Back to the table

115 Advanced Co Continuity Bible Series & Universe

Continuity Bible

The record that keeps a long series consistent.

What it is

A continuity bible is the file you open when you can't remember what you decided. That doesn't mean what you intended, what would be best, or what you'd choose if you were starting now. It means what you decided, and then printed, and so owe. The file exists because your memory of book one will fail before your readers' memory does, and it won't fail as a clean blank. You'll half-remember that the harbor was called something beginning with A, invent a likely replacement, and publish it. Nothing on the page will look wrong to you at any point along the way.

That trap explains why careful writers still get caught. Careful people don't forget facts outright. They reconstruct them, confidently and badly.

Everything else on this page hangs on one difference, and getting it wrong is the most common reason a writer's "series bible" stops being opened by book three. A continuity bible is a reference, not a plan. An outline is written beforehand, to decide. A beat sheet is written beforehand, to see the shape on one page. A series arc is a long-range intention about where the whole thing is going. All three look forward, all three are provisional, and you can overwrite any of them without penalty. If chapter nineteen of your outline turns out to be wrong, you delete it and nobody is harmed.

A continuity bible runs the other way. You write it during and after, and every entry in it is a liability. Think of it as a ledger of debts rather than a design. Each line says this is now true, I cannot quietly change it, and here is where it became true. You don't negotiate with it. You either obey it, or you make a deliberate, dated decision to break it and accept what that costs.

A few practical results follow straight away. You can't usefully write it in advance, because a continuity bible built before book one is a wishlist. It'll hold thirty facts, and four of them will survive contact with the draft. Worse, it trains you to distrust the file, and that's fatal. A reference you half-trust is worse than none, because you'll look things up in it and then go and check anyway. What you can do in advance is create the empty headings, which takes four minutes and is the right amount of preparation. Entries get added, never planned. You write a line at the moment "I have thought of something" turns into "I have committed to something". That happens when you've written the name on the page, given the distance in dialogue, or let a character say the fire was nine years ago. And the manuscript outranks the file, except when it doesn't. When the two disagree, something has gone wrong, and one of them has to change today. Usually the manuscript wins, because that's what readers will see. Now and then the file wins, because the manuscript drifted and you've caught it before publication, which is the best outcome you can hope for. What you can never do is leave them disagreeing and decide later. Later is how a wrong fact reaches print.

Writers who merge the plan and the record end up with one bloated document. It's too clogged with reference material to glance at and too speculative to trust, so they stop opening it. Keep two files for two jobs, facing two directions in time.

A character bible and a continuity bible are different documents, and the overlap is where writers lose an afternoon arguing with themselves. The test is one sentence: would this fact still be true if the entire cast were removed? If yes, it belongs here. That covers the distance from the harbor to the pass, the name of the newspaper, and what a loaf costs. It covers which year the treaty was signed, the rule about what magic can't do, and the fact that the cathedral has a west door and no south one. If no, because the fact belongs to a person, it goes in the character bible: eye color, date of birth, the five speech habits, the scar and where it came from, who is whose half-sister.

The boundary needs a little refining before you can use it. A few facts are shared, and they get stored once, here, with a pointer from the character side. A character's address is a character fact. How long it takes to get from that address to the city is a world fact. Her regiment is hers, and the regiment's history is the world's. Store each fact once, in whichever file owns it, and cross-reference. A fact kept in two places will sooner or later disagree with itself, and you won't notice which copy is out of date. Also, these are names for sections, and they don't have to be separate files. Most working novelists keep one document with a cast section and a world section, and that's completely fine. The distinction matters because it tells you what each section is for, and because it stops the world material swelling until you can't find the cast entries. If one file stays searchable, and one screen gets you to one subject, keep one file.

This page is only about the record, and it leaves some ground to other elements on purpose. Worldbuilding owns the act of invention and the argument about how much to build. Series owns the form itself: how long, what kind, how the books connect. Series Arc owns the overarching story.

One habit turns a pile of notes into a reference work, and almost nobody does it. Every entry carries a provenance line, the book and chapter where the fact became public. (Provenance just means where a thing came from.) That's where a reader can find it, which isn't always where you decided it.

Harbor at Thornlow: tidal, dries at low water. Bk 2 ch 4 (Dev's first interview).

That short tail does several jobs at once, and you'd pay later for each one if it weren't there. It tells you whether the fact is committed or private. A fact with no provenance line has only ever existed in your notes, so you're still free to change it without consequence. A fact with a provenance line is in print, and you're no longer free. No other distinction in the whole file is worth as much, and you can't rebuild it later from memory. It makes a contradiction findable. When book five disagrees with the file, you can go straight to the page and see which version was actually published, which takes ten seconds instead of an afternoon. It lets somebody else check the record, so a copyeditor, a co-writer or a continuity reader can test your file against the text instead of taking your word for it. And it quietly records when a reader was told, which turns out to matter a great deal for the knowledge problem below. Add a second short tail where it applies: superseded by (meaning replaced by) or corrected in. A line that reads Harbor dries at low water (bk 2 ch 4): contradicted bk 4 ch 11, bk 4 is wrong, fix at next revision is doing more work than any tidy entry could.

What follows is the list of categories that get contradicted in practice, with the reason each one earns its place. Don't treat it as a form to fill in. If a category never bites in your kind of book, leave it out.

Almost nobody keeps the secrets register, and leaving it out causes the most expensive errors. Every other entry in the file is a property. A name has a spelling, and a road has a length. Information works differently. A fact doesn't just sit in your story. It has a distribution, meaning it's spread across certain people, and that spread changes on particular pages. One character has known since before book one. Another works it out at the midpoint of book three. A third is told off the page, between chapters, by someone else. Two never find out, and one believes a version that isn't quite true. Then there's the column people forget: the reader knows or doesn't know, on a schedule of its own.

So the register is organized by fact rather than by person, and it has six columns:

Add a standing row underneath for the reverse: who holds a false version, and which version. A great deal of subtext and most plot twists live in that row.

The register takes about twenty minutes, and it repays them in two ways. A knowledge error does structural damage, far beyond a blemish. A wrong eye color is embarrassing. When a character mentions something she hasn't been told, the revelation you were saving was never a revelation. Worse, one of your scenes now has no reason to exist. You can't repair that with a word. It takes a scene, or three. And this kind of error slips past every check you have. Your spellchecker won't blink. Your copyeditor, working on one manuscript, has no way to know what was established in book two. Your beta readers will often absorb it and report only a vague sense that the ending came too easily. The people who reliably catch it are the readers who have just finished all six books in two weeks, and they catch it in public.

Be clear about what the register is not. It doesn't decide when the reader should find out. Those decisions belong to foreshadowing, exposition and the way you stage reveals. The register is the machinery underneath that makes them safe to carry out. The cross-reference works in the other direction too. The character bible keeps the per-person view: what one named person knows at any point, which is what you need while writing her chapters. The register is the fact-first view across the whole series, which you need when you're deciding whether a reveal is still open to you in book five. Keep the register as the authority, and let the character entries point to it. In a multi-POV series, told through several characters' eyes, the problem multiplies, because every head carries its own version of the facts. In close third, where the narration stays tight inside one character's head, it becomes a limit on every line. The narration can't know what the viewpoint character doesn't.

A series piles up two kinds of liability: facts, which you must not contradict, and promises, which you must pay sooner or later. A promise is anything you've put in front of a reader that hints at a future. It might be an object described twice, a prophecy, a name dropped and not explained, or a letter that hasn't been opened. It might be a character who leaves saying she'll be back, a question a detective asks that nobody answers, or a scar whose origin you've held back. Readers keep their own version of this list in their heads whether you do or not, and in a series they keep it for years.

So the bible gets an owed list, a page with four columns: the promise, where it was made, what it implies, and whether it's been paid. Its uses run from the obvious to the most valuable. It stops you dropping a thread. It stops you paying the same debt twice, which is a real and oddly common failure in long series: two different books explaining the same mystery slightly differently. The best use is the one that justifies the whole file. Read the owed list before you plot the next book, and about a third of your plotting is already done. An unpaid promise from book two is free material with a built-in audience, and it beats a fresh invention because readers have been waiting for it. Checking promises against payoffs across the whole series belongs here, because an unfired gun in book one becomes an expectation across five. Foreshadowing owns the craft of planting and firing, and makes the case for checking plants against payoffs within one book. A fourth column is optional and useful: deliberately unpaid. Some promises are atmosphere and should stay open, so write down which ones. Then in three years you can tell restraint from forgetting.

Arithmetic is where series break in public, because it's the one thing a reader can check without rereading, and a certain kind of reader enjoys checking it. The failure is always some version of the same thing. Say a character is fifty-five in book two, which works. Then in book four you need him at a particular siege, and you write him as twelve at the time, because you were thinking about the siege and not his birthday. Someone does the subtraction and posts it.

The fix is a date spine: one chronological list, kept in one place and treated as the authority, covering the whole series rather than one book. Here's how to build it:

That last point needs a word, because writers back into it by accident. Your characters can age in real time, on a slowed clock, or not at all. All that belongs here is the rule that the decision must exist as a dated line in the file, and never as an assumption. Every other number on the spine depends on it. An unwritten policy produces the classic failure: a lead who is forty-three in three books in a row and then suddenly has a twenty-year-old daughter. Series owns the choice itself and what each option costs the format. Notice how the work splits with Ticking Clock. That page owns the within-book arithmetic of a deadline: hours gone by, who checked the time, whether a journey fits in the time given. That arithmetic is a revision tool for one manuscript. The date spine is the tool that works across books, measured in years, and the two should agree with each other.

The form matters less than writers hope, and they usually choose it for the wrong reason. Described as kinds of tool rather than as products, there are five, and each one does something better than the rest.

The question to ask is which one you'll actually open with a sentence half-written. You use a continuity bible in the two seconds between making a decision and forgetting it. If the container takes longer than that to open, the file will have gaps, and a record with gaps that you trust is more dangerous than none.

That brings up the trap. Migration, moving a mature file from one kind of container to another, is where bibles die, and it always takes longer than you expect. The data usually survives. What gets lost is the provenance: the chapter references, the superseded notes, the corrections log, the little "check this" in the margin. Those are the parts that made it a reference and not just a list. If you have to migrate, move one section, use it for a month, and only then move the rest. Whatever you choose, keep it in a format you can export as plain text. One day you'll need to hand a slice of it to an editor, a narrator or a translator who uses something else entirely.

Nobody keeps a perfect continuity bible, not one writer, and the professionals who come closest pay somebody to do it full time. Any advice that assumes otherwise will be dropped by week three, so here's the version that survives.

The habit that works is capturing a fact the moment you invent it, instead of sweeping for facts afterward. Virtue has nothing to do with it. In a later sweep you'd have to recognize which facts are new, and you can't, because by then they all feel equally familiar. At the moment of invention, the fact stands out, and the entry takes four seconds. The practical trick that makes this stick is to write the note in the manuscript itself, instead of in the bible. Invent a marker you'll never type by accident, and put it right in the line:

She gave the address as @@ 14 Caradoc Street, top floor @@ and waited.

Then, once a chapter, search for the marker, move the facts into the file with their provenance lines, and delete the markers. This works where a separate window doesn't, because you don't have to switch tasks or make a decision. The marker sits in the draft reproaching you, which a closed file never does. If even that won't stick, use a fixed slot instead: the last five minutes of each session, or a pass at the end of each chapter. Either way, the bible trails the draft, because it's a log, and logs are written afterward.

To keep the file small enough to survive, follow two rules. The simplest version that works beats the complete version you abandon. That's more than a consolation, because other fields where records prevent errors have found the same thing. Atul Gawande's The Checklist Manifesto (2009) spends a chapter on why most checklists fail, and his answers are length, vagueness, and no fixed moment of use. A short record checked at a reliable pause point beats a thorough one nobody opens, every time. Your pause points are the end of a chapter and the start of a new book. And record only what can be contradicted. If a fact can't be gotten wrong, it isn't a continuity entry, however much you enjoyed deciding it. Atmosphere, intentions, interpretations and lovely unused ideas belong somewhere else. A bible that holds everything is a bible nobody consults, and a document nobody consults does nothing at all.

One thing to avoid is leaving the whole job until the first draft is finished. Building the file afterward means rereading the manuscript, and that's the very cost the file exists to save you.

Why it matters

Memory is lopsided, and the gap gets worse with every book. Writing down that the harbor dries at low water takes four seconds. Now suppose you discover, in your third round of proofs, that you've described it two ways across two published books. That costs a corrected file, a re-upload, an uneasy week wondering how many copies went out wrong, and a review that mentions it. You wrote book one years ago, once, slowly, and you've written three hundred thousand words since. More and more, your readers read the whole series inside two weeks, often back to back, often while posting about it. They hold it as one continuous experience. You hold it as a filing system, so they'll remember the name of the pathologist from book two, and you won't.

It makes a backlist safe to extend. Most preparation gets used up by the book it was made for. An outline is spent, and so is a beat sheet. The continuity bible is the one document in your process that gains value instead. It's worth more at the end of book one than at the start, more in book four, and more again the moment anybody else needs to work on your material. For a writer publishing several books a year, that growing value is a plain fact of production. A world you can re-enter without a week of rereading is a world you can write a fourth series in. In practice, a Storycosm needs that above all: a record good enough to make invention safe, far more than it needs more invention. A crossover is simply the moment two records must agree.

Sooner or later you'll find a contradiction that's already in print. Writers panic about this part, so here it is plainly. You have four options, one of them is usually right, and readers are more forgiving than you expect.

Silence means changing nothing, saying nothing, and keeping writing. Silence is usually the right answer, and the hardest to accept. Most readers never notice, and of those who do, most enjoy noticing. When someone complains, the complaint costs you less than a clumsy repair would. Arthur Conan Doyle moved Dr. Watson's war wound from shoulder to leg between two novels, and the Holmes stories still became the most durable popular fiction in English. Readers made a hundred-year parlor game out of explaining away slips like it. That generosity is real, and you can't build a strategy on it, so don't plan on it. Once the error is already published, though, you can rely on it. The character bible sets out that case in full.

Quiet correction means fixing the file, fixing any text still to come, and fixing the published book too if the format allows it. Ebooks and print-on-demand make this cheaper than it's ever been, and uploading a corrected file is perfectly legitimate. One caution: readers who already own the old version now have a third variant to argue about. Correct real errors, and leave trivia alone.

An in-world explanation is the most elegant option when it works, and the most embarrassing when it doesn't. It works when your world already holds unreliable sources, such as a biased chronicle, a character who lies or a legend nobody has checked. Then the disagreement reads as texture rather than as a repair. It fails when the contradicting voice was the narrator, because a narrator who turns out to be wrong about a measurable fact costs you more trust than the error did. The useful lesson is that building some unreliability in early is a kind of insurance. Lore makes the broader case.

Acknowledging it is rare, and surprisingly charming when the author has the standing to pull it off. You write the note, in a preface or an afterword, and let the reader in on the problem. Patrick O'Brian did this when his series outgrew the war it was set in, and the case is worked through below. You can use this option too. It has to be a decision rather than a shrug, and it only works when what you're owning up to is a constraint rather than carelessness.

The one thing you can't do is the retcon by stealth. (A retcon, short for retroactive continuity, rewrites what a story had already established.) Doing it by stealth means quietly writing the new version as though the old one never existed, without correcting anything. It's the reflex, and it's what the word retcon usually means in practice. It's also the one move that turns an error into a credibility problem, because the readers who notice are the very ones who've bought everything. There's also a structural answer, and it scales where the others don't. When two published books disagree and neither can be withdrawn, you can rank which one has authority instead of resolving the contradiction. Lucasfilm's continuity database did this for years. It sorted Star Wars material into tiers, so that higher-tier and newer work overrode lower-tier work wherever they clashed, and a named person kept the database. That answer was built for a shared universe, and the idea shrinks down well for one novelist. Write at the top of your file which text wins when two texts disagree, and you've turned a recurring argument into a one-line rule.

The moment anyone else touches the material, the bible becomes the interface. It stops being housekeeping and becomes the place where your work and theirs meet. Here are four situations, in rising order of how formal the document has to be, and then adaptation, which follows different rules.

Four kinds of professional will want slices of it, and each wants a different slice. Most writers picture an audience of one for this document, which is why most versions of it are no use to anybody else.

All four are different views of one file, so keep one master record and pull slices out of it. Don't keep a separate narrator sheet, translator sheet and editor sheet. All three will drift out of step with each other and with the books. There's a fifth reader too, and thinking of this one makes people write better entries: the stranger. Write as though a new co-writer will use the file, and you'll state the fact instead of gesturing at a memory. Most bad entries are notes to self that only work while you still remember what they refer to.

Real-world examples

Patrick O'Brian's Aubrey-Maturin novels (1969 to 1999): the series that ran out of calendar

The clearest published example of what a date spine does is a twenty-volume literary sea story. It also shows what happens when the spine is tied to real history and the series outgrows it.

O'Brian began with Master and Commander in 1969 and finished the twentieth novel, Blue at the Mizzen, in 1999, and he left a twenty-first unfinished at his death. The books are anchored to real naval history and real dates. The sequence runs from 1800 to late 1815, and the early volumes move briskly, covering about twelve years in the first six.

Then the arithmetic turned on him. He'd tied his series to the Napoleonic Wars, and the wars are a fixed quantity. By the tenth novel there was more story left than there was history to put it in.

What he did about it is the part to learn from, and it came in two halves. First, he told the reader. The Far Side of the World (1984) opens with a note about the problem. It says that if he'd known how many books would follow the first, he'd have started the sequence much earlier in real time. Then it asks readers to accept volumes placed in hypothetical years, "an 1812a as it were or even an 1812b", like the hypothetical moons used to calculate Easter. The note is a small masterpiece of candor from an author, and nobody minded, because it was a decision and not a shrug. It's option four in the list above, done well.

Second, the spine then had to carry a huge load. The period from June to December 1813 is stretched to hold events that would properly fill five or six years. Read that as a continuity fact rather than a curiosity. For the rest of a very long series, every seasonal detail, every letter, every mention of how the war was going and every character's age had to fit a calendar that had been stopped on purpose. Only a written record makes that possible. You can't hold it in your head.

A writer with three books out can take three lessons from this. Anchoring to real history gives you free realism and a hard ceiling. If you set a series in World War II, or against a real election cycle, or after a real disaster, the research pays for itself, and the calendar won't bend. Count the years you have before you commit to the series length. It's the same trade as anchoring to a real institution. Series covers how that choice affects how long a series can run, including Ian Rankin's Rebus novels, where a real police force's retirement age removed the lead on a date the author couldn't choose. A suspended calendar has to be written down, or it won't hold. Once the time inside your story stops matching any outside clock, your file is the only authority. That makes O'Brian the strongest possible argument for the date spine, since his problem is the extreme version of every series writer's problem. And say it, or don't do it at all. The device worked because O'Brian declared it. Done silently, the same move would have been the retcon by stealth described above, and readers would have caught it within a volume.

Setting uses these novels for a different argument altogether: how a confined space such as a ship can outperform a large one. The two pages make different points about O'Brian's work.

The Baby-Sitters Club (1986 to 2000): the bible built with a highlighter, then published

The Baby-Sitters Club is the clearest documented case of a continuity bible retrofitted, or built after the fact, out of books already published. It also shows why that job exists at all.

Scholastic published the series from 1986 to 2000, and it reached 213 novels and more than 190 million copies. Ann M. Martin wrote the early books herself. Then, as the schedule became impossible, ghostwriters such as Peter Lerangis, Ellen Miles and Nola Thacker wrote from outlines developed by Martin and her editor, and Martin read the final drafts. She has said she had almost no choice in the matter, and that she outlines everything.

That setup creates a continuity problem with a particular shape. You have several writers, none of whom have read all the books, turning out titles at about two a month, all set in one small town with a large recurring cast. The answer was a reference document. The person given the job was David Levithan, who arrived at Scholastic as a nineteen-year-old intern in the early 1990s and went on to edit the franchise. As quoted in Mental Floss, he has summed up the work in one line: he was the man on the subway reading the books "with a highlighter to keep track of who spoke French, who had green eyes, and so on."

Look closely at that sentence, because it teaches two lessons. The record was built by reading the text, not by remembering it. Once the facts are in print, no other method works, and it's the one you'll use if you come to this page three books in. Nobody rebuilt the Stoneybrook facts from the outlines or from Martin's memory. Somebody read the novels with a marker in hand. And the job was to track, never to interpret. Who spoke French, and who had green eyes: those are just the kind of facts that can be contradicted. Keeping the job narrow on purpose is what made the document possible to finish.

The ending makes series writers smile, because the bible grew until it could be published in its own right. The Complete Guide to the Baby-Sitters Club (Scholastic, 1996) is a 347-page reference credited to Martin. Its eight sections cover the club, its members, their families, their clients, other people, places and things, plus a map of Stoneybrook. Together they hold more or less every person, place and thing mentioned in the series up to that point. A continuity bible is one of the few production documents readers will buy.

Sarah J. Maas's House of Sky and Breath (2022): the moment two series became one record

This example hinges on the ending of a 2022 novel, so if you're partway through the series, skip to the next heading.

Maas was running three large fantasy-romance series at once: Throne of Glass, A Court of Thorns and Roses, and Crescent City. At the end of House of Sky and Breath, the second Crescent City book, her protagonist Bryce Quinlan opens a portal meant for one destination and arrives somewhere else entirely: Prythian, the world of the A Court of Thorns and Roses books. In House of Flame and Shadow (2024) she spends a large part of the book there, among that series' established cast, and the book lays out how the worlds connect. Maas has said publicly that she'd been planting hints of a shared universe across all her books.

As craft, the move is a masterstroke of reader promise. It's also a continuity event of the first order, which is the part that concerns this page.

Before that scene, Maas had three records that only had to agree with themselves. After it, she has one record. Every future Crescent City book is now bound by what's true in Prythian. Every future A Court of Thorns and Roses book is bound by what happened to Bryce, by what her characters now know, and by the fact that travel between worlds is possible and some of them have seen it. The secrets register doubles in size, because the question "who knows the worlds are connected?" now has an answer for each character in each series. The owed list doubles too.

Plenty of self-publishing writers have two series in one world, and the practical lesson for them is this. A crossover is a decision about your filing system as much as about your plot. It can be a very good decision, and the moment to merge your records is before you write the scene, not after readers have started asking. Crossover owns the craft of it, and keeping the bible straight is what it costs you. (The same author's A Court of Mist and Fury, 2016, appears on wound for an unrelated reason: how trauma is written as symptom rather than explained. The two pages make different points about her work.)

Lee Child and Andrew Child (2020 onward): the handover that makes the record the product

In January 2020 Lee Child announced that he would stop writing the Jack Reacher novels and hand the series to his younger brother, the thriller novelist Andrew Grant, who publishes as Andrew Child. The transfer didn't happen all at once. Twenty-four Reacher novels were Lee Child's alone. From The Sentinel (2020) the books carry both names, and they've kept appearing under the joint byline through 2026.

The handover is the most visible current example of a question every series writer faces in the end, even writers with no plan to retire: could anybody else write this?

There's nothing sentimental about the question. It's practical, and it comes up in three ways. If you work with a collaborator, now or later, the record is what you share. If you're ill, or just unavailable for six months, the record is what lets a book get finished. If your estate one day licenses continuations, the record is the asset, because the manuscripts alone leave every unstated decision to guesswork.

Here's a blunt test you can borrow, and it takes an afternoon. Hand your bible and the outline of one unwritten chapter to a writer you trust, and see what they get wrong. Whatever they get wrong is what's missing from the file, and you'll find it's rarely the big things. More often it's the register a character speaks in (how formal or casual they sound), the fact that nobody in your town locks a back door, the pronunciation of a surname, and which of two half-stated rules is the one that counts.

The Reacher handover also says something about scale. A series doesn't have to be a shared universe for this to matter. Reacher is one man in a mostly ordinary world, and the continuity load is still real: a stated birth date, a military career with dates in it, and a backlist of events that happened in named places.

An everyday scenario: three books into the Pellow & Sawle mysteries

Here's the ordinary version, because almost nobody comes to this page before book one.

You write cozy mysteries, and your detective is Tansy Pellow, a valuer at a small auction house, Pellow & Sawle, in the market town of Thornlow. Three books are out, the fourth is half-drafted, and two things have happened in the same week.

A reader has emailed, politely, to say that the sale where the fake Dutch landscape was sold is dated 2014 in book one and 2016 in book three. And your copyeditor has queried the spelling of a dealer's name, Gethin Mawr, because book two spells it Mawe twice.

Neither is a disaster. Both are symptoms of the underlying problem. You have no record, only a memory and a half-finished notes file full of character questionnaires from 2023.

So you spend a day, not a month. You open a spreadsheet with eight tabs and work backward from the published text, because the published text is the only authority now.

You start with proper nouns, because you can search for them: every place, firm, street, pub and person, with the spelling as printed and the book and chapter beside it. The Mawr/Mawe split turns up at once, and so do two others you didn't know about. You decide Mawr is correct, record Mawe as the rejected variant, and make a note to fix book two's file.

Then comes the date spine: one column of years, 2011 onward. You put the three books' spans on it, then every dated event anyone has mentioned out loud, with chapter references. The 2014/2016 problem sorts itself out in ten minutes. Book one says 2014 in narration, and book three says 2016 in dialogue spoken by a character who wasn't there. You decide 2014 is canon, the official version, and note that the book three line is wrong. While you're in there, you find something more useful. You've said Tansy's apprenticeship at the firm lasted four years and ended before the fake landscape sale. If she was nineteen when it started and is thirty-four now, in 2027, it ran from 2012 to 2016, so it can't have ended before the 2014 sale. There's a two-year hole in your protagonist's biography that nobody has found yet.

Next comes the secrets register, and here the day pays for itself. You list the four facts the series turns on, and for each one, you note every character, when they learned it, and how. Halfway down the third row you stop, because Lindiwe Odiah, the restorer, has known since book two that the landscape was overpainted. Yet in your half-drafted chapter nine, Tansy tells her for the first time, in a scene you were rather proud of. The scene can't stand. That's worse news and better news at once, because the scene the book needs instead is the one where Lindiwe has to decide whether to admit she's known for two years. It's a far stronger chapter than the one you had.

Then you do the owed list, which takes twenty minutes and hands you most of book five. Three promises are outstanding: the letter in the desk that nobody has opened, the reason Dev Sekhon left the Metropolitan Police, and Hubert Quennell's missing inventory. You'd forgotten two of the three. Readers hadn't.

By the end of the day you have an eight-tab file, three real errors found (one of them structural), and a chapter nine that's better than the one you lost. The whole thing exists because two people noticed something, and that's almost always how this starts.

Try it yourself

There are three procedures here. Steps 1 to 12 start a bible on a series you're writing now. Steps 13 to 19 retrofit one from books already published. Steps 20 and 21 are the pass you run before each manuscript leaves the house. The first two are afternoon jobs, not projects.

  1. Make the empty file before you make any entries. Eight headings on one screen, using the template below. Resist every urge to fill them in from imagination. The headings are the preparation, and the contents are the record.
  2. Choose the container by one criterion: can you open it in two seconds with a sentence half-written? If the answer is no, choose something else, however good it looks. Prefer something you can export as plain text.
  3. Adopt an inline marker. Pick a string of characters you'd never type by accident (@@ works), and wrap every invented fact in it as you draft. You aren't interrupting the sentence, just tagging it.
  4. Harvest once per chapter. Search for the marker, move each fact into the file with its provenance line (book and chapter), and delete the marker. It takes ten minutes. Do it at a fixed point, so it becomes automatic instead of a test of virtue.
  5. Write the provenance line every single time. First the fact, then where a reader can find it. An entry without one is a rumor, and in two years you won't be able to tell whether you're free to change it.
  6. Build the date spine on the day you first mention a year. Anchor it, give each book a span, put every birth year on it, and record stated intervals word for word. For invented worlds, use the fixed zero-year method and keep the subtraction accurate. The technique belongs to History, which explains it properly.
  7. Record your aging policy in writing, before book three. Your characters can age in real time, on a slowed clock, or not at all, and Series sets out the trade-offs. Write one dated line in the file, so that every age on the spine has something to be consistent with.
  8. Start the secrets register at the first secret, not at the first contradiction. Use four columns at least, and six if you can manage it. If your plot turns on four facts, the table has four rows, and it'll still catch something.
  9. Start the owed list at the first unexplained thing. Every object mentioned twice, every unanswered question, every departure with a promise in it.
  10. Keep the world and the cast apart. Run the test (would it still be true with the cast removed?) and send the facts about people to your character bible. Store each fact once.
  11. Review before each new book, not during. Read the bible cover to cover before you plot the next one. At that pause point the file earns its keep, and the owed list writes a third of your plot for you.
  12. Date every change. The corrections log is the page that settles arguments.
  13. Retrofitting: accept the order of authority first. The published text wins over your memory, and the earlier book wins over the later one unless you decide otherwise. Write that rule at the top of the file before you start, or you'll spend the day arguing each fact all over again on its merits.
  14. Sweep one: proper nouns. Search each manuscript file for the capitalized words you care about, or read the whole series with a marker if the files can't be searched. Give each name one row: the spelling as printed, every variant you found, and the book and chapter for each. This sweep alone usually finds two or three real errors in a three-book series. It's also the sweep most likely to find the error a copyeditor would otherwise find for you.
  15. Sweep two: numbers and time words. Search for every digit, and for years, months, weeks, days, ago, since, when I was, the year of, last summer, before the war. Each hit goes on the date spine with its chapter. Afterward, read the spine as a single column, and the breaks will be obvious.
  16. Sweep three: stated rules and capabilities. Find every sentence where anybody says what is or isn't possible, whether about the magic, the technology, the law, the institution or the geography. Record what was said, word for word, and not what you meant.
  17. Sweep four: the secrets register, built from the reveals. List the facts the series turns on, then go through each book marking the page where each character learns each one. Expect this sweep to find a real problem. That's what it's for.
  18. Where two books already contradict each other, do this. Write both versions down with their provenance lines. Then ask these questions, in order. Which one has more text built on it? Which one does the plot need? Which one did more readers see? (A version in a first chapter or a sample was seen by everybody, and one in chapter twenty-two wasn't.) Then choose a winner, write the decision and the date in the corrections log, and mark the loser superseded. Decide separately whether to touch the published file. Don't leave it undecided, and don't fix it twice. The most common mistake here is picking the version you like better instead of the version the series is standing on.
  19. Then stop. You're building a reference, not an index of your own work. Once the file answers the questions you actually ask, it's finished enough. If you never consult a section in a year of writing, delete it without ceremony.
  20. Before every manuscript goes out, run the continuity pass. You don't need the whole bible, just four checks, in the order readers tend to catch things. Check names and spellings. Check the date spine against this book's events. Check the secrets register against anything anybody knows in this book. Check stated rules against anything your climax relies on. Then make one last check that's easy to forget: every fact this book has newly committed to is now in the file, with its provenance line. Skip that last step, and your bible ends up accurate up to book three and silent about book four, which is the state most abandoned bibles are in.
  21. Send it with the manuscript. Send it in full to the copyeditor, and to a continuity reader with the one-paragraph brief. Pull the narrow versions out of the one file, such as pronunciations for a narrator and name origins for a translator, instead of keeping three separate files.

The eight headings, to paste into an empty file:

Spellings: word (rejected variant) · where it first appeared Calendar: anchor year, then one line per book with its span Ages: name · birth year Distances and times: from · to · how long · by what means · seasonal closures Rules: stated (with chapter) / known but unstated / deliberately undecided Secrets: fact · who knows · when learned · how · told whom · reader told? Owed: promise · made in · what it implies · paid? Corrections: date · was · is now · why

Common mistakes and misunderstandings

The bible that was finished before book one started. A reference work compiled from imagination is a wishlist with headings. Most of it will be wrong within forty thousand words, and you can't see which parts, so you'll go on consulting a file that no longer describes your book. The cure is to build the empty structure in four minutes, and let the contents arrive from the draft.

Treating it as a plan. Decisions about what will happen get written into the same document as the record of what already has. The file gets too speculative to trust and too crowded to search, and within two books it's dead. The plan is an outline and the bible is a ledger, and they point in opposite directions in time.

Entries with no provenance. Without the chapter reference, you can't tell a committed fact from an idea, so you can't tell what you're free to change. It's the most common structural flaw in real writers' bibles, and the easiest to fix.

Two live copies. One on the laptop, one in the cloud, one in a note app, one pasted into an email to your editor. The day they disagree, every lookup turns into a research task. Keep only one copy live, and mark everything else clearly as a backup.

The file isn't updated after revision. Revision is where facts change most and where the bible is most often forgotten, because you're deep in the prose and the file feels like admin. A bible that describes the draft instead of the published book is actively dangerous. You'll trust it at two in the morning, and so will your copyeditor and whoever writes book seven.

Confusing thoroughness with usefulness. Forty pages, color-coded, opened twice. The test isn't "is this complete?" The test is "when did I last look something up in it?" If the true answer is never, shrink it until it's the thing you reach for.

Migration swallows a weekend and quietly loses the provenance. Moving the file to a better system feels like progress, and it's also the surest way to destroy a mature bible. The parts that don't transfer cleanly are the reference tails, the superseded notes and the queries in the margins. Move one section, live with it for a month, then decide.

The arithmetic is done once and never again. A date spine built for book two and not touched since is worse than none, because you'll trust it. Every book adds dated statements, and each one limits all the others.

Interpretation is recorded instead of fact. "The town resents the harbor trust" is an observation about your own book. It can't be contradicted, it'll keep shifting as you write, and having it there teaches you that the file is a place for musing. Keep this document boring, and put the thinking where the thinking belongs.

Readers are assumed not to notice. They notice, they keep lists, and in a long series they compare notes with each other. The sensible response is a file. Fear won't help. The error that really costs you is the structural one that undermines a reveal, far more than the small one a reader spots, and that's the kind a record prevents.

Silence is mistaken for cowardice. Once they've found a contradiction in print, writers reach for a repair: a retcon, an awkward line of explanation in the next book, a note to readers. Most of the time the right move is to fix the file, write it correctly from now on, and say nothing. Readers forgive far more than writers expect, and they forgive an uncorrected slip more readily than a visible patch.

Nobody owns it. On a co-written series, or one with an assistant, a record that everybody maintains is a record nobody maintains. You need one named owner, and a rule about who may add a line.

Expecting it to fix a thin world. A perfect record of a dull place is a perfect record of a dull place. The bible is a tool for consistency, and it can't invent anything for you. If the note from your readers is that the setting felt generic, that work happens in worldbuilding, setting and culture, not in a spreadsheet.

Related elements