Chunking
Grouping related information into meaningful units can make it easier to scan, understand and remember.
Definition
Chunking is the process of recoding separate pieces of information into a larger, familiar unit. A useful chunk reflects a relationship the person can recognise; drawing boxes around unrelated content does not create one. In an interface, meaningful grouping may support comprehension and navigation, but memory research does not imply that a menu must contain a fixed number of items.
Current interpretation
What this may help explain.
George Miller’s 1956 paper was a theoretical review of several perception and immediate-memory studies, not a single experiment conducted by Miller at Harvard. It distinguished limits in judging simple stimuli from limits in remembering sequences. The most direct recoding demonstration it described came from Sidney Smith: 20 participants had baseline spans of about nine binary digits and seven octal digits, then five people learned each method for translating groups of binary digits into larger named units.
Every recoding method increased binary-digit span, but a few minutes of learning did not produce the full predicted benefit; the translation needed to become familiar enough that it did not consume attention itself. Miller also warned that the recurring number seven could be coincidence. A 2001 review argued for an average capacity nearer three to five chunks only under tightly controlled conditions that limit rehearsal and long-term-memory strategies. Four experiments published in 2019 found that familiar chunks reduced working-memory load, but not as a fixed number of equal “slots”: benefit varied with chunk size, overlapping elements and where the chunk appeared in the list.
Use this as a starting point for a hypothesis, then check it with your own users and context.
Source check
What this guidance is based on.
3 sources have been checked by The UXologist. Findings and limitations are shown together so you can judge how well they fit your situation.
How Does Chunking Help Working Memory?
Research-assisted source review
- What it supports
- Chunking improved recall of chunked material and sometimes freed capacity for other information. Benefit was not independent of chunk size when chunks overlapped, persisted under articulatory suppression and differed by list position; working memory was therefore not limited to a fixed number of chunks regardless of their size.
- Who or what was studied
- Adults took part in four controlled working-memory experiments. The indexed abstract identifies male, female and young-adult participants but does not provide one combined sample count.
- Study setting
- Participants recalled lists containing familiar chunks alongside unchunked material. Experiments varied whether chunks had unique or overlapping elements, suppressed rehearsal and changed a chunk’s serial position.
- Where it may not transfer
- The studies used tightly controlled list-recall tasks and familiar learned chunks. They do not test visual card layouts, navigation hierarchies or search behaviour, and grouping unfamiliar interface content may not create the compact long-term-memory representation that produced the laboratory benefit.
The Magical Number 4 in Short-Term Memory: A Reconsideration of Mental Storage Capacity
Research-assisted source review
- What it supports
- The reviewed evidence was more consistent with an average central capacity of about four chunks, often described as three to five, when chunks could be identified under carefully constrained conditions. Cowan stressed that strategies and long-term knowledge can change performance in less controlled tasks.
- Who or what was studied
- A target review article synthesising a wide range of verbal, visual and attentional studies rather than reporting one new participant sample.
- Study setting
- Examined estimates of short-term-memory capacity under conditions designed to prevent rehearsal, recoding into larger chunks, sensory-memory refill and other strategies.
- Where it may not transfer
- The exact capacity and underlying mechanism remain theoretically contested, and the estimate applies only under specified experimental conditions. It is not a count for menu items, content cards or interface sections, where scanning and external information reduce the need to hold every option in memory.
The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information
Research-assisted source review
- What it supports
- All recoding schemes increased binary-digit span, although brief training produced less benefit than predicted. With extensive practice, Smith recalled about 40 binary digits by recoding them into larger familiar units. Miller argued that chunking depends on learned organisation and cautioned against treating every appearance of seven as the same limit.
- Who or what was studied
- A theoretical review of multiple earlier perception and memory studies. Its direct recoding demonstration, originally reported by Sidney Smith, used 20 participants for baseline binary and octal spans and five participants per recoding method; Smith later trained on the harder methods himself.
- Study setting
- Compared absolute judgments and immediate recall across different stimulus types. The recoding demonstration taught people to translate groups of binary digits into larger base-four, octal or decimal units before retesting binary-digit span.
- Where it may not transfer
- This is a wide-ranging theoretical paper built from heterogeneous mid-century tasks and brief study reports, not a modern interface experiment. Digit recall does not establish an ideal number of navigation options, and the paper itself says that what counts as a chunk depends on learned familiarity.
Practical takeaways
- Group information by a relationship users already understand: task, topic, sequence or decision. Give each group a specific heading and keep its items visually connected without making the page hierarchy unnecessarily deep.
- Do not use 7±2, four, or any other memory number as a menu limit. Choose navigation length from findability, label clarity, task frequency and device constraints; use search, filtering or progressive disclosure only when those tools make the structure easier to understand.
- Compare grouped and ungrouped versions with the intended audience. Measure whether people can find, explain and use the information—not recall alone—and check newcomers, screen-reader users and people working under interruption. Change the grouping if labels are misunderstood or needed items become harder to reach.
Seen in real products
How teams put it to work.
Examples identified in our interface teardowns, linked back to the full product journey.
