Where "seven plus or minus two" came from

In 1956 the psychologist George Miller published a paper with one of the best titles in the history of psychology: The Magical Number Seven, Plus or Minus Two. He surveyed a range of experiments — how many tones people could tell apart, how many dots they could count at a glance, how many unrelated items they could repeat back — and noticed the same number appearing near the ceiling: about seven, give or take two.

The number escaped into popular culture and hardened into a fact it was never quite claiming.

Why modern estimates are lower

Later work, notably by Nelson Cowan in 2001, argued that Miller's seven was inflated. It included the tricks people quietly use to stretch the limit — silently repeating items back to themselves, grouping them, using rhythm. Strip those away and measure raw capacity, and you get something closer to four independent items for most adults.

That is strikingly small. Four things. And "thing" here means an unfamiliar, unconnected item — the kind you cannot group with anything else you know.

The escape hatch: chunking

The limit applies to units, and units are negotiable. This is what chunking means, and it is the single largest factor in why experts appear to have superhuman memory for their domain.

The classic demonstration comes from chess. Show a chess master a real board position for five seconds and they can reproduce it almost perfectly. Show them a board with the same number of pieces placed randomly, and they do no better than a beginner. The master is not remembering more individual pieces; they are storing four or five familiar patterns instead of twenty unrelated locations.

You do the same thing constantly without noticing:

  • A phone number is ten digits, or three chunks: area code, prefix, line.
  • A password of four dictionary words beats a string of random characters, and you can recall it.
  • A chess opening, a guitar chord, a coding idiom — each is one unit built from many parts.

Which leads to the uncomfortable corollary: the way to hold more in your head is not to train harder, it is to know the domain better. Experts have bigger chunks, not bigger heads.

It is not one big storage box

Alan Baddeley's influential model splits working memory into parts: a verbal store that holds sounds and words, a visuospatial store that holds images and positions, and a central system that decides what to attend to and juggles the pieces.

This has a practical consequence. Doing two things at once hurts far more when both need the same part. Reading an article while someone talks to you is a collision in the verbal store, so both suffer badly. Listening to instrumental music while doing a visual task is far less damaging, because the work is happening in a different channel. The popular advice that "background music helps you focus" and the advice that it "destroys your focus" are both true, depending on whether the music has words.

Multitasking is not doing two things at once. It is switching between two things fast enough that both get done worse.

Why this shows up in puzzles

Working memory is the bottleneck in most of the games people play for five minutes:

  • 2048 loads the visuospatial store. You are holding a picture of where the next few merges should land and mentally rehearsing a sequence of slides before committing to one.
  • Minesweeper loads it differently: you hold several number clues at once and reconcile them against each other.
  • Flow Free is almost entirely planning under a spatial memory load — every cell has to be filled, so a path that wanders has to be tracked end to end.

This is also why these games feel exhausting long before they feel difficult. The challenge is not complexity, it is holding four or five things steady while you reason about them.

Four ways to work with the limit

  • Write things down. Getting an item out of your head and onto paper is not a crutch, it is correctly matching the task to the hardware.
  • Group before you memorise. Convert a list of twelve items into three categories of four, and the twelve fit where they did not before.
  • Reduce switching, not effort. Every switch pays a cost as the old context is rebuilt. Batching similar work beats interleaving unrelated work.
  • Build chunks deliberately. The reason an expert seems to remember everything is that they have spent years turning many small things into few large ones.

Try it yourself

Ten mines on an 8×8 board is enough to keep several clues in mind at once without tipping over into frustration — a decent place to feel the limit rather than read about it.

Play Minesweeper  →

Editorial note: This article is for general educational purposes. Cognitive effects vary between individuals, and claims about specific benefits should be based on evidence from well-designed studies.