What is chunking in memory?

Chunking is the cognitive process of grouping individual pieces of information into meaningful units — chunks — that are treated as single items in working memory. Miller (1956) demonstrated that working memory holds approximately 7 (± 2) chunks regardless of how much information each chunk contains. By organising information into meaningful chunks — a phone number into three groups, a list of items into a narrative, separate facts into a category — learners can store substantially more information than the 7-item limit on raw items would suggest.

What is Miller's Law?

Miller's Law is the finding, published by George Miller in 'The Magical Number Seven, Plus or Minus Two' (1956), that human working memory can hold approximately 7 (± 2) chunks of information simultaneously. Crucially, a chunk is a meaningful unit — and chunk size is flexible. A chess grandmaster's chunk might be 'king's Indian defence position with queenside pawn break,' representing 15 pieces and their positions. A novice's chunk is one piece. Miller's Law does not limit the total information that can be held — it limits the number of chunks.

How does pattern recognition expand memory capacity?

Pattern recognition allows large amounts of information to be perceived and stored as a single chunk rather than multiple items. Chase and Simon (1973) demonstrated this in chess: grandmasters could recall entire board positions from a 5-second glance because they perceived the board in terms of recognised patterns (5–10 chunks), while novices perceived individual pieces (25–30 items). The grandmaster's working memory was not larger — their chunk size was vastly larger. Learning any domain to the point of pattern recognition effectively multiplies working memory capacity for that domain.

How can I use chunking to study more effectively?

Three chunking strategies improve study efficiency: (1) Categorical chunking — group related items by category before memorising, and build a mnemonic for the categories rather than the items (e.g., remember 'mammals, birds, reptiles' then recall items in each group). (2) Sequential chunking — identify logical or causal flow in the material; sequences with internal logic are perceived as fewer, larger chunks than arbitrary sequences. (3) Pattern-finding — spend 5–10 minutes identifying the organising principle of the material before building mnemonics; genuine understanding often reduces or eliminates the need for mnemonics.

Is chunking the same as memorisation?

No. Chunking is a perceptual and organisational process that precedes memorisation and makes it more efficient. You chunk information by recognising patterns and grouping related items. You then memorise the chunks — the meaningful units — rather than the individual items. Chunking without memorisation (understanding the structure without encoding) produces better comprehension but not reliable recall. Memorisation without chunking (rote rehearsal of individual items) is inefficient and produces fragile recall. The combination — perceiving patterns, then using mnemonics to encode the chunked structure — is the most effective approach.

← Blog

Chunking and pattern recognition: Miller's Law applied to learning

May 16, 20268 min readBy warpread.app

In 1956, George Miller published what became one of the most cited papers in cognitive psychology: "The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information." Its central claim was simple: human working memory can hold approximately 7 (± 2) chunks of information simultaneously.

The key word is chunks. Not items. Not bits of information. Chunks — meaningful units that the mind treats as single entities. Miller showed that the 7-item limit applied regardless of chunk complexity: 7 random letters, 7 syllables, 7 words, or 7 sentences all fit into the same number of slots. What varied was how much raw information each slot could hold, depending on how richly the material was patterned.

This single observation is the foundation of every effective mnemonic technique.

Chunks, not items: why chunk size is not fixed

Miller's insight was not that people can hold exactly 7 things — it was that "things" is determined by the learner's existing knowledge, not by the structure of the material. A phone number presented as "07832174963" requires 11 working memory slots (one per digit) for someone with no pattern. The same number presented as "07832 174 963" requires 3 slots (three groups), each group read as a single chunk.

Chase and Simon (1973) demonstrated this most dramatically in chess. They showed grandmasters and novices chess board positions for 5 seconds, then asked them to reconstruct the positions. Grandmasters recalled near-perfectly; novices recalled 4–5 pieces. When the test was repeated with randomly arranged pieces (no meaningful patterns), grandmasters performed identically to novices.

The conclusion was unambiguous: grandmasters did not have larger working memories or better general memory. They perceived the board differently. Where a novice saw 25–30 individual pieces, a grandmaster saw 5–10 meaningful configurations — recognised patterns from thousands of hours of study. Each pattern was a chunk; and 5–10 chunks fits within Miller's limit, while 25–30 individual items does not.

The same chunking process operates in every expert domain. A radiologist reading an X-ray perceives "consolidation in the right lower lobe consistent with bacterial pneumonia" — one chunk — where a medical student perceives dozens of individual features. A musician reading a score perceives harmonic progressions and rhythmic patterns — chunks — not individual notes. Domain expertise is, at its core, the accumulation of a large library of domain-specific chunks.

How chunking applies to learning new material

For the learner without domain expertise, the chunk library does not yet exist. Material that an expert perceives as 5 chunks is perceived as 50 items — far exceeding working memory's capacity. This is why complex new material feels overwhelming: it is not that the material is inherently unmanageable, but that the chunks that would make it manageable have not yet been built.

Mnemonics create artificial chunks — imposed patterns that group otherwise unrelated material into a single retrievable unit. An acrostic phrase is a chunk: "Plastic Meat Aint Tasty" is one thing to hold in working memory, and that one thing expands on demand into four items (Prophase, Metaphase, Anaphase, Telophase) in the correct order.

The expansion property is what makes chunks genuinely powerful. A chunk is not just a compressed list — it is a retrievable structure with internal organisation. The acrostic phrase encodes both the content (the first letter of each item) and the order (left to right maps to sequence). The chunk stores more than the items alone.

Finding patterns before building mnemonics

The most efficient chunking strategy is not to impose artificial patterns but to find genuine ones. Most bodies of knowledge have inherent structure: taxonomic hierarchies, causal chains, spatial arrangements, sequential processes, analogical relationships. When that structure is perceived and understood, it becomes the chunk — and the need for artificial mnemonics is reduced or eliminated.

Before building any mnemonic, spend 5–10 minutes asking: what organises this material? Why are these items in this order? What do they have in common? What distinguishes them from each other? This pattern-finding effort serves two functions. First, genuine understanding produces memory through semantic encoding (Craik & Lockhart, 1972) — the deepest processing level and the most durable encoding pathway. Second, if a mnemonic is still needed after understanding the structure, it will be far stronger because it connects to existing pattern knowledge rather than encoding isolated items.

The phases of mitosis illustrate both approaches. A rote approach memorises PMAT as four arbitrary items. A structural approach understands: Prophase (chromosomes condense and become visible), Metaphase (chromosomes align at the cell's middle — note the M), Anaphase (chromosomes pull Apart — note the A), Telophase (chromosomes reach the cell's two poles — cell starts to divide). With structural understanding, the sequence has logic; M for "middle" and A for "apart" provide internal retrieval cues. A mnemonic supplements this understanding rather than replacing it.

Categorical and hierarchical chunking

For long lists without obvious sequence logic, categorical chunking organises items into groups by shared property, then builds mnemonics for the category labels rather than all items simultaneously.

For example, memorising 20 countries by their continental regions: first chunk into 5 groups of 4 (Africa, Asia, Europe, Americas, Oceania). Build a mnemonic for the 5 continental labels. Within each group, build a mnemonic for the 4 countries. This two-level hierarchical mnemonic holds 20 items in 5 working memory slots (the continental mnemonics), each of which expands on demand into 4 slots (the country mnemonics).

Ericsson and Kintsch (1995) described this as a retrieval structure — an organised framework in long-term memory that allows experts to rapidly access and expand chunks relevant to a task. The categorical chunking approach builds a minimal retrieval structure even for novices, by imposing an organisational hierarchy before encoding.

The working memory and mnemonic connection

Every mnemonic technique works by reducing the number of items working memory must hold simultaneously:

A first-letter acrostic compresses a 6-item list into 1 chunk (the phrase)
A memory palace sequence compresses a 20-item list into 1 chunk (the route), with location-cued expansion
Categorical chunking compresses a 20-item list into 5 chunks (the categories), each with internal sub-structure

In each case, the mnemonic creates artificial chunking that brings the total number of working memory units within Miller's 7 ± 2 limit, making encoding, maintenance, and retrieval feasible.

Practice chunking: The Mnemonic Builder shows you the first-letter chunk structure of any list and helps you build an acrostic that compresses the full list into a single memorable phrase. Free, no account required.

The Mnemonics & Pattern Memory course covers chunking in Lesson 2, including the Chase and Simon chess research and the Ericsson and Kintsch long-term working memory model. Six evidence-based lessons, free, no account required.

References

Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4(1), 55–81. https://doi.org/10.1016/0010-0285(73)90004-2
Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(6), 671–684. https://doi.org/10.1016/S0022-5371(72)80001-X
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review, 102(2), 211–245. https://doi.org/10.1037/0033-295X.102.2.211
Miller, G. A. (1956). The magical number seven, plus or minus two. Psychological Review, 63(2), 81–97. https://doi.org/10.1037/h0043158

Topics

chunking memoryMiller's lawworking memory chunkspattern recognition memorycognitive chunkingmemory capacityhow to improve working memorychunking learning technique

Build your first mnemonic now

Paste any list into the Mnemonic Builder to extract first letters, build an acrostic phrase, and test your recall — or take the free 6-lesson course on mnemonics and memory patterns.

Try the Mnemonic Builder →Start the free course →