Salience captures attention but does not guarantee comprehension
Aliases: capture-comprehension dissociation · orienting vs identification · N2pc
What it is
Something sufficiently salient — high contrast, a sudden onset, an abrupt change — automatically pulls attention toward it. This is attentional capture. But capture only guarantees that attention was oriented there; it does not guarantee the content was understood.
This is the piece of selective attention most often glossed over: orienting and identification are separate processing stages. Orienting happening does not mean identification follows. Treating "the user noticed it" as equivalent to "the user understood it" is where many alert designs quietly fail.
Why it happens
Orienting is a fast, bottom-up, stimulus-driven response. It shows up in EEG as an early component such as N2pc, typically completing within one to two hundred milliseconds of stimulus onset, and it is almost impossible to voluntarily suppress. What orienting accomplishes is limited: it allocates a scarce attentional resource to a location or object, opening a processing window for whatever comes next.
Identification is what happens inside that window: it needs the window held open long enough, and it needs the sensory input matched against existing knowledge and the current task goal — a slower step that draws on working memory. If the window closes early — cut short by the next capture event, or by high perceptual load already occupying the observer — attention withdraws before identification finishes. Capture happened; identification did not.
This is why a high-intensity alert (large red text, aggressive flashing) reliably gets users to look toward it, but does not guarantee they came away understanding the content: capture solves "where to look," not "what was seen there."
Studying it
Common paradigms:
- Pop-out effects in visual search, using flat reaction-time curves across increasing numbers of distractors to establish that orienting is automatic.
- Cueing paradigms, comparing identification accuracy for the target content after capture against the orienting reaction time itself, recorded separately.
- Perceptual load manipulations, testing whether capture still occurs under high task load while identification fails — the core evidence line behind Lavie's perceptual load theory.
Typical independent variables: stimulus salience (contrast, magnitude of change), the length of the processing window available after target onset (SOA), and the perceptual load of the primary task. Dependent variables are deliberately split into two families: orienting measures (first-fixation latency, N2pc latency) and identification measures (content accuracy, confirmation reaction time) — a study measuring only one cannot reveal that capture and comprehension have already come apart.
In interface research, this method is mainly used to evaluate alert and notification design: whether "noticed" and "understood" are the same event, rather than assuming the former entails the latter.
One methodological caveat: most usability tests only log whether a user "noticed" a prompt, conflating noticing with understanding. Answering the design question requires separate evidence for orienting (eye tracking, reaction time) and for identification (content recall, or whether the follow-up action was correct).
Where it stops holding
- The dissociation is clearest under high perceptual load or when multiple stimuli compete for processing; in a quiet, single-stimulus interface the gap between the two stages is small and easy to mistake for "noticed therefore understood."
- It does not hold when users are actively searching for the target — orienting there is goal-driven (top-down), and the gap between capture and identification is inherently small, so no meaningful dissociation shows up.
- Most of the evidence comes from static-image or simple-search laboratory paradigms. Real interfaces stack multiple simultaneous events (a popup, a motion effect, a background change at once), which compresses the identification window further — the dissociation magnitude measured in the lab should not be read as the real-world failure rate.
Applying it
- Do not accept "the user noticed the prompt" as the acceptance criterion; require "the user can recall what it said" or "the user acted on it correctly."
- High-salience devices (red, flashing, sudden appearance) are good at producing orienting, but the content itself needs time to be identified — keep key text or icons stable and legible for a readable duration instead of dismissing them the instant capture occurs.
- When several high-salience elements appear together, their identification windows compress and compete; stagger them instead of firing them simultaneously.
- Verification: in usability testing, ask "what did that prompt say" in addition to "did you notice it." A gap between the two scores means capture and comprehension have decoupled in this design.