Change Detection Test

What this measures

This measures visual working memory in this task: how many items you can hold across a blink, and how reliably you can tell whether the one that comes back is the one that left.

Play the measured block for your capacity in items, climb the ladder, or sprint the clock — nothing here is timed but the flash itself.

How to get better

You mostly get better at this task — spreading your attention over the whole lattice instead of fixating one corner, and resisting the urge to answer before you have checked what you actually held. That is a real, honest thing to practise. It makes you better at this task on this screen, and the honest reading of your capacity is the one you get when you are not rushing.

Questions

What is change detection?

An array of small items appears for a quarter of a second, the screen goes blank for about a second, and then one of the items comes back alone in the position it held. You say whether it is the same item or whether one of its features has changed. Steven Luck and Edward Vogel published the paradigm in Nature in 1997 and it is the standard way visual working memory capacity is measured.

What is Cowan's K, and where does my number come from?

K is the set size multiplied by your hit rate plus your correct-rejection rate minus one. In plain terms: how often you caught a change, plus how often you correctly said nothing had changed, minus the credit you would get for guessing. The screen prints both rates and the K each one produces, so you can do the arithmetic yourself. It is an estimate with a wide band, not a count of slots in your head, and it moves several tenths between runs.

Why does the quick test use two different array sizes?

Because one size measures badly at one end. Four items is too easy for someone with a large capacity and their score sits at the ceiling; eight items is noisy for someone with a smaller one. The mean of the two estimates is the practical compromise, and both are shown on the result screen so you can see where the headline came from.

Does my device speed affect this test?

No, and that is the point of it. The response is untimed — the probe waits as long as you do — and only the flash and the blank interval are on a clock, both fixed. A slow phone cannot move this number, which is unusual on a bench where most records are reaction times. Records are still kept per device class, because the array is a different size in your field of view on a phone than on a laptop.

I can't easily tell the colours apart — can I still play?

Yes, and at full strength. Every item carries a shape as well as a colour, and a change alters exactly one of the two, half and half, so half the changes are already shape changes. The menu also has a Shapes only toggle that makes every change a shape change. That is a different task, so it is kept as its own record rather than mixed in with the other.

Why does the sprint subtract for a wrong answer?

Because half the trials are changes, so a count of correct calls would reward tapping SAME on everything and scoring about half. Net calls — right minus wrong — puts a guesser at zero, which is what a guess is worth. It is the single most important honesty decision in the game and the result screen says it out loud.