Research for a novel is not the collecting of trivia; it is the protection of an illusion. Fiction works because of a contract: the reader suspends disbelief, and in exchange the author gives them no reason to reinstate it. A factual error, a wrongly described firearm, a plant flowering in the wrong month, a train on a line that does not exist, tears that contract up in a second. Worse, it tears it up retroactively: a reader who has caught the author once starts checking everything.
John Gardner called fiction a “vivid and continuous dream” and argued that the writer’s chief job is not to interrupt it. Most craft guides apply that to style: the clumsy sentence, the metaphor that breaks the rhythm. But nothing interrupts the dream more brutally than falsehood. A clumsy sentence gets forgiven, because the reader can see the author trying. A false fact takes away something else: the belief that the author knows what they are talking about. And the whole rest of the book stands on that belief.
Why one error costs so much
The mechanism is asymmetrical. A hundred accurate details build trust slowly and invisibly, because an accurate detail is transparent: the reader does not register it, they simply believe the scene. One error, by contrast, is as visible as a stain on a tablecloth, but only to those who know the subject. And there lies the trap: every reader knows some subject well. A doctor will catch the mishandled symptoms, a lawyer the impossible trial, a sailor the wind blowing against the plot, a Bristol local the street that does not cross the other one.
So the author is not writing for an average reader with a little knowledge of everything. They are writing for the sum of specialists, each of whom will check their own slice. The question is never “will anyone notice” but “who will notice, and what will they then think of the rest”. Reader reviews are merciless on this point: a factual error migrates into the review as evidence that “the author didn’t do their homework”, even when it concerns three sentences out of four hundred pages.
The most common factual errors in novels
Decades of editors’ and beta readers’ work settle into a repeatable list. It is worth knowing, because each type of error has a different method of verification.
Anachronisms. Things present in a scene before they existed or long after they vanished: technology, vocabulary, brands, prices, customs. The most frequent error in historical fiction, but it happens in contemporary novels too, whenever the action steps back a decade. A character in 2005 is not scrolling social media on a phone. The sneakiest are linguistic anachronisms: a word that sounds old but entered the language a century after the novel’s period.
Geography and topography. Distances that do not match travel times, rivers flowing uphill of the plot, neighbourhoods glued onto the wrong cities. A local reader knows their city better than the author, and locals are the largest group of error-spotting reviewers there is. If the action takes place somewhere real, the map is part of the craft.
Nature. Species outside their range, trees fruiting in the wrong season, animals active at an hour when they sleep, constellations on the wrong hemisphere. This type of error stings particularly hard, because nature description usually exists to build atmosphere, and atmosphere built on falsehood only works until the first reader with binoculars.
Procedures: medicine, law, police, military. Professions with their own rigours are a minefield. An autopsy that takes an hour, an interrogation with no legal basis, military ranks confused with posts. An encyclopedia is not enough here, because the error is rarely about a fact and usually about a procedure: order, competence, duration.
Numbers and measures. Speeds, payloads, ranges, doses, calibres. Everything with a unit can be checked, so everything with a unit will be checked.
A story world inconsistent with itself. A separate category: errors against facts the novel itself established. Eye colour changing between chapters, a journey that took a week in book one and a day in book three. That is the domain of internal consistency and the story bible, and it obeys different rules than research, but readers punish it identically.
How to research a novel without drowning
Research has two failure modes. The first is neglect, covered above. The second, less obvious, is excess: months of documentation instead of writing, followed by a compulsion to show the reader all that work. You can recognise over-researched novels by scenes in which characters lecture each other on things they both already know. Why that kills a scene is a separate piece: infodump and exposition.
The practical compromise looks like this:
- Draft first, document second. At first-draft stage, limit research to the things the plot’s construction depends on: does this poison even work in that timeframe, is this journey feasible in a week. Everything else gets a to-check marker in the text and you keep writing. Stopping a scene to spend two hours reading train timetables is the most expensive form of procrastination there is.
- Keep a claims register. A separate file: every claim about the real world that made it into the text, plus the source it came from. It sounds bureaucratic, but at editing time, and when a publisher’s queries arrive, this file is priceless. It is the same principle as a story bible, pointed outwards from the book.
- Talk to people in the field. An hour with a paramedic gives you more than a week of reading, because a specialist knows what everyone gets wrong about their profession. Crime writers have relied on police and medical consultants for decades. One reliably good question: “what do films about your job always get wrong?”
- Rank your sources. Official databases and institutional publications before encyclopedias, encyclopedias before blogs, everything before your own memory. Check numerical facts at the data source, not in a text quoting it: numbers decay a little with every retelling.
- Verify before submission, not after. Fact-checking is a self-editing stage, just like pacing work or sentence polish. Put it on the checklist we describe in is my book ready to publish, because after publication every error is public.
How much realism does fiction need
A fair question: it is fiction, so how much of this actually has to be true? The answer depends on the contract with the reader, and that contract is signed by the genre. A realist novel set in present-day Edinburgh promises Edinburgh and will be held to that promise. Fantasy promises coherence instead of accuracy: dragons may exist, but they must exist consistently, which is the subject of our guide to worldbuilding a consistent world.
The most interesting case is the deliberate departure. An author has every right to move a street, invent a small town, compress a procedure that would really take too long for the pace of the plot. The difference between an error and poetic licence is that licence is a decision: the author knows what they are changing and why. Many writers add an author’s note at the back honestly listing their departures from fact. Readers take this surprisingly well, because the note tells them exactly what they wanted to know: the author knows the truth and changed it on purpose.
The problem, then, is not invention. The problem is unconscious invention: the moment when the author merely believes they know. Against that there is no weapon except verification, because by definition you cannot see your own error from the inside. It is only visible to someone checking claim after claim.
That is exactly how fact-checking works in Vellam: the analysis goes through a chapter, picks out claims about the real world, from geography through nature to numbers and dates, and tests them against sources. You get a list: what has support, what is contested, what could not be verified, each with a reference so you can weigh it yourself. It does not replace research; it closes the gap where memory and confidence fail.
Frequently asked questions
How do you research a novel?
In two passes. Before writing, check only the facts the plot’s construction depends on: the feasibility of a journey, how a poison works, the timeline of historical events. While drafting, mark doubtful spots with a to-check tag and do not interrupt scenes. After the draft is done, systematically verify every claim about the real world, ideally keeping a register of claims along with their sources.
What factual errors most often end up in novels?
Anachronisms (objects and words from outside the period), geographical errors (distances, the topography of real cities), nature errors (species out of range, wrong flowering seasons and animal activity), procedural errors (medicine, law, police, military) and wrong numbers and measures. Contradictions with facts the novel itself established are a separate category.
Can you change facts in fiction?
Yes, and writers always have, provided it is a conscious decision rather than an oversight. An author may move a street or compress a procedure for the sake of pace; many list such departures in an author’s note. The boundary is the genre contract: a realist novel promises accuracy to the world, fantasy promises the internal coherence of its own.
How do you fact-check a book before publication?
Systematically, claim by claim, not by feel. Verify numerical facts in source databases, consult professional procedures with practitioners, geography with a map, nature with range atlases. A claims register built during writing helps, as do tools that automatically pick out real-world claims in the text and test them against sources.
Do readers really notice factual errors?
Yes, because every book lands on a sum of specialists: a doctor checks the symptoms, a lawyer the trial, a resident of the described city the streets. No single reader catches everything, but it is enough that each catches their own slice. Factual errors appear regularly in reader reviews as an argument against the credibility of the whole book.