Working memory tests
Holding a phone number in your head while you cross the room — and rewriting it the moment somebody reads you a different one — is working memory doing its job. The classic ways of measuring it run free and online in your browser here, with no account, and this page says what a span score is actually worth.
What working memory is
Working memory is the small amount of information you can hold in mind and work on at the same time. It is not the store where last year lives — that one is enormous and it lasts. This is the desk, not the archive: a handful of items, held only while you are using them, gone the moment your attention moves.
It is not one store either. Verbal material and visual-spatial material behave like separate systems — people can be good at repeating digits and poor at repeating a tapped sequence, and the other way around. That split is a large part of why the tests below use deliberately different material: digits arriving one at a time, letters arriving in a stream, and blocks flashing on a board.
What it is not: a photographic recall that adults reliably have (they do not), the same thing as intelligence (related, not identical), or a screening tool for anything. A short span in a browser tab is a short span in a browser tab, on one evening, on one device.
What these tests are not
- Not an IQ score
- Span tasks correlate with reasoning tests, which is exactly why they get sold as IQ tests. They are not. Nothing here is standardized, administered under controlled conditions or scored against a normative sample, and no number on this site is an IQ.
- Not a screen for anything
- These tasks sit inside neuropsychological batteries, where they are one measure among many and read by a clinician alongside a history. On their own they screen for nothing — not dementia, not ADHD, not the after-effects of an injury. If something about your recall genuinely worries you, that belongs with a doctor, not with a web page.
- Not a percentile
- There is no "you beat X% of people" here, because there is nothing honest to compute it from: we publish no norms and hold no representative sample. The only comparison offered is with your own earlier runs.
How working memory is measured
Two questions have shaped the whole field. How much can be held at once — which gives you span tasks, grown one item at a time until they break. And what happens when it has to be updated while you hold it — which gives you the n-back.
- 1887
The first span
Joseph Jacobs asks schoolchildren to repeat back strings of digits, one digit longer each round, and records where each child stops. The method — grow the list until it breaks, then call the last success the span — is still exactly how a span is measured today.
- 1958
Kirchner needs a moving target
Wayne Kirchner wants to measure how people handle information that keeps changing, not a list that sits still. His answer is to keep a stream running and ask whether the current item matches the one n steps back — so the answer has to be updated on every single step.
- 1972
Corsi builds the spatial twin
Philip Corsi, working on his doctorate at McGill, needs a visual-spatial version of the familiar "repeat these digits" task. His answer is a board of nine blocks tapped in sequence — no words, no numbers, nothing that can be rehearsed under your breath.
- 1974
The working memory model
Alan Baddeley and Graham Hitch propose that holding information in mind is not a single buffer but a system: a verbal loop, a visual-spatial sketchpad, and a controller sharing attention between them. It is the reason nobody expects one span score to cover both kinds of material.
- Since
The claim that did not hold up
A 2008 study by Jaeggi and colleagues reported that training on a harder dual n-back raised scores on tests of fluid reasoning, and an industry grew on it. Larger reviews and replications since have found little or no transfer beyond the trained task itself. People do get better at n-back; whether that carries anywhere else is, at best, unsettled.
Free working memory tests you can take here
Three working memory tests are published so far — N-Back Test, Corsi Block Test and Digit Span Test. Each is free, runs online in this browser, and shows your own result the moment you finish.
All of them run client-side, need no signup and upload nothing. Take them in any order — they use different material on purpose, so a result on one does not predict the others.
How to read your result
A span is a small integer and an n-back result is a handful of counts. Both invite far more meaning than they hold.
-
The comparison that works
Your run against your own earlier runs, on the same device, under similar conditions. Comparing a span to a number you read somewhere is comparing two different tasks that happen to share a name.
-
Hits are only half of it
On an n-back, pressing Match on everything produces a perfect hit rate and means nothing at all. What a run is worth is hits and false alarms read together — the wrong presses tell you as much as the right ones.
-
One run is noisy
A span can move by a whole item between two runs on the same evening, from attention and from the luck of the sequence. Take several before you believe any of them.
-
It is not a diagnosis
These tasks appear in clinical batteries, but always beside other measures and always interpreted by someone qualified. A short run here describes a short run here.
Working memory in the NEURA app
NEURA on iPhone works this domain on purpose rather than by accident. Four of the 19 games are built for it, two more touch it while measuring something else, and the daily session carries two task formats that land there.
- Games built for it
- Corsi Blocks
- N-Back
- Word Recall
- Word Sweep
- Games that also work it
- Pair Match
- Tower Moves
- Daily-session tasks
- Grid recall
- Symbol span
A daily session is 20 tasks, about five minutes, and every answer is scored into the domain its task belongs to. Those domain results roll up into one number, which opens once you have completed three daily sessions — see Focus Index for what that number is and what it is not. A game’s primary domain takes the full weight of its result and the domains it also touches take a share. The exact weighting is not published, because the useful reading is the direction your number moves over weeks — not the arithmetic that made it.
Questions about working memory
What is a working memory test?
A working memory test is a short task that loads a small amount of information into your head and asks for it back, usually growing the load until you fail. The two classic shapes are span tasks, where a sequence gets one item longer each round until you can no longer repeat it, and updating tasks, where a stream keeps running and you report whether the current item matches one from n steps back. The digit span and the Corsi block test are the first kind — one verbal, one spatial; the n-back is the second.
What is the difference between working memory and short-term memory?
Short-term memory is the holding — keeping a few items available for a few seconds. Working memory is holding them AND doing something to them at the same time: reordering, updating, comparing them against whatever just arrived. A plain span task leans on the holding — which is why a forward digit span is closer to short-term memory than to working memory, and why turning the list around in backward mode is the part that earns the name. An n-back leans on the work throughout, because every new letter forces the answer to be recomputed. That is why the tests on this page do not measure quite the same thing.
What is a good span score?
There is no single good span score, and anyone quoting one for a browser test is guessing. A span depends on the material, the timing, how many failures end the run, your input device and how tired you are — change any of those and the figure moves without your recall changing at all. What is meaningful is your own span across several runs on the same setup.
Which working memory test should I take first?
Start with the Digit Span Test if you want the oldest and plainest of them: digits arrive one a second and you type them back, forward or in reverse, and the sequence grows until you lose it. Take the Corsi Block Test for the same ladder without any words in it — flashing blocks instead of digits, which is the visual-spatial side of the same question. Take the N-Back Test if you want the harder and more famous one, where nothing sits still and the answer has to be updated on every letter. They use different material deliberately, so doing well on one says little about the others.
Can working memory be trained?
You get better at the working memory task you practice — that part is not in doubt. Whether it generalizes is: the 2008 n-back training result made this the most argued-over claim in the field, and larger replications since have found little transfer beyond the trained task. NEURA trains this domain every day because a measured habit is worth having, not because anyone can promise you a bigger span.
One evening’s span is a number. A month of them is a picture.
A browser span measures one evening. NEURA on iPhone works this domain every day, adapts the difficulty to you, and turns the runs into a Focus Index you can watch move over weeks. Your first session is free.
