What Is Working Memory? Definition, Examples & Baddeley's Model

Everyone assumes the people who juggle more information have a bigger workbench. The evidence points somewhere less flattering and far more useful.

Test Your Working Memory Below (N-Back Test) ↓

Multiply 23 by 17 in your head.

Somewhere around the middle — holding 161 while you work out 230, then adding them without dropping either — you'll feel a specific kind of strain. Not the strain of not knowing how. The strain of not having room.

That's working memory: the system that holds information and operates on it at the same time. Where short-term memory keeps a phone number intact across a room, working memory recites it backwards, checks it against another, or dials it mid-conversation. Storage plus manipulation, on a surface with very little room.

Baddeley's Model: A Team, Not a Box

The dominant account came from Alan Baddeley and Graham Hitch in 1974, and it replaced the single-box picture with a small staff.

The phonological loop handles sound and language — the inner voice rehearsing a number, or reading this sentence to you now. Its contents fade within seconds unless that voice keeps refreshing them.

The visuospatial sketchpad is the mind's eye: mentally rotating furniture, retracing a route, holding a layout. It runs in parallel with the loop, which is why maintaining an image alongside a sentence is often easier than maintaining two competing verbal streams. The Spatial Span Test probes this component fairly directly.

Above both sits the central executive — the manager that allocates attention, switches between tasks, and blocks interruptions. Baddeley later added the episodic buffer, where material from the loop, the sketchpad, and long-term memory gets bound into integrated representations.

Notice which part of that staff has no storage of its own. The executive doesn't hold anything. It decides. That turns out to matter enormously.

The Limit Everyone Feels

Every part of this system is capacity-limited, and the limits are low — a few chunks, competing for one pool of attention. The failures are so familiar they barely register as failures: losing the thread of a sentence with too many clauses, forgetting the third instruction while executing the first, arriving at the fridge with no idea why.

What's striking is how much this one measure predicts. People with higher working memory capacity tend to do better on reading comprehension, mental arithmetic, and fluid reasoning — solving problems they've never seen before. In schools it's among the stronger predictors of academic performance, and children with working memory difficulties often struggle despite entirely normal intelligence, a pattern our guide to working memory and ADHD takes apart in detail.

Which raises the obvious question: what exactly do high-capacity people have more of?

The Doorman

The intuitive answer is space. More room on the bench, more items held.

In 2005, Edward Vogel, Andrew McCollough and Maro Machizawa tested that directly. Using a neural signal that tracks how much is being held in visual working memory moment to moment, they gave people displays containing both relevant items and irrelevant ones, and watched what each person actually stored.

High-capacity individuals held the relevant items and largely left the rest. Low-capacity individuals took in the distractors too — so much so that under some conditions they were holding more total information than the high performers. Not more of the useful kind.

Filtering efficiency varied a great deal between people, and it tracked their measured capacity closely. That study looked at visual short-term memory under distraction, so it isn't the last word on every part of the system — and the bench really is small either way. But it shifts where the interesting variation lives: effective capacity depends not only on how much can be held, but on how efficiently irrelevant material is kept from consuming space that was never plentiful. A doorman, working a room with a hard limit on occupancy.

Storage, updating, and comparison, all at once. See how many steps back you can track ↓

Why the N-Back Is Widely Used

Simple span tasks — repeating digits, as in the Digit Span Test — mostly measure passive holding. They capture the shelf rather than the workshop.

The N-back captures one demanding part of the job: continuous updating and comparison. Items stream past one at a time and you respond whenever the current one matches the item from N steps back. Doing that means holding the last N items, comparing each arrival against the oldest, then dropping that one and shifting everything along — with no pause to catch up and nothing allowed to linger.

At 2-back it's demanding. At 3-back it's humbling. That's why it became a workhorse in cognitive neuroscience, and it's the test embedded below — though it's worth knowing that it isn't the whole picture. Complex-span, backward-recall and visual-capacity tasks each tap partly different things, and performance on one doesn't always predict performance on another.

Can You Train It?

Here the doorman idea earns its keep, because it explains a twenty-year disappointment.

A 2008 study by Susanne Jaeggi and colleagues reported that weeks of dual N-back training improved fluid reasoning, and an entire industry grew out of that result. Replications since have been mixed. A 2016 meta-analytic review by Melby-Lervåg, Redick and Hulme, covering 87 publications, found reliable gains on working memory measures themselves but no convincing evidence of transfer to reasoning, verbal ability, reading comprehension or arithmetic once trained groups were compared against active controls.

That pattern makes sense if capacity was never a container to be stretched. You can get very good at running a particular door, and that skill doesn't obviously generalise to every other door in your life.

What does help is working with the constraint rather than against it. Chunk information into meaningful units. Write intermediate results down instead of holding them. Cut the background load — the notification, the second conversation, the open tab — because those are exactly the items your doorman is spending capacity on. Single-task anything that matters. Experts don't hold more raw items than novices; their units are simply bigger, so fewer of them are needed. Cognitive Train's free brain training exercises load this system deliberately, which is a reasonable way to get familiar with where your own limit sits.

If you want to see how this system sits alongside the others it works with, the types of memory guide maps the full architecture, the Memory & Recall section collects the tests that target each piece, and the wider brain tests collection covers the abilities that lean on it.

Test Your Working Memory

The N-back below starts gently and scales. You'll watch a stream of items and flag each one that matches the item N positions earlier — 1-back first, then 2-back and beyond as you improve.

Expect it to feel harder than it sounds. That difficulty is the executive, the loop, and the updating machinery all running at once, which is the point. And pay attention to what happens when something distracts you mid-stream — that moment is the doorman failing in real time.

For the passive-storage side of the system, measure your span with the Short Term Memory Test — or see the two systems set directly against each other in working memory vs. short-term memory.