structure. Thirty fields on one screen with no grouping ask the user to hold the whole form in mind.
Reducing it
Recognition over recall. Show the options; show the previous value; show the reference number where it is needed. Autocomplete, recently used items, and visible state all convert recall into recognition.
Chunking. Group fields, steps and information into meaningful sets of three to five, with headings. A phone number split into recognisable parts is easier than a string of digits. A checkout in three labelled stages is easier than one long page.
Defaults and recommendations. Pre-fill what can be inferred, pre-select the common choice, mark the recommended plan. Each default is a decision the user no longer has to make.
Remove decisions that do not matter. Every option that is rarely chosen is a candidate for removal or demotion.
Progressive disclosure. Show what is needed now; reveal detail on request. Advanced settings behind a toggle; the second half of a form after the first half is done.
Consistency. One pattern for each kind of thing, everywhere, so the user learns it once. This is what a design system buys.
Plain language. The user's words, not the system's. Test labels on someone outside the team.
Integrate related information. Put the instruction beside the field, the error beside the input, the label on the chart rather than in a legend.
Show progress and state. Where the user is, how much remains, what has been saved. Uncertainty is load.
Cut noise. Every element that is not helping the task is costing attention. Whitespace is the cheapest load reduction there is.
Spotting overload in testing
In a usability test, overload shows as participants re-reading, scrolling back to check something, saying "wait" or "where was it", hesitating over choices, making errors on steps they understood, and expressing fatigue or frustration out of proportion to the task. Time on task rises and completion falls. Participants often blame themselves, which is a reliable sign the interface is at fault.
The fix is usually not to simplify the task but to simplify its presentation: fewer things on screen at once, more shown and less remembered, more decided by default.
The trade-off
Reducing load is not the same as removing capability. Expert users doing complex work want density and options, and an interface that hides everything from them adds load of a different kind: the effort of finding what was hidden. The goal is to match demand to the user and the moment: minimal for newcomers and simple tasks, richer for experts who have built the mental model. A product that does this well feels easy at first and powerful later, which is the pattern behind most software people describe as a pleasure to use. If yours feels exhausting and the team cannot see why, book a call; a screen-by-screen load review usually finds a short list of culprits.
Common questions
What is cognitive load in UX?
The mental effort an interface demands: everything the user must remember, work out and decide to complete a task, measured against a working memory that holds roughly four items at once. When a screen demands more, users slow down, make errors and give up. Good design keeps the demand below the limit.
What are the three types of cognitive load?
Intrinsic load, inherent in the task itself and not removable; extraneous load, created by how the interface presents the task and which design should eliminate; and germane load, the useful effort of understanding and forming a mental model. Reducing extraneous load leaves capacity for the other two.
How do you reduce cognitive load in a design?
Show information rather than requiring users to remember it, group content into chunks of three to five, set sensible defaults, remove decisions that rarely matter, reveal complexity progressively, keep controls consistent, use plain language, place instructions and errors beside the fields they concern, show progress and state, and cut every element that does not help the task.
What are signs of cognitive overload in usability testing?
Participants re-reading, scrolling back to check something, hesitating over choices, making errors on steps they understood, saying they are lost, and blaming themselves. Time on task rises and completion falls. The remedy is usually to simplify the presentation, with fewer things on screen at once and more decided by default, rather than the task.
Does reducing cognitive load mean making an interface simpler?
It means making the presentation demand less, which is not the same as removing capability. Experts doing complex work want density and options, and hiding everything from them adds the load of searching. The aim is to match demand to the user and the moment: minimal for newcomers, richer for experienced users who have learned the product.
