When someone tells you a phone number and you repeat it in your head while looking for a pen, you are using working memory. It is the small, temporary workspace in which you keep a few pieces of information active while you use them. It is what lets you follow a sentence, do mental arithmetic and compare two options. Everything in this workspace disappears quickly if you are interrupted.
The magical number seven
In 1956 the psychologist George Miller published a paper with the playful title "The Magical Number Seven, Plus or Minus Two". He observed that people could remember about seven items in a series of digits, letters or words. The number became famous, and it is still quoted in design guides for phone numbers and menus. But Miller's own point was subtler. What counts is not the number of digits but the number of chunks. A person who sees the letters F-B-I-C-I-A-N-A-S-A as three familiar groups, FBI, CIA and NASA, has three things to remember, not ten.
When researchers took chunking out of the equation, with items that cannot easily be combined, the limit fell. In a widely cited review in 2001, Nelson Cowan argued that the true capacity of the focus of attention is about four chunks, with a range of three to five for most adults. Seven is reached with the help of rehearsal and grouping.
Models of working memory
Alan Baddeley and Graham Hitch proposed a model in 1974 that is still the starting point. It has a central executive that controls attention, a phonological loop for sounds and words, and a visuospatial sketchpad for images and positions. A later addition was an episodic buffer that links information to long-term memory. The model explains why you can say a word to yourself and track a moving dot at the same time with little interference, whereas two verbal tasks at once interfere heavily.
How span is measured
- Digit span. Repeat a growing series of digits, forward or backward. Typical adults manage about seven forward and five backward.
- Corsi block-tapping. Philip Corsi introduced it in 1972. Blocks on a board light up in sequence and you repeat the order by tapping. Typical spans are around five to six. Our Grid Pattern game follows this principle.
- N-back. Wayne Kirchner used it in 1958. You see a stream of items and say when the current one matches the one from n steps earlier. It requires continuous updating, not just storage, and is considered a demanding measure of working memory.
- Complex span. Tasks that mix storage with processing, such as remembering letters while checking equations, which correlate with reading comprehension and reasoning.
Why it matters
Working memory capacity is related to a number of everyday abilities: following instructions, learning to read, mental arithmetic and complex reasoning. It varies between people and also within a person during the day. Sleep deprivation, stress, anxiety and multitasking all reduce it for the moment. It also develops through childhood and declines slowly with age.
Can it be trained?
This is a field where claims run ahead of evidence. Training on a specific task makes you better at that task and often at similar tasks. The question is whether it transfers to general intelligence or everyday performance. A well-known 2008 study by Jaeggi and colleagues reported transfer from n-back training to fluid intelligence, but several later studies and meta-analyses found small or no lasting effects. We therefore do not claim that these games raise your IQ. What does work well is learning strategies, which are skills in their own right:
- Chunk. Combine items into groups of two or three with a meaning or a rhythm.
- Use space. In position tasks, imagine a path or a shape through the dots instead of remembering each separately.
- Use names. A colour or shape that you give a word is easier to keep than a pure image, for tasks where that is possible.
- Reduce interference. Look at the full display, then look away at once; do not stare at a different part of the screen.
Our memory games are scored on accuracy, and the levels increase the number of items slowly, so you can find the level at which you start to make mistakes. That level is your current span for that task, a number that is more informative than a score.