The Blink Attention Takes

A field note on attention ·

The Blink Attention Takes

When your mind catches something that matters, it goes blind for about half a second. A second surprise arriving inside that window is never seen. AI is about to deliver surprises faster than any mind can file them, and the order they arrive in will decide which ones you live.

Catch one, and the gate closes behind it.

A stream of letters passing through a glowing gate; the first gold letter is caught and a second target behind it fades to grey.
100 ms900 ms

Gap of 300 ms: inside the blink. The second surprise is usually never seen.

Letters flash past one at a time. The gold one is the first surprise. The blue one is the second. Move the gap and watch whether the gate is open when the second one arrives. The timing follows the shape of the classic result; it is an illustration, not data.

The banner says the offer came through. You read it twice, standing in the kitchen, grinning at your phone. The message from your sister that landed right behind it, the one about your father, you will not see until almost seven.

It was on the screen. It was in the same stack, under the same thumb. You did not ignore it. You never had it. Your mind had just caught something, and for a while after a catch, the gate is shut.

That gate is one of the most reliable facts in the study of attention, and almost nobody building AI is thinking about it.

The half second after

In 1992 three psychologists, Jane Raymond, Kimron Shapiro and Karen Arnell, flashed letters at people in a single spot on a screen, about ten a second. The task was simple. Name the one white letter, then say whether an X appeared somewhere after it.

If the X came well after the white letter, people saw it. If it came between roughly two hundred and five hundred milliseconds after, they very often missed it. It was right there, as bright as every other letter, and it did not register. Stranger still, an X arriving immediately after the white letter was often caught, as if it slipped through the door before it closed.

They called it the attentional blink. The usual explanation is that catching something is work. A flicker on the retina has to be turned into a thing you can report, and that turning takes time and occupies the gate. Whatever arrives while the gate is busy flickers and fades, unreported and, as far as you will ever know, unseen.

THE BLINK slips through seen missed TIME SINCE THE FIRST CATCH → (ABOUT 0 TO 800 MS)
The shape of the result, not its data: the second thing is seen right away, lost in the middle, and seen again once the gate reopens.

Noticing is not free. Every catch closes the gate behind it.

Two ways through a closed gate

The blink has exceptions, and the exceptions are the most useful part.

The first is emotion. In 2001 Adam Anderson and Elizabeth Phelps showed that emotionally charged words get through the blink far more often than neutral ones, and that people with damage to the amygdala lose most of that advantage. The gate has a side door, and the body keeps the key to it.

The second is grip. In 2005 Christian Olivers and Sander Nieuwenhuis found that the blink got smaller when people listened to music or let their minds drift a little during the task. Two years later, a team led by Heleen Slagter found that three months of intensive meditation training shrank it too. The blink is not only a limit of the machinery. Part of it is over-investment: attention clutching the first catch so hard that nothing else can get in.

In my own terms, a surprise is a comparison. An Actual arrives, the comparison machine brings an Expectation, and the felt jolt comes from the ratio between them. While the machine is still filing the first Actual into history, the second one arrives with no comparison ready for it. No comparison, no surprise. It happens, and it is not had.

The lab findings are the record. Reading the blink as a busy comparison machine is my argument, not theirs.

The blink has bigger cousins

The lab blink lasts half a second. I think the same shape repeats wherever filing a surprise takes time. A diagnosis in the morning makes the afternoon’s email land on no one. A big week buries the next piece of news. People whose job is timing have always known this. On September 11, 2001, a British government adviser emailed colleagues that it was a very good day to get out anything they wanted to bury. She was condemned for it, and she was right about the mechanism.

That is the principle under everything that follows. A mind can only have one surprise at a time, and each one it has costs it a window in which the next one is lost. Which surprise you have is decided less by importance than by order.

What AI changes: the spacing

Here is the input that moves. For most of history, the world spaced your surprises for you. The mail came once a day. A phone call came when someone found the time to make it. A reply took a week. Friction was a scheduler, and it was free.

Agents remove the friction. Forty errands handed off on Monday can all resolve inside the same ten minutes on Friday. The refund, the appointment, the contract, the school, the offer. They do not trickle back. They land together.

The consequence most people miss is that the problem is not volume. You can survive a flood of trivia. The problem is that inside a burst, the first catch wins and the next few disappear, and the one that escapes the blink is the one that shouts, not the one that matters. Without spacing, the order of arrival quietly becomes the order of your life.

Your delegate will not decide what you know. It will decide what you notice, by deciding what comes second.

Your phone shows one line: your sister needs to talk about Dad, and she is free now. You call. Twenty minutes later, when you set the phone down, it shows you the offer.

Five predictions

  1. 01

    Spacing becomes the delegate’s most valuable skill.

    A good delegate will hand you one consequential surprise at a time and then leave a gap. It will be judged by what you actually caught, not by what it delivered.

  2. 02

    The daily digest becomes a sequence.

    A list of ten results delivered at five o’clock is ten surprises inside one blink. The top item eats the rest. Digests will be rebuilt as an order with pauses, read one line at a time.

  3. 03

    Burying becomes automatic, and so does digging.

    Fee changes, recalls and new terms will be timed to land just behind good news. Part of your agent’s job will be to notice what arrived inside your blink and bring it back when the gate is open.

  4. 04

    Everything will be dressed as urgent, and it will stop working.

    Emotional things get through a closed gate, so senders will make everything feel emotional. The side door will jam with traffic. The only thing left that can tell what truly matters will be a delegate that knows you.

  5. 05

    A loose grip becomes a way of seeing more.

    The blink shrinks when attention holds its first catch lightly. People who can take good news without clutching it will catch more of what comes second. That will read less like calm and more like perception.

Where I am already living it

I told my own AI that it decides, because every question it sent back pulled me into the middle of things I had handed over. I wrote about why handing things over only helps once I can stop holding them in Permission to Forget, and about how only the ending of a delegated task survives in memory in The Ending Attention Keeps. The blink adds one more rule to the same list. Endings are what I keep, but I can only keep one at a time. A delegate that finishes everything at once has not saved me attention. It has chosen, by accident, which of my endings I get to have.

The claim ledger

Already true
The attentional blink is among the most repeated findings in attention research. Emotional items get through it, and it shrinks when attention is relaxed or trained. Agents already finish many tasks in parallel and report back within minutes of each other.
What has to happen
The half-second gate has to have longer cousins: filing a consequential surprise has to cost minutes or hours of reduced noticing, in the same shape. And agents have to become a main source of the news that matters in a person’s day.
Where I am probably wrong
If the blink stays a half-second effect and people file real-life surprises side by side without loss, then bursts cost nothing and order does not matter. Then a delegate’s job is delivery, not choreography, and this whole piece is about a lab curiosity.

Back to the kitchen

Friday, 4:15. The offer was real. So was the message about your father. Both arrived. Only one was had, and not because it mattered more. It was first.

The second thing does not need to be louder. It needs a gap.

Give it one.

Background

  • Raymond, J. E., Shapiro, K. L., and Arnell, K. M. (1992). Temporary suppression of visual processing in an RSVP task: an attentional blink? Journal of Experimental Psychology: Human Perception and Performance.
  • Anderson, A. K., and Phelps, E. A. (2001). Lesions of the human amygdala impair enhanced perception of emotionally salient events. Nature.
  • Olivers, C. N. L., and Nieuwenhuis, S. (2005). The beneficial effect of concurrent task-irrelevant mental activity on temporal attention. Psychological Science.
  • Slagter, H. A., and colleagues (2007). Mental training affects distribution of limited brain resources. PLoS Biology.
  • John Rector, The Reality Equation.
  • John Rector, Surprise Has a Noise Floor.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.