This is the 2007 version. Click here for the 2017 chapter 06 table of contents.

The Magical Number Seven plus or minus Two

One of the best-documented characteristics of working memory is its limited capacity. The short-term storage process of working memory can hold only about seven items at a time. To deal with more information than that, the information must be organized into larger chunks. For example, words can be combined into sentences or stories; then more than seven words can be held in working memory.

Psychologist George Miller pointed out the seven item limitation of working memory in a classic 1956 article, "The magical number seven, plus or minus two: Some limits on our capacity for processing information." As you can see from date, this journal article was published in the early days of the encoding fact, some people say this article started the whole idea of using computer concepts like information processing to understand human memory.

What is a chunk?

The magic number seven is the number of chunks of information a person can hold in working memory at the same time. A chunk is a unit of some kind. It could be a letter, a word, or a short sentence. Think of it as a box or container in memory. Miller examined short-term memory tasks and found that typical subjects could hold about 7 chunks in memory at once. This was true whether the subjects were holding 7 letters in memory at once, 7 numbers at once, or 7 words at once. Miller wrote in a humorous tone that he was being "persecuted by an integer" (the number 7) in these studies.

Old-time psychologists, before the encoding revolution, probably would have assumed that fewer words could be held in memory than letters, because each word contains many letters. But this is not the case. Miller's big discovery was that an organized whole (a chunk) functions as one item in primary memory.

What was Miller's "big discovery?"

Miller realized the profound implications of this simple insight. If items can be grouped and treated as chunks in memory, then the capacity of memory can be increased by organizing and grouping things. To demonstrate this to yourself, try holding the following sequence of numbers in memory, all at once.

How can you reduce the string of 10 numbers to 4 chunks?

7 4 1 4 9 2 1 9 4 5

If you interpret this as a string of ten separate numbers, it exceeds the capacity of working memory. Ten chunks are too many to hold at one time in primary memory. But if you recognize two meaningful dates in the string of digits, you have only four chunks, and you easily hold the string of 10 digits in working memory.

Chunking points to the importance of organization in overcoming the limits of memory. If short term, working memory is limited to about 7 chunks, the only way to improve its capacity is to organize larger chunks. This turns out to be a common theme in memory research. Memory is improved by organizing little pieces into larger wholes.

How did Smith quadruple his memory capacity?

In his original article, Miller described a 1954 experiment by psychologist Sidney Smith. Smith memorized sets of four binary digits, which are always 1s or 0s (e.g. 0 0 1 0). Each four-number set of binary numbers is equivalent to one decimal digit (0 0 1 0 equals the number 2). This meant that a string of 16 binary numbers could be converted into 4 decimal numbers. Once Smith learned to make the 4-to-1 conversion easily and automatically, his memory span for binary digits increased from 10 to about 40. In other words, he could memorize 10 decimal numbers in a row, then convert them back into 0's and 1's to reconstruct a list of 40 binary numbers.

How did an undergraduate use recoding to improve his memory for digits, in an experiment lasting more than a year?

Ericsson, Chase and Faloon (1980) decided to see how far this "recoding" idea could be pushed. They had an undergraduate student memorize random strings of decimal digits an hour a day, 3 to 5 days a week, for more than a year and a half. (Presumably they paid him well for this effort!) At the end of this period his memory span had increased from 7 to 79 digits. In other words, he could repeat back a string of 79 random digits immediately after hearing it without any error. His long-term memory for the digits also improved. By the end of the experiment, he often remembered many sequences from previous days.

The subject was not instructed in any particular coding scheme; he invented his own. Being a runner, he found it easiest to translate number sequences into running times. The number 3492 was recoded as "3 minutes and 49 point 2 seconds, near world-record mile time." Later this was supplemented with ages; e.g. 893 became 89 point 3, very old man. (Ericsson, Chase & Faloon, 1980, p.1181).

What sort of "burden" seems to improve, rather than harm, memory?

This should remind you of example from Miller, Galanter, and Pribram described earlier, involving a subject who memorized "BOF" and "MIB" by composing a sentence about a man named BOF who was in "false misery" (MIB). There is an important principle lurking here: Human memory seems to work better, not worse, when a person adds elaborate encoding schemes, as long as they provide organization to aid memory retrieval.

Write to Dr. Dewey at

Don't see what you need? Psych Web has over 1,000 pages, so it may be elsewhere on the site. Do a site-specific Google search using the box below.

Custom Search

Copyright © 2007-2011 Russ Dewey