Working memory: what it is, what moves it, and what the evidence actually shows
Working memory is how much you can hold in mind and use at once — not what you have filed away, but what you are actively keeping hold of right now. This page covers what it is, what genuinely moves it day to day, and what the research does and does not support about improving it.
We have tried to be straight about which parts of this are settled and which are not. Where the evidence is weak, this page says so rather than rounding up.
What working memory actually is
Think about the last time someone read a phone number aloud and you had to carry it as far as the keypad. Or the moment in a recipe where you are holding "two eggs, 180 grams, twelve minutes" while your hands are busy. Or reading a long sentence and keeping its beginning intact until you reach the end.
That is working memory: information held in an active, usable state while you do something with it. It is not a storage cupboard. It is closer to a desk — limited surface area, and everything on it needs actively keeping there.
Two things follow from that, and both matter more than they sound.
It is small. Not small compared with a computer — small compared with what people expect of themselves. Most of the frustration people feel about their memory in daily life is not a defect; it is the ordinary size of the desk.
It is fragile in a specific way. Because the material is being actively held rather than stored, an interruption does not just delay you. It can empty the desk. This is why being spoken to mid-calculation loses the calculation entirely, rather than pausing it.
What working memory is not
- Not your memory in general. Remembering faces, names, where you left your keys, or what happened last summer are different systems with different properties. Being poor at one says little about the others.
- Not intelligence. The two are related in the research literature but they are not the same thing, and neither substitutes for the other.
- Not a fixed quantity you were issued at birth. It varies between people, and it varies within the same person from day to day — enough that a single measurement is a snapshot of a day, not a portrait of a person.
What moves it day to day
Established
The single most useful thing to understand about working memory is how much of its day-to-day variation has nothing to do with your capacity, and everything to do with the conditions you are measuring it in.
Interruption is the big one, and it is worth separating from "distraction" in general. Material being actively held is vulnerable in a way that stored material is not. A notification arriving mid-thought does not cost you the seconds it takes to read it; it can cost you the thing you were holding.
Doing two things that compete for the same resource. Holding a sentence in mind while someone talks to you is harder than holding it while music plays without words, because the two verbal streams compete directly.
Sleep, stress, illness and time of day all change how much you can hold. This is why people notice their memory is "worse" during a hard month and conclude something has changed about them, when what changed was the month.
Familiarity with the material. Holding seven random digits and holding seven digits that happen to be a date you know are not the same task, even though they look identical. Meaning does part of the work for you.
Why "just remember more" does not work — and what does
Established
You cannot make the desk bigger by trying harder. What you can do is put things on it more efficiently, which is why the same person can hold three unrelated items and also hold a much longer string that has structure in it.
This is the difference between holding 4 9 1 7 6 2 8 3 and holding 49 17 62 83 — the same digits, reorganised into fewer units. People who do well on span tasks at the harder lengths are almost always doing something like this, whether or not they could describe it.
The practical version of that, outside a test: externalise. Write the number down. This sounds like giving up, and it is the opposite — the desk is small by design, and the useful skill is knowing what does not need to be on it.
Does brain training improve working memory?
Established — this is the best-evidenced part of this page, and the most commonly misrepresented
This is the question most people arrive with, and it has a clearer answer than the marketing around it suggests.
The most comprehensive review of the brain-training literature — Simons and colleagues, 2016, published in Psychological Science in the Public Interest — examined the published peer-reviewed studies cited by leading brain-training companies themselves, on the reasoning that these represent the evidence best supporting their claims. Its finding, in the authors' own framing, was a three-tier one:
- Extensive evidence that brain-training improves performance on the trained tasks.
- Less evidence that it improves performance on closely related tasks.
- Little evidence that it improves performance on distantly related tasks, or on everyday cognitive performance.
In plain terms: practise a memory game and you will get better at that memory game. Whether that carries over to anything you actually care about — remembering what your colleague just said, keeping track of a recipe, following an argument — is where the evidence thins out sharply.
The review also found that many of the published studies had shortcomings in design or analysis that prevented firm conclusions, and that none of the cited studies met all of the best practices the authors identify as necessary to draw clear conclusions about everyday benefit.
None of this means training does nothing. It means the specific claim that it makes you generally sharper in daily life is not currently well supported — and that is a different statement from "it has been disproved".
Why this matters commercially, not just scientifically
Established
In January 2016 Lumos Labs, the company behind Lumosity, paid $2 million in redress to settle Federal Trade Commission charges of deceptive advertising. A $50 million judgment was suspended based on the company's financial condition.
The charges concerned advertising claims — among them that the product could stave off memory loss, dementia and Alzheimer's disease. The FTC's position was that the company did not have the science to support what its advertising said.
It is worth being precise about this, because it is often repeated loosely: this was a settlement over marketing, not a finding that the games were harmful, and "fined $2 million" is not an accurate description of what happened.
If you are worried about your own memory, an online test is not the tool for that question — and neither is an article. A doctor is. Nothing on this page can tell you whether something is medically wrong, and a low score on any test here does not indicate a condition.
Measuring your own working memory
There is a reason we built a test before we wrote this page, and a reason the test is deliberately narrow.
Digit span — hearing or seeing a sequence and giving it back in order — is one of the oldest and best-established ways of measuring this. It is not a proxy invented for a website. It also has an unusual property for an online measure: it is robust to the things that ruin most browser-based tests, because it measures how much you held rather than how fast you clicked. Screen refresh rate and input lag do not meaningfully affect it.
About two minutes. No account, no email. Nothing is stored — it runs in your browser and your result never reaches us. The scoring rule and the exclusion rules are published in full on the page.
It will give you a number and an explanation of what that number does and does not mean. It will not tell you how you compare with other people, because we have no reference sample and will not pretend otherwise. Our methodology page explains exactly where we are on that.
What we do not know
Open questions — stated because leaving them out would make this page look more certain than the evidence is
- Whether working-memory training transfers to anything meaningful in daily life. This is genuinely unsettled rather than settled-and-negative. The evidence for far transfer is weak; that is not the same as evidence of no effect.
- How much of the day-to-day variation is capacity and how much is attention. These are hard to separate cleanly, and a bad day on a span task could be either.
- What a single measurement is worth. One test on one day, in one set of conditions, is a noisy estimate. We would rather say that than sell it as a portrait.
Sources
- Simons, D. J., Boot, W. R., Charness, N., Gathercole, S. E., Chabris, C. F., Hambrick, D. Z., & Stine-Morrow, E. A. L. (2016). Do "Brain-Training" Programs Work? Psychological Science in the Public Interest, 17(3), 103–186. doi:10.1177/1529100616661983
- Federal Trade Commission (5 January 2016). Lumosity to Pay $2 Million to Settle FTC Deceptive Advertising Charges for Its "Brain Training" Program. FTC case 132-3212, including the complaint and stipulated final judgment.