Miller's law and the 7±2 myth: how much information fits in memory

Miller's law says memory handles about seven items. Why that number is misread, what Cowan corrects (about four chunks) and how to group information.

info Quick Definition
Miller’s law comes from a 1956 article by George A. Miller, which concluded that immediate memory is limited to around seven items, plus or minus two. Over time the number became a design rule: “never show more than seven options”. That reading is a simplification. What Miller described is a capacity measured in meaningful chunks, and its real value for design lies in grouping information.

What did Miller actually say?

The article is titled “The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information” and was published in Psychological Review (1956, volume 63, pages 81-97). Miller compared two different things: absolute judgment (telling apart, for example, tones) and immediate memory (repeating back a list). Both seemed to hover around seven items, but he separates their causes: absolute judgment is limited by the amount of information, and immediate memory by the number of items.

The most useful idea for design is in the second part. Miller explains that, because immediate memory holds a nearly fixed number of chunks, you can increase the information it contains by grouping and recoding items into larger chunks. Illustrative example: for someone who knows the date, “1492” takes up a single chunk; for someone who does not, it is four loose digits.

On the figure itself, Miller was cautious. When he reviews the many “sevens” in culture (the days of the week, the notes of the musical scale), he writes that he suspects the pattern is only a “pernicious, Pythagorean coincidence”. The version in which there is a “magic number” of seven items became popular later; it did not come from him.

What did Cowan correct?

In 2001, Nelson Cowan published “The magical number 4 in short-term memory: A reconsideration of mental storage capacity” in Behavioral and Brain Sciences (volume 24, pages 87-114). His proposal is a central limit of about four chunks, and he specifies the conditions in which it shows up: when information overload limits chunks to individual items, or when the recoding of items into larger chunks is blocked. In other words, four describes a “pure” capacity, without grouping tricks.

There are other nuances. According to Wikipedia, span is around seven for digits, six for letters and five for words, and it is lower for long words than for short ones. A chunk also depends on what the person knows: a word is one chunk for someone who speaks the language and several for someone who does not. And Gobet and Clarkson, from Brunel University London, found that in more than half of the conditions in their experiment only about two chunks were recalled. That is why taking any single number as a universal limit is a mistake.

What do NN/g and Laws of UX say?

Kate Moran of NN/g summarizes in a 2016 article on chunking that Miller found that most people can remember about seven chunks of information, and that other researchers have suggested anywhere from three to six. Her practical recommendation is to pack information into meaningful chunks and not to ask users to hold more than a few pieces of information at once.

On menus, the same source is explicit: the seven-item limit is a misreading of Miller, menus rely on recognition rather than recall, and so limiting items to seven brings no usability gains. They can remain easy to use with more options if those options are structured in a meaningful way.

Laws of UX carries the same idea: do not use the “magical number” to justify unnecessary design limitations, organize content into smaller chunks, and remember that capacity varies by person, prior knowledge and context.

How does it apply to design?

The practical application is not counting elements but designing groups.

  • Forms. Grouping personal, contact and payment details into three sections reduces what the user must retain while completing each one.
  • Navigation. Clear categories with subcategories let people recognize their route without memorizing it.
  • Multi-step processes. Showing where the person is prevents them from having to remember how much is left.
  • Formatted fields. NN/g recommends autoformatting fields such as phone or card numbers, so people do not have to type or remember the separators.

To see how steps are organized in a flow, read UI flow chunking.

Example

Illustrative example: an insurance app asks people for their vehicle details on a single screen with twelve fields. Several people abandon halfway. The team reorganizes the screen into three groups: vehicle details, policyholder details and type of cover. Each group has a title and shows a progress indicator. There are no longer twelve fields to hold in mind at once; there are three groups, and each is completed before moving on. The total number of fields has not changed, but each step demands less to remember.

For your next screen: count how many items the person must remember at the same moment. If there are many, look for a criterion to group them before removing anything.

When should you NOT use Miller’s law?

  • As an excuse to remove necessary information. A high number of fields may be needed if each one is essential.
  • As a fixed rule of seven options. A well-organized menu can have more.
  • To decide complexity without data. If the question is how much someone can take in at a glance, a test with users answers it better.
  • When the person already knows the system. With practice, working memory leans on recognition and experience.

Common mistakes

  • Treating 7±2 as a universal limit and cutting necessary content because of it.
  • Forgetting that it is a theory about memory, not about preferences.
  • Grouping at random, without a criterion that gives each group meaning.
  • Ignoring recognition, which usually demands less memory than recall.
  • Quoting Cowan’s four as a new fixed rule: it describes controlled conditions, not a design cap.
  • Not testing with users when the question is whether an organization is understandable.

For the underlying problem, cognitive load explains why each element the person must retain has a cost. Hick’s law covers decision time, another effect of the number of options. And Jakob’s law shows how familiar patterns reduce what has to be learned.

Sources

Frequently asked questions

What does Miller's law say?
George Miller published in 1956, in Psychological Review (volume 63, pages 81-97), that immediate memory is limited to about seven items, plus or minus two. Laws of UX summarizes it by saying the average person can only keep 7 (plus or minus 2) items in working memory.
Does it mean a menu can only have seven options?
No. NN/g argues that the seven-item limit is a misreading of Miller and that menus can be easy to use with more than seven options if they are structured in a meaningful way, because they rely on recognition rather than recall.
What is the revision of Miller's law?
Nelson Cowan proposed in 2001, in Behavioral and Brain Sciences, a central limit of about four chunks when grouping stimuli into larger chunks is prevented. The number varies by person, material and familiarity.
How is Miller's law applied in design?
By grouping information into meaningful chunks, not asking users to hold more than a few items at once and relying on recognition of visible options. The number seven should not be used as a hard limit.