Chunking is grouping small pieces of information into larger, meaningful units, called chunks, so they can be held in memory as one. Short-term memory can only keep a handful of things in mind at once, but it counts chunks, not letters or digits. That is why a ten-digit phone number is easier as three groups, and why experts can take in at a glance what a beginner has to work through piece by piece. The idea comes from George Miller’s 1956 paper on the “magical number seven”.

Chunking at a glance

  • What a chunk is: any unit you already know as a whole: a word, an abbreviation, a date, a familiar pattern of chess pieces or musical notes.
  • Why it works: the limit of working memory is about the number of units, so making each unit bigger lets more information through.
  • How many: about four chunks at once on modern estimates, not Miller’s famous seven.
  • Where chunks come from: long-term memory. You can only chunk material whose patterns you have learned.
  • Two kinds: deliberate grouping you choose to do, and automatic chunking that builds up with experience.

Where the idea comes from: Miller’s magical number seven

In 1956 the psychologist George A. Miller noticed that people could repeat back about the same number of items whatever the items were. In the experiments he reviewed, people could repeat about eight decimal digits but only about nine binary digits (0s and 1s), even though a decimal digit carries far more information. His conclusion was that immediate memory is limited by the number of chunks, not by the amount of information in them.

The demonstration he described came from an experiment by Sidney Smith. Smith learned to recode strings of binary digits into octal digits (each group of three 0s and 1s becomes one number from 0 to 7). He could remember about 12 octal digits. With pairs recoded, those 12 chunks were worth 24 binary digits; with groups of three, 36; with groups of four or five, about 40. Miller’s point was that this kind of recoding, grouping input and giving the group a new name, is something people do constantly, most of all with language.

How many chunks can you hold?

Fewer than seven. In a 2001 review, Nelson Cowan argued that Miller’s number had been more a rough estimate than a real limit. When the evidence is restricted to conditions where people cannot rehearse items or combine them into bigger chunks, the limit averages about four chunks. A 2012 study by Fabien Mathy and Jacob Feldman squared the two numbers. They varied how compressible sequences of digits were and found that memory depended on the number of distinct chunks left after compression: the true limit was about three or four chunks, equivalent to about seven items of typical compressibility. Miller’s seven, in other words, already had some chunking built in.

A 2016 study by Mustapha Chekaf, Nelson Cowan and Fabien Mathy showed how fast this happens. Untrained people recalled about three items when a list had no structure, and more as structure was added, which suggests new chunks can form within a single trial.

Everyday examples of chunking

  • Numbers: phone numbers, card numbers and postcodes are printed in groups because three or four groups are easier than a dozen digits.
  • Letters and words: FBI-CIA-USA is three chunks, not nine letters. Fluent readers read whole words, not letters.
  • Dates: 1-9-4-5 is four digits to a stranger and a single chunk to someone who knows the year the Second World War ended.
  • Skills: a musician reads a chord, a driver handles a junction, a programmer recognises a loop, each as one unit.
  • Studying: splitting a topic into a few named parts and learning each part as a whole is chunking used on purpose.

Chunking and expertise: chess masters and a 79-digit memory

The strongest evidence that chunks are learned comes from experts. The Dutch psychologist Adriaan de Groot showed in the 1940s that chess masters could reconstruct a real game position after seeing it for a few seconds far better than weaker players. In 1973 William Chase and Herbert Simon showed why: the master’s advantage appeared for positions from real games, which are full of familiar patterns, and largely disappeared for pieces placed at random. Masters were not remembering more pieces; they were remembering bigger chunks. A 1996 review by Fernand Gobet and Simon added a correction: strong players keep a small edge even with random positions, though much smaller than with real ones. Our page on whether chess makes you smarter covers what that skill does and does not transfer to.

In 1980 K. Anders Ericsson, William Chase and Steve Faloon reported on a student, known as SF, who practised remembering random digits. After more than 230 hours in the laboratory, his digit span rose from 7 to 79 digits. He did it by grouping digits into meaningful chunks he already knew, and his performance on other digit tests matched that of lifelong memory experts. The same idea sits behind classic techniques like the memory palace.

How chunking helps working memory

Why does a chunk help? A 2019 study by Mirko Thalmann, Alessandra Souza and Klaus Oberauer at the University of Zurich tested it directly. A familiar chunk improved memory not only for itself but also for other, unrelated information held at the same time, which means chunking reduces the load on working memory rather than just packing items more neatly. Their explanation: the chunk is retrieved from long-term memory as one compact unit that replaces its separate elements, freeing capacity for whatever comes next. Two details matter. The benefit was largest for chunks early in a list, and capacity was not a fixed number of chunks regardless of their size, so bigger chunks are not free.

Gobet and colleagues’ 2001 review distinguished two forms: deliberate chunking, a strategy you choose, such as grouping a phone number, and automatic chunking, a continuous process that builds up as you gain experience with a kind of material. Expertise is mostly the second kind. The idea also explains why overloaded learners struggle, as our page on cognitive load describes: what is one chunk to an expert is many separate pieces to a beginner.

How to use chunking

  • Group into threes and fours. For lists, numbers and codes, groups of three or four suit the capacity limit.
  • Make groups meaningful. A group you can name or link to something you know (a date, a word, a rule) works better than an arbitrary split.
  • Build the patterns first. Chunks come from knowledge, so learning the basic patterns of a subject is what makes the advanced material manageable.
  • Expect it to be specific. Chunks learned on digits help with digits. They do not make you better at remembering everything.

Chunking and IQ

Working memory is strongly related to fluid intelligence, the ability to solve new problems that reasoning tests try to capture. A 2018 study by Mustapha Chekaf, Nicolas Gauvrit, Alessandro Guida and Fabien Mathy looked at why. They measured how compressible each list was and found that compressibility predicted working memory performance and that people’s ability to compress information predicted their fluid intelligence well. Their reading is that the ability to find structure and chunk it is part of what links working memory to intelligence.

That is not a promise that chunking practice raises IQ. Expert chunks are tied to the material they were learned on, and broad gains from memory training have not held up, as our pages on dual n-back and how to increase IQ explain. Our guide to fluid and crystallized intelligence covers what IQ scores do measure.

Trust and scope notes

This page summarises published research for general information, with figures taken from the studies listed below. It is not medical advice: memory problems that are new, worsening or worrying deserve a conversation with a doctor, and no online test can assess them. The IQ Revealed test is an online reasoning test of 40 original items for adults aged 18 and over; it is not a clinical assessment and it does not measure chunking or memory span. Our guide to how IQ tests work explains what an online test can and cannot tell you.

IQ Revealed is not affiliated with, endorsed by or connected to the researchers, universities, journals or publishers named on this page. Names are used only to identify the research being described.