“My problem is that I have been persecuted by an integer.”

That’s the opening line of George Miller’s 1956 paper, and 70 years later the integer is still being quoted at me in design reviews. Seven, plus or minus two. Usually as the reason a menu has 7 items instead of 9.

The number is too high and the rule was never a rule. The working figure is about 4, and Miller wasn’t writing about interfaces at all.

The short version

  • The 7-item rule rests on a paper whose author didn’t trust the number.
  • The measured limit is nearer 4, in a lab. Your reader isn’t in one.
  • Showing every option on screen doesn’t help. The choosing still happens in their head.
  • Cutting items is usually the wrong fix. There’s a better one, and it removes nothing.

What the paper actually says

Miller reviewed 2 separate findings that happened to land near the same number. The span of absolute judgement, meaning how many levels of a single stimulus you can reliably tell apart. And the span of immediate memory: how many items you can hold and repeat back.

He said plainly they aren’t the same limit. “In spite of the coincidence that the magical number seven appears in both places, the span of absolute judgment and the span of immediate memory are quite different kinds of limitations that are imposed on our ability to process information.”

Then he said what he thought of the seven itself. “But I suspect that it is only a pernicious, Pythagorean coincidence.”

That’s the man everyone cites for the 7-item menu, telling you in his own paper that he doesn’t believe the number means anything.

The number came down

Nelson Cowan revisited it in 2001, in “The magical number 4 in short-term memory: a reconsideration of mental storage capacity”, Behavioral and Brain Sciences 24(1), 87–114.

His argument is about boundary conditions. Miller’s 7 was measured in situations where people could rehearse, group items themselves and lean on things they already knew. Strip those supports out and the figure drops.

Cowan’s conclusion: “A single, central capacity limit averaging about four chunks is implicated.”

Four. And that’s the clean laboratory ceiling, measured on someone concentrating on purpose. Your reader is on a phone, on a train, halfway through something else.

Why this bites even when the page is right in front of them

A menu isn’t a memory test. The items are on screen. Nobody has to recall them.

But choosing from a menu is a memory test, because comparison happens in working memory. You read option 3, hold it as a candidate, read 4 through 9 and then try to remember whether 3 was better than 6. The screen holds the list. Your head holds the shortlist.

The shortlist is where the 4 bites.

So the constraint isn’t how many things are on the page. It’s how many things are inside a single decision.

Grouping, not deletion

The instinct is to cut. Nine nav items, delete 4. That’s usually wrong, because the 4 you delete were doing a job for somebody.

Miller’s own answer was recoding. “Since the memory span is a fixed number of chunks, we can increase the number of bits of information that it contains simply by building larger and larger chunks, each chunk containing more information than before.”

Nine items in 3 groups of 3 is 3 things to compare, then 3 more. Both sit inside the budget. Nothing was removed.

Where it applies:

  • Navigation. Nine top-level links is 9 candidates in one decision. Three headings with 3 links under each is 3, then 3.
  • Pricing. Five plans across 20 feature rows is a comparison running in 2 directions at once. Three plans, features in 3 named groups, and the comparison becomes readable.
  • Forms. Twenty-two fields on one screen reads as an interrogation. The same 22 in 4 labelled sections reads as 4 questions.
  • Dashboards. Fourteen metrics on a screen means nobody carries a picture of the business away from it. Pick the 4 that would change a decision this week. The rest can be one click deeper.
  • Decks. A slide with 6 bullets is 6 slides that haven’t been separated yet. One claim per slide, and the claim goes in the heading.

The test is identical everywhere. Count the options inside one decision, not the elements on the page. If it’s more than 4, you’re asking for something the reader can’t give.

And when the cost of choosing exceeds the value of choosing, people don’t choose badly. They leave.

The honest version of the title

Miller wasn’t wrong. He was tentative, he was careful and he wrote his own doubts into the paper.

What was wrong is what the industry did with him. It took two findings about tones and digit strings, dropped every caveat and turned them into a rule about navigation menus. Then defended that rule for 70 years on the authority of a man who called his own number a coincidence.

The correction isn’t complicated. Design for 4, group the rest, and stop deleting things to hit an arbitrary number.

Miller was persecuted by an integer. The industry has spent 70 years designing for the wrong one.

Sources: George A. Miller, “The magical number seven, plus or minus two: some limits on our capacity for processing information”, Psychological Review 63(2), 1956. Nelson Cowan, “The magical number 4 in short-term memory: a reconsideration of mental storage capacity”, Behavioral and Brain Sciences 24(1), 87–114, 2001 (https://pubmed.ncbi.nlm.nih.gov/11515286/, DOI: https://doi.org/10.1017/S0140525X01003922)