Working Memory Explained: Your Brain's Mental Scratchpad

Working memory is the small mental workspace that keeps a few pieces of information ready while you use them: the running total as you add up a bill, the question a colleague just asked while you hunt for the answer, the start of a long sentence while you read its end. It is used in most kinds of communication and problem solving, and differences in it go along with differences in reading and reasoning.¹

The workspace sits between the senses and everything you already know. Graham Hitch, Richard Allen and Alan Baddeley, two of whom proposed the best-known model of it in 1974, describe its short-term stores as a bridge between fast streams of incoming information and slower internal thought.² Because the space is small, the most useful everyday habit is to put less in it: a total jotted on paper or a note beside the screen is one thing fewer to hold.

Working memory means holding information while you use it

Working memory is the part of the mind that holds a limited amount of information in a state of heightened availability so it can be used in the task at hand. That is the “generic” definition in a 2017 review by Nelson Cowan, a psychologist at the University of Missouri who has studied working memory for more than 40 years. Researchers do not all mean the same thing by the term: Cowan counted nine distinct definitions in use, which is why two studies of “working memory” can measure different things.¹

Alan Baddeley’s own 1992 definition put manipulation, not just holding, at its center:

Working memory may be defined as the system for the temporary maintenance and manipulation of information, necessary for the performance of such complex cognitive activities as comprehension, learning, and reasoning

Repeating a phone number just long enough to dial it is holding. Working out a tip on a bill while the waiter waits is holding plus processing, and that combination is what many researchers now reserve the name for. Cowan traces the shift to a 1980 reading test that demanded both at once, after which tasks that only asked people to store something came to be called short-term memory tasks.¹

A practical way to use the definition: when a task feels harder than it should, list what you are keeping in your head while you work, such as a figure, a name or the next step. The strain often comes from the holding, not the thinking, and each item you write down frees a little room.

Three models, more alike than their diagrams

Three models dominate: Baddeley and Hitch’s multicomponent model, with separate stores for sounds and images run by an attention controller; Cowan’s own model, in which working memory is the currently active part of long-term memory; and Randall Engle’s attention-control view, which centers on holding goals against distraction.¹

The multicomponent model grew out of a simple experiment. Baddeley and Hitch had healthy adults keep a string of items in mind while they reasoned, read or learned: a full memory load caused only moderate interference, and a light one very little. Holding things, in other words, did not use up the whole system. They proposed a limited “central executive” that directs attention, served by a phonological loop that keeps words alive through silent rehearsal and a visuospatial sketchpad for images and locations. In 2000, Baddeley added an episodic buffer, a limited store that combines information from different senses and from long-term memory into integrated wholes.²

The evidence for separate stores came from pairing tasks. In one set of experiments, keeping a stylus on a moving spot of light disrupted a visual imagery task but not a verbal one.² Two tasks of the same kind clash; two of different kinds clash less, although both still draw on the same limited attention.

Cowan’s alternative drops the boxes: inside whatever part of long-term memory is currently active sits a small focus of attention. Engle’s group narrowed the idea to attention itself: using attention to keep goals active and to block distractions, alongside simple stores that need no attention.¹ How few items fit in that focus is the subject of the chunk limit on short-term memory.

Looking back after 50 years, Hitch, Allen and Baddeley conclude that the competing models agree on the basic findings to be explained and show more similarities than differences, and that the most important open question is how executive control works.² All three give attention a central place, which is the useful part day to day: whatever else competes for your attention takes room from what you are holding.

How working memory is tested, and why the test matters

Researchers measure working memory capacity mainly with complex span tasks, which make people store items while doing something else, then count how much they recall. The classic is reading span, introduced by Meredyth Daneman and Patricia Carpenter in 1980 and described in Cowan’s 2017 review: you read a series of sentences, judging each one, while remembering the last word of every sentence in order. Operation span, developed later by Engle and colleagues, swaps the sentences for simple arithmetic problems, each followed by a word or letter to remember. The juggling is the point, because reading and reasoning demand it too. Daneman and Carpenter’s reading and listening spans predicted reading ability far better than simple digit or word spans did.¹

Another popular task, the n-back, shows a stream of letters or pictures and asks whether each one matches the item shown a set number of steps earlier. But Thomas Redick and Dakota Lindsey, in a 2013 meta-analysismeta-analysis: A study that combines the results of earlier studies on the same question into one overall estimate. Pooling makes the estimate more precise, but it cannot repair the studies it pools: a meta-analysis of surveys is still survey evidence.Full entry in the glossary pooling 20 samples of mostly young adults, found that n-back and complex span scores were only weakly correlated, and concluded the two cannot be used interchangeably as measures of working memory.³

The lesson for anyone who has taken an online “working memory test”: a score on one game describes that game. People who practice n-back tasks consistently get better at other versions of the n-back, Redick and Lindsey note.³ Whether that spreads to everyday life is the question large trials of brain-training games and everyday memory set out to answer.

What a bigger workspace predicts, from reading to a wandering mind

People who score higher on working memory tasks tend to understand what they read better, score higher on intelligence tests and, when trying to concentrate, drift off-task less in daily life, according to two meta-analyses and a 2017 daily-life study. These are correlations, mostly in young adults: they show which abilities travel together, not that working memory causes the rest.

Reading comprehension is the classic example. A 1996 meta-analysis by Meredyth Daneman and Philip Merikle pooled 77 studies and found that tasks combining processing and storage predicted comprehension better than tasks that measured storage alone, and that math-based versions were good predictors too.⁴ Holding the start of an argument while you process its middle is exactly what reading asks of you.

Intelligence is linked, but not the same thing. A 2005 meta-analysis of 86 samples by Phillip Ackerman and colleagues put the average correlation between working memory and general intelligence at about 0.5, substantially less than a perfect match of 1.⁵ The two are related abilities, not one ability measured twice.

In a 2017 US study led by Michael Kane, 274 undergraduates in North Carolina did lab tests of working memory and then, for a week, were prompted eight times a day to report whether their thoughts had wandered. Working memory did not predict how often people’s minds wandered overall. It mattered through concentration: when people said they were trying hard to focus, those with higher capacity stayed on task more; when they were not trying, they wandered more.⁶

An illustration: of two colleagues reading a dense contract, the one who can keep clause 3 in mind while reading clause 9 spots the conflict. A one-page list of key terms beside the contract would do some of that holding for both of them.

Stress, anxiety and lost sleep shrink it for a while

Working memory is not a fixed amount. Meta-analyses find lower scores under acute stress, with higher self-reported anxiety and after a night or more without sleep, and brain recordings show interruptions pushing out what was held. Most of these effects are small to moderate.

The study

Moderate evidence

How much brief lab stress costs working memory

Grant Shields, Matthew Sazma and Andrew Yonelinas of the University of California, Davis pooled experiments that exposed healthy adults to a stressor, such as a mock job interview in front of judges or a hand in ice water, and compared their working memory with a no-stress condition. Stress lowered working memory by a small amount (a standardized effect of roughly 0.2), and the effect was consistent across studies. In a secondary analysis, the drop was clearer on high-load tasks and close to zero on lower-load ones.⁷

A small effect sounds trivial, but it lands on the hardest part of a stressful day, when the task is already heavy. The caveat is the setting: short stressors staged for an experiment, which say little about months of pressure at work.

Anxiety follows a similar pattern. Pooling 177 samples in a 2016 meta-analysis, the psychologist Tim Moran found that higher self-reported anxiety went with poorer working memory scores. The gap is small to moderate: visible when groups are compared, not a verdict on a single anxious person.⁸ Why an exam answer you know can vanish is explained in what test anxiety does to your thinking.

Going without sleep brings clearer drops. Julian Lim and David Dinges, whose 2010 meta-analysis pooled 70 articles, found moderate drops in working memory after 24 to 48 hours awake.⁹ Shorter nights add up too, as how sleep debt builds explains.

Interruptions work differently: they evict what you were holding. In two small 2010 US experiments by Wesley Clapp and colleagues, young adults lost a picture they were holding in mind when an interruption demanded their attention, and had to call it up again afterward.¹⁰ The wider toll at work is weighed in what interruptions cost your focus.

Myth

Your working memory is a fixed size you bring to every task.

Fact

Scores dip with acute stress, lost sleep and interruptions, and run lower with anxiety, so the same person can have better and worse days.

An everyday version: after a bad night and a tense meeting, you read a six-step instruction, turn to your screen and lose step three. That is the moment to write the steps down rather than try harder. Lapses of a different kind, such as confusion in familiar places or trouble with familiar tasks, appear among the early signs of dementia listed by the World Health Organization, and are worth discussing with a doctor or your local health service, wherever you are.¹¹

Work around the limit instead of stretching it

The approaches with research behind them work around working memory rather than enlarging it: move information out of your head, arrange material so it need not be stitched together mentally, and let knowledge turn many pieces into one.

Offloading is the everyday fix. In a 2016 review, Evan Risko and Sam Gilbert define cognitive offloadingcognitive offloading: Using something outside your head, such as a note, a calendar alert, a search engine or a chatbot, to cut the mental work a task needs. It usually helps you do the task, though you may remember less of what you handed over.Full entry in the glossary as physical action that cuts the mental demand of a task, from tilting your head to read a rotated image to setting a phone reminder, and note that it has improved performance in areas from memory to arithmetic.¹² Whether you then remember less of what you handed over is weighed in research on AI and cognitive offloading.

Layout matters too. Cognitive load theory, summarized in 2019 by John Sweller and colleagues, describes a split-attention effect: when a diagram and its explanation can only be understood together, learners must hold one while reading the other, which loads working memory and hampers learning. Putting the explanation on the diagram removes that work.¹³ The same applies to a spreadsheet whose notes live on another tab.

Knowledge changes the limit itself. The same review notes that once information is well stored in long-term memory, the capacity limits of working memory stop applying to it, and it can be brought in as a single unit.¹³ That helps explain why an expert reads a balance sheet at a glance while a newcomer tracks every line.

The bottom line

Working memory is the small, easily crowded space where you hold information while you use it, and researchers define and measure it in more than one way. It varies from person to person and from day to day: stress, anxiety, lost sleep and interruptions all push it down for a while. The best-supported way to do better is to ask less of it: write things down, keep what belongs together side by side and protect the moments that need your full attention.