Structural patterns form the skeleton of readable documentation. They determine how quickly readers can find information and how well they retain it.
Heading Hierarchy
The 3-Level Rule
Effective documentation rarely needs more than 3 heading levels:
- H2 — Major sections (concepts, features, APIs)
- H3 — Sub-sections (specific methods, use cases)
- H4 — Rarely needed (parameter groups, edge cases)
Quality Headings
Good headings are:
- Descriptive: “Prerequisites for OAuth Setup”
- Specific: “Configure Database Connection Pool”
- Action-oriented: “Deploy to Production”
- Scannable: Key info in first 2-3 words
Bad headings:
- Generic: “Overview”, “Introduction”, “Details”
- Vague: “Some Things to Know”
- Buried keywords: “A Few Notes About the Configuration Settings”
Front-Load Keywords
Users scan the left edge (F-pattern). Put keywords first:
- ✅ “Authentication: OAuth 2.0 vs API Keys”
- ❌ “A Comparison Between OAuth 2.0 and API Keys for Authentication”
Chunking
Chunking groups related content into digestible units. Working memory holds ~4 chunks—respect this limit.
Optimal Chunk Size
| Element | Size | Why |
|---|---|---|
| Paragraphs | 3-5 sentences | Beyond 5, readers lose thread |
| Sections | 2-4 paragraphs | Before next heading |
| Lists | 3-7 items | Beyond 7, chunk into sublists |
| Pages | 5-10 min read | 1500-3000 words |
Chunking Long Lists
Before (9 items = overload):
## Configuration
- API key
- Base URL
- Timeout
- Retries
- Cache enabled
- Log level
- User agent
- Max connections
- SSL verify
After (3 chunks):
## Configuration
### Connection
- Base URL
- API key
- Timeout
### Reliability
- Retries
- Cache enabled
- Max connections
### Debugging
- Log level
- User agent
- SSL verify
Lists vs. Prose
When to Use Lists
- 3+ related items
- Steps in a sequence
- Options or choices
- Quick reference information
- Scannable content
When to Use Prose
- Explaining relationships between concepts
- Providing context or narrative
- Building understanding progressively
- Connecting cause and effect
Three List Types
Numbered lists: When sequence matters (steps, priorities, rankings)
Bulleted lists: When items are equal/parallel (features, requirements)
Description lists: When you have term-definition pairs (parameters, glossaries)
The Scan Pattern
Readers scan in F-patterns:
- Read first heading fully
- Scan subsequent headings
- Read first sentence of relevant sections
- Decide whether to commit
Design for scanning, not reading.
Structural Checklist
- No more than 3 heading levels
- Headings are descriptive (not “Overview”, “Details”)
- Keywords front-loaded in headings
- Paragraphs under 5 sentences
- Lists chunked to 3-7 items
- Clear H2 structure visible in outline
- Each section answers one question
Sources
- Google Developer Documentation Style Guide: Heading hierarchy recommendations
- Cognitive research: Working memory capacity (Cowan, ~4 chunks)
- Nielsen Norman Group: F-pattern scanning behavior
Deep Dives
Structural Patterns
Heading hierarchy, chunking, lists, and information architecture for documentation
Reading Flow Patterns
Progressive disclosure, entry points, and navigation for documentation
Visual Hierarchy
Typography, whitespace, callouts, and code formatting patterns
Cognitive Load
Working memory limits, chunking strategies, and reducing mental overhead in documentation
Scannability
F-pattern optimization, the 60-second test, and making documentation instantly graspable
Progressive Disclosure
Layered information architecture, expandable sections, and showing depth on demand