Attention is a narrow spotlight, not a floodlight. Whatever falls outside it, even a gorilla, can vanish completely. So being in front of someone is not the same as being seen.
Attention is a filter, and a strict one. Colin Cherry's 1953 "cocktail party" experiments showed the basic shape: play a different message into each ear, ask someone to repeat one of them, and they can do it, but they retain almost nothing of the other, often not even the language it was in. Early models pictured this as a gate that lets one channel through and blocks the rest (Broadbent, 1958). Decades of argument about whether the gate sits early or late in processing were largely settled by Lavie (1995), who showed it depends on load: when the main task is demanding, irrelevant things get filtered out; when it is easy, they leak in. Posner (1980) added that the filter is movable, a spotlight you can shift around a scene without moving your eyes, speeding up perception wherever it lands.
The startling part is what happens to everything the spotlight misses. In Simons and Chabris's (1999) study, viewers counting passes in a basketball game routinely failed to see a person in a gorilla suit walk into shot, thump their chest, and walk off. About half missed it entirely. This is inattentional blindness: with attention committed elsewhere, even a large, strange, fully visible object simply does not reach awareness (Mack & Rock, 1998). A close cousin is change blindness. Rensink and colleagues (1997) found that if you put a brief blank between two versions of a scene, people struggle to spot even big, repeated changes. Simons and Levin (1998) took it into the street: an experimenter asking for directions was swapped for a different person while a door passed between them, and only half of people noticed they were now talking to a stranger. The conclusion across all of it is humbling: we perceive and remember only what we attend to, and the vivid sense of seeing everything is mostly an illusion.
What this says about the mind's inner picture is still argued. One reading is that we barely represent the unattended world at all. A softer reading is that we represent more than we can later access or compare, so the failure is one of memory and comparison, not raw perception. The feeling of a rich, complete visual field (sometimes called the "grand illusion") is genuine; its cause is not settled. Perceptual load theory, though influential, also has its limits and does not tidy up every finding.
The memorable figures are not fixed laws. "Half miss the gorilla" moves with how hard the counting task is and exactly how people are instructed. And counting-passes tasks are artificial. The reason the work is taken seriously anyway is that it holds up outside the lab, as the real-world door study shows.
The open question is how much of an unattended scene we actually encode, and where the line really sits between something never perceived and something perceived but instantly forgotten.
This is the mechanism sitting under a quiet, expensive assumption: that if you put something in front of people, they have seen it. They have not. Exposure and attention are different things, and most of what you place in someone's visual field never reaches them.
Being present is not being seen. The ad in the feed, the banner on the page, the message on the pack: most of it is filtered out by people focused on something else, the effect online advertisers learned to call banner blindness. Two consequences. First, your reach and impression numbers overstate real attention, often massively, so do not treat "delivered" as "noticed." Second, you have to earn the glance, with relevance, with distinctiveness, with something that breaks the viewer's current task set, rather than assuming exposure does the work.
A message "delivered" to a scrolling voter is mostly unseen. Impressions are not attention. What survives inattention is simple, distinctive, and repeated: the same recognisable face, phrase, or colour, often enough that it lands on one of the rare moments attention is actually pointed your way.
Safety and health information dropped into the environment, the warning label, the sign, the leaflet, is routinely not noticed by people attending to something else. "We told them" is not "they saw it." Critical information has to be designed to capture attention at the moment it matters, and tested to confirm people actually register it, not just that it was technically present.
Separate exposure from attention everywhere, in your metrics and in your design. Assume, as a default, that most of what you put in front of people is never consciously processed, and build accordingly: fewer elements, sharper contrast, genuine relevance, and a test that checks whether people noticed rather than whether you shipped it.
.png)