The current task goal determines which features get prioritized in search

What it is

When searching through a cluttered display, people do not scan every location evenly — they carry a working description of what the target looks like, called a target template, held in working memory by the current task goal. The template specifies the target's color, shape, or size, and the visual system uses it to weight attention toward locations matching those features. This is top-down attentional guidance.

It is not the same as "processing whatever the eyes land on." Given the identical display, setting the template to "find red" versus "find round" produces a completely different order of attentional deployment — the scene hasn't changed, the goal has.

Why it happens

Visual search is not simple item-by-item checking. Early in processing, the whole visual field undergoes a parallel feature-weighting step: any location matching the template's features gets its activation boosted, and attentional resources flow preferentially there. This is the core idea behind guided search.

The template comes from a representation maintained in working memory for the current task, which means it consumes working-memory resources — and also means it can be swapped in real time as tasks change: searching for a red icon one moment and a square button the next causes the weighting map to be recomputed immediately.

The template's precision determines how well it guides. When the target feature is singular and stands out strongly from the background (e.g. "the only red dot"), the weighted target nearly pops out on its own, and search time barely increases with the number of distractors. When the target feature is vague or overlaps heavily with distractors, the weighting is diluted and search degrades into item-by-item serial checking, with search time increasing linearly with distractor count.

Studying it

The standard paradigm is the visual search task: a target is placed among distractors, with the target either defined by a single distinctive feature (feature search) or requiring a conjunction of features to identify (conjunction search). The dependent measure is the slope of reaction time against the number of distractors (ms/item). A slope near zero indicates near-parallel search with effective template guidance; a clearly positive slope indicates serial search.

Common independent variables: number and salience of target-defining features, whether the target is known in advance, and feature similarity between target and distractors. Common dependent variables: the RT-by-set-size slope, error rate, and number of fixations before the hit (requires eye tracking).

In interface research this paradigm is commonly used to evaluate the discriminability of searchable elements such as icons and buttons: given a well-defined find task, the search slope under different visual designs is measured — a lower slope indicates a design that is more effectively guided by the target template.

A methodological caution: lab paradigms typically tell participants exactly what the target is in advance, which is itself constructing a precise template artificially. In real interfaces, users' memory of the target is often vague ("something blue, I think"), so guidance is weaker than lab data would suggest.

Where it stops holding

• The template needs a clearly definable target feature. If the target itself is fuzzy (e.g. "information that looks important"), no effective template can form, and guidance degrades into essentially random scanning.
• Guidance weakens as distractors get closer to the target's features. When an interface's icons share a highly consistent style, guidance based on color or shape alone is noticeably degraded.
• Novices are weaker guides than experts. Experts hold a more precise, stable mental representation of the target, yielding a higher-quality template; novices may not remember exactly what the target should look like, so the template itself is imprecise.
• This describes search efficiency when the target is already known. It does not apply to browsing where the user has no defined target — that relies on a different mode of attentional allocation.

Applying it

• Give frequent find-tasks a distinct, single defining feature: elements users need to locate quickly (an "unread" badge, an error icon) should stand on one independently sufficient visual feature — color, shape, or position — rather than requiring a conjunction of features to confirm.
• Avoid letting functionally different elements share the same visual feature set: if multiple icons use the same color and similar shape, each one's target template gets diluted by the others, and users are more likely to fall back to serial checking.
• For high-frequency find tasks, preview the target's appearance beforehand: showing users what they're looking for before the search (e.g. "new messages are marked with a red dot") measurably sharpens the template and shortens search time.
• How to check: time users on a "find X in this interface" task while randomly varying the number of other on-screen elements, then plot search time against element count. A near-flat curve means the element is being located through effective guidance; a clearly rising curve means users are checking items one by one.

Related

• Same group: A5.12.2 Expectations and prior knowledge bias the interpretation of input early in processing · A5.12.3 Voluntary attention shifts are slower to initiate than stimulus-driven capture · A5.12.4 A strong task goal can partially suppress bottom-up salience capture, but cannot eliminate it · A5.12.5 A misspecified goal makes top-down guidance systematically miss relevant information outside the target set
• Nearby: A5.07 Attentional capture (the bottom-up counterpart) · A5.10 Attentional capacity and bottlenecks
• Search terms: target template · guided search · attentional set · visual search