The mind's workbench

Three kinds of remembering

People use "memory" for very different systems. Keeping them apart saves a lot of confusion later, because they behave differently, they are tested differently, and they respond differently to practice.

Long-term memory

The large, durable store of everything you have learned: facts, formulas, faces, how to ride a bicycle. Its capacity has no practical limit, but getting new material into it takes repetition and time.

Short-term memory

Passive, temporary storage: holding a short list exactly as it arrived and repeating it back. Reciting digits in the order you heard them is the classic case.

Working memory

Short-term storage plus control: holding information while you transform it, update it, or protect it from interruption. Saying the digits backwards, keeping a running sum, or remembering a position while you judge something else all need working memory.

The distinction is not just vocabulary. In a well-known study of 133 adults, Engle and colleagues found that short-term and working memory were separate but closely related abilities, and that working memory was strongly linked to fluid intelligence while short-term memory was not. The ability that mattered was keeping a representation active in the face of interference and distraction, not raw storage.

The classic model: one manager, several stores

In 1974 Alan Baddeley and Graham Hitch replaced the idea of a single short-term store with a team of parts. Baddeley added a fourth part in 2000. The model is still the most common way to describe working memory, and it maps neatly onto the tasks you will meet in assessments.

The central executive is the manager. It stores little itself; it decides what gets attention, switches between jobs, and pushes old information out when new information arrives. When a task makes you update or protect what you hold, this is the part under strain.

The phonological loop holds speech-like material, the voice in your head that repeats a number to itself. Its best-known signature is the word-length effect: lists of long words are recalled less well than lists of short words, because rehearsal takes longer and the trace fades while you get round the loop.

The visuospatial sketchpad holds what things look like and where they are. Later work split it further: remembering a visual pattern and remembering a sequence of locations can be damaged separately in patients and disrupted by different kinds of interference. That is why our library has one game for positions in sequence (GridRecall) and another for a static picture (Puzzle Reconstruction).

The episodic buffer, added by Baddeley in 2000, is a limited store that binds information from the other parts and from long-term memory into a single episode. It explains why meaningful material, such as a sentence or a familiar abbreviation, is held far more easily than its parts would predict.

1. 1956

   George Miller describes the span of immediate memory as about seven items, and introduces chunks.

2. 1958

   Wayne Kirchner reports the first n-back task, the ancestor of Flashback.

3. 1974

   Baddeley and Hitch propose working memory as a multicomponent system.

4. 1980

   Daneman and Carpenter publish reading span, the first complex span task.

5. 1999

   Engle and colleagues show working memory, not short-term memory, tracks fluid intelligence.

6. 2000

   Baddeley adds the episodic buffer.

7. 2001

   Nelson Cowan argues the true limit is about four chunks.

How much fits: seven items, or four chunks

Miller's 1956 paper, one of the most cited in psychology, opens with a famous line.

My problem is that I have been persecuted by an integer.

The integer was seven. But Miller's more important point was about chunks. He noticed that the span of immediate memory hardly depends on how much information each item carries: people hold about as many words as letters, even though a word carries far more. What is limited is the number of units, and a unit can be as big as your knowledge lets you make it. He called the trick recoding: turning many small pieces into a few larger ones.

Try it. Watch the letters below, one at a time, then see how many you can say back.

Ordering exerciseSign in to try it

In 2001 Nelson Cowan reviewed decades of experiments and concluded that seven overstates the real limit. When people are prevented from rehearsing or chunking, adults hold about four chunks, typically three to five. Seven items is what you get when rehearsal and grouping quietly help out. Both numbers are right about different conditions, and the gap between them is exactly the room that strategy gives you.

Holding is not the whole job

If working memory were only a box, one number (its size) would describe it. In practice the demands come in several flavours, and a person can be strong at one and ordinary at another. Our taxonomy splits memory into four facets, each measured by particular games.

A sixth game on the shelf, Sequence Transformer, is mostly a rule-finding task, but you must hold the example sequences in mind while you test your guess about the rule, so it contributes a smaller share to the memory score.

The difference between the second and third columns is the heart of modern working memory research. A simple span task (Digit Span forwards) mostly asks how much you can hold when nothing competes. A complex span task interleaves storage with an unrelated processing job, so it measures holding despite interference. The researchers who built complex span tasks found that this second kind of score, not the first, is the one that tracks reasoning ability. The next lesson looks at that evidence.

Where you use it every day

Working memory is invisible when it copes and very visible when it overflows. Some familiar moments:

• Mental arithmetic. Adding 47 and 38 in your head means holding a partial result (80) while you work out the rest (15). Lose the 80 and you start again.
• Reading a dense paragraph. You hold the subject of a long sentence until its verb arrives. Research on reading span grew from exactly this observation.
• Following instructions. "Open the config, change the port to 8081, restart the service, then clear the cache" is four steps. Most people who get it wrong drop the last one.
• Data interpretation sets. A CAT or bank-exam DI set asks you to combine two or three numbers from different tables. Each number you hold while you look for the next is a slot in use.

At work the same demands show up as juggling a customer's details while typing into a form, tracing a variable through several functions while debugging, keeping three constraints in view while you plan a schedule, or answering the second half of an interviewer's question after you have dealt with the first half. None of these is a memory test, and all of them go better when you know how to keep the load small: chunk, write things down when you are allowed, and deal with interruptions quickly so they do not overwrite what you are holding.

Try it now

A quick forward digit span. Digits appear one per second; type them back in order. Miller put the typical figure at about seven, give or take two. Note your result: the digit span lesson teaches a grouping strategy you can compare it against.

Practice drillSign in to try it

What to take away

• Working memory is a small, active workspace: storage plus control. Short-term memory is the storage part alone.
• Baddeley's model has a central executive and three limited stores: the phonological loop, the visuospatial sketchpad and the episodic buffer.
• Capacity is about four chunks when grouping is blocked, and closer to seven items when rehearsal and chunking help. The gap is where strategy lives.
• Our memory score has four facets: span capacity, memory updating, working under load, and visual memory.
• Holding information despite interference, not raw storage, is the part most closely tied to reasoning.

Sources

1. Baddeley, A. D., and Hitch, G. (1974). Working memory. Psychology of Learning and Motivation, 8, 47–89. doi:10.1016/S0079-7421(08)60452-1
2. Baddeley, A. (2000). The episodic buffer: a new component of working memory? Trends in Cognitive Sciences, 4(11), 417–423. doi:10.1016/S1364-6613(00)01538-2
3. Baddeley, A. (2012). Working memory: theories, models, and controversies. Annual Review of Psychology, 63, 1–29. doi:10.1146/annurev-psych-120710-100422
4. Baddeley, A. D., Thomson, N., and Buchanan, M. (1975). Word length and the structure of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14, 575–589. doi:10.1016/S0022-5371(75)80045-4
5. Miller, G. A. (1956). The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. Full text
6. Cowan, N. (2001). The magical number 4 in short-term memory: a reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114. doi:10.1017/S0140525X01003922
7. Engle, R. W., Tuholski, S. W., Laughlin, J. E., and Conway, A. R. A. (1999). Working memory, short-term memory, and general fluid intelligence: a latent-variable approach. Journal of Experimental Psychology: General, 128(3), 309–331. doi:10.1037/0096-3445.128.3.309
8. Della Sala, S., Gray, C., Baddeley, A., Allamano, N., and Wilson, L. (1999). Pattern span: a tool for unwelding visuo-spatial memory. Neuropsychologia, 37(10), 1189–1199. doi:10.1016/S0028-3932(98)00159-6
9. Kirchner, W. K. (1958). Age differences in short-term retention of rapidly changing information. Journal of Experimental Psychology, 55(4), 352–358. doi:10.1037/h0043688
10. Daneman, M., and Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450–466. doi:10.1016/S0022-5371(80)90312-6