Before anybody reads a page, they search it with their eyes.

Visual search is the task of scanning the visual environment for a target among other items. It needs attention, because not everything can be processed at once. Looking for a result on a page is visual search.

The target is the object or feature the observer is looking for. Its properties, and how they differ from their surroundings, decide how fast it is found. On a page, the target is whatever answers the person's need.

Distractors are the non-target items that compete for attention. The more they resemble the target, the harder the search. Every advertisement, box and link on a result page is a distractor for someone.

Reaction time is the time an observer needs to find or report the target. It is the main behavioural measure of search efficiency. Small differences in milliseconds reveal how the search is organised.

Eye movements show where attention goes as a person searches, through fixations and the jumps between them. They make the search visible. They also show what was looked at and still not recognised.

Feature search is the search for a target that differs from all distractors by one simple feature, such as colour or orientation. A red item among green ones is found almost at once. Search time hardly grows with the number of items.

Conjunction search is the search for a target defined by a combination of features that distractors share partly, such as a red vertical bar among red horizontal and green vertical bars. Search time grows with every added item. Most real targets on a page are conjunctions.

If every target popped out, each extra item on the display would add only a few milliseconds. For a target defined by a combination of colour and shape, each item added about 29 ms when the target was present and 67 ms when it was absent.

The pop-out effect is the rapid, seemingly automatic detection of a target with a unique feature, regardless of how many distractors surround it. The target seems to jump out. Designers use it, and so does advertising.

Bottom-up processing is driven by the stimulus: strong contrasts and unique features attract attention on their own. It works before any intention. It decides what grabs the eye first.

Top-down processing is driven by goals and knowledge: the searcher's idea of the target directs attention. It speeds search when the target is known. It depends on what the person expects to find.

Parallel processing analyses many items across the visual field at the same time. Simple feature search works this way. It is why a single red item among green ones is found no matter how many there are.

Serial search moves attention from item to item. It characterises difficult searches such as conjunction search. The time needed grows with each item to be inspected.

Preattentive processes analyse simple features across the whole field before focused attention is applied. They provide the raw map that attention then works on. What they cannot separate, attention has to inspect one by one.

Selective attention is the capacity to process relevant input while ignoring the rest. Visual search is one of its clearest cases. What is ignored may never be perceived at all.

Feature Integration Theory, proposed by Anne Treisman and Garry Gelade in 1980, holds that simple features are registered in parallel and that focused attention is needed to bind them into objects. It explains the difference between feature and conjunction search. It shaped the field for decades.

The Guided Search model, developed by Jeremy Wolfe, holds that bottom-up salience and top-down knowledge together build a priority map that guides attention to likely targets. Search is neither purely parallel nor purely serial. It is guided.

Endogenous orienting is the voluntary direction of attention according to goals or instructions. A searcher who knows where results usually are looks there first. It is slow to deploy and can be held.

Microsaccades are tiny involuntary eye movements during fixation. Their direction tends to follow where attention is covertly directed. They show attention moving while the eyes seem still.

The distractor ratio is the proportion of different distractor types in a display. It changes how quickly a target is found, because searchers can restrict their search to the smaller group. The composition of a page matters as much as its size.

Focal spatial attention is attention concentrated on a region of space, which binds the features there into objects. Outside its focus, features stay loosely combined. What is not in focus is not fully seen.

An illusory conjunction is the perception of an object that combines features from different objects, such as a red X seen where there was a red O and a green X. It happens when attention is overloaded or diverted. People can see things on a page that are not there.

Set size is the number of items in a display. Plotting reaction time against set size shows how efficient a search is. More items on a page raise the cost of every search that is not pop-out.

What pulls the eye without permission, and what gets missed. The usual treatment explains features, conjunctions and the two theories. It leaves out attentional capture by salient distractors, the cost per added item, the effects of clutter and rare targets, learning from repeated layouts, the measures in the brain, and the professionals whose misses cost the most. For a page, this is where being present and being found come apart.

Exogenous orienting is the reflexive shift of attention to a sudden or salient event, such as a flash or an abrupt onset. It happens fast and without intention. Moving elements and pop-ups exploit it.

Attentional capture is the involuntary drawing of attention by a salient item, even when it is irrelevant to the goal, as Jan Theeuwes showed. A searcher who knows what they want still loses time to the loudest element. Pages compete with their own decoration.

Inhibition of return, described by Michael Posner and Yoav Cohen in 1984, is the tendency to be slower to return attention to a location just inspected. It keeps search moving forward. A person who has scanned a region once is unlikely to look there again.

Contextual cueing, shown by Marvin Chun and Yuhong Jiang in 1998, is the faster search in layouts that have been seen before, learned without awareness. Repeated structure guides the eye. Consistent page layouts help people find things faster.

The visual search slope is the increase in search time for each item added to the display, measured in milliseconds per item. A flat slope means efficient search, a steep one means item-by-item inspection. It turns clutter into a cost.

An attentional template is the representation of the target held in working memory during search. John Duncan and Glyn Humphreys made it central to their account. A searcher with the wrong template overlooks the right answer.

A visual saliency map is a topographic representation of how conspicuous each location is, proposed by Christof Koch and Shimon Ullman and computed by Laurent Itti and Koch. It predicts where the eye will go first. It can be calculated for any page.

The N2pc is a brain wave recorded over the side of the head opposite to an attended item, about 200 milliseconds after a display appears. Steven Luck and Steven Hillyard linked it to the deployment of attention. It shows selection before any response.

Covert attention is attention shifted to a location without moving the eyes. Eye tracking alone cannot see it. A person can notice something at the edge of a page without looking at it.

Visual crowding is the impaired recognition of an object in peripheral vision when other objects surround it. Herman Bouma described its spacing rule in 1970. Dense layouts make items unrecognisable before they are looked at directly.

Stimulus onset asynchrony is the time between the onset of one stimulus, such as a cue, and the onset of the next. It controls how much time attention has to prepare. Timing decides whether a cue helps or goes unnoticed.

Hybrid visual search is searching a display for any of several targets held in memory, a paradigm developed by Jeremy Wolfe. It resembles scanning a page for any of the things one needs. Search time grows with both the display and the memory set.

The target prevalence effect is the finding that rare targets are missed far more often than common ones, shown by Jeremy Wolfe, Todd Horowitz and Naomi Kenner in 2005. Observers stop searching sooner when targets are usually absent. What people rarely find, they stop looking for.

If target frequency had no effect, the error rate would stay level across the three conditions. It rose from 7 % to 30 % as the targets became rare, and the errors were almost all misses.

The useful field of view is the area from which information can be taken in a single glance, without moving the eyes or head. It shrinks with age and with load. It limits how much of a page registers at once.

A visual foraging task asks observers to collect several targets from a display, as in natural search, a paradigm developed by Árni Kristjánsson and colleagues. Observers switch strategies as targets deplete. Browsing a page for several useful items works this way.

The premotor theory of attention, proposed by Giacomo Rizzolatti and colleagues in 1987, holds that shifting attention is closely linked to planning an eye movement. Attention and action are prepared together. Where attention goes, the eyes are ready to follow.

A visual clutter score quantifies how cluttered a display is, for example through the feature congestion measure by Ruth Rosenholtz and colleagues. More clutter predicts slower search. Clutter can be measured before a page goes live.

The Posner cueing task, introduced by Michael Posner in 1980, measures how a cue speeds or slows detection at cued and uncued locations. It separates endogenous from exogenous orienting. It is the standard experiment of spatial attention.

Security screeners search baggage images for rare threats under time pressure. Their work combines low target prevalence with heavy clutter. Research on their errors shows how professional search fails.

Radiologists search medical images for abnormalities that are often rare and subtle. Expertise changes how they scan. It does not protect them from missing what they do not expect.

If trained eyes caught whatever was on the screen, a gorilla 48 times the size of a nodule would have been reported. 20 of 24 radiologists did not report it, and 12 of those looked directly at it.

A result or a page is found only if it wins a visual search against everything around it. Salience, clutter, familiar layout and expectation decide which answer is seen. Being present on the page and being found by the eye are separate achievements.

Make the answer the most conspicuous element, reduce clutter around it, keep layouts consistent, and place what people expect where they expect it. Mark rare but important information so that it cannot be passed over. A page that asks what the visitor is looking for can bring the target to them, which is the idea of Supervised Search.