Cognitive Load in UX Design: What Research Says (2026)

A user lands on your product, glances at the screen, and within a couple of seconds decides whether this feels easy or exhausting. That decision is not really about how your design looks. It is about how much thinking your interface asks them to do. That mental effort has a name in the research literature: cognitive load. If you understand where it comes from, you can find and remove it, and your product starts to feel obvious instead of tiring.

This is not a soft, feel-good idea. Cognitive load is one of the most studied concepts in psychology and instructional design, and a handful of its findings map directly onto the screens you ship every week. Here is what the research actually says, and what to do with it.

What cognitive load means

Cognitive load is the amount of working memory a task demands. Working memory is the small, short-term store your brain uses to hold and manipulate information right now, and it is famously limited. When a task asks for more than that store can hold, people slow down, make errors, or give up.

The framework most designers borrow from is Cognitive Load Theory, developed by educational psychologist John Sweller in the late 1980s. Sweller split load into three kinds, and the split is the useful part:

• Intrinsic load is the difficulty inherent in the task itself. Filing a complex tax return is hard no matter how well you design the form. You cannot remove intrinsic load, but you can break it into smaller pieces.
• Extraneous load is the effort your design adds on top, through clutter, confusing labels, inconsistent patterns, and anything that makes the task harder than it needs to be. This is the load you are paid to eliminate.
• Germane load is the good kind: the effort that goes into actually learning and building a mental model of how something works. You want to protect this by clearing the other two out of the way.

Almost every UX improvement worth making is a reduction in extraneous load. When people say an interface is intuitive, what they usually mean is that its extraneous load is close to zero.

Miller's Law, and why "7 items" is a myth worth understanding

You have probably heard that people can only hold seven things in their head at once. That comes from George Miller's 1956 paper, "The Magical Number Seven, Plus or Minus Two." It is one of the most cited claims in all of psychology, and it is also widely misapplied.

Two things are worth getting right. First, Miller was writing about the capacity of working memory in controlled experiments, not prescribing a limit on menu items. Later research, particularly work by Nelson Cowan, suggests the practical number for most people is closer to four items when they cannot rehearse or group them. Second, the real lesson is not the exact number. It is that working memory is small and fills up fast.

So the design takeaway is not "never show more than seven options." It is: do not force people to hold information in their heads when the interface could hold it for them. Show the order summary on the payment screen instead of asking users to remember it. Keep the field label visible while they type, rather than using placeholder text that vanishes on focus. Every time you offload memory from the person to the screen, you cut load.

Hick's Law and the cost of choice

The other principle that shows up constantly is Hick's Law, named after psychologists William Hick and Ray Hyman. It says the time it takes to make a decision grows as the number and complexity of choices grow. More options, slower decisions, and past a point, no decision at all.

You see this everywhere in products that stall. A pricing page with nine plans. A settings screen with forty toggles and no grouping. A first-run experience that presents every feature at once. None of these are broken in a technical sense. They just ask the user to evaluate too much before acting, and evaluation is expensive.

Hick's Law does not mean fewer choices are always better. It means unstructured choices are costly. You can present many options as long as you structure them so the user only ever compares a few at a time. Group related settings under clear headings. Recommend a default plan so most people can stop deciding. Reveal advanced options progressively instead of all at once. The goal is to keep the number of live comparisons small even when the total number of options is large.

Where extraneous load hides in real products

The research is tidy. Real interfaces are not. Here is where the avoidable load usually accumulates, based on the patterns that come up again and again in UX reviews:

• Redundant decisions. Asking users to choose something the system could reasonably infer or default. Every optional field that looks required adds a small tax.
• Inconsistent patterns. A button that behaves one way on one screen and differently on the next forces people to relearn your product each time. Consistency is a load reducer, which is why it is one of Nielsen's ten usability heuristics.
• Vague or jargon labels. A menu item that means something specific to your team and nothing to an outsider makes the user stop and translate. That pause is load.
• Visual noise. Competing colours, too many font weights, and dense layouts make people work to find the one thing that matters. Clear visual hierarchy is not decoration. It is a way of doing some of the user's thinking for them.
• Hidden system status. When a user cannot tell whether something worked, whether it is loading, or where they are in a flow, they have to hold that uncertainty in their head. Good feedback dumps that uncertainty back out.

If you want to see how these compound, watch a first-time user attempt one core task in your product without helping them. The places they pause, squint, or scroll back up are almost always extraneous load you stopped noticing a long time ago.

How to measure and reduce it without a lab

You do not need eye-tracking equipment to work on cognitive load. A few practical moves get you most of the way.

Start by counting. For any key flow, count the decisions, the fields, and the number of distinct things on screen at each step. High counts are a signal, not a verdict, but they point you at the steps worth scrutinising. A signup that asks for eight fields when three would do is carrying load you can delete today.

Next, apply progressive disclosure. Show the minimum needed for the current step, and reveal the rest only when it becomes relevant. This is the single most reliable way to lower load on a complex screen, because it shrinks the number of things competing for attention at any one moment. Onboarding flows benefit from this more than almost anywhere, which is closely tied to how much friction users hit before they reach value. We wrote more about that in our guide to finding and fixing conversion leaks in SaaS onboarding.

Then check the mobile version specifically. Small screens amplify load because less fits in view and every scroll is a small memory test. A layout that feels calm on a wide monitor can feel overwhelming on a phone. Our mobile UX audit guide covers what to test there.

Finally, read your own labels as if you had never seen the product. If a word makes you pause, it makes a new user pause harder.

The takeaway

Cognitive load is the quiet reason good-looking products still feel hard to use. The research gives you a clear target: leave the intrinsic difficulty of the task alone, protect the effort people spend learning, and ruthlessly cut everything your design adds on top. Fewer live decisions, less to remember, clearer feedback, and only the relevant thing on screen at the relevant time. Do that and your interface stops feeling clever and starts feeling effortless in the one sense that matters, which is that people barely notice it.

If you want a fast read on where your own product is asking users to think too hard, run a free AI UX audit. It scores seven dimensions, including visual hierarchy, navigation, and content clarity, which are exactly the areas where extraneous load tends to pile up. It is a useful starting point for teams building SaaS products that need to convert.

Related reading: We Audited 10 SaaS Landing Pages. Here's What Most Get Wrong.

Next Article