Working Memory and Chunking: What You Can Actually Hold in Mind

In 1956, cognitive psychologist George Miller published one of the most cited papers in psychology: "The Magical Number Seven, Plus or Minus Two." It proposed that human short-term memory can hold roughly 7 chunks of information at once. Six decades later, the picture is more nuanced — and the good news is that chunking, not raw capacity, is what makes experts look superhuman.

The updated picture

Modern research (especially Cowan's 2001 work) suggests the true number of independent items most adults can hold is closer to 4 ± 1. The "7" Miller observed was actually chunks, not items — and chunks are learned, not innate. A chess grandmaster doesn't remember 25 individual pieces; they remember 5–6 chunks, each representing a familiar pattern of 4–5 pieces.

What a chunk is

A chunk is any grouping that your brain already knows how to treat as a single unit. Examples: the year "1776" is one chunk to an American (but four digits to someone unfamiliar with US history); a phone number with dashes is 3 chunks instead of 10; "ASAP" is one chunk, four letters to a non-English speaker. Chunking works because it leverages long-term memory to compress what working memory has to hold.

Four chunking strategies that work

• Group digits in pairs or triplets. 583729 becomes "58-37-29" or "583-729" — suddenly two or three chunks instead of six. Try it on the Number Memory test.
• Attach meaning. 1492 → Columbus. 2001 → the movie, the year, or September 11. Anything that links to existing knowledge compresses dramatically.
• Use rhythm. Saying numbers with a beat ("fifty-eight, thirty-seven, twenty-nine") recruits auditory working memory, which is a separate channel from visual.
• Build a story. For a list of words, weave them into a bizarre mental scene. The weirder the image, the more durable — emotional salience is a strong memory anchor.

Why most "brain training" disappoints

Commercial brain-training apps reliably improve your score on their own game. What they almost never do is produce far transfer — improvement on unrelated real-world tasks. A 2017 review of 375 studies concluded that the transfer effects of most apps were "close to zero." The exception: physical exercise and learning genuinely new skills, which improve working memory as a side effect of building the underlying neural infrastructure.

What you can actually improve

Your raw working memory capacity (the ~4 items) is fairly stable across adulthood. What changes dramatically with practice is your chunk library — the set of patterns your brain recognizes as single units. That's why a programmer can hold a 20-line function in mind but a beginner can't: it's not better memory, it's better chunks. Train by deliberately practicing in your domain, not by playing unrelated memory games.

Put it into practice

Use the TimeSpeedy memory test as a clean benchmark: your current level shows your untrained capacity. Then practice chunking deliberately — you'll likely gain 1–2 levels per month of short daily practice, purely from better encoding.