How to research a novel (without disappearing into it)
· The StoryProp team
There are two ways research kills a novel, and they look nothing alike. The first is a book that gets things wrong in ways a reader can feel — a character who pays for something with money that would not have existed, a hospital scene where nobody behaves like hospital staff, a boat that turns the wrong way. The second is a book that never gets written because its author spent most of a year becoming an amateur authority on canal locks. Both failures come from the same missing skill, and the skill is not knowing more. It is knowing when to look, how much to look, and what to do with what you find so it stays useful three hundred pages later.
Research debt and research procrastination
Research debt is what you take on when you draft past something you do not know. You write "she filled out the intake form" without knowing what an intake form asks, and you keep going. That is often the right call. Debt is not failure; it is a decision to pay later in exchange for momentum now. The danger is only that debt goes unrecorded, and unrecorded debt compounds into a revision pass where you cannot remember which details you checked and which you waved through.
Research procrastination is the opposite trade and the more common one. It feels virtuous. You are working — there are tabs open, there are notes, there is a document titled RESEARCH that is now longer than your draft. But the tell is simple: procrastination research answers questions your scenes have not asked yet. If you cannot name the sentence in the manuscript that the answer will go into, you are not researching. You are avoiding.
The diagnostic I use is a single question, asked out loud: what scene does this go in? If the answer is a specific scene — the argument in the parking lot, the funeral, chapter nine — keep reading. If the answer is "it might come up," close the tab and note the question for later.
Just-in-time beats pre-loading
The pre-loading instinct says: learn the world first, then write in it confidently. It fails for a reason that has nothing to do with discipline. Before you draft, you do not know what your book needs. You think you are writing about a woman who inherits her mother's house, so you read about probate. Then you draft, and the book turns out to be about the neighbor, and the probate is one line of dialogue, and the thing you actually needed was how long a house can sit with the power off before the pipes are a problem.
Drafting generates the questions. That is most of what a first draft is for. So the sequence that works is: write the scene, mark the holes, then research the holes — in a batch, after the scene exists, when you can see exactly which facts have weight.
Marking a hole should be fast and consistent. Pick one notation and never vary it, so a single search finds every one of them later; some writers use a bracketed word, some use a rare character string no ordinary prose would contain. The form does not matter. In a draft it looks like this: "He signed the form — [CHECK: does the coroner sign, or a deputy?] — and handed it back." You have not stopped, and you have not lied to yourself. In a week, when you sit down for a research session, you have a list of real questions instead of a vague sense that you should learn about coroners.
The rabbit-hole budget
Curiosity is not the enemy; unbounded curiosity is. So bound it. Before a research session, decide how long it gets and what it is for, and treat the boundary as real. When the time is up, you stop, whether or not you found the perfect source — and the things you did not resolve go back on the list.
The budget has a second and better use: it gives you permission to actually enjoy the wandering inside it. A writer who has decided that Tuesday morning belongs to lighthouse keeping can follow that subject sideways with a clear conscience, because Tuesday afternoon belongs to chapter four regardless. Guilt-free rabbit holes are shorter than guilty ones — guilt makes people linger, because stopping means admitting the time was stolen. One more rule: bring back less than you found. A session that produces a dozen pages of notes has mostly produced future work. Close it by writing three or four sentences of what you actually learned, in your own words, and let the rest go.
Keep the source attached to the claim
Here is the failure mode that costs the most, late, when it is most expensive to fix. You wrote a detail months ago. You are almost certain you checked it. You cannot remember where, or when, or whether the thing you read was authoritative or a forum post by someone confident. Now you either re-research from zero or ship on faith.
The fix is unglamorous and total: a fact and its source travel together, always, from the first moment you write the fact down. Not in a separate bibliography — attached. Where the claim lives, the source lives, with the date you found it. Dates matter more than writers expect, because a surprising number of facts are perishable: procedures change, institutions reorganize, a thing that was true when you checked it quietly stops being true while your draft sits.
The second half of that discipline is keeping evidence separate from canon. What a source says and what is true in your novel are different categories, and collapsing them causes real damage. A source tells you a certain kind of trial takes months. Your book compresses it to weeks, on purpose, and you know why. That decision belongs in your story's canon with its reason recorded. The source belongs in your evidence. If those two live in the same undifferentiated pile of notes, a later revision will "correct" your deliberate compression and you will have lost something you chose.
This is one place a project-aware tool earns its keep. In StoryProp, research is saved with its sources and dates, and evidence is kept separate from canon — so a fact stays checkable months later, and a deliberate departure from the record does not get mistaken for an error.
Invented, verified, and the honesty in between
Every novel contains three kinds of detail. Verified: you checked it, you know where. Invented: you made it up and you know you made it up. And the dangerous third category — details you absorbed from somewhere, cannot source, and have started to believe. That third pile is where most factual embarrassments come from, because you defend those details as confidently as the checked ones. So mark them. When you write a detail you are not sure about, say so in the margin at the moment of writing, because that is the only moment you still know the difference. A week later the sentence will read as equally solid whether you researched it or dreamed it.
Take the demo manuscript we use around here. In The Quiet House, Mara drinks from a chipped cup that belonged to her mother. Nothing needs checking — it is invented, and inventing it is the job. But if the same scene says the house was condemned in a specific year under a specific municipal process, that is a claim, and claims want sources. The craft skill is telling those two sentences apart while you are writing them, when they feel identical.
Invented detail is not a lesser grade of research. In fiction it is often the stronger choice, because a made-up detail can be shaped to the scene while a real one has to be accommodated. The rule is only that you know which is which — a real institution named correctly, an imaginary town on a real river, a procedure that resembles the true one and departs where the story needs it to. Entirely invented systems are a different discipline with its own traps; that is the territory of worldbuilding without the wiki trap, and the failure mode there is building more world than the book will ever use.
What good enough looks like
Readers are not fact-checkers, but they are pattern-detectors, and what they detect is not error so much as unfamiliarity. The writer who has never been in a courtroom gives away the game not by getting a rule wrong but by describing the room the way television describes it. So aim your research at texture more than at facts: what the place smells like, what the boring parts are, what the people who work there find tedious or funny, what nobody outside would think to mention. One person who has done the job for a decade, in an unhurried conversation, is worth a great deal of reading.
And accept a threshold. Somewhere short of certainty there is a point where more research stops improving the book and starts protecting you from writing it. You will recognize it: the questions get finer, the answers stop changing any sentence, and the reading is enjoyable in a way the drafting is not. That is the signal. Note the remaining holes, put them on the list, and go write the scene. The list will be there Tuesday.
Research is only as good as your ability to find it again — and to tell a checked fact from a chosen invention. StoryProp keeps research with its sources and dates, and keeps that evidence separate from what your story has decided is true.