Cognitive Load: Why Less Is Often More in Interface Design

Picture the booking screen of a gym app. At the top, a banner with the deal of the month. Below it, twelve filters, a weekly view with 60 classes in six colors, and small icons for “beginner friendly,” “equipment needed” and “outdoor” whose meaning is never explained. If you want a spot in the Thursday evening yoga class, you first have to remember which studio it takes place in, because that only appears on the next screen. And you have to keep the class number from the invitation email in your head, because the search field asks for it.

None of this is a serious mistake on its own. Together, though, these details turn a simple booking into hard work. People give up, book the wrong slot or just call the gym instead. The problem has a name: too much cognitive load.

This article explains what the term means, how to tell when there is too much of it and which levers help you reduce it. The gym app’s booking screen serves as a running example throughout. It is made up, but typical.

What cognitive load means

Cognitive load describes how much mental effort it takes to understand and use an interface. The Nielsen Norman Group (NN/g) informally calls it the “brainpower” someone has to spend to reach a goal. That brainpower is limited, and whatever the interface uses up is no longer available for the actual task.

The concept comes from learning research. In the late 1980s, Australian educational psychologist John Sweller developed cognitive load theory to explain why some learning materials hinder understanding rather than help it. His key paper appeared in the journal Cognitive Science in 1988. The core idea: working memory, the part of memory we actively think with, has very little room. When it is overloaded, understanding and performance drop.

How much fits into working memory?

At this point, the number “seven, plus or minus two” usually comes up. It goes back to psychologist George A. Miller, who wrote in 1956 that people can hold roughly seven items in mind at once. Later research paints a stricter picture. In a widely cited 2001 review, Nelson Cowan concluded that the central capacity limit averages about four chunks of information once you account for tricks like silently rehearsing items.

In practice, the exact number matters less than the order of magnitude. Working memory is small, and it fills up fast. That does not, however, justify a rule like “no more than seven menu items.” A menu you can see does not need to be memorized. The strain begins when people have to carry information in their heads across several steps.

The three types of cognitive load

Cognitive load theory distinguishes three types of load, and they translate well to interfaces:

• Intrinsic load is the difficulty built into the task itself. Booking a class is easy, filing a tax return is not. Design cannot remove this load, only spread it out well.
• Extraneous load comes from the way information is presented: unclear icons, scattered details, distractions. It helps nobody, and it is the part design can actually fix.
• Germane load is the effort that goes into genuine understanding, for example when someone grasps a new concept. It was added to the theory later and remains debated among researchers. For UX work, distinguishing intrinsic from extraneous load is usually enough.

NN/g sums this up in practical terms: intrinsic load is not a bad thing, because people come to a website to get something done or to learn something. Extraneous load, on the other hand, eats up attention without giving anything back. Good design does not remove all effort. It removes the unnecessary kind.

How to tell when cognitive load is too high

Cognitive load is invisible, but its effects are not. Usability tests reveal typical signals:

• People pause, scroll back and forth and look for information that is actually on the screen.
• They read things aloud, repeat terms or jot something down on paper.
• They ask what icons, abbreviations or technical terms mean.
• They pick an option that “will probably do” instead of the best one.
• They lose track when switching between two screens.

Usage data offers clues, too: high drop-off rates at specific steps, lots of jumping back and forth between pages, recurring support requests about the same steps, or input errors piling up in one field.

For the gym app, a short test with five people might look something like this: several ask what the leaf icon means. Some go back to the weekly view because they no longer remember whether the class started at 6:00 or 6:30 pm. Someone mistypes the class number from the email and ends up in a different class. Each of these patterns points to a spot where the interface demands mental work it could be doing itself.

Seven levers for reducing cognitive load

1. Remove visual noise

Every element on a screen has to be processed, even if it contributes nothing to the task. NN/g names redundant links, images without a purpose and decorative typography as typical culprits. That does not mean interfaces have to be bare. It means every element should earn its place.

A simple exercise: go through the screen element by element and ask whether it helps the person at this moment. In the booking screen example, the deal-of-the-month banner moves from the booking view to the news section. The six calendar colors shrink to two: one for classes with open spots, one for fully booked ones. The remaining categories appear as text in the class entry, where they make sense without a legend.

Consistency reduces noise as well. If the same button has three different colors in three places, the brain has to check every time whether it does the same thing.

2. Recognition rather than recall

Recognizing something is much easier than retrieving it from memory. Jakob Nielsen made this one of his ten usability heuristics: “Recognition rather than recall.” In essence: minimize memory load by making elements, actions and options visible. Nobody should have to remember information from one part of the interface to use another.

The booking screen breaks this rule twice: the studio only appears on the next screen, and the class number has to be copied from an email. The fix:

• Before: A class card reading “Yoga Flow, 6:00 pm.” Studio and instructor only after tapping.
• After: A class card reading “Yoga Flow, Thu 6:00 to 7:00 pm, Downtown studio, 4 spots left.” The confirmation screen repeats the same summary.

The link in the invitation email leads straight to the right class instead of giving a number someone has to type in. And when people open the app, they see their recently booked classes at the top instead of searching for them every time.

3. Use familiar patterns

People bring expectations they have learned from other websites and apps. Experts call these mental models: internal ideas of how something works. An interface that meets these expectations does not have to be learned first. That is why NN/g recommends labels and layouts people already know.

Originality has its place, but rarely in navigation or standard tasks. A shopping cart icon in the top right, a logo that leads to the home page, a calendar that starts on the day your audience expects: break these conventions and you ask everyone who uses the product to learn something new. Each instance is small, but together they cost attention.

The gym app had invented its own swipe gesture for the calendar that switched between studios instead of weeks. After the redesign, swiping changes the week, as in almost every calendar app. The studio is chosen through a clearly labeled selector at the top.

4. Make decisions easier

Every decision takes effort. Hick’s law states that the time it takes to choose increases with the number of equally likely options, but logarithmically rather than linearly. Twice as many options do not mean twice the time. One important caveat: the law applies to choices between known options. When someone has to scan an unsorted list first, the effort tends to grow with every item.

Then there is the famous jam experiment. In 2000, Sheena Iyengar and Mark Lepper set up a tasting table in a supermarket with either 24 or 6 kinds of jam. The large display drew more people, but those who tasted at the small one were much more likely to buy. The finding is often cited as the “paradox of choice.” However, a 2010 meta-analysis by Benjamin Scheibehenne and colleagues covering 50 experiments found an average effect of virtually zero, with large differences between studies. Too much choice can overwhelm people, but it does not have to. It depends on how easily people can compare the options and how clearly they know what they want.

This leads to three practical approaches:

• Sensible defaults: If one option suits most people, preselect it. NN/g explicitly lists smart defaults as a way to take work off people’s hands. Defaults must not sneak in choices that work against the person’s interests, though.
• Organized options: Group, sort and label options so they are easy to compare. A longer list with clear groups is often easier to scan than a short but disorganized one.
• Recommendations: A highlighted suggestion helps people who are unsure without limiting the choice for everyone else.

In the app, this means the twelve filters shrink to three visible ones: studio, day and class type. The rest sit behind “More filters.” The most recently visited studio is preselected.

5. Reveal things step by step

Not everything has to be visible at once. Progressive disclosure means showing information and features only when they are needed. The default view stays clear, while details and rarely used options are one click away.

A closely related idea is chunking: grouping information into meaningful units that can be processed as one. A long bank account number split into groups of four is easier to check than an unbroken string of digits. A long form with clearly labeled sections usually feels more manageable than the same fields without structure. The article Forms Nobody Abandons shows what this looks like in practice.

The limit is reached where people need information to make a decision. Finding out after three clicks that a class costs extra does not mean less work, it means more. Only hide what is genuinely secondary for the decision at hand.

In the booking view, the class card therefore shows the time, studio, open spots and, where applicable, a price. The description, instructor profile and equipment notes open in a detail view when needed.

6. Clear language

Text often creates more cognitive load than layout. Jargon, internal abbreviations and convoluted sentences force people to translate before they can act. A few ground rules help:

• Use words from the world of the people who use the product, not from the org chart. “Cancel class” instead of “Release reservation.”
• One term per thing, used consistently everywhere. If the same action is called “booking” in one place, “registration” in another and “reservation” in a third, people wonder whether these are three different things.
• Lead with what matters most. A note like “Free cancellation up to 2 hours before class” works better than a paragraph that ends with this information.
• Label your icons. An icon with text is usually easier to understand than an icon on its own, especially when it is not one of the few universally recognized ones.

In the example, the three cryptic icons get short text labels: “Good for beginners,” “Equipment provided” and “Outdoors.” Nobody has to ask what they mean anymore.

7. Let the system do the work

Anything the software can reliably handle itself, it should. NN/g calls this offloading tasks: calculating, comparing, remembering and transferring information are jobs computers do better than people.

Typical examples include previously entered data that reappears in the next step, automatically calculated totals and deadlines, input fields that accept spaces or different formats, and reminders that arrive at the right moment. In the gym app, the booking confirmation calculates the cancellation deadline directly (“Free cancellation until Thursday, 4:00 pm”) instead of just stating the rule “up to 2 hours before.” One tap adds the class to your own calendar.

When less is not more

“Less is more” is a good rule of thumb, but not a law. Three pitfalls come up again and again in practice.

Hiding too much. Tucking every piece of information behind menus, tabs and accordions creates a tidy surface, but also lots of clicks and searching. People then have to remember where something was instead of seeing it. Cognitive load does not go down. It just moves somewhere else.

Treating expert tools like everything else. People who work with an analytics tool, a point-of-sale system or a control room every day often benefit from a dense display with many numbers at a glance. For them, constantly expanding and collapsing panels is the bigger burden. What looks cluttered to occasional users can be exactly right for professionals. Here, research and testing with the actual audience help more than general rules.

Trying to design away intrinsic load. Some tasks are complex, like comparing insurance policies or deciding on a loan. Simplifying them artificially by leaving out important information does not save people work. It takes away the basis for a good decision. The goal is to structure complexity well, not to hide it.

Measuring cognitive load

You can check whether a redesign actually reduces the load. You do not need a lab for that.

The simplest approach is a moderated usability test with thinking aloud. Five people complete the same task in the old and the new version. You watch for the signals described above: pauses, questions, jumping back and forth, errors. You can also record success rate, number of errors and time on task.

For structured self-assessment, there are questionnaires. A widely used one is the NASA Task Load Index (NASA-TLX), developed by a research group at NASA’s Ames Research Center in the 1980s. It asks about six dimensions, including mental demand, time pressure, effort and frustration. Many studies use a simplified, unweighted version known as Raw TLX. For a product team, it is often enough to ask only about mental demand after a task and compare the answers between two versions.

What matters is that you compare versions using the same task. A single score says little on its own. Only the comparison shows whether a change made the work easier.

Conclusion

Cognitive load is the mental work an interface demands. Part of it belongs to the task and cannot be avoided. Another part is created by design, and that part you can deliberately reduce: less visual noise, visible rather than remembered information, familiar patterns, well-prepared decisions, step-by-step disclosure, clear language and a system that does the work.

After the redesign, the gym app’s booking screen does not look dramatically different. It has fewer colors, fewer visible filters, labeled icons and class cards that show all the important details at a glance. A booking that used to require concentration now happens in passing. And that is the point: good design saves people mental effort so they can spend their attention on what matters to them.

Checklist

• Every element on the screen helps with the current task. Everything else has been removed or moved.
• Colors, icons and buttons are consistent and understandable without a legend.
• Information needed for a decision appears where the decision is made.
• Nobody has to remember or retype anything from one screen to the next.
• Navigation, gestures and terms follow familiar conventions.
• Options are sorted and grouped, and sensible defaults are set.
• Rarely used features and details are revealed step by step, but nothing decisive is hidden.
• Copy uses the audience’s language and one term per thing.
• The system calculates, remembers and transfers information wherever it can do so reliably.
• The effect has been checked with a test or a short survey.

Sources

• Kathryn Whitenton, Nielsen Norman Group: Minimize Cognitive Load to Maximize Usability
• Jakob Nielsen, Nielsen Norman Group: 10 Usability Heuristics for User Interface Design
• John Sweller (1988): Cognitive Load During Problem Solving: Effects on Learning. Cognitive Science, 12(2), pp. 257 to 285
• Nelson Cowan (2001): The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1)
• Wikipedia: Hick’s law
• Sheena Iyengar and Mark Lepper (2000): When Choice Is Demotivating. Journal of Personality and Social Psychology, 79(6)
• Benjamin Scheibehenne, Rainer Greifeneder and Peter M. Todd (2010): Can There Ever Be Too Many Options? Journal of Consumer Research, 37(3)
• Wikipedia: NASA-TLX
• Jon Yablonski, Laws of UX: Cognitive Load