How much in one chunk
| Chunk | How much | Why |
|---|---|---|
| Sentence | one thought, up to 20–25 words | by the second subordinate clause the subject is lost; two facts — two sentences |
| Paragraph | one thought, 2–5 sentences, 3–5 lines on a computer | on a phone lines are half as long, and 5 lines become 10–11 |
| List item | a line or two | a bullet promises something short; longer is a paragraph |
| List | 3–7 items | more and the items blur; split into groups with subheadings |
| Manual step | one action and what should happen | two actions in a step — the second gets skipped |
| Section | one reader question, 1–3 screens | longer and it already holds two questions, so it needs an h3 |
| Table | 2–5 columns | at 360 px only 2–3 columns fit in a row; the rest scrolls off |
| Screen | one idea | what’s visible without scrolling answers one question (One screen, one idea) |
The numbers are rules of thumb. George Miller’s “7 ± 2” (1956) is the capacity of short-term memory, and applying it to lists on a screen is a mistake: a list isn’t memorized, it’s right in front of you. A list’s limit is something else — as long as the items are visibly different from one another.
Where to cut: signs in the text
| What you see in the text | How to cut |
|---|---|
| “also”, “in addition”, “besides” in the middle of a paragraph | a new thought — a new paragraph |
| three or more items separated by commas | a bulleted list |
| “first… then… after that” | a numbered list of steps |
| two or three things compared by attributes | a table: things in rows, attributes in columns |
| “if…, then…” more than twice | an “If… — Do” table |
| the paragraph’s topic changes halfway | a subheading before the new topic |
| a command, a path or a piece of config inside a sentence | a code block on its own line |
| “important”, “note that” | one warning per page, the rest as plain text |
Where not to cut
| Case | Why |
|---|---|
| an argument with “because”, “so”, “but then” | the connectors are the content; in bullets they vanish |
| a cause next to its effect | “ufw doesn’t close a container’s port” without “because the packet goes to forward” is a baffling rule |
| a warning next to the step it belongs to | a warning moved to the end gets read after things are broken |
| a one-item list | that’s a sentence; the bullet is superfluous |
| a caveat or limit: a version, “not tested on ARM” | chunking cuts it first, and without it the text can’t be trusted |
If it doesn’t work
A 15-item list gets scrolled past unread
- Cause
- A long uniform list gives nothing to hold on to: the items blur together, and the difference between them doesn’t show.
- Fix
- Group them by 3–7 under subheadings, or turn them into a table with an attribute column.
The list items are three sentences each, and it’s heavier than paragraphs
- Cause
- A list was chosen for looks, but inside there’s argument: a bullet promises a short line and doesn’t deliver.
- Fix
- Argument goes in paragraphs; only enumerations go in a list.
On a phone a paragraph fills the whole screen
- Cause
- A 6-line paragraph at 714 px is about 13 lines at 360 px: a line holds 35 characters instead of 76.
- Fix
- A paragraph is one thought and up to 3–5 lines on a computer; if it’s longer, look for the second thought.