Back to the table

83 Intermediate Hi History Worldbuilding

History

The timeline of events before the story begins.

What it is

History is the sequence of events that actually happened in your world before the story begins. Who conquered whom. In what year. What the settlement did to the map. The word that matters in that definition is actually. Plenty of people in your world have opinions about the past. History is what really happened, and only you know all of it.

That makes it a private document. Almost none of it belongs in the book. Of everything in the world section of this table, history is the one part that has to agree with itself, because dates have no poetry in them. Someone was born either before the war or after it. There is no third option, and a reader with a pencil will find out which one you picked.

So what goes in the file? Only the events that are still doing damage. A border drawn in the wrong place. A family hiding what it owes. A law passed during a famine that has outlived every memory of the famine. The rule, in one line: keep a past event only when you can name what it is doing to someone today. If you cannot name it, cut it.

Most teaching runs the other way. Build the chronology, then write the plot. History works better in reverse. Start with the friction you want deep in the book, around page two hundred, and trace it backward to the event that caused it. The reason is what readers actually want. They do not want completeness. They want the feeling that this place has a past. Four or five well-placed disagreements between characters will give them that feeling. Nine hundred years of dated entries will not.

Before any of that works, you need to separate history from three elements that sit right next to it. Blur them and each one stops doing its job.

Lore is what the people in your world believe about the past. The songs. The official version. The version taught in school. The rumor in the tavern. Belief can be false, and the interesting kind usually is. History is what you, the author, know underneath all that. The gap between what people believe and what happened is territory the two elements share, and it is the most productive vein in the whole world section. Anything about belief goes to Lore. The timeline itself lives here.

Backstory is one character's past. History is the world's. The difference matters more than it looks. A character's past has an owner. She can answer questions about it, refuse to answer, or lie. A world's past has no owner. Everyone uses it, and it outlives every witness.

Worldbuilding holds the larger questions: how much to build, why readers police consistency so hard, and why the info-dump is the writer's failure to leave material under the surface. Worldbuilding is the front door to this page. Read it first. Nothing from it is repeated here.

A fourth page, Exposition, covers how to deliver any information, what each method costs, and how to do it without stopping the story. Here you get only the delivery problems that belong to history alone. The rest go back to Exposition.

Here is the test that decides what gets into the file. A past event earns a place in your notes only when it has an effect you can name on page one. Two reasons get rejected on sight: "it explains the culture" and "it's interesting". You need a specific, current effect. Someone is still angry. A boundary sits in the wrong place. An office still exists for a purpose everyone has forgotten. A family debt is still unpaid. A word is still used as an insult.

This test is what separates working history from chronology as a hobby. You can run it on a finished draft in minutes. Beside every entry in your timeline, write "Because of this, today, ______." and fill the blank. If the blank stays empty, that entry is furniture. Move it to a file called maybe and ignore it.

Two qualifiers. First, atmosphere counts as an effect, as long as the event is "still causing something". An aqueduct that broke and was never repaired, because nobody knows how anymore, means the city is poorer and less capable than the one that built it. That passes. An event whose whole output is one descriptive paragraph fails. Second, the reader does not have to see the effect in chapter one. You have to see it by the end. A war that pays off in the last fifty pages passes. A war that never pays off fails, however often it gets mentioned.

Almost every piece of invented history that works takes one of three shapes. When you are stuck, pick the shape first and then invent an event to fit it.

A real wrong boundary shows how much one afternoon can do. On May 10, 1871, the Treaty of Frankfurt moved Alsace and part of Lorraine from France to the German Empire, after the French defeat in the Franco-Prussian War. At the close of the First World War the land went back to France, and the Treaty of Versailles ratified the return in 1919. Two changes of sovereignty inside fifty years. The people stayed exactly where they were. What changed was the language their children were schooled in, which army conscripted a young man, and which law governed a marriage. All of it from one afternoon of diplomacy.

Timothy Hickson's On Writing and Worldbuilding: Volume II, listed below, has a chapter titled World Histories, and the chapters near it, on civil wars, place names, and how class systems are kept up and break down, show the same thing from several angles: how a dated list turns into current trouble.

None of this belongs to fantasy alone. Contemporary fiction runs on history just as hard. A town whose factory shut in 2009. A family novel where two sisters have not spoken since 1987 and nobody says why. A crime novel where a case from the 1970s was closed badly by the man who now runs the department as chief. The mechanism is identical every time. Something happened before page one. It still presses on people. They know it only in part, and each party knows it differently.

Genre changes only the source of the events. In secondary-world fantasy, where the whole world is made up, you invent them. In historical fiction, and in any fiction set in the real recent past, you research them, you cannot alter them, and you take on a separate craft problem of records that stay silent, which is Historical Fiction's territory. In contemporary fiction the national events are real and the family events are invented. If you write realism and were about to skip the whole world block of this table, this is the page about your book.

Almost every guide, template and questionnaire on world history runs forward. Founding, then dynasties, then the present. Chronology first is a legitimate method, and it is the wrong default. It generates bulk you cannot test, because the early entries cannot be wrong yet; nothing exists for them to contradict. And it produces a past that is about itself: tidy, unconnected to the book's themes, and stuffed with names nobody needs.

The backward method has five steps. First, start from the friction you need, and make it a scene rather than an era. A woman wants to sell her late mother's house and learns that she cannot. Two guilds will not sit in one room. A soldier will not take orders from anyone with a certain accent. Second, ask what must have happened to produce exactly that. Vagueness fails here. "She cannot sell the house" needs a precise legal or customary cause, and that cause needs an origin. Third, give the event a rough date and ask who is still alive. Inside living memory, you get witnesses who quarrel. Two generations back, you get families holding positions without knowing the reasons. Three centuries back, you get a rule nobody can justify and nobody can repeal. Fourth, name two side effects. They are free texture, and one of them will probably turn into a subplot. Fifth, stop. Do not fill in the century before.

Every entry built this way arrives with its effect already attached, so it passes the test without being asked. You also get a bonus: the slightly shabby texture of real history, where a big rule traces back to something small and unflattering.

Chronology first is the right choice in three cases. You are building a world for a decade of writing, a storycosm rather than one novel, so a spine of major periods repays the effort and laying it out forward shows up impossibilities early. You are writing a saga across many generations, where the chronology is the structure itself. Or you simply enjoy it, which is fine as long as you admit it is fun rather than preparation. Worldbuilding has the test for telling those two apart.

A good hybrid: one morning forward, in the roughest strokes, five or six named periods with no dates. Then stop, and build everything else backward from the plot.

Now the question everyone actually asks: "how much history do I need to invent". The answer is that the reader has to feel a past. That feeling is a repeatable effect, a few signals create it, and it needs neither a chronology nor completeness. No reader has ever closed a novel wishing the world had three hundred more years of backstory.

Four signals carry almost the whole effect.

The first is offhand disagreement. Two characters touch the same fact with different assumptions and neither explains. One says the old regime built the road. The other says the old regime only finished it. No argument follows. The talk moves on. The reader works out that these two people remember the past differently. Of the four, this one is the strongest, and it costs one line.

The second is patchy survival. Absurd detail here, nothing at all there, and no pattern to it. A parish ledger that is complete from its first year and blank before. A guild that knows the price of iron in a year that did not matter but has no idea who its founder was. Real archives look exactly like this, so it is the quickest way to make invented records convincing. It also gives you honest blanks to hide plot in.

The third is offhand dating. Nobody in conversation says "the war of 1871". People say "before the war", "the bad winter", "when the mill was still open", "my mother's time". When a private landmark is used as a date, the event has grown into a unit of time, and that tells the reader more than any year could.

The fourth is names of the forgotten. A street, a bridge, a disease, a measure, a knot, a coin, a month, each named after someone nobody in the room can identify. Real places are thick with these names. No signal costs less: a proper noun and a shrug.

Notice what all four have in common. None of them states a historical fact. Each shows a relationship between people now and events then, and the reader rebuilds the events. Visible depth is bought in hints, and the price is per line, never per page.

Worldbuilding's rule that you should know more than you show holds here too, with one correction. The multiplier is two or three, never two hundred. Twenty entries that all cause something radiate more history than four hundred that mostly sit idle. A world feels old because of how strongly the past connects to the scene, never because of how much of it there is.

The most common technical fault in invented history is a timeline that behaves nothing like a real one. Once you can see it, it is simple to fix.

Six symptoms give it away. Ten thousand years in which almost nothing changes. A single empire that rules unbroken for two thousand years and then collapses within a year. One named event per century, evenly spaced. Technology that never moves, so the swords and plows of the founding are still in use. Language that never moves, so a modern character reads a nine-hundred-year-old document without trouble. And a "dark age" sized exactly to erase whatever the plot needs erased.

Readers feel the falseness without being able to name it, because every one of these breaks something they know, without any schooling, about their own world. A sense of scale helps. The Great Pyramid at Giza was finished around 2560 BC. Cleopatra VII lived from 69 BC to 30 BC. Neil Armstrong walked on the Moon in 1969. Count the years and Cleopatra was closer to the Moon landing than to the pyramid standing in her own kingdom. In her day the pyramids were already ancient monuments that drew Greek and Roman sightseers. Ten thousand years is about four of those gaps piled on top of each other, and inside each one everything should have changed at least once.

Six repairs, none of them expensive.

Shrink the number. The commonest fix, and the best. A thousand recorded years with two collapses in them beats ten thousand unbroken years, both for belief and for use. Hardly any plot needs more depth than a reader can hold in mind.

Put in gaps. Real long histories alternate ordered periods with breakdowns, where the center fails and the records stop. Historians carve up the dynastic history of Ancient Egypt with intermediate periods for exactly that reason. Two or three collapses are worth more than any amount of detail, and each one hands you an honest blank.

Move the language. The oath everyone swears is in a form of the tongue nobody now pronounces properly. A legal term has slid away from its original meaning. An old inscription can only be read by a specialist. One sentence, large return.

Move the technology, in fantasy too. Something was invented after the founding: a better plow, printing, a drug, a way of making larger windows. Say what it displaced and whose livelihood it ended. A thousand years with no invention means a world with no economic history, and readers feel that airlessness even when they cannot name it.

Make the periods uneven. Twelve years of frantic change, then eighty years whose only records are two failed harvests. Real history bunches up. Regular spacing exposes the template underneath.

Distrust the hard-times model. Pop history says "hard times make strong men": poor, hardened people from the margins inevitably overrun rich, softened centers. A great deal of invented history rests on that belief. The historian Bret Devereaux took it apart in the Fremen Mirage, a four-part series that began in January 2020 on his blog A Collection of Unmitigated Pedantry, and argued from archaeology and written sources that the folk model mostly runs backward: organized states are far better than their fringe neighbors at raising and using armed force. You can still have a barbarian conquest. Build it on a reason instead of on folk belief.

Then add at least one avoidable event. A near miss that a named person got wrong, which could plausibly have gone the other way. A timeline in which everything was inevitable reads as scenery. One bad decision makes the past a place where people lived.

After the consequence test, the most useful idea on this page is the next one, and it is where invented history stops being a spreadsheet. Every fact about the past reaches the reader through a channel. A document, a lesson, a song, a boast, a court record, a grandmother, a museum caption, the plaque on a statue. Every channel carries a bias. Even a detached third-person narrator telling you about a war is a channel with a point of view.

Here the boundary with Lore is clean. Lore is belief, and the staging of belief. History is the record: what got written down, by whom, for what purpose, and what survived. Belief and record are different things, and a world convinces more when the two strain against each other, say a character holding a document that contradicts what everyone knows.

The best model for this comes from the historian Michel-Rolph Trouillot in Silencing the Past (Beacon Press, first edition 1995, 20th anniversary edition 2015), listed below as the strongest treatment of record-keeping as power. He describes four points where silence enters the making of history: when the facts are created (sources), when they are assembled (archives), when they are retrieved (narratives), and when they are given significance after the fact (history as such). Those are four different kinds of silence with four different repairs. A ledger that left something out in 1700 is lost in one way. A document that sits unread in an archive is lost in another. You can use his model as a worldbuilding tool almost unchanged, and a free extract of the chapter that lays it out, hosted by NYU Manifold, is also listed below.

Each of his four points gives you a question to ask of every event in your timeline.

Who could have recorded it, and why would they bother? Most records that survive exist for dull administrative reasons. William the Conqueror commissioned the Domesday survey at his Christmas court at Gloucester in 1085, and it was finished by 1086. He wanted to know what could be taxed and how rich his barons were. Today it is the richest single source for eleventh-century England, and it leaves out several major towns, London and Winchester among them, because they fell outside its purpose. Your world's best records will be court transcripts, guild minutes, ships' manifests and tax rolls. Chronicles rank below all of them. And every record has holes wherever its purpose stopped.

Where did it end up, and who keeps the door? An archive is a building, a doorway and a gatekeeper. Getting in, what it costs, and what has quietly gone missing from the collection are all plot.

Who read it later? A record nobody has read is not yet history. Fiction has no stronger excuse for a scholar hero. It also makes a discovery in chapter thirty feel earned rather than convenient: the document was there all along, miscataloged.

Who decided it mattered? An event can sit ignored for a century and then get promoted to founding story when some later regime needs it. A state that edits its records usually burns nothing. It just changes which anniversary gets a holiday.

Three practical habits go with this.

Give bias a motive rather than a personality. "The chronicle is biased" bores the reader. "The chronicle was written by the abbey that received the disputed land, forty years after the grant, while the case was being heard" tells the reader what to discount, and in which direction.

Let two records clash on something checkable: a number, a date, a name. When meanings clash, you are in Lore. When documents clash on a fact, you are in history, and nothing proves faster that the uncertainty is real rather than decorative.

Rule privately on which record is right, and write the ruling in your notes. A writer who never decides will eventually contradict herself, and readers read that as carelessness rather than ambiguity.

Beyond the general problems of exposition, history brings three delivery problems of its own. Exposition covers the methods and their costs, and nothing from it is repeated here.

Dates do not stick. A reader holds on to a relationship, a grievance or an effect, never a year, and never the order of two similar years. So never let a feeling depend on the reader remembering a date. Where order matters, say it as a relationship: "her grandfather's war", "the one after the flood". The numbers stay in your private file.

Everyone already knows. A shared past is the hardest thing to deliver, because it is exactly what the locals have in common, so nobody in the room has a reason to ask. There are two familiar escapes. Put a genuine outsider in the room and give them a want, so they are more than a funnel for facts. Or, and this suits history better, put two insiders in the room who disagree about what it meant. That delivers the event, the politics of memory and characterization in one go.

The appendix is tempting. A timeline, a family tree and a list of reigns in the back matter. Some readerships want them, and Epic Fantasy covers that. The rule here is narrow: nothing load-bearing may exist only in the back matter. Appendices serve readers who are already invested and never replace placing a fact in the story. If a chapter relies on one, the chapter is failing at its job.

Real history turns everything above inside out. The craft belongs to Historical Fiction, so only briefly: the events are given rather than invented, though you still choose and weight them. Consistency is checked against the record rather than your private notes, and readers can check it too. Research is the worldbuilding, which is Worldbuilding's own line on realism. And the real craft question turns out to be the same one, met from the other side. Either the record is silent, or it is loud and says the wrong thing for your story.

Two answers are respectable. Write into the silences, the enormous unrecorded zones of private thought, speech and feeling, which is where most historical novels live. Or alter deliberately and say so, usually in an author's note. Genre readers do not forgive an undeclared change to something checkable, because it turns every other fact in the book into a maybe.

There is a third route, and Guy Gavriel Kay owns it: move the whole story sideways into an invented world, near enough that readers feel the history and far enough that you are free. One of the examples below shows how.

Why it matters

A past event makes conflict that nobody chose and nobody can end. In a typical first draft, history is atmosphere. An old war gets mentioned so the world feels old. A fallen empire gets mentioned so there can be ruins. None of it does any work. Turn it around and history becomes one of the best sources of plot you have. A protagonist can walk away from an argument. She cannot walk away from the fact that her family's claim to the mill rests on a law that was struck off forty years ago. Three standing setups are worth keeping on file.

The inherited claim is the most durable conflict there is. Two parties, one thing that cannot be divided, and both claims look fair because one event created both of them. Nobody is lying. It lasts because reasonable people cannot settle it.

The unresolved treaty gives every side a sincere case. Both parties signed it. Each understood it differently. Sometimes the text itself differs from one language to the other. The Treaty of Waitangi, below, is the case in point. A disputed agreement serves a novelist better than a broken one, because every party's case is honest.

The unadmitted wrong carries the heaviest moral weight. Something was taken or done. The heirs of the people who took it are prospering on it. The official line is that nothing happened. Of the three setups this one pulls hardest toward theme, which is why literary fiction loves it.

History hands over the originating event and nothing else. Factions explains why organized groups are worth inventing. Plot handles what the protagonist does about it.

A long series will trip on a date, and this file is what catches it. The plain, unglamorous reason the file exists is to keep a long series consistent. The failure is arithmetic, and it is almost always the same one. In book one, the war was "thirty years ago". In book two, a veteran of it is fifty-five, which fits. Book four, written three years later, needs that veteran at a siege, and makes him twelve at the time, because you were thinking about the siege and forgot his birthday. A reader with a spreadsheet catches it. Several do. They post.

The slips come in five predictable forms. A character's age against a dated event they supposedly saw. An interval that was a round number in book A and implies something different in book B. Travel and seasons: a journey that took three weeks in book one takes four days in book five, or a harvest-to-harvest trip fits inside one month. The length of a reign against the number of events assigned to it. Technology and institutions: a novelty in book three is already there at the founding by book six.

Chronology is a structural commitment, and changing it costs. George R.R. Martin planned a jump of about five years after A Storm of Swords (2000), so the younger characters could age into the roles the story needed. He abandoned it after roughly a year of work. Where a lot had happened in the gap, it all turned into flashback and explanation. Where little had happened, the result was a book about nothing much. A Feast for Crows (2005) went ahead without the skip, and the material was split between two books by where the characters were rather than by time. Skips can work. The moral is that a series' chronology bears weight from book one, and the price of revising it climbs steeply.

Readers who love a series will do the reconciling for you, and you must never count on it. Arthur Conan Doyle put Watson's war wound in his shoulder in A Study in Scarlet (1887) and in his leg in The Sign of the Four (1890). Nobody stopped reading. Readers founded a parlor game instead, now a century old, the Sherlockian "Game": it treats Holmes and Watson as historical people and Doyle as their literary agent, and explains the slips from there. Affection did that, never strategy.

Working writers protect themselves with three unexotic habits. First, one authoritative chronology in one place, never scattered across manuscripts or kept in your head. Where that file lives and what else goes in it is the Continuity Bible's territory. Second, every date counted from a fixed zero, never stated vaguely. Pick one year, a founding, a treaty, a disaster, and call it 0. Number every event. Give every character a birth number. Any age at any event is then subtraction, which cannot deceive you the way "about thirty years ago" in prose can. This one habit prevents most of the slips above. Third, an external check. On the Wheel of Time, Robert Jordan's editorial assistants Maria Simons and Alan Romanczuk kept continuity from his notes. After Jordan died, Brandon Sanderson wrote the final three volumes. Simons and Romanczuk later compiled The Wheel of Time Companion (Tor, 2015) with Harriet McDougal. Self-publishers have no assistants but face the same job: assign one particular reader to it, hire a continuity pass, or slow-read the bible on purpose before each book. Intending to remember does not work.

Real-world examples

Jade City (2017): the war is won, and the resistance's two halves now run the city

Fonda Lee's Jade City (Orbit, November 2017) won the World Fantasy Award for Best Novel in 2018, and it is the clearest modern secondary-world demonstration of the consequence test. A gangster family saga rather than an epic fantasy, it hangs a whole series on one historical event.

Kekon is an island nation that used to be occupied by a foreign empire. The Many Nations War sits where the real Second World War sits. The One Mountain Society, a national liberation movement, fought the occupiers during that war and won. The story opens one generation later.

Here is the core. After the victory the Society fractured, and two of its fragments became the ruling clans of the capital: the Mountain, headed by the Ayt family and isolationist, and No Peak, headed by the Kauls and internationalist. The driving conflict sets the two halves of the liberation movement against each other over what an independent Kekon is for. It never sets Kekonese against their former occupiers.

Three things to borrow. Recency makes it personal. One generation is the right distance for a great deal of fiction: the people who made the decisions are old but alive, their children hold positions without the reasons, and the quarrel has hardened into institutions without turning academic. Economic fact and historical fact coincide. Only Kekon has jade. Jade grants enhanced abilities, and only Kekonese can use it safely over the long term. That one fact explains the occupation, the war, why independence mattered, and why outsiders are courting Kekon now. A history holds together when the thing that was fought over is still the thing people want. And victory as catastrophe. A won liberation war opens the next conflict instead of closing anything, because the winning coalition shared exactly one goal. For any invented past, asking "who was on the same side, and what happened to them afterward" gets you further than asking "who won".

The Safekeep (2024): a house, and a wrong nobody has owned up to

Yael van der Wouden's debut The Safekeep (United States edition from Avid Reader Press, a Simon & Schuster imprint, May 2024) was shortlisted for the 2024 Booker Prize and won the 2025 Women's Prize for Fiction. A quiet, domestic novel, and nearly everything in it is driven by one event twenty years before the opening. Fair warning: the central reveal holds up the whole book, and the next paragraphs give it away.

Isabel lives alone in her family's house in Overijssel, a Dutch province, in 1961, and keeps everything exactly as it was. Her brother asks her to put up his girlfriend, Eva, for the summer. Isabel's dislike of Eva turns into its opposite. Meanwhile Eva pockets small things: a fork, a spoon, a baking dish.

The reveal: the house belonged to Eva's family. They were Jewish and were deported during the occupation, and Isabel's family moved in. Eva's father died in the camps. Her mother died after the war. Eva came back once as a child and was turned away at the door. The things she has been pocketing are hers. The whole visit was engineered: meet the brother, attach herself to him, come for the summer.

Three lessons. Be absolutely specific. There is no line saying "the war left scars". There is a house with a name, a province with a name, one particular front door and one particular drawer of cutlery. The broad fact is that Jewish property was stolen all across occupied Europe. One instance of it powers the entire novel, and that singleness is what hurts. Nobody admits anything, and that is the plot. Isabel is no villain and tells no lie. She simply never asked where a family's furniture came from. Everyday incuriosity keeps the wrong alive, with no conspiracy required, and it beats a cover-up for both truth and drama. The history stays invisible until it fills the room. Most of the novel reads as domestic claustrophobia and a slow-burn love story. All the way through, the past lives in objects, in tension and in small refusals, and nothing is spelled out. All four depth signals run inside a novel with no invented world at all. If you write contemporary or literary fiction and want to see how far one unacknowledged historical fact can carry a book, this is the current model.

Trust (2022): one life, four tellings, and who held the pen

Hernan Diaz's Trust (Riverhead Books, 2022) won the 2023 Pulitzer Prize for Fiction jointly with Barbara Kingsolver's Demon Copperhead, the first time the fiction prize had ever been shared. Four documents, one couple: a New York financier of the 1920s and his wife.

The four: a novel that puts the couple behind thin disguises; the financier's own unfinished autobiography; a memoir by the woman he hired as his ghostwriter, written decades later; and the wife's diary. Each contradicts the ones before it, and each has a plain motive. The novelist wanted a story. The financier wanted a legacy. The ghostwriter was paid. The diary was never meant to be read. Nobody tells the reader the truth. The reader has to weigh the sources.

Diaz does to one marriage what this page asks you to do to a world. The events are fixed. What varies is who recorded them, when, for whom, and what they stood to gain. Trouillot's four moments, written as a novel.

Two lessons, and neither needs an experimental novel. The later a document was written, the more it reflects its own moment. The ghostwriter's memoir says more about the decade it was written in than the decade it describes. The same is true of real sources, and it is a free characterization device. So give every invented document a date, then ask what its author's life looked like in that year. And the most trustworthy account is the one never meant for publication. Private records, a diary, an account book, an unsent letter, carry authority because they had no audience to manage. Put the version of the past you want the reader to trust in the least public document you have.

The Three-Body Problem (2006): real history driving a science fiction trilogy

Cixin Liu's The Three-Body Problem was serialized in Chinese in 2006. Tor Books published the English edition in 2014, translated by Ken Liu, and it won the 2015 Hugo Award for Best Novel, the first time an Asian writer had taken that award with a novel.

It opens in 1967, at a Cultural Revolution struggle session: a public meeting where one person is denounced and physically attacked. Ye Zhetai, a physicist at Tsinghua University, dies in front of a crowd, charged with teaching relativity, quantum mechanics and the big bang. His daughter Ye Wenjie is watching. Everything in the trilogy descends from that afternoon. Ye Wenjie is recruited into a secret military project, faces a choice about a signal from space, and the fate of humanity follows from what she decides.

The example is here for two reasons. The first is the consequence test at full strength, on real history. The struggle session is a documented national event. Liu had to build nothing except one person's reaction to it, which he then let spread. If you write contemporary or near-future fiction you can do the same with any real event your character stood beside: a plant closure, a flood, a bank failure, a referendum. The craft is in the spreading. The invention matters far less. The second reason is scale. One afternoon of violence in 1967 feeds three novels. There is no stronger case against inventing broad history, a dynasty, an age, an era, where a particular hour would do. Wide-scale history explains a world. A precise hour gives you an agent.

The Lions of Al-Rassan (1995): a quarter turn, and 350 years into twenty

If you are stuck between "invent everything" and "get the real history right", Guy Gavriel Kay's approach is the most useful third way there is, and he has explained it better than anyone. The Lions of Al-Rassan (1995) is set on a peninsula that is transparently eleventh-century Iberia. The fragmented south is modeled on Muslim Iberia. The north, bearing down on it, is modeled on Christian Iberia. A minority modeled on Iberia's Jews lives between them. The soldier Rodrigo Belmonte is modeled on El Cid. The names, faiths and geography are all invented and all recognizable.

In a 2013 interview Kay called the effect a "quarter turn to the fantastic", a phrase a reviewer coined and he liked enough to take over. The quarter turn buys freedom to reshape. Holly E. Ordway's essay on Kay's world-building, hosted at BrightWeavings, the fan site for his work, lists two particular changes. El Cid had daughters; Kay gave Belmonte sons, because the story needed them. And a reconquest that took about three and a half centuries is squeezed into roughly twenty years, so readers watch the whole arc land on people they care about.

The compression is the technique to take away, and it works in fully invented worlds too. Historical time is longer than novel time, and closing that gap is your job. Societies change on a generational timescale. Fiction has to land the change on an individual. Two solutions: compress, as Kay does, or pick the single decade in which the change visibly turned and set the book inside it. Asking a reader to feel three centuries does not work.

Compression has to be declared, at least to yourself. The invented peninsula is Kay's license. Because Al-Rassan is invented, the novel claims nothing about real eleventh-century Spain, and nothing in it can be checked against the record. Straight historical fiction has no such license, which is why its confession of compression goes in the author's note at the end.

The Treaty of Waitangi (1840): one agreement, two texts, and a tribunal still sitting

Here is the best single proof, from outside fiction, that a disputed document beats a broken one. The Treaty of Waitangi was signed at Waitangi, in the Bay of Islands, on February 6, 1840, by William Hobson for Britain and a number of Māori rangatira, or chiefs. It exists in two versions, English and te reo Māori, and the Māori version is not a translation of the English. The biggest divergence is in the first article. In the English text, Māori cede sovereignty to the Crown. In the Māori text the word is kāwanatanga, governorship, a narrower idea that carried no European notion of sovereignty for the people signing. In the second article the Māori taonga reaches far beyond the English list of property. The Waitangi Tribunal's website presents both versions, Sir Hugh Kawharu's translation of the Māori text, and footnotes on exactly these terms.

The Treaty of Waitangi Act 1975 created the Tribunal to rule on the Treaty's meaning and effect and to investigate claims under it. In 1985 its jurisdiction was pushed back to cover Crown acts and omissions since 1840. Claims are still being heard. Nearly two centuries on, one afternoon in 1840 still operates inside a country's law.

Three lessons. One agreement with two meanings is worth more than ten broken ones. A broken treaty gives you one wronged party and one guilty one. Two sincere readings give you two wronged parties, each with a case, each quoting the text, and no settlement that costs nobody. The "inherited claim" setup has no purer form. The act of translating is itself a historical event. A world with more than one language, and it ought to have more than one, keeps its key documents in more than one version, and somebody chose the words. Who translated, and which side they served, holds a plot. And effects persist for centuries through institutions. What carries 1840 into the present is an Act, a tribunal, a pile of findings and a settlement process. Memory comes a distant second. The machinery is dull, and the machinery is what keeps the past alive. A grievance that has to stay live in your world needs a machine of its own, a court, a register, an office, an annual payment, and never just sustained anger.

An everyday scenario: a timeline spanning nine centuries and a workshop with no permitted heir

Picture a writer called Vesna. Her timeline covers nine hundred years: four named ages, two dynasties each with its own succession list, one schism, three dated wars and one calendar reform. It took her eleven weeks and it is the finest thing in her folder. Her novel is about a heroine in her thirties who is blocked from inheriting her late mother's glass workshop.

Vesna has absorbed the know-more-than-you-show rule, so the timeline's absence from the draft does not bother her. After one bad month she sees the real problem. Nothing connects. The second dynasty could vanish and no scene would change. The novel and the timeline are two files about one place with no cross-talk between them.

So she starts again from the other end, and it takes one afternoon. The friction: no daughter may inherit the workshop. Why not: the glassmakers' guild transfers a license only to a member, and the heroine is not one, because her mother never enrolled her. Why a mother would skip enrolling her own daughter: enrollment requires a bond, and the bond was raised past what one workshop can afford. When the bond rose, and who raised it: the guild, after a fire seventy years ago, wanting money and wanting fewer members. Who survives: the fire is beyond living memory, the bond is still levied, and the guild lives off it.

Five questions, one afternoon, and she has a history that is the plot. She also has three bonuses she never asked for. The fire explains why one quarter of the city is newer than the rest, which is a piece of geography. The guild's income from the bond is a financial motive for keeping membership small, which gives her an antagonist with a motive instead of a temper. And a fire nobody remembers but everybody pays for hands a character a bitter line in chapter four at no cost in exposition. The whole yield is two entries, about ninety words of notes, with a date on each.

She keeps the nine centuries in a separate second file. Across the next two books she pulls four items from it, each one at a scene's demand. The calendar reform gets real use in book two, for a reason she could not have foreseen when she invented it. So overbuilding does have an honest defense, and Vesna's four items in two books are also an honest picture of how often it pays.

Try it yourself

Steps 1 to 4 take an afternoon. The rest take about an hour. You need a book in draft, or a world you have been building for longer than you have been writing.

  1. Run the consequence test on everything. Open your timeline. Beside each entry, fill in "Because of this, today, ______." with something a character could notice: a person, an object, a rule, a price, a refusal. Flag every entry whose blank stays empty. Delete nothing. Move the flagged entries to a file called unused and count what survives. Usually far fewer. The survivors get more of your attention, and they get better.
  2. Do one proper backward trace. Take the biggest unexplained obstacle in your plot. Ask what must have happened to produce exactly that. Date it. Name who is still alive who was there. List two side effects. Keep the whole thing under a hundred words. If you run over, you invented an era where you wanted an event.
  3. Fix year 0. Choose one event, a founding, a treaty, a disaster, the day the factory closed, and call it year 0. Number every entry in the timeline from it. Give every named character a birth number. Then check each character's age against every event they supposedly witnessed. Twenty minutes. At least one error will surface.
  4. Seed three scenes with depth signals. Take three scenes you have already drafted and add a single signal to each. Pick from the menu of four: two characters crossing one past event on different assumptions and neither pausing; a record that survives for one period and is missing on either side of it; a private landmark used as a date instead of a number; a thing named for someone nobody present can place. One line each. Reread. Watch the world age with no fact stated.
  5. Write one event as three documents. Take the most important past event and write it three times, as three surviving records of 60 words each: one contemporary and administrative; one written a generation later by an interested party; one private and never meant to be read. Rule on which one is true. Write the ruling where you can find it again.
  6. Build a machine for one grievance. Pick a historical wrong that is still live. Replace the "people are still angry" version with an institution that enforces it: a court, a register, a license, a levy, a yearly ceremony, an office with clerks in it. Decide who staffs it and how they feel about it.
  7. Run the plausibility pass. Test the timeline against the six giveaways: a span far beyond what the plot needs, stability all the way through with no collapses, even spacing, static technology, static language, and a dark age of convenient size. Apply the six repairs. Halving the span almost always helps.
  8. Find the near miss. Identify one moment with one particular person, one particular choice, and a plausible other outcome. Where everything reads as the tide, put in one blunder.
  9. Start the bible now. It holds the chronology, the zero year, the birth numbers, the in-world documents, and which document is true. How to lay it out, and what else goes in it, is the Continuity Bible's territory. In book one it costs nothing. By book six it is a week's work. Use this template for each entry:

Event: __________ (year, relative to zero: ____) Because of this, today: __________ Shape: grudge / wrong boundary / debt Who is still alive who was there: __________ Recorded by: __________ (when: ____, why: __________) What that record leaves out: __________ Contradicted by: __________. The true version is: __________ Institution that keeps it in force: __________ First reaches the reader in chapter __, as: __________

Common mistakes and misunderstandings

A timeline attached to nothing. The commonest failure, and the root of all the others. Four hundred years of dated entries, and none of them changes a character's options on page one. The fix is the consequence test, applied without mercy. The payoff is that the survivors get more attention and get better.

Delivering the past as a lecture because nobody present has a reason to ask. History, of all material, most often produces two natives narrating their own war at each other. Fix it with a genuine outsider who wants something, or two insiders who disagree about what the event meant. Exposition has the complete menu.

One file holding both the record and the belief. A pair of documents, a pair of elements. What happened goes here, and it stays consistent. What people claim happened goes to Lore, and it does not have to be consistent. Keep only one and you get one of two worlds: one where everyone is right, which is flat, or one where nothing can be known, which is exhausting.

Ten thousand changeless years. Typing deep time costs nothing. Making it credible costs plenty. A world sitting at one technological and political level since its founding needs either a hard cut to the span or the collapses, inventions and drift in language that such a span would produce.

Reaching for a dynasty when an afternoon would do. Writers grab for scale where a single incident would do the job. Pin down one hour in which one known person did one clear thing. More plot than a named age, and far easier to remember.

Confusing the world's history with a person's. A personal past is Backstory. World history explains the institutions, borders and rules around the character. The two interlock, since private wounds often have public causes, but they are separate documents, built separately.

A monochrome past. Either a golden age that was lost or a dark age that was escaped. Both are usable and both are usually lazy, because each one dictates the reader's view of an entire era. Real periods are mixtures, and mixture breeds disagreement: one person's nostalgia for an era that was hell for someone else in the same room.

A book that depends on its appendix. Back matter serves readers who are already committed. It never delivers the story. If chapter nineteen only makes sense with the back-matter chronology open, chapter nineteen is the thing to rewrite.

Believing realism needs none of this. Contemporary novels are full of history: a closure, a merger, a scandal, the year a street was rezoned, the silence between two relatives. The realist failure mode is weightlessness, a present with nothing behind it, rather than implausibility. The fix is the same.

Altering a checkable fact and not confessing. Historical Fiction readers forgive a great deal that is confessed and very little that is hidden. One undeclared change throws doubt on every other fact in the book. Move dates when the story demands it. Own up in the note at the end.

Infallible archives. Every document was produced by someone with a purpose, and most of them had a reason to slant it. Accurate archives mean a past without politics, which throws away the best tool on this page.

Related elements