/research

Research Practice

What needs checking

Check anything a reader with first-hand knowledge could catch:

• Dates, sequences, and durations of real events
• How a job, weapon, vehicle, tool, or procedure actually works
• Medicine: injuries, recovery times, drugs, and what a body can survive
• Law and police procedure for the setting's country and era
• Geography, travel times, weather, tides, daylight hours for the date
• Period detail: money, clothing, food, speech, technology available
• Language and names for the culture, region, and era

Invented worlds still borrow physics, biology, and human behaviour; check those when the plot turns on them.

Forming questions

Turn a topic into questions the prose actually answers. "Victorian medicine" cannot be verified; "Would a London surgeon in 1872 have used carbolic spray?" can. One note may hold several related questions; list each under ## Question so findings can answer them one by one.

Search plan

Write the plan into the note before searching so the work can be resumed or handed to the user. The examples in this file are illustrative: the sources, figures, and URLs show the format and are not verified facts.

## Search Plan

- Primary: Environment Agency tide tables for 2024; Severn Bore
  Trust timetable (official data)
- Secondary: Smith, *The Severn Estuary* (2020), ch. 4
- People: a local surfer or river warden (interview)
- Place: Minsterworth bank at a spring tide (site visit)

Search the primary sources first, then use secondary sources to interpret them. Record dead ends too ("no 1950s timetable online; archive request needed") so they are not repeated.

Judging sources

Prefer, in order: primary sources (records, first-hand accounts, official data), expert secondary sources (academic work, professional references), then general references. Note the date of each source; practice changes. When two sources disagree, record both and mark the note disputed until the user decides which the story follows.

A primary source is evidence from the time or the thing itself: a record, a diary, a photograph, a statute, a manual, the place, a person who was there. A secondary source interprets primary ones. Treat encyclopaedias, forums, and AI summaries as leads to a real source, never as the source.

Recording findings

Write findings as short factual statements, each with its citation: author or institution, title, date, and page, section, or URL.

The figures below are illustrative placeholders; verify real values against the sources before using them.

## Findings

- The Severn bore is largest near the spring equinox, two to three days
  after a new or full moon. (Smith, *The Severn Estuary*, 2020, p. 112)
- It travels upstream at about 16 km/h. (Environment Agency tide notes,
  https://example.gov.uk/severn-bore, accessed 2026-03-02)
- > "The bore is at its best when the tide exceeds 9.5 metres at
  > Sharpness." (Severn Bore Trust timetable 2024, p. 2)

Quote exact wording when the prose depends on it: a term of art, a regulation, a period phrase. Keep quotes short and always give the page or URL. Add each source to the sources frontmatter list as well.

Confidence

Set confidence from the evidence, and revise it as findings change:

Confidence is separate from status. A note can be verified at medium confidence when its one good source is enough for how the prose uses it.

Accuracy levels

Risk and reviewers

Set risk whenever a mistake could cause real harm or liability:

• legal - law, courts, contracts, police powers presented as how things work
• medical - symptoms, dosages, treatment, survivability
• weapons - how weapons function or are made
• safety - anything a reader might copy (climbing, diving, chemistry, fire)
• cultural - portraying a culture, faith, community, or identity
• defamation - real, identifiable people or organisations
• technical - engineering, security, aviation, or other specialist systems

A risky note needs a qualified human reviewer before the chapters using it are final. Record them in reviewed-by, a list of names or roles that story add research has no flag for, so edit the frontmatter by hand:

reviewed-by:
  - "Dr A. Patel, A&E consultant"

Research gathers facts and flags concerns; it never substitutes for professional advice, and the agent never gives legal, medical, weapons, or safety advice. Omit operational detail that would let a reader cause harm; the scene rarely needs it. Route defamation and sensitivity questions to the editorial-review skill.

Deliberate departures

Fiction may bend facts. Record every intentional change under ## Story Use with the reason ("moved the bore a week earlier so it lands on the festival"). This stops a later fact-check from "fixing" a choice.