Context cardTransitional topic
Working memory and procedure steps: why "seven plus or minus two" is the wrong rule
Does working memory research justify a fixed maximum number of steps in a procedure, such as seven plus or minus two?
The short answer
No. George A. Miller's 1956 paper was about the limits of absolute judgment and immediate memory span and about recoding information into larger chunks; it set no rule for documents. Later work, notably Nelson Cowan's 2001 review, puts the capacity of working memory at about four chunks, but a reader of a procedure does not need to hold all steps in mind, because the steps stay on the page.
For: Technical writers and editors who write procedures and style guides
Key points
- Miller (1956) reviewed span limits in absolute judgment and immediate memory and showed that recoding into chunks extends what can be held; he treated the recurring seven with caution.
- Cowan (2001) estimates the capacity of working memory at about four chunks (roughly three to five) when rehearsal and chunking are controlled.
- A procedure on a page or screen is an external memory: the reader holds the current step, the goal and the place in the list, not all steps.
- What loads working memory within a step is what must be held while acting: several actions, values to carry over, conditions to remember, a look away from the text.
- No step-count rule is derived from this research; grouping, one action per step and visible progress follow from the task analysis and testing.
The context
The myth
Style guides and training slides often say that a procedure may have at most "seven plus or minus two" steps "because of short-term memory". The rule sounds scientific, but it rests on a misreading of one paper and ignores what a procedure is.
What the research shows
Miller (1956). George A. Miller's paper "The magical number seven, plus or minus two" reviewed two kinds of experiments: absolute judgment (how many categories of a single stimulus dimension, such as tones of different pitch, people can tell apart) and immediate memory span (how many items people can repeat at once). Both came out at around seven, but Miller treated the coincidence with explicit caution. His more lasting point was recoding: grouping items into larger units, or chunks, lets people hold far more information than the bare span suggests.
Working memory. Alan Baddeley and Graham Hitch (1974) replaced the idea of a single short-term store with a working memory of several components that hold and manipulate information during a task. The theory glossary describes working memory, working memory capacity and chunking as terms of their own.
About four chunks. Nelson Cowan (2001) reviewed studies that prevent rehearsal and chunking and concluded that the capacity of working memory is about four chunks, roughly three to five. The number depends on what counts as a chunk, which in turn depends on the reader's knowledge: an expert holds a whole familiar sequence as one chunk.
What it means for documentation
A numbered procedure is printed or displayed while the reader works. It is an external memory, so the reader does not need to keep the steps in mind. What working memory has to carry is the current step, the goal, the place in the list and anything that must be remembered while the eyes are on the machine. That moves the design question from the length of the list to the content of each step:
- One action per step, or a small group of actions that belong together, so nothing has to be held across a look away.
- Values and conditions where they are used: a torque value, a setting or a precondition in the step that needs it, not three steps earlier.
- Visible structure: numbering, short stage headings for long procedures, and a clear result statement, so the reader finds the place again after an interruption. Interruptions also call on prospective memory and cause a resumption lag, both described in the glossaries.
- Chunking by meaning: long tasks split into sub-tasks that match how the work is organised, as in a DITA task, not cut at an arbitrary count.
How many steps a task needs follows from the task itself, from the task analysis and from testing with users. Minimalist instruction (John M. Carroll, 1990) argued for short, action-oriented instruction from studies of real users, not from a span limit.
Limits
The capacity estimates come from laboratory tasks with lists of unrelated items. They support the direction of the advice (do not make the reader hold things) but not any number. The usability of a procedure, and of the safety information in it, is checked with users. No cognitive argument replaces the risk assessment or legally required safety information: a safety-relevant step is never dropped or merged away to reach a target length, and short steps do not make instructions for use compliant.
Questions readers ask next
- So is there a maximum number of steps?
- Not from memory research. A long procedure can be easier to follow than a short one if each step holds one action and the structure shows where the reader is. Split long procedures by meaning, into sub-tasks, not by a count.
- Is the magical number now four instead of seven?
- Cowan's estimate of about four chunks describes working memory when rehearsal and chunking are prevented. It is not a rule for documents either; it explains why a single step should not ask the reader to hold several new items at once.
- Does chunking mean grouping steps into blocks?
- In the research, a chunk is a unit the reader already knows, such as a familiar word or sequence. In documentation, grouping steps under meaningful sub-headings helps because it matches how the reader thinks about the task, not because it reaches a number.
Sources
- George A. Miller (1956), The magical number seven, plus or minus two: Some limits on our capacity for processing information — Psychological Review, 5 October 2026
- Nelson Cowan (2001), The magical number 4 in short-term memory: A reconsideration of mental storage capacity — Behavioral and Brain Sciences, 5 October 2026
- Alan D. Baddeley and Graham Hitch (1974), Working memory — Psychology of Learning and Motivation, Vol. 8, 5 October 2026
- John M. Carroll (1990), The Nurnberg Funnel: Designing Minimalist Instruction for Practical Computer Skill — MIT Press, 5 October 2026
Review log and changes
Every context card is checked against its sources before it is published, and again whenever it changes; the date under the byline is the last review. Corrections (something was wrong) and additions (something was missing) are logged below with date and time (Berlin time). Typos, formatting and link fixes are not listed.
Reviewed
No corrections or additions since publication.