Working memory: what it is and how much it holds

A model with parts

Until the 1960s, psychology talked mostly about short-term memory, a store where information stayed for a few seconds. In 1974, Alan Baddeley and Graham Hitch proposed a different reading: this space exists so that information can be used, and storing it is only part of the job. Doing a sum in your head, following directions or understanding a long sentence requires keeping a few things active while you work with them.

Their model had three components:

• Phonological loop: holds sounds and words, as when you repeat a phone number under your breath so you won't forget it.
• Visuospatial sketchpad: holds images and positions, like the path between two rooms that you picture in your mind.
• Central executive: directs attention and decides what comes in, what goes out and what to do with what is there.

In 2000, Baddeley added a fourth component, the episodic buffer, which binds information of different kinds, and from long-term memory, into a single episode. The model has been revised many times, and the central idea still stands: there is an active, limited space where thinking happens.

How much it holds

In 1956, George Miller published a paper that became famous for its title, about the “magical number seven, plus or minus two.” He noticed that across various tasks, people could keep about seven items in mind. Miller also showed that the limit depends on how the information is grouped: a long string of digits becomes fewer items when you read it as three or four chunks.

Decades later, Nelson Cowan reviewed the studies and argued, in 2001, that the real limit is smaller, somewhere around four chunks, when people are prevented from rehearsing and grouping. Miller's “seven” includes help from strategy; Cowan's “four” tries to measure the space without that help. Both numbers describe the same thing under different conditions: the space is small, and what you do with it matters.

Without strategy to help, working memory holds somewhere around four chunks of information.

Why working memory goes hand in hand with reasoning

The most demanding working memory tasks, called complex span tasks, ask you to hold items while doing something else: solving simple arithmetic and remembering the last word of each sentence, for example. In 2003, Andrew Conway, Michael Kane and Randall Engle reviewed the studies linking these tasks to general intelligence and found a strong, consistent link, especially with fluid reasoning.

The most widely accepted explanation is that solving a new problem requires holding several pieces at once: the rules already discovered, the hypotheses being tested, what is left to check. People who keep more things active, and control their attention over them better, tend to get further. This shows up clearly in reasoning matrices, where the hardest items combine several rules.

A strong link does not mean they are the same thing. The authors themselves make clear that working memory and general intelligence overlap a great deal without being identical.

Digit span, forward and backward

The oldest way to measure this space is digit span: you see or hear a sequence of numbers and have to reproduce it. In forward span, in the same order; in backward span, in reverse. Forward span depends more on storage. Backward span forces you to store and rearrange at the same time, so it places more demand on the central executive.

On the IQNorth test, the memory area accounts for 7 of the 37 scored questions, with digit sequences and recall questions. In daily training, the memory exercises use forward span of 4 to 8 digits and backward span of 3 to 7, with a new sequence each time so it cannot be memorized.

What memory training changes

Training span improves span. People who practice digit sequences every day can almost always hold longer sequences after a few weeks, often because they learn to group the numbers into chunks. The hard question is whether that carries over to reasoning.

The 2013 meta-analysis by Monica Melby-Lervåg and Charles Hulme answered carefully: working memory training produces gains on tasks similar to the trained ones, which tend to fade over time, and shows no convincing effect on nonverbal reasoning, reading or arithmetic. Later reviews reached the same point, as the article on training and IQ explains.

That is why IQNorth's training presents the memory exercises as practice with the format, with a level from 1 to 5 that goes up and down with your answers. The level shows how you are doing on that exercise. It is not IQ, and we do not promise that training will increase IQ.

See also: Working memory on the test.