A strong task goal can partially suppress bottom-up salience capture, but cannot eliminate it

What it is

A high-contrast, suddenly-appearing element, even when completely irrelevant to the current task, still pulls a small amount of attention away from a user focused on a clear search goal — even when the user is entirely unaware of it. This shows that a task goal's suppression of salience capture is incomplete: the stronger and clearer the goal, the better the suppression, but it never drives capture to zero.

This is easy to misread as "focus hard enough and interference is fully blocked." The actual conclusion is the opposite — suppression has a ceiling, and residual capture always remains, differing only in strength and in whether it becomes noticeable.

Why it happens

The key is distinguishing two links in the capture chain: whether the signal occurs, and whether the signal gets used.

The salience signal itself is generated by an early, bottom-up pathway that is barely modulated by the current task goal — as long as something is physically abrupt or high-contrast enough, this pathway produces an orienting response toward that location almost automatically and obligatorily. What the task goal can influence is what happens after the signal occurs: whether that initial orienting gets carried forward into further processing, whether it draws on downstream identification and decision resources, and whether it turns into an observable behavioral shift or conscious percept.

In other words, what a strong task goal suppresses was never the signal itself — it is the stage where the signal gets used. The stronger the goal, the more cleanly that stage gets severed: behaviorally this shows up as a tiny RT cost, and users may not even report having seen the element. But with fine enough measurement — a small RT delay, an early ERP component — the occurrence of the signal can almost always be detected.

This is the same distinction as "being visible" versus "being seen" in inattentional blindness: the capture signal corresponds to "visible," its conversion into further processing and response corresponds to "seen," and the gate in between is the only place a task goal can actually exert control.

Studying it

The classic paradigm is the additional singleton paradigm: the main task is finding a target of a specific shape among homogeneous distractors, while a color singleton that is entirely task-irrelevant is added. Even when participants are explicitly told to ignore it, its presence still slows reaction time — that cost is the behavioral marker of residual capture.

A second approach is the contingent capture paradigm: it manipulates whether a distractor shares a defining feature with the current target template, testing whether capture strength varies with whether the distractor matches the task set. RT costs and eye-movement/ERP measures (such as the N2pc component) are used together to separately index "whether orienting occurred" and "whether it affected final behavior."

Common independent variables: distractor salience, feature overlap between distractor and target, strength of task instructions (whether ignoring is explicitly required). Common dependent variables: RT cost, error rate, and electrophysiological markers of early orienting (measured separately from the behavioral outcome).

In interface research, this paradigm is commonly used to test whether a user saying "I didn't notice that ad/pop-up" is the same as "that element consumed no attentional resources at all" — the two are not equivalent, and self-reported behavior routinely underestimates the capture that actually occurred.

Where it stops holding

• The degree of suppression depends on the strength and clarity of the task goal. When the goal is vague, the task itself is simple, or motivation is low, suppression is markedly weaker and capture is more likely to surface as observable distraction.
• Under high load or high stress, this suppression stage itself can fail wholesale — not just weaken a little, but the gate that normally cuts off the signal's downstream use largely stops working, so a peripheral high-salience item that does generate a capture signal never reaches further processing at all. This is exactly the situation in which attention narrows to a single information source under stress and peripheral warnings get ignored, and it does not contradict the claim here that "the signal still occurs" — what would be contradictory is treating "the signal occurs" as equivalent to "it will be acted on."
• Suppression cannot be trained away to zero. Practice can lower the proportion of distractors that turn into behavior, but "experienced users" should not be expected to become fully immune to high-salience elements.
• This describes the dissociation between the capture signal and its downstream use, not how long it takes attention to return to the original task after capture — that is a separate stage of capture.

Applying it

• Do not use "the user's goal is clear, so they should automatically ignore it" as a reason to leave an irrelevant, high-salience element in place. Even with a clear task, an irrelevant high-contrast or animated element still draws on some attentional resources — the user simply may not notice or be able to report that draw.
• Reduce an irrelevant element's salience at the source, rather than relying on users' suppression ability. Cutting unnecessary high-contrast colors, motion, and sudden overlays is the only way to actually reduce the capture signal itself.
• In contexts that require sustained high focus (monitoring, review, driving-related interfaces), consolidate non-task alerts and lower their presentation intensity, since the higher the required focus, the more the hidden cost of residual capture can matter, even when it stays invisible.
• How to check: while users perform the main task, measure whether reaction time shows a small but consistent difference with versus without an irrelevant, high-salience distractor present — even if users later report not noticing it. Whether that difference exists is the direct evidence for whether the element's salience needs to be reduced.

Related

• Same group: A5.12.1 The current task goal determines which features get prioritized in search · 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 · A5.12.5 A misspecified goal makes top-down guidance systematically miss relevant information outside the target set
• Nearby: A5.07 Attentional capture · A5.09 Cognitive tunneling · A5.05 Inattentional blindness
• Search terms: contingent capture · additional singleton paradigm · salience capture · residual capture