How many things can you hold in mind at a glance?
About three, for most people. In the visual working memory test data that Balaban, Fukuda and Luria published in 2019, the 2,040 people scored here held a median of about 2.7 items after one brief look, the middle half between roughly 2.1 and 3.3, and only the top 5% more than about 4.1. What you see in a blink has to fit in that small store before you can compare it with anything; the rest is simply gone.
This test uses the change-detection task that Steven Luck and Edward Vogel introduced in 1997. A tin of coloured buttons opens for about 0.4 seconds and shuts. After a one-second wait it opens again with one button ringed, and you answer Same or Changed: has that button swapped to a different colour? You start with 2 buttons and every level adds one, with two lids at each level and a fair coin deciding each time whether the ringed button changed. There is no top to aim for: you climb until your second wrong answer, about 11 lids on average, and nine runs in ten are over within 17. From 9 buttons on, the buttons get a little smaller and a colour can appear twice, once with four holes and once with two. Your score is the capacity K that best explains all of your answers, under the standard model behind Cowan’s K: holding K of N items, you know the probed one with chance K/N and otherwise guess.
It is not an IQ test, and it doesn’t test attention, check for memory problems or diagnose anything. A run of about 11 lids is short. To rank you fairly, each of the 2,040 people in the comparison is put through this same ladder under the same model, and you are ranked among those runs. On those runs the score lands within about 0.45 of the person’s full lab score half the time and within 1.12 nine times in ten, and short runs pull scores toward the middle: even clearing 10 buttons without a miss reads 3.5, which is more than 97% of runs. The study flashed coloured squares for 0.15 seconds over 120 trials in a lab, while here it is buttons, a longer look and whatever screen you have. Red, green and orange buttons are harder to tell apart for colour-blind players.
What the test measures
This test measures one thing: how many coloured items you can keep in visual working memory after a single glance.
The task
- Single-probe change detection (Luck & Vogel, 1997): a tin of buttons opens for about 0.4 seconds, stays shut for one second, then reopens with one button ringed
- Answer Same or Changed: has the ringed button turned a different colour?
- Start at 2 buttons, one more each level, two lids a level, no fixed top; whether the ringed button changed is a fair coin every time
- Every wrong answer is a miss, and the second miss ends the run
- Nine clearly different colours, so a change is a new colour, not a shade
- After every answer you see whether you were right and what the button really did
Your result
- Your capacity K: how many things you held at a glance, to one decimal
- Where you stand among 2,040 people from Balaban, Fukuda & Luria (2019), each run through this same ladder: the share of them you held more than, on a chart of those runs
- Your levels one by one: which lids you got right and where the run ended
- A plain note on how far a score from so few lids can wobble
- A shareable card and a link that challenges a friend to hold more
Before you start
- About a minute: about 11 lids on average, nine runs in ten within 17
- Works on a phone or a computer; on a keyboard, the left and right arrow keys answer
- Sound is optional: it only marks the lid and right or wrong
- Free, with no sign-up
- If you are colour-blind, changes between red, green and orange will be harder
Visual working memory FAQ
Is this test free? Do I need to sign up to see my result?
Yes, it’s free, and there’s nothing to sign up for: no account, no email, no payment. Your result appears on this page the moment you finish, and you can share it or play again straight away.
How many things can the average person hold in working memory?
About three things after a brief look. In Balaban, Fukuda and Luria’s 2019 change-detection data, the 2,040 people scored with the standard capacity formula had a median capacity of about 2.7 coloured items and a mean of 2.71; the middle half held roughly 2.1 to 3.3. People did slightly better with 4 items on screen (mean 2.81) than with 8 (mean 2.60).
Isn’t short-term memory seven items, plus or minus two?
Not when you can’t group or rehearse what you see. George Miller’s 1956 “seven, plus or minus two” was, as Nelson Cowan argued in a 2001 review, more a rough estimate than a real capacity limit. When people can’t chunk items or repeat them to themselves, the limit Cowan found across many studies averages about four chunks. Brief flashes of colour, as in this test, are one of those conditions.
How is a visual working memory test scored?
Most change-detection studies use capacity K: the number of items on screen times the difference between the share of real changes spotted and the share of “changed” answers when nothing had changed. That needs many trials at each set size. With about 11 lids in a run, this test estimates the same K from every answer under the model behind that formula: holding K of N items, you know the probed one with chance K/N and otherwise guess. The 2,040 people it compares you with are run through the same ladder under the same model.
Why did I get a different score the second time?
A run is about 11 lids, so scores move from run to run. When the study’s own participants are put through this ladder under the same model, the score lands within about 0.45 of their full 120-trial lab score half the time and within 1.12 nine times in ten, and their rank moves by under 17 percentile points half the time. A few runs tell you more than one, and a different screen or a distraction can shift the number too.
Does this memory test measure IQ?
No. The button-tin test measures one narrow thing, how many colours you keep in mind after a 0.4-second look, and it is not an intelligence test, an attention test or a check for memory problems. A low score from one short run on a phone says little about how your memory works in everyday life; anyone worried about their memory should talk to a doctor.
How long does the visual working memory test take, and does it work on a phone?
About a minute, about 11 lids on average, on a phone or a computer, with no sign-up. Sound is optional. On a keyboard the left and right arrow keys answer Same and Changed. Colour-blind players will find changes between red, green and orange buttons harder, which lowers the score for reasons that have nothing to do with memory.
Where the numbers come from
Your rank compares you with 2,040 people from the open data of Balaban, Fukuda & Luria (2019), scored there as the mean of K with 4 and with 8 items. Each of them is rebuilt from the study’s percentile table and put through this ladder under the same change-detection model, and you are ranked among those runs. Eight of the study’s experiment groups, whose logs look as if the response keys were coded the other way round, are left out as whole groups. Their set-up (coloured squares shown for 0.15 seconds, 120 trials) differs from this one. The site does not publish its own players’ data.
- Balaban, Fukuda & Luria (2019), Cognition 191, 103984 — What can half a million change detection trials tell us about visual working memory?
- Balaban, Fukuda & Luria — open trial data for the 2019 study (OSF)
- Luck & Vogel (1997), Nature 390, 279–281 — The capacity of visual working memory for features and conjunctions
- Cowan (2001), Behavioral and Brain Sciences 24, 87–114 — The magical number 4 in short-term memory: a reconsideration of mental storage capacity
How the score works
Change detection
A brief look at a set of items, a short blank, then one item to judge: same or different. Because you can’t know which item will be asked about, you have to hold the whole set.
Capacity K
Holding K of N buttons, the ringed one is in memory with chance K/N; otherwise you guess, right half the time. Your K is the value that best explains all your rights and wrongs: the average K of the 2,040 study participants, each weighted by how likely they were to answer exactly as you did.
Why guessing doesn’t help
A guess is right half the time, so luck can carry you a level or two, but no further: every wrong answer counts toward the two that end the run, and the score weighs every answer, not the level you reached. Answering at random gives a low K.