Back to the table

40 Intermediate Cb Character Bible Character

Character Bible

One reference document holding everything about a character.

What it is

A character bible is where you store your decisions about people, so you can't quietly change them by accident. The name comes from television. A series bible is the document handed to a new writer so they can write an episode without having watched every one. Hold on to that origin, because it tells you what the document is for. A series bible isn't a deep reflection on who these people are, just a reference work, made for looking things up by somebody who's in a hurry and doesn't share your memory.

The novelist's version is the same idea, at a size one person can keep up. One file, or one folder, or one set of linked notes, with an entry for each character. Each entry holds the facts a later chapter could contradict, and nothing else. That last part is the whole discipline. A character bible is defined by what it leaves out.

Before a fact goes in, ask one question: could I contradict this later without noticing? If the answer is yes, it goes in the bible. If it's no, it stays out, however interesting it is and however much you enjoyed deciding it.

Try the test on a few examples. Eye color: yes, because it comes up again and again, and it's the most-reported continuity slip in fiction. Favorite breakfast: only if somebody eats breakfast twice in the book. The spelling of a surname with a silent letter in it: definitely yes. Her deepest fear: no. You can't contradict a fear, because it's an interpretation you'll keep refining. It belongs in your work on flaw, wound and want vs need, not in a reference table.

The test sounds narrow, and it's meant to. A character bible that holds everything is one nobody looks at, and a document nobody looks at does nothing at all.

The fields below are the ones that get contradicted in practice. Don't treat this list as a template to fill in.

Every other field on that list is a fixed fact about the character. The last one is a timeline, which is why it gets left out of the tidy table and then goes wrong.

Take one piece of information, say that a certain death wasn't an accident. In a well-run manuscript, that fact has a history for every person in the cast. One character has known since before page one. Another learns it at the midpoint. One is told off the page by a third party, between two chapters. Two never find out. One is told a version of it that isn't quite true.

Now picture writing chapter twenty-six out of order, eleven weeks after you wrote the midpoint. You let a character mention something she hasn't been told yet, and the line reads perfectly and sounds just like her. Your spellchecker won't blink. Your beta readers will probably take it in without noticing. A reader on a forum will find it within a week of publication, because readers track who knew what better than anything else.

This kind of continuity error does the most damage because no automatic check can see it, and most human ones miss it too. A wrong eye color is a blemish. A character who knows something she hasn't learned breaks the chain of cause and effect the book runs on. The reveal you were saving was never a reveal, and usually one of your scenes now has no reason to exist. It's also the hardest error to fix, because the repair is a whole scene, not a word.

So the bible needs a knowledge record, with five columns instead of one:

Add a sixth column for the useful flip side: who believes a false version, and which version. That column is where a lot of subtext and most plot twists actually live.

This record doesn't decide when the reader should find out. The knowledge record is the nuts and bolts that make those decisions safe to carry out. Backstory decides the reveal, and the bible makes sure nobody in the cast jumps the gun. The whole question of what to reveal and when belongs to backstory.

People mix up four documents, and keeping them apart saves a lot of time.

The most common way a character bible fails is by being finished before the book starts. A document you fill in before you start has two problems. The document is mostly wrong, because the character who comes out of forty thousand words of drafting isn't the one you designed, yet the file still describes that one. And filling it in feels just like progress. The fields get answers and the answers pile up, and at no point are you writing prose.

The other way is dull, and it works. Start with almost nothing: a name, a spelling, an age, a job, three speech rules. Then keep the file open beside the draft, and add a line whenever you invent a fact, not whenever you plan one. You just gave her a sister, so write the sister down. You just decided the scar came from a dog, so write the dog down. It takes seconds, it happens while the decision is fresh, and you end up with a document that describes the book you're actually writing.

If that habit won't stick, fall back on a fixed slot. Take fifteen minutes at the end of each writing session, or do a pass at the end of each chapter, and pull the new facts out of what you just wrote. Either way, the bible follows the draft instead of leading it. The bible is a log, and logs get written afterward.

The one thing you shouldn't do is leave it until the end of the first draft. Building the whole document after the fact means rereading the whole manuscript, which is the very cost the bible is there to save you. You'll be piecing decisions back together instead of recording them.

Why it matters

Prevention takes minutes. Correction takes days. Writing down that her mother's name is spelled with a ph takes four seconds. Now say you find out after release that it's spelled two ways in your published novel. That costs you a corrected file, a re-upload, a nervous week wondering how many copies went out wrong, and a review that mentions it. The gap between those two costs is huge, and it always runs the same way.

It's the only preparation that keeps paying. The draft uses up most of your prep. An outline gets spent once. A character arc plan gets soaked up by the chapters that carry it out. The bible is different, worth more at the end of the book than at the start, more in the second book than the first, and more again when somebody else has to work on your material. It's the only document in your process that gains value over time.

The series problem is the case that wins over skeptics. By book six you've written several hundred thousand words about these people, and you wrote book one years ago. You can't remember whether the brother was older, or whether the bar was called the Anchor or the Anvil, or whether she ever actually said out loud that she'd been married. Meanwhile, readers have often read all six books within two weeks, sometimes back to back, and they remember better than you do. That gap is built into how series work, and it gets worse with every book. The way it goes wrong is specific. You don't forget something entirely. You half-remember it and make up a believable replacement. Believable replacements are the ones that reach print, because nothing about them looks wrong on the page. Everything a series needs from this document gets more valuable the longer the series runs: the cast index, the ages set against a fixed calendar, the knowledge record, and the list of things a returning character isn't supposed to know yet. If you also write connected standalones, or share a world across pen names, then the series arc and the crossover start depending on the bible instead of your memory. That's why the professionals who do this at scale hire someone to keep the file.

Once your book stops being yours alone, the bible is how everyone else plugs into it. Not everyone needs the same part of it. A developmental editor doesn't need it. A copyeditor does. Send it, and they go from someone who can only catch contradictions inside this book to someone who can catch contradictions with your other six. A proofreader working on a late file needs the name spellings above all, because that's the kind of error that survives every earlier pass. A co-writer needs the whole thing, and needs it to be the final word. Two people writing from two private memories of the same cast always end up with a rewrite. A narrator needs a smaller slice, and needs it early: who's in the book, how each name is said, each character's age and way of speaking, which accent is which, and which characters listeners must be able to tell apart by ear. A translator needs the parts that carry meaning: what a nickname implies, which form of address signals distance, and which speech habit is on purpose and not an error to tidy away. Each of those is a different view of the same file, so keep one master document and pull from it. Don't keep a separate narrator sheet, editor sheet and co-writer sheet. They'll all drift out of step.

It changes what revision costs. In revision, you answer the question "is this consistent?" either by looking something up or by rereading. One takes ten seconds and the other takes a day. A bible you keep up also lets you make big changes on purpose. Say you decide in the third draft that the daughter should be eleven instead of fourteen. The bible shows you every fact that depends on it (the school, the phone, the way the mother speaks to her, the two flashbacks), so you don't have to find them by luck.

Real-world examples

Arthur Conan Doyle's Dr. Watson (1887 to 1891): a wound that moved and a first name that changed

The character bible that teaches the most in the history of fiction is one nobody kept. In A Study in Scarlet (1887), Watson explains that a Jezail bullet hit him in the shoulder, shattering the bone and grazing an artery. In The Sign of Four (1890) he mentions, twice and very clearly, a wounded leg with a Jezail bullet through it. Same war, same bullet, different limb.

It gets better. Watson is introduced as John H. Watson, and the stories never say what the H stands for. Then, in "The Man with the Twisted Lip" (1891), his wife calls him James. The crime novelist Dorothy L. Sayers wrote a short essay, reprinted in her collection Unpopular Opinions (1946), suggesting that the H stood for Hamish. Hamish is the Gaelic form of James, so the wife and the narrator could both be right. It's a neat repair, and it's still a repair. Conan Doyle simply wrote the wrong name.

These are the very fields listed above, a physical detail that comes back and a name in one of its forms, and a one-page record prevents just this kind of error. There's a more comforting point too. Sherlock Holmes is one of the longest-lasting bodies of popular fiction ever written, and the contradictions didn't sink it. Readers noticed, enjoyed noticing, and spent a century turning the job of reconciling them into a parlor game. Protect your continuity because errors cost you some credibility and a little goodwill, not because they're fatal. If you find one in a published book, fix it in the file and move on.

Agatha Christie's notebooks and Poirot's age (1920 to 1975): the one field she couldn't fix

Christie's working papers are unusually well documented. Seventy-three handwritten notebooks were found among her belongings, and the scholar John Curran later deciphered them and published them as Agatha Christie's Secret Notebooks (2009). They aren't a tidy reference work. They're working notes: lists, alternatives, plot ideas, often undated, often jumping between books written years apart. That's what an author's actual record looks like: messy, current, and used.

She also made something closer to a bible decision, on the biggest scale there is. She wrote Curtain, Poirot's last case, in the early 1940s, because she didn't expect to survive the bombing of London. Then she had it locked away for three decades, and it was published in 1975. The ending of a series was a recorded fact for thirty years before readers saw it.

And then there's the field she couldn't fix. Poirot arrives in The Mysterious Affair at Styles (1920) already retired from the Belgian police, so he's already old, and he's still working fifty-five years later. Christie wrote her autobiography between 1950 and 1965, and it was published in 1977. Looking back in it, she said that starting him so old had been a terrible mistake. She felt she should have dropped him after the first three or four books and started again with somebody younger.

That's the ages-and-dates field, and it shows something the field list can't. Some entries are decisions that grow like compound interest. A date of birth you set casually in a first book limits every book after it, and just writing it down won't save you from the mistake. What saves you is writing it down and looking at it before you commit to a long series. Then an assumption you never checked becomes a choice you can see.

Robert Jordan's The Wheel of Time (1990 to 2013): the record that outlived the author

Robert Jordan died in 2007 with his fourteen-volume series unfinished, and another novelist, Brandon Sanderson, completed it. He worked from Jordan's notes and outlines, alongside Jordan's widow and editor, Harriet McDougal. He also worked with the two editorial assistants who had spent years keeping track of the series' continuity, Maria Simons and Alan Romanczuk.

In 2015, after the series was finished, that built-up record became a published book, The Wheel of Time Companion. It's credited to Jordan, McDougal, Romanczuk and Simons, and it draws on Jordan's unpublished notes. It's an alphabetical reference, with character biographies, histories, and a dictionary of the invented language. In other words, it's a character bible and a continuity bible, tidied up and sold.

You don't need to aim to publish yours. The lesson is that a continuity record kept by named people, as their job, is what made an unfinishable series finishable. The same rule works at every size, right down to one author with a spreadsheet. The document is worth as much as you'd lose if the only copy of these facts were in somebody's head.

You can see the same machinery running in public today. Sanderson's own publishing operation keeps an internal continuity wiki and a detailed timeline, kept up by a member of the editorial team. It also keeps an archive of audio clips from earlier audiobooks, so narrators can match how a name was said several books ago. You could steal that idea. The archive is the bible's narrator view, made permanent.

The Snowflake Method's step seven and the Proust Questionnaire: two documents that look alike

Randy Ingermanson's Snowflake Method is a ten-step design process that grows a novel outward from a single sentence. He set it out on his own site and later in a book of the same name. Step seven asks you to take a week and expand your character descriptions into full character charts "detailing everything there is to know about each character". That means birth date, description, history, motivation, goal, and how the character changes by the end. Ingermanson's reason is clear and sensible. Time spent here is time saved later.

Now put that next to the Proust Questionnaire. It grew out of Victorian confession albums, a parlor game where you answered a set of personal questions in a friend's album. Marcel Proust filled in two of these as a young man, in the 1880s and again in the early 1890s. One album turned up again in 1924 and was published. The TV host Bernard Pivot used a version of it to close episodes of his French literary talk show, and Vanity Fair made a version of it a regular feature on its back page. As an interview tool, it's great. It was never designed to produce a reference work, and it doesn't.

On the page, the two documents look almost the same: a name, a list of fields, and answers underneath. They do opposite jobs. Ingermanson's chart is preparation, with a stated purpose and a fixed end point. The parlor-game questionnaire is for sparking ideas, and its value is in the surprising answer, not in the record.

Things go wrong when you do the second while believing you're doing the first. Say you answer two hundred questions about a character's favorite breakfast, her childhood pet, what's in her purse and how she feels about jazz. You'll get pages of answers and a warm feeling of being prepared. You'll also get a document you can't use as a reference and will never open again. Call it avoidance with a filing system. The trap is tempting because it isn't idle. You really did work, and there really is something to show for it, yet nothing is at stake in any of it.

So here's the plain difference. A questionnaire is worth answering when you're stuck on a character and need to be surprised. What you get out of it is the two or three answers that changed how you'll write her, and those two or three go in the bible. The other hundred and ninety-seven get deleted, and deleting them is the skill.

Star Trek series bibles (1987 onward): the version written to be handed over

Television has done this longer and more carefully than publishing, and its documents are public. The official Star Trek site has published the series bibles for The Next Generation (1987), Deep Space Nine (1992), Voyager (1995) and Enterprise (2001). Any novelist would get a lot out of an hour with them.

What stands out is how little they look like a character questionnaire, and how much they look like a briefing. They describe the show's themes and feel, sketch each regular character in a way a stranger could write from, and show where things are heading. They're also clearly drafts that changed later. There are planned characters who never appeared, a lounge that ended up as two different sets, and descriptions written before casting that read oddly afterward.

They teach you something you can use. A bible is written for a reader who is not you, and that limit makes it better, because it forces you to state the fact instead of waving at a memory. Write your entries as if a new co-writer will use them, and you'll write better entries, even if nobody else ever opens the file.

They also show that the document is a snapshot of decisions, not scripture. When the manuscript and the bible disagree, one of them has to be corrected, on purpose, that same day. That raises a question that matters more than any field list: which one wins? Usually the manuscript, because that's what readers will see, so you change the file. Sometimes the bible wins, because the manuscript is what drifted. What you must never do is leave them disagreeing and decide later. "Later" is how a wrong fact gets published. There's a warning case on record. An error crept late into Harry Potter and the Goblet of Fire (2000), when the order in which two characters appeared was changed during a rushed edit. It was corrected in later editions, after the author confirmed the original manuscript had been right. When nobody can point to a record everyone trusts, the loudest recent opinion wins the argument.

An everyday scenario: the field nobody filled in

Say you write police procedurals. Your detective is Perpetua Loach, four books in, and you keep a decent bible. It has the spellings, ages set against a real calendar, the sergeant's motorcycle, and the fact that Loach never uses contractions when she's testifying.

Book five has a subplot about a witness, Dosia Halloran, who saw something at a gas station in 2019. In March you write chapter nine, where Loach's sergeant, Nessim Farag, interviews Halloran and gets nothing. In June you write chapter twenty-two, where Farag tells Loach that Halloran recognized the car. It's a good scene, in fact the best scene in the book.

Except Halloran told him nothing in chapter nine, and nobody has spoken to her since. You know Farag knows because you know, and from the inside, you can't see the gap between what the author knows and what the character has been told. Your copyeditor, working on this file only, has no reason to question it. Two beta readers feel the ending comes a little too easily, and they can't say why.

The bible had thirty-one fields for Halloran: height, hair, the dog, her shift pattern at the gas station. It had no row saying the car, Farag, not yet, chapter 9 interview produced nothing. One row, ten words, and you couldn't have written chapter twenty-two wrong.

Then there's a second cost. Because the reveal came free, one scene never got written: Farag going back to the gas station and earning it. That's the scene the book actually needed, and it would have raised the stakes on his own position. Missing continuity doesn't just create errors. It quietly removes scenes.

Try it yourself

  1. Start from the manuscript, not from a template. Open whatever you've written and make one entry for each character who appears more than twice. Fill in only what the text already commits you to. Give it ten minutes per character, not an afternoon.
  2. Write the name block first. For each character, write down the full name, the exact spelling, what each other character calls them, and any variant you've used by accident. Then search the manuscript for each variant. This step alone usually finds two or three errors in a novel-length draft.
  3. Build a dated spine. Make one column of years, from the earliest event in anyone's past to the end of the book. Put each character's birth year on it, then each big event in their past. Math errors show up right away, and you'll find at least one.
  4. List recurring physical details only. Go through the draft and note any physical fact you've mentioned more than once. Those are your constraints. Ignore the single mentions. They're free.
  5. Write five speech rules per speaking character. Make them rules you can hear and check: "Never swears", "calls everyone boss", "says I suppose when she disagrees". If you can't name five, you may have a voice problem and not a record problem, and that's a matter for character voice.
  6. Add the knowledge record. List the four or five facts your plot turns on. For each one, list every character and when they learned it. Expect this step to find a real problem in an existing draft. Finding one is what the step is for.
  7. Decide the split with the world. Anything that would still be true with the cast removed goes into your continuity bible instead. Don't store it twice.
  8. Choose a home you will actually open. The requirements are boring. It must be searchable, and it must let you see one character on one screen. You must be able to edit it in the ten seconds between writing a sentence and forgetting what you decided. There must be only one live copy. A spreadsheet, a plain-text folder, a wiki-style note app or the character sheets built into novel-writing software all qualify. Features and pricing change all the time, so check current terms before you commit to anything paid. Pick a format you can export, because you may be handing it to an editor or a co-writer who uses something else.
  9. Give it a maintenance slot. End of session, or end of chapter. Add only facts you invented, not facts you plan.
  10. Extract views when you need them. Before a copyedit, send the whole thing. Before an audiobook, pull out the smaller narrator sheet: cast list, pronunciation, age, way of speaking, accent, and which voices listeners must be able to tell apart. Keep pulling from the one file, and never keep two.

Here's a copy-and-paste outline for one character:

Name: ______ (exact spelling: ______ ); called ______ by ______, ______ by ______ Born: ______ · Age at chapter 1: ______ Recurring physical: ______ Injuries / permanent marks, and when acquired: ______ Relationships: ______ (also recorded under: ______ ) Job, address, vehicle, phone, pets: ______ Five speech rules: 1. ______ 2. ______ 3. ______ 4. ______ 5. ______ Knows / when learned / how / has told: ______ Believes falsely: ______ First appears: ch. ______ · Last appears: ch. ______ Changed on: ______ (date, and what changed)

And for the plot's knowledge record, one line per fact:

Fact: ______ Knows from the start: ______ · Learns in ch. ______: ______ (how: ______ ) · Never learns: ______ · Believes a false version: ______ (which version: ______ )

Common mistakes and misunderstandings

The beautiful unused document. By far the most common failure is forty pages per character, color-coded, with a mood board, built over three weeks and opened twice. Don't ask "is this thorough?" Ask "when did I last look something up in it?" If the answer is never, the document is a hobby. Shrink it until it's the thing you reach for, which usually means one screen per character and a search box.

The bible that contradicts the manuscript. A record that disagrees with the book is worse than no record, because people will trust it: you at two in the morning, your copyeditor, your co-writer. Every session that changes a fact must change the file the same day. Date every entry you change, too, so that when two versions turn up you can tell which one is current.

Building it instead of writing. Preparation is the most comfortable way to avoid writing, because it gives you something to show for it. If you've spent longer on the cast document than on the chapter, the document is where you're hiding. The fix is simple. Limit the up-front version to one screen per character, start drafting, and let the file grow out of the draft.

Mistaking a questionnaire for preparation. Answering two hundred questions about a character's favorite breakfast feels like work, and it's almost never preparation. The questions that matter are the ones whose answers change a decision on the page. If knowing her breakfast won't change a single sentence you write, it's trivia. Use a questionnaire on purpose, as a tool for getting unstuck, and keep only the answers that moved something.

Recording interpretation instead of fact. "Deep down she is afraid of being ordinary" doesn't belong in a reference table. You can't contradict it, it'll keep changing as you draft, and having it there teaches you to stop trusting the file. Keep the psychology where you're working it out, in your notes on the lie, the wound and the character arc. Keep the bible factual and boring.

Skipping the walk-ons. The entry that saves you is rarely the protagonist's. You won't misremember your lead's sister. You'll certainly misremember the name of the pathologist who appeared for two pages in book two and comes back in book five, and that's the error a returning reader spots first. Everyone with a name gets a line. Your supporting cast is where the bible earns most of its keep.

No index, and no per-character view. One long document you have to scroll through is really a filing cabinet with the labels torn off. What you need is to get to one character in one step. That's the whole practical difference between a bible that survives book four and one that doesn't.

Not recording the spelling decisions themselves. Writing down "Halloran" isn't enough if you spent twenty minutes choosing it over "Hallohan". Record the rejected version too. That's the one that will turn up in chapter eighteen, and a search for the right spelling will never find it.

Assuming your editor has it. A copyeditor knows the file you sent. Without your bible they can't see a contradiction with book two, and they may helpfully "correct" a choice you made on purpose three books ago. The only thing you must do here is actually attach the document. How that whole exchange should work is covered under editing.

Treating it as the world's record too. Bolt geography, calendars, magic rules and institutions onto the character file, and you get something too big to look things up in and impossible to hand to a narrator. Split it, and put the world in the continuity bible.

Expecting it to fix a thin character. A complete record of a flat person is still a record of a flat person. The bible keeps things consistent. It doesn't build character. If your beta readers say nobody cared about her, no number of fields will help. That work happens in flaw, motivation, want vs need and character arc.

Related elements