After processing one target, the next is hard to process for a short window
Aliases: RSVP · lag-1 sparing · two-stage processing model
What it is
Among rapidly successive targets, once the first one has just been identified, a second target appearing roughly two to five hundred milliseconds later shows a clear drop in report accuracy. This is the attentional blink. The classic paradigm is rapid serial visual presentation (RSVP): two digits are embedded as targets in a stream of letters flashing by at roughly one hundred milliseconds per frame, and participants must report both digits. When the two digits are separated by around two to three hundred milliseconds, the second is frequently missed — even though it would be trivially easy to identify on its own.
Why it happens
The dominant explanation is a two-stage processing model: identifying the first target requires a consolidation process — moving from initial detection into working memory — and this consolidation occupies central processing resources for a stretch of time. If the second target happens to fall within this consolidation window, its initial representation does form, but because it cannot get the resources to complete consolidation, it decays rapidly from short-term memory before it can be reported. This is not a failure to keep up at the sensory level — the signal is generated normally — it is a temporal bottleneck specifically at the identification-and-consolidation step.
An interesting exception is lag-1 sparing: if the second target follows almost immediately after the first (a very short gap, on the order of tens of milliseconds), it is often reported correctly along with the first, because the two are treated as a single processing event before the competition for consolidation resources even begins.
Studying it
The standard method is the RSVP paradigm: the independent variable is the time gap between the two targets (lag, typically expressed as the number of intervening items or in milliseconds); the dependent variable is the second target's report accuracy, plotted as a function of lag. The typical curve is U-shaped — high at lag-1 due to sparing, dropping to its lowest around lag 2–3, and recovering to baseline by around lag 7–8.
In interface research, this method is mainly used to evaluate whether, in a stream of consecutive alerts or prompts, one following closely on another gets systematically missed.
Methodological caveat: the effect is sensitive to presentation rate. Classic RSVP rates run around 8–10 items per second; at slower rates the blink weakens or disappears. Real interfaces typically present prompts far more slowly than RSVP, so before applying lab lag numbers, confirm the presentation rate actually falls within the range where the effect reliably appears.
Where it stops holding
- The blink appears reliably only in tasks requiring conscious identification of the first target, not mere detection; if the first target does not need to be reported, the second is unaffected.
- The effect's magnitude shrinks somewhat with specialized training, but does not disappear.
- This describes a processing bottleneck specific to identification; it is a different mechanism from other attentional limits found in visual search or motion tracking tasks, and the two should not be conflated.
Applying it
- When users must identify consecutive prompts one at a time, do not assume shorter gaps are always processed faster — identification has its own processing time window, and a gap landing right in the two-to-five-hundred-millisecond range is actually the most dangerous combination for being missed.
- Verification: measure the second prompt's detection rate as a function of gap under real prompting cadence, focusing on whether a clear dip appears in the two- to five-hundred-millisecond range.