What this measures
Three people each show their preferences for five seconds. Then you have to name the one thing they all agree on.
This is working memory under a binding load. Remembering three lists is hard enough; the task actually requires keeping each list attached to the right person while you intersect them. Drop the binding and you can recall every item and still fail.
Why five seconds is not much
Three sets of preferences is well past the four-or-so items most people hold reliably in working memory. You cannot store it all — so the task is really asking what you choose to encode.
The efficient answer is not to memorise three lists. It is to compute the intersection as you read, which is a much smaller thing to hold.
Strategy
- Read the first person as your candidate set. Everything they like is still possible.
- On the second person, eliminate. Drop anything the first liked that the second does not. Your list shrinks immediately, usually to two or three items.
- On the third, you are checking, not memorising. Often only one candidate survives before the display even finishes.
- Do not go back and forth. Trying to re-check person one after seeing person three is where the five seconds disappear.
- If you lose the thread, guess from what you kept. A partial intersection still beats a random choice.
Done this way, you never hold more than about three items — comfortably inside capacity. Done by memorising, it is close to impossible. That gap between strategy and effort is the point of the task.
Why the task is built this way
Set intersection under time pressure appears in cognitive assessments because it looks like a memory test but is actually a strategy test. Two people with identical memory spans will score very differently depending on whether they realised they could filter incrementally.
It also has an obvious real-world analogue — working out what a group can agree on while the information is still arriving — which is why it tends to be framed as a social scenario rather than as abstract sets.
Frequently asked
Q. Is there always exactly one answer?
A. Yes. Rounds are generated so the intersection is unique — if two items both worked,
the round would be unfair, so those are filtered out.
Q. Should I try to memorise all three?
A. No, and this is the main thing the task teaches. Filtering as you go is both easier
and more accurate.
Q. I remember the items but not who liked what.
A. That is exactly the binding failure the task is designed to produce. Switching to
incremental elimination removes the need to bind at all.
For pure span without the binding load, try flash memory.