Inattentional blindness: why users look straight at it and never see it

What it is

Arien Mack and Irvin Rock gave the phenomenon its name in their 1998 book Inattentional Blindness, reporting that observers concentrating on a judgement about a briefly flashed cross frequently failed to report an additional shape presented at the same time. The failure was not a limit of the eye. The unexpected shape fell on functioning retina, at readable size, and the observers simply had no experience of it.

The best-known demonstration is Daniel Simons and Christopher Chabris’s 1999 study, published in Perception. Observers watched a short video of two teams passing basketballs and were asked to count the passes made by one team. Partway through, a person in a gorilla suit — or, in other conditions, a woman carrying an umbrella — walked through the middle of the scene. Across 192 observers and all conditions, 46 per cent failed to report the unexpected event. Noticing varied with how the video was filmed and with how demanding the counting task was, which is the part designers should care about: the miss rate is not a fixed property of people, it is a function of the load you put them under.

It is not a novice effect either. Trafton Drew, Melissa Võ and Jeremy Wolfe reported in Psychological Science in 2013 that when a gorilla image roughly forty-eight times the size of an average lung nodule was inserted into the final case of a nodule-detection task, twenty of twenty-four radiologists — 83 per cent — did not report it, and eye tracking showed that most of those who missed it had fixated its location. Expertise sharpened the search; it did not widen it.

The level in this game reproduces the structure rather than the imagery. You are told to count the stars scrolling past in a feed, and a plausible-looking block of legal fine print sits below the form. Inside that fine print, in the same size and colour as the surrounding boilerplate about emoji accuracy and ISO appendices, is the actual instruction: type the word GORILLA into the count field. The counting task is real and effortful, which is precisely why almost nobody reads the paragraph they have already classified as boilerplate.

Why it works on people

Attention is allocated by task, not by pixel prominence. Once a user has adopted a goal — count these, find that, get through checkout — perception is tuned to the features that serve the goal, and material that does not match the current set frequently fails to reach report even when it is fixated. This is why "but it was right there on the screen" is not a defence: fixation and awareness are separable, and the eye-tracking data from the radiology study makes that separation concrete.

Load is the dial. In the original basketball study, noticing dropped when the counting task was made harder, and in Chabris, Weinberger, Fontaine and Simons’s 2011 field experiment, where participants jogged after a confederate past a staged three-person fight, only 35 per cent noticed the fight at night against 56 per cent in daylight. Anything that consumes capacity — poor visibility, time pressure, a form with unclear requirements, an unfamiliar flow — raises the miss rate for everything you did not ask the user to attend to.

Similarity to the attended set also governs noticing. Users looking for one kind of thing are more likely to spot unexpected items that resemble it. A disclosure styled as legal furniture is categorised as legal furniture in peripheral vision and never promoted to the foreground; a disclosure that looks like part of the form the user is currently filling in has a far better chance.

The practical consequence is uncomfortable. A designer can honestly render a warning, place it above the fold, give it a border and an icon, and still have a large fraction of users complete the flow with no memory of it. Compliance with "we showed it" is not the same as the user having seen it, and only the second one changes behaviour.

Where you meet it

• Subscription confirmation screens where the renewal price and the auto-renew date sit in a paragraph beside a large, well-styled Continue button that is the entire object of the user’s attention.
• Fine print used as a compliance artefact rather than a communication: rendered at small sizes and low contrast, positioned after the point of decision, and written in a register that signals "skip me" within the first three words.
• Error and warning banners that appear at the top of a long form after the user has scrolled to the submit button at the bottom, so the message is technically present and practically invisible.
• Deployment and admin consoles where a destructive scope selector ("apply to all environments") sits outside the region the operator is concentrating on while typing a resource name.
• Price or quantity changes applied silently between the cart page and the payment page, where the user’s attention is committed to card fields.
• Cookie and consent dialogs where the meaningful choice is a text link outside the button cluster the user is scanning.
• Onboarding coach marks fired during a first data-entry task, which are cheap to build, reliably ignored, and then cited internally as evidence that the feature was explained.

Designing around it

• Put material information on the task path, not near it. The reliable placement is between the user’s current focus and the control they are reaching for — inside the same visual container as the submit button, in the same type size and weight as the form content. Adjacency to the action beats size, colour or animation.
• State the consequence in the control itself. "Subscribe — £96 charged today, then £96 every 12 months" carries the disclosure through the one element you know has attention. This is a legitimate technique precisely because it does not depend on the user electing to read anything extra.
• Reduce the primary load before you add a second message. If the flow is difficult, adding a banner will not fix comprehension; simplifying the field validation, splitting the step, or removing a decision will do more for noticing than any amount of visual emphasis.
• Never rely on contrast reduction to make a disclosure unobtrusive, and treat low-contrast fine print as a defect rather than a style. WCAG 2.2 success criterion 1.4.3 sets 4.5:1 for normal text and 3:1 for large text as the floor; text below that is not merely hard to read, it is pre-classified as ignorable.
• Time the message to the decision. A warning shown on the previous screen, or during a loading state, has no reliable relationship to the moment the user commits. Show it where the commitment happens.
• Test for recall, not for presence. Asking "did you see the notice?" after the fact invites a false positive, because people reconstruct. Run the task, then ask an open question — "what will you be charged, and when?" — and count how many can answer without looking back. That number, not the render log, is your comprehension rate.
• Instrument the disclosure honestly. Log viewport intersection plus dwell time rather than "rendered", and treat a disclosure that is seen for under a second by most users as unread.
• Limit the number of simultaneous novel elements during high-load steps. Every extra toast, tooltip and badge competing for the same moment lowers the probability that the one that matters is the one that registers.

Questions

What is inattentional blindness?

It is the failure to consciously perceive a clearly visible object or event because attention was engaged elsewhere. The classic demonstration is Simons and Chabris’s 1999 study, in which 46 per cent of observers counting basketball passes did not report a person in a gorilla suit walking through the scene.

How is inattentional blindness different from change blindness?

Inattentional blindness is failing to notice something that is present; change blindness is failing to notice that something has altered between two views. They share a cause — attention is limited and selective — but the design implications differ: the first argues for placing information on the task path, the second for making transitions explicit rather than swapping content silently.

Does showing a warning count as informing the user?

Legally it sometimes does; behaviourally it often does not. Rendering a notice and having a user perceive, understand and act on it are separate events, and a substantial fraction of users will complete a flow with no recall of a notice they fixated. If the goal is comprehension rather than a compliance record, measure recall after the task instead of counting impressions.

Can you design an interface so that people stop missing things?

You can shift the odds substantially, but not to certainty. The two levers with the most evidence behind them are lowering the load of the primary task and moving the message onto the path the user is already travelling — ideally into the label of the control they are about to press. Increasing size, colour or motion helps far less than designers expect, because the problem is attentional allocation rather than visibility.

Related patterns

• The cocktail party effect: picking one channel out of a noisy screen
• Signifiers: an affordance nobody can see is not a feature
• Forced consent: unreadable terms and the ritual of agreeing
• Cookie consent dark patterns: double negatives and asymmetric choice