Research for Fiction

The signal is fluency. You have enough research when you can invent plausible details in the right register without stopping to check. If you’re pausing mid-sentence to verify a fact, the material hasn’t become instinct yet. If you’re inventing freely and the invented detail feels right — right enough to keep moving — you’re there. The research has migrated from notes to inhabitation. This is not permission to be sloppy: fluency built on wrong information produces confident errors. The threshold assumes you’ve read enough that your invented details land in the right register even when you’re not consciously checking.

Secondary sources — histories, biographies, academic surveys — are pre-interpreted. The author has already decided what’s significant, what’s causal, what the period felt like. Reading only secondary sources means inheriting someone else’s framing. Primary sources give you the texture that secondary accounts smooth away: memoirs catch the period idiom; diaries record what people noticed; court records preserve verbatim testimony, including its contradictions; letters carry the actual register of speech. For historical fiction, these aren’t optional enrichment — they’re the difference between a story that sounds like it’s about the past and one that sounds like it comes from inside it. Hard SF has an equivalent problem: peer-reviewed papers don’t read like science journalism, and the novelist who researches from magazines rather than papers knows something different and writes something different. Secondary sources work when atmosphere matters more than specificity. For psychological particulars, go to the source.

Hilary Mantel spent years with Tudor primary documents before writing Wolf Hall, but she also visited the physical locations. That combination matters. Book knowledge tells you what happened; embodied knowledge tells you how light moves through a particular window, what a space does to the people inside it. Walking the ground, handling period objects, cooking period recipes — these produce details no amount of reading replicates. The specific, unexpected sensory fact that doesn’t appear in any book is usually what makes a scene feel lived-in.

Expert interviews carry a specific hazard. The interview mode of attention is wrong for fiction writing. A journalist transcribes; a novelist absorbs. Take sparse notes, then sit with the conversation: what you remember without the notes is what landed. The rest was data, not texture. The deeper hazard is that a compelling real story becomes a constraint. Once you’ve heard it, deviating feels like betrayal. The research stops serving the fiction and starts replacing it.

This is where craft separates from information. The question every researched detail must pass: would a character who lives in this world already know this? If yes, they don’t explain it. They use it, refer to it obliquely, bump against it as part of the furniture of their existence. See Show Don’t Tell and The Iceberg Principle for the underlying technique.

Werner, the radio technician in All the Light We Cannot See, doesn’t explain how radio waves propagate. He hears things in the static. He calibrates. He worries about precision. The reader absorbs radio physics through his relationship with the equipment, not through his interior monologue about electromagnetic theory. Doerr knows the physics completely; Werner lives it. That’s the distinction.

Deployed correctly, research produces specificity that creates authority. The specific brand of cigarette, the exact ration allocation, the precise social protocol for a given encounter — these details signal that the writer knows this world. The reader relaxes into the story. See Setting in Realist Fiction for how specific detail functions as texture rather than information.

Over-research has a characteristic failure mode: scenes that exist to justify the research rather than advance the story. The history lecture. The expert character who explains everything to the protagonist, who asks obligingly naive questions. The technical briefing where the novel stops moving so the writer can show their work.

This is a drafting problem with a structural origin. The writer has spent months learning something and feels compelled to prove it. The proof happens in the text. Readers experience it as friction — a story that has suddenly stopped being a story and started being an educational module. See Exposition for how to handle necessary information without halting narrative momentum.

The fix isn’t to cut the research — it’s to trust the iceberg. Nine-tenths of what you know should never appear on the page. It’s there to make the visible tenth feel solid. If you’ve researched the economics of the Dutch art market in the seventeenth century, you don’t need a scene explaining it. You need Tartt’s Hobie knowing exactly what a painting is worth and exactly what a buyer is offering, and the gap between those two numbers driving the scene.

Historical Fiction has two authenticity requirements that are often confused: period accuracy and period feel. Accuracy means getting facts right — events, dates, technology, social structures. Feel means making the past seem inhabited by people who didn’t think like us. The second is harder and more important. The specific pitfall is making historical characters think like contemporary people in period costume. A medieval peasant’s relationship to disease, death, hierarchy, and the divine is not a modern relationship with different clothes. Getting feel requires immersion in primary sources — the actual writing and speech of people from the period.

The authenticity-versus-accuracy tradeoff is real. Some period-accurate language is opaque to modern readers; some period-accurate attitudes require handling. Language is usually the most consequential call: a character who speaks in contemporary idiom in a 16th-century setting breaks the spell regardless of surrounding accuracy.

Speculative fiction requires extrapolation rather than discovery — you can’t verify anything against an external record. Hard SF has the clearest framework: start from real, primary-source science and extend logically. Le Guin didn’t need physics papers for The Left Hand of Darkness, but she needed to understand human gender sociology well enough to extrapolate what a world without fixed gender would look like at the level of language, social texture, and psychology. The research was sociology and anthropology. Secondary-world fantasy draws from everywhere — feudal economics, metallurgy, military logistics, religious structures — and the research is therefore comparative rather than deep: enough from enough fields that the world coheres without any single domain receiving the primary-source treatment historical fiction demands. See Fantasy and Sci-Fi World-Building and Magic and Technology Systems.

Comprehensive knowledge becomes a liability when it produces the compulsion to be comprehensive on the page. Patrick O’Brian’s Aubrey-Maturin novels are saturated with Napoleonic-era naval detail — rigging, tactics, ship medicine, prize law — and they work because O’Brian trusts readers to follow competent people doing difficult things without full explanation. The research is absorbed. But O’Brian could have written annotated sailing manuals, and some writers with equivalent knowledge do exactly that. The diagnostic: does this scene advance what the characters want, fear, or have at stake? If yes, the research is doing its job. If the scene’s primary function is to prove the writer knows something, it’s misusing the research. Cut the explanation. Keep the texture. The reader doesn’t need to understand Dutch Golden Age painting to feel the weight of Theo Decker’s obsession with The Goldfinch. They need to feel the painting the way he does. See Story Development for how this applies to scene function.