What is working memory? Everyday examples and the model behind the term
· 6 min read · Tyamka team
Working memory explained with everyday examples, the Baddeley and Hitch model, how small its limits are, and which three games on this site rely on it.
Working memory is the small, short-lived store you use to keep something in mind while you do something with it. You use it when you read a code from a text message and type it into another app, when you add up a bill in your head, or when someone gives you directions and you keep "second left, then right after the pharmacy" in mind while walking. It holds only a few things, and they fade within seconds unless you keep repeating them or paying attention to them. This article explains the idea, the best known model, the limits research has found and how they are measured. If you want to try a task that relies on it, the sequence memory test is one of three on this site.
Working memory and short-term memory
The two terms are often used as if they meant the same thing, and in casual speech that is fine. Researchers usually draw a line. Short-term memory is about holding information for a short time, for example repeating a list of digits right after hearing it. Working memory adds the "working" part. You hold something and use it at the same time, such as keeping a number in mind while you add another to it, or remembering the start of a sentence while you read the end.
Long-term memory is something else again. Your home address, the rules of chess and the plot of a film you liked are stored there, and they do not fade in seconds. Working memory often pulls things in from long-term memory. When you remember 1984 as a year rather than four digits, you are using something you already know to make the load smaller.
The Baddeley and Hitch model
In 1974 Alan Baddeley and Graham Hitch published a chapter called "Working memory" in the book series The Psychology of Learning and Motivation. They proposed that working memory is not one store but several parts that work together.
- The phonological loop. Holds sounds and words, and keeps them fresh when you repeat them to yourself, like a phone number you mutter on the way to the phone.
- The visuospatial sketchpad. Holds images and positions, such as where the chairs are in a dark room or the shape of a route on a map.
- The central executive. Directs attention between the other two and decides what to focus on. It is the least understood part of the model.
In 2000 Baddeley added a fourth part, the episodic buffer, in a paper in Trends in Cognitive Sciences. It is meant to link information from the other parts and from long-term memory into single episodes, for example a sentence together with the face of the person who said it.
The model is not the only one, and researchers still argue about how the parts fit together. It remains the most widely taught description and a useful way to think about everyday examples.
What makes it harder
The model explains some everyday experiences well. Holding a phone number while someone talks to you is hard, because the talk competes with the numbers for the phonological loop. Holding a picture of a route while you talk may be easier, because the two rely more on different parts.
Researchers use the same logic in the lab. If people repeat a word like "the the the" out loud while memorising a list, their inner rehearsal is blocked and they remember less. Word length matters too. In a 1975 study, Alan Baddeley, Neil Thomson and Mary Buchanan found that people recalled fewer long words than short ones, as if the loop held a certain amount of speaking time rather than a fixed number of items.
Interruptions, noise, tiredness and trying to do two things at once all show up quickly in tasks like these. That is one reason a single round of a memory game says little on its own.
How small it is
The best known number is seven. In 1956 George Miller reviewed many experiments and found that adults could usually repeat about seven items, plus or minus two. Miller also noted that people get past the limit by grouping items into chunks, like reading 1984 as one year.
Later research put the limit lower when grouping and rehearsal are made difficult. In a 2001 review, Nelson Cowan argued for about three to five chunks. For visual material, Steven Luck and Edward Vogel reported in 1997 that people held about four coloured squares after a brief glimpse, and that each of those could carry several features at once, such as colour and angle.
These numbers come from specific lab tasks, and they differ because the tasks differ. They are not targets for a person, and there is no test that gives you "your" working memory as one number.
How researchers measure it
A few tasks come up again and again.
- Digit span. Listen to a list of digits and repeat it, forwards or backwards. The list grows until you make mistakes.
- Corsi blocks. The examiner touches blocks on a board in a sequence, and you touch them in the same order. It is the spatial counterpart of digit span.
- Change detection. Look at a few coloured squares, wait a moment, and say whether one of them has changed. This is the task from Luck and Vogel.
- N-back. Watch a stream of items and say whether each one matches the one shown a set number of steps earlier.
Each task leans on different parts of the model. Digit span relies heavily on the phonological loop, Corsi and change detection on the visuospatial side, and backward span or n-back add more work for the central executive. That is why a person's results on different tasks do not line up perfectly.
Which games on this site use it
Three of our games are built on the same kind of task.
- Hidden Numbers. Numbers appear on tiles, then hide, and you tap the tiles in order from the smallest. You hold positions and their order. It follows a task from a chimpanzee study, covered in the article on the number memory test average.
- Grid Trail. Cards on a grid light up one by one, and you tap them in the same order, or on one level in reverse. It is closest to Corsi blocks. More in the article on visual working memory.
- Button Chain. A sequence of buttons grows by one each round, and you repeat it. On later levels you repeat it backwards, or the buttons change places after the show and you follow their numbers. See the article on the sequence memory test average.
In all three a mistake costs one of three lives, and points depend on correct taps and speed. After a level you see your points and your previous best on that level, not a comparison with other people.
What the games do not claim
Practising one of these games makes you better at that game. We do not claim that the practice carries over to daily life. The research on that question is in does brain training work. Your results can also vary between attempts for many reasons, so a single bad round says little.
If you want to see how the idea looks in practice, try a few rounds of Button Chain, then read the articles linked above for what each lab task measured. Other games are on the all games page.