Cognitive Psychology
A stream of letters, one at a time, fast. Somewhere in it there is a digit, and somewhere after it there may be an X. Reporting the digit costs you the X — but only for a few hundred milliseconds, and not straight away.
Illustrative browser timing — a demonstration, not a measurement of anybody
By the end you should be able to read a lag function, and say why second-target accuracy has to be conditioned on the first.
About 25 minutes. Nothing you do here is saved or sent anywhere.
The stream is rapid text on an unchanging background. Only the character changes; the background, the size and the position stay put, so there is no full-screen brightness change of the kind associated with photosensitive reactions. Even so, rapid presentation does not suit everybody.
Nothing is compulsory. Each trial waits indefinitely for your answers, you can stop between trials, and the worked example reaches every conclusion on this page without running a single stream. If your system asks for reduced motion, the tool says so and points you there.
Letters appear one at a time in the same place, about ten a second. One of them is a digit — that is the first target, and you report it. On most trials the letter X appears somewhere after it: that is the second target, and you say whether you saw it. The gap between them, measured in stream positions, is called the lag.
Watch the middle of the box. When the stream finishes, two questions appear and wait as long as you need.
Four slow practice trials with feedback, then a block of 25 scored trials. Results are withheld until the block ends.
Status Not started.
Answer the question above to unlock the experiment.
| Lag | Gap | Trials | Digit correct | X seen, all trials | X seen, digit correct only |
|---|
Several accounts of the attentional blink are live at once. They agree about the basic lag function and disagree elsewhere. For each statement below, decide whether it is something the data on this page can support.
The interesting thing about the attentional blink is not that reporting two things is harder than reporting one. It is the shape of the cost over time: a dip that is deepest two or three items after the first target, roughly two to five hundred milliseconds later, and gone by half a second or so. A general "doing two things is hard" account predicts no particular shape, so the shape is the finding.
If the second target arrives immediately after the first — the very next item — accuracy is often as good as at long lags, or nearly so. This is lag-1 sparing, and it is the single most awkward fact for any account in which the first target simply blocks a channel for a while. The usual reading is that two items arriving in quick succession can be taken in together, as one episode, and that the cost begins once that window has closed.
If you count second-target accuracy on trials where the first target was missed, you are mixing two quite different situations. In one, attention was engaged by the first target; in the other it was not. The convention is to report second-target accuracy given that the first was correct, and this tool reports both so you can see how much the choice matters.
Two-stage accounts say the first target occupies a limited consolidation process, leaving the second target in a fragile early representation that decays. Attentional-control accounts say the system over-invests in the first target or fails to suppress the item following it. Others treat the blink as a temporary loss of control over what gets selected, and some argue it reflects the cost of switching between task settings. All produce a dip. The demonstration on this page cannot separate them, and it would be a mistake to leave believing it had.