What Six Months of Research Notes Taught Me
I spent the first half of last year researching a novel set among the telegraph clerks of the 1880s. By June I had 340 pages of notes, 61 photographs, a spreadsheet of wage rates and precisely zero pages of fiction.
That ratio is normal. What is not normal — what I got badly wrong and want to warn you about — is what I thought the notes were for.
Research is not evidence
I had assumed the purpose of research was accuracy: get the details right so nobody writes in to complain. That is a real but minor benefit. The actual purpose is generation. You read until the period starts producing your plot for you.
The moment it happened for me was a line in a staff handbook about clerks being forbidden to discuss the contents of messages even with each other. That single rule created my protagonist’s entire problem: a woman who knows something and has no legal person to tell it to. I would never have invented that constraint. I could only find it.
The two kinds of note
Once I understood that, I split my notes into two files, and I recommend this to anyone doing period work:
facts.md— dates, prices, procedures, distances. Consulted while writing, never read for pleasure.frictions.md— anything that would make a life difficult. Rules, shortages, indignities, waiting times.
The second file is where the novel lives. Mine ended up at eleven pages, and nine of those eleven made it into the book in some form. The facts file is 300 pages and contributed maybe forty scattered details.
Accuracy keeps a reader from leaving. Friction is what makes them stay.
What to do with the rest of it
Most of what you read is not going in the book. Accepting this early is the difference between a novel and a very long essay wearing a plot.
My rule now is a hard budget, applied in revision:
- No more than one researched detail per page, on average.
- No detail that a character would not notice.
- No detail explained by the narrator that a character could explain worse.
- Anything that survives three drafts is probably load-bearing. Anything I keep defending in editorial notes is probably vanity.
Rule four cost me the tariff subplot, which I loved, and which slowed the middle of the book to a crawl. My editor was right. She usually is, and I usually take a month to admit it.
The practical setup
Nothing sophisticated. Plain text, one file per source, a consistent header so I can find things with grep, and a single page at the front of the folder listing the six questions I am actually trying to answer. That page gets rewritten monthly, and comparing versions is its own kind of progress report — the questions get narrower and better.
I also keep a running list of things I could not find out. Gaps are useful. Where the record is silent, the novelist is free, and knowing exactly where the silence begins tells you where you may invent without lying.
The part that surprised me
Six months in, I could no longer tell which details I had read and which I had imagined. This is alarming and, it turns out, the goal — but it means the fact-check pass at the end is not optional. I hired someone. She found nine errors, one of which was a train that did not exist yet, and she was worth every penny.
If you take one thing from this: read for friction, budget your details ruthlessly, and pay a stranger to check your trains.
