Back to the table

114 Advanced Sa Series Arc Series & Universe

Series Arc

The larger story that stretches across all the books.

What it is

A series arc is the one story that runs across every book in a series. A big question opens in book one, the books in between carry it forward, and the final book closes it. The arc sits under each volume's own plot. It never takes the place of that plot, and believing it does is the commonest misunderstanding about series fiction.

So every volume owes the reader two things. The first is a complete story: a problem raised, met and settled inside that one book. The second is one real step on the big question. Leave out the first and readers call the book "half a book." Leave out the second and the series starts to feel repetitive. Both duties, every volume. A volume that does only one of them is the fault readers spot soonest.

In return, the arc gives you the single thing only series fiction can do. You can set something up in chapter three of book one and cash it in four books later. A standalone novel cannot do that, because a standalone ends.

The simplest way to hold the idea is this: the arc is the story the complete set tells. The world, the recurring cast and the brand make a series recognizable. The arc makes it a story rather than a collection. In practice that comes down to one big question, raised early, pushed forward in every volume and answered in the last one, plus the chain of cause and consequence that links the raising to the answering.

A single novel works differently. By its final chapter, the dramatic question is closed. A series question is engineered to outlive book one, and that gives it three requirements. It must be too big for one book. It must be specific enough that an answer exists. And it must be conspicuous enough in book one that a reader who quits there can feel something is still pending. The Dramatic Question page covers the single open question of one book and gives the fuller account of the difference.

Under all of this sits a contract with two layers. The contract is the spine of the element, and nearly every series failure breaks it somewhere.

Layer one: each volume gets its own payoff. The crime is solved. The couple arrives somewhere. The siege is lifted or lost. The reader can close the book and report the result. An open question and a withheld answer are different things, and the difference is non-negotiable. Having a series plan changes nothing about it. The Resolution page covers closing a volume, how much may fairly stay open, and how to tell the two apart.

Layer two: the big question moves in each volume. Getting near it does not count. Hinting at it does not count. By the end of the volume something is newly known, newly possible or newly lost. The test for movement is simple: the reader's best guess at the series ending is different after the volume than it was before.

The paid-and-moved test turns both layers into two minutes of work per volume. Complete these two sentences plainly, with no hedging:

"This book paid off ______."

"This book moved the series question from ______ to ______."

If you cannot write the first sentence, you have a chapter sold as a book. Readers notice at once, and the reviews that say so persist for the life of the series. If you cannot write the second, you have written an installment rather than a volume. Do that repeatedly and you get what might be called the stack: four competent books that could be read in any order, by an author who seems to be waiting for something. One lapse is survivable. Two are not. If a deadline forces you to risk one breach, know which is worse. A volume that only advances the arc breaks a promise. A volume that only resolves its own story is a flat patch. So resolve the book.

Two impostors are routinely mistaken for a series question. One is the premise. Premise is subject matter: a wizard police constable in London, or a society kept emotionless by a conditioning protocol. The other is the theme, which is a recurring argument: competence is not goodness, or people are shaped by the place they left. Both are useful. Neither carries an arc, because neither can be answered.

A real series question has three properties, and it needs all three.

Answerable. You can write the answer now, in one sentence, and it reads as an answer rather than a mood. "Will the Seldon Plan hold?" passes. "What is the true nature of magic?" fails, because that one is a topic with no stopping point. A series built on it ends when the author gets tired, and readers can sense it.

Refusable. "No" has to be possible, and you have to be able to write that no as a book. If only a good ending is possible, the question is decoration. Most vague questions die here. "will good triumph over evil" has only one commercially plausible answer, so the reader never doubts, and an arc nobody doubts weighs nothing. The Stakes page covers what makes a refusal costly.

Statable. A reader who has read only book one can say the question in one sentence, even if no character has said it. Better if none has. The question should assemble in the reader's head out of what is withheld. A character who announces it turns a discovery into an explanation. The Dramatic Question page makes this point, and the Foreshadowing page has the technique.

Three strong questions, to show the standard: Can the protocol that holds this society together be broken, and who pays for it? Who is the man behind the killings, and what is he actually trying to build? What happened to the friend this narrator has spent sixty years measuring herself against? Three weak ones: Can she find herself? Will the kingdom survive? What is the prophecy? Each weak one fails at least two of the three tests, and the symptom in the manuscript is always the same: a book three with nothing particular to do.

Arcs come in a few shapes. None is better than the others, and each has its own typical failure, so knowing your shape means knowing what to watch for.

In the single escalating threat, a danger or an antagonist grows until it has to be confronted. This shape is the easiest to plan and the easiest to sell, because every volume has an obvious job. Its failure is escalation inflation: a city is threatened in book three, a continent in book four, and by book five the reader feels no difference. The fix is to go narrower instead of bigger. Escalate intimacy rather than scale, so the endangered thing gets closer instead of larger.

In the mystery in installments, a withheld answer is released piece by piece. The reader learns something real in every volume and sees the whole only at the end. Ben Aaronovitch's Rivers of London books take this shape, across seven novels from 2011 to Lies Sleeping (2018), where the arc closes. The shape fails in two ways. The first is installment starvation, when a piece is too small to feel. Readers will tolerate a long wait for an answer and almost no wait for nothing. The second is an answer too thin for seven books of waiting, which comes up again under endings below. The Series page covers what became of Aaronovitch's series after its arc closed; the failure modes stay here.

In a character's long transformation, one person becomes another across eight books. This shape costs the least to run, since it needs no world-scale escalation, and costs the most to recover from, because a change, once made, is spent. The standard commercial fix is to keep the lead's core conviction fixed and hand the transformations to a rotating supporting cast. The Character Arc page covers the flat arc and why series leads usually have one.

In a generational or world-historical arc, a society, a family, an institution or a species changes, and the individual characters are its instruments. Its strength is durability: it outlives any character's death and runs as long as the history does. Its weakness is that readers bond with people, and nobody bonds with a century, so every cast swap spends some of that bond. The good practice is one continuity thread that runs the whole way: a place, a house, an object, a named institution.

The retrofitted arc is the commonest real situation. The book sold, the publisher asked for three more, and now an arc is needed. There is no shame in it. Some of the longest-lived series are retrofits. Retrofitting is a specific craft of its own, and the Foundation example below shows how it is done.

The deliberately thin arc has near-zero arc by design: the classic detective series, many cozies, most long-running children's series, much adventure fiction. A fixed lead, a fixed world, a fresh case each time, the same furniture. The shape is legitimate. A reader can start anywhere, it strands nobody, and it is the only shape that lasts forty years without its bookkeeping collapsing. Thin is not zero, though. The good ones drift: the lead ages, people leave, there is a marriage, there is a death. Ten books on, the world has measurably changed, and nobody ever announced it. You plan a thin arc as a direction rather than a question.

Four neighboring elements share this ground, so it helps to know where each one stops. Series is the form and the decision: the kinds of series, their length, reading order, whether a reader can come in at book four, and the business case for writing several books about the same people. If you have not decided whether to write a series at all, start there. Everything here assumes that decision is made. The Continuity Bible is the record. Every item you are told to write down here lives there: the arc is the plan, and the bible is the filing. The Beat Sheet is the single-page working sheet, and that page shows a series version beside the book version. Foreshadowing is the plant itself, Chekhov's gun, the setup that reads as texture now and inevitability later. What the series arc adds to that is scale. Planting at chapter scale and planting at series scale are different disciplines, for one blunt reason: book one is frozen.

Why it matters

The arc is what makes a series a series. Take it away and what remains is a format: same lead, same setting, new problem. A format is commercially fine; the thin arc is one. But a format is a different product, and readers price it differently. In a format, the reader buys the next book if this one pleased them. No need is involved, only liking. The arc converts liking into needing. An unanswered question gives the reader a reason to return even after a weak volume, and every long series has weak volumes.

Most of the money side is covered under Series. Read-through, meaning readers carrying on to the next book, plus pricing and sales patterns, are treated on that page. They come up here only where they change a structural choice.

An arc is a debt, and readers keep a careful count. Every question you open is an answer you have promised. Eleven questions opened and two closed is debt rather than suspense, and the reviews start saying so around book four.

In a standalone novel, a plant can go in afterward. A plant is an early detail set up to pay off later. You write the climax and find the reader needed the locksmith from chapter two, so you go back and insert her. Most good foreshadowing secretly rests on that kind of retroactive insertion.

In a series, book one is already printed. There is no revision pass. Whatever book four needs either sits in book one or it does not, and nothing added later can fairly supply it. That one fact is why this element exists and why the arc gets decided early. Tidiness has nothing to do with it. The first book is the only one you can never edit.

The planting work divides into three parts. The first is what you write down before book one ships. The second is what you keep vague on purpose. The third is the discipline that lets the other two coexist.

Part one: write these down before book one ships. The series answer, in one sentence, kept private. The two or three facts the answer rests on, each one checked and found present in book one. Every clock you have started: an age, a date, a term of years, a pregnancy, a prison sentence. Clocks are what writers most often start and then contradict. And the one rule of your world you will never break, since the answer probably depends on it.

Part two: keep these vague on purpose. The mechanism of the answer. The wording of anything predictive. The identity of anyone who might need swapping later. The number of books. Vagueness here is not laziness. With it, book three can want something different and the plan survives.

Part three is the discipline: plant options, and save promises for later. A promise is a plant with one fair payoff: a sealed letter holding one possible thing, a scar with one possible cause. An option is a plant that can be read three ways and paid off in any of them: a logbook page in a second hand, a man who left town fast, a key that fails the door it sat beside. They take the same page space, and both read as texture the first time through. The option preserves your freedom. The promise removes it. So early volumes get options, and you convert an option into a promise only once you are committed, which normally means once the book that pays it off exists in draft and not merely in a plan. That rule prevents a familiar failure: a detail in book one that is obviously significant, stranded forever, because the one payoff it permitted, meant for book four, turned out to be wrong.

The list of what is planted and what has fired lives in the continuity bible. At series scale the continuity bible gains a column headed still live. The Foreshadowing page shows how to test a single plant.

The sagging middle is the central craft problem of this element. Think of it as the series-scale version of the sagging second act, with the same cause and worse consequences. The danger zone is book three of five. The question is no longer novel, because books one and two spent the novelty. It cannot be answered yet, because that is book five's job. The easy per-book premises are used up. The cast has grown. And the author, who knows the ending, ends up administering a gap rather than writing a story. The reader's verdict is exact and damning: things keep happening and nothing changes.

Two well-documented cases show the sag, and both are useful because the authors chose it on purpose. Neither one is an accident you can shrug off as bad luck. Each author looked at the choice, made it, and later said so.

Robert Jordan chose a simultaneous book. Crossroads of Twilight (Tor, January 7, 2003) is the tenth of what became fourteen books in The Wheel of Time, and it runs largely simultaneous with the book before it rather than after it. Jordan admitted the choice in an April 2004 interview with the fan site Wotmania, preserved in the Theoryland archive. Starting every major segment of the book on the same day had seemed terrific at first, and he recognized a better way only when he was too committed to rewrite. The reader response is recorded in Leigh Butler's long-running Wheel of Time reread on Tor.com. She had expected the fallout from the previous climax and got a long catch-up instead, with whole chapters that could be summarized in a sentence or two.

George R.R. Martin dropped a five-year gap. In A Song of Ice and Fire, roughly five years were supposed to pass between books three and four. Martin spent about a year trying to write across that gap and dropped it, because the leap needed too much flashback and exposition. The material that replaced the gap outgrew one book, so he divided it by geography rather than chronology: A Feast for Crows (2005) and A Dance with Dragons (2011) run at the same time with separate casts.

The common thread is the most useful diagnostic on this page: a sagging middle is time that stopped. Readers gauge progress by elapsed time and by irreversible change. They do not count pages or events. A volume that runs simultaneous with its predecessor reads as "nothing happened," however much did.

Six concrete repairs exist, and all six work mid-series. You do not have to wait for the next book to use any of them.

Repair one: promote a sub-answer into a payoff. Take something you were saving for the finale that can be answered now without spoiling the ending. A series question is normally three or four stacked questions, and you can afford to answer one of them early. Of the six, this one is the most effective and the most resisted.

Repair two: close the thread that has expired. A will-they-won't-they or a withheld confession that has sat unchanged since book two has stopped being suspense and become routine. The Subplot page covers threads that span volumes, including Elly Griffiths's account of how she judged the moment a long-running romantic thread needed resolving. Closing a thread costs something. Postponing it costs more.

Repair three: narrow the cast for one book. Casts only ever grow. Three viewpoints after nine is focus rather than retreat, and it turns the middle volume into a story instead of an administrative roundup.

Repair four: change the mode. A shorter book, a new place, a new viewpoint, a new kind of problem. The sag is partly sameness, and Pacing is the lever.

Repair five: pull a book-four payoff into book three. Then invent a replacement for book four. Almost always possible, almost always right. Payoffs are not a finite supply, and the ones invented under pressure often beat the ones planned at leisure.

Repair six comes first: check the timeline. If the middle volume overlaps its predecessor, or spans three days, or ends with the characters where they started, in place and in feeling, you have located the sag. You may not need the other five repairs at all.

Planned arcs and discovered arcs both work. The evidence runs both ways, and anyone who claims otherwise is selling a method. On the planned side: Rick Riordan fixed the length of his first series in advance, because a series needs a strong ending, and the result is a five-book arc that lands, with a finale that pays off a book-one setup designed for exactly that. J.K. Rowling's seven-book question is the other famous planned case, and the Dramatic Question page explains how that question is visible in book one without ever being spoken. On the discovered side: Isaac Asimov's Foundation stories were written one at a time over nine years, from 1941 to 1950, by an author who kept wanting to quit, and the series has been in print since May 1942. Terry Pratchett's Discworld grew its coherence along the way. And plenty of long-lived crime series found their long thread somewhere near book four, when the author noticed what was actually bringing readers back.

Over-planning has its own failure, and it is less visible. Once the plan specifies book three chapter by chapter, drafting book three turns into following orders. The symptom is scenes that are present because the plan says so, with nobody in them wanting anything. The day you defend a scene by citing the outline, the plan has stopped helping. The Outline page has the general version of this problem.

The practical middle fits on one page. Before book two, write down five things and no sixth. The one-sentence answer. Every fact in book one that the answer depends on. Which open threads carry weight and which are decoration (there is more decoration than you expect). The rough shape, meaning the number of volumes and what the last book is about, though not its events. And the one rule you will not break. All of it on one page. No plan and no outlines for books three to six. Everything beyond that page is optional, and most of it is better discovered. Keep the page in the continuity bible and update it truthfully every time you change your mind, because you will. K.M. Weiland's article on outlining a series, at Helping Writers Become Authors and listed below, is the clearest free version of this middle position: fix the series ending and its shape, and detail only the book in hand.

A final volume owes four things, and they are different from a standalone's debts. Each one is a place where a finale that felt fine in draft gets caught by readers.

First, the answer must have been available from book one. Consistency is not enough. The reader has to be able to look back and see that the answer was possible all along. An answer that needs information the series never supplied is a cheat, however logical it is.

Second, pay the oldest debts first. The first promise is the one readers remember longest. A finale that resolves only the previous volume's setups and nothing from book one has the order backward.

Third, the per-book plot is the vehicle. Layer one is still owed in the last book, and what that plot resolves should carry the series answer. An unrelated case running beside the finale is padding, literally.

Fourth, take a last look at the people. Writers commonly cut this for pace, and they should not. Six books with a cast earn a few pages on everyone's fate. The Resolution page covers the mechanics and how long you may take over it.

The hardest problem is a question that has stayed open for a decade. A question like that gathers answers. Every reader who has waited builds a private ending, and the best of those private versions are very good, because thousands of people have had years to work on them with no obligation to fit anything. Your real answer has to compete with all of them. Readers are not being unreasonable. The sheer numbers do it. This problem is structural to long arcs, and no single book can be blamed for it.

Softener one: write an answer that costs rather than explains. Readers forgive pain quickly and thinness slowly. An explanation competes with every cleverer version a reader has imagined. A cost competes with nothing, because the private versions were about the mystery, never about the bill.

Softener two: answer in installments. Make the last book the final step rather than the whole staircase. Three real answers already delivered teach readers to expect answers of that size. Everything withheld teaches them to expect a revelation big enough to justify ten years, and no such revelation exists. Television faces the same problem every season, which is why Neil Landau's The TV Showrunner's Roadmap is listed below: one chapter treats a multi-season show's central question or central mystery as a design job, rationing the answers per installment and keeping the eventual answer worth the wait, with interviews from showrunners who sustained an arc across five seasons.

A Song of Ice and Fire is the clearest and most delicate case, and it should be read as structure rather than verdict. The question is enormous. The answer has been postponed across five long volumes, and the sixth, The Winds of Winter, had not been published at the time of writing; check the current position for yourself. Martin's own published record of the five-year gap, plus the Meereenese tangle (the knot around the city of Meereen), is unusually candid: a structural knot in mid-series ended up dictating the schedule. The lesson has nothing to do with one author's pace. An arc bears only so much, and the strain shows about two volumes after it is created.

Arcs fail to finish in three different ways, and none of them is rare. A series is abandoned, usually over sales. An author is outlived by the series. Or the series stalls, each book harder than the last. All three happen regularly, and there are documented cases of each. Jordan died in September 2007 with the intended final volume of The Wheel of Time underway, and left notes and recordings detailed enough for Brandon Sanderson to finish it, as three books from 2009 to 2013. Nothing makes a stronger practical case for keeping a real continuity bible, and that page covers the record itself. Patrick Rothfuss's Kingkiller trilogy, whose first volume is covered on the backstory page, had its second volume in 2011 and no third since. And every month, unknown authors quietly abandon series. Abandonment by an unknown author is the commonest case of all and the reason this section exists. The Series page treats stalls and abandonments as a matter of form.

The structural answer is to build every book to outlive an unfinished series. Four cheap measures do it, and none of them costs you anything in the books that do get written.

Measure one: resolve every volume for real. Layer one is your insurance. A reader who quits after the third book of five, with the series abandoned, has still read three whole books.

Measure two: put no essential payoff in an unwritten book. If book one's mystery is bearable only because book four will explain it, book one is held hostage to a book that may never exist.

Measure three: never end a volume on a reversal that only the sequel can justify. The Cliffhanger page covers what readers accept, and why the same ending is praised on volume four and punished on volume one. The more incomplete a volume is, the more of the series' survival is bet on the author continuing.

Measure four: when the risk is real, pick a shape that degrades gracefully. The installment mystery and the thin arc degrade well, because a reader who quits has had answers. The single escalating threat degrades worst, because an unfaced threat is the one thing it cannot survive.

The pure-foresight version belongs to Agatha Christie. Write the last book first and store it safely. The Curtain example below shows how she did it.

Real-world examples

Elena Ferrante's Neapolitan Novels (2011 to 2015): a plant paid off roughly 1,600 pages later

The Neapolitan Novels are four books: My Brilliant Friend, The Story of a New Name, Those Who Leave and Those Who Stay and The Story of the Lost Child, published in Italian from 2011 to 2014 and in English from 2012 to 2015, translated by Ann Goldstein for Europa Editions. Within modern literary fiction they are the clearest instance of a series question that passes all three tests, and they contain the longest-range planting anyone attempts. The Series page looks at the same quartet from the opposite side and asks whether it is a series at all or a single novel in four parts. What matters here is the question and the object that answers it.

Book one opens on a woman in her sixties taking a phone call from her oldest friend's son. The friend has vanished, deliberately and totally, and has even cut herself out of the family photographs. The narrator decides to write it all down. So the arc is framed before the story proper begins. The question is answerable, because it concerns a specific event and a specific person. The question is refusable, because one possible answer is that the friend is gone for good and there is nothing more. And the question is statable: fifteen pages in, any reader can say it in a line.

Both layers are honored. Each volume covers a stretch of two lives and settles something: a marriage, a store, a book, leaving the city. The bigger question, what each woman has done to the other, advances a step. A reader could stop after any volume, complete, and would not want to.

The plant sits in the opening section of My Brilliant Friend. Two small girls lose their dolls down a cellar grate. The scene takes a few pages and reads as childhood, the kind of thing an editor would call texture. If you have not read the quartet, skip the next two sentences. In the last pages of The Story of the Lost Child, roughly 1,600 pages on, a package arrives with no sender and no address, and the two dolls are inside. The arc is answered by an object last seen six decades of story earlier, and the object says what no explanation could.

Look at what the plant cost: a few sentences in an early chapter. And notice that it was an option rather than a promise. Lost dolls could mean a dozen things, and book one commits to none of them. The answer costs rather than explains. The reader is left unsure whether the package is a gift or a rebuke, and an ending like that competes with no private theory.

Percy Jackson and the Olympians (2005 to 2009): a prophecy written as an option

Rick Riordan's first series runs five books, from The Lightning Thief (Miramax, June 28, 2005) to The Last Olympian (Hyperion, May 5, 2009). It teaches two lessons. One is that these are among the cleanest two-layer series in children's publishing. The other is that fiction's most dangerous planting device, the prophecy, is handled here with rare discipline. The Series page covers the same books as decisions: how long to run, the companion series, coming back after years away. Here the subject is only how the arc was built.

The two layers are simple, almost mechanical, and that is praise. Each book is a quest with a stated goal, a deadline and an outcome. The stolen object is recovered or it is not. The destination is reached or it is not. A nine-year-old can tell you how each book came out. Beneath the quests runs a prediction: one child of the three senior gods will decide whether Olympus stands or falls. The prophecy surfaces in book two, is pressed in every later volume, and is recited whole only in book five.

A prophecy is the device likeliest to turn an arc into a trap, because it is the purest promise there is: published text the author must honor. Riordan wrote his as an option. The subject is a category rather than a name, since several living children qualify, and the books make sure readers know it. The decisive term is a choice rather than an action. And the wording gives the shape of an event with no sequence of events. So the ending was not forced at the moment of planting, and whoever makes the decisive choice supplies the answer.

If you have not read the 2009 book, skip the next sentence. The one who chooses is the series' long-running betrayer rather than the hero, the choice is self-destruction, and the prophecy's wording permitted it from the start. The answer was available in hindsight, which is the test every ending has to pass.

The planning lesson comes from Riordan himself. The FAQ on his own site, listed below, says flatly that a series needs a strong ending and that Percy's story was always going to be five books. He accepted the shape as a limit before he needed it. Five books is long enough for a book-two prophecy to stay live through book four and short enough that it never fades.

Nalini Singh's Psy-Changeling (2006 onward): the toughest per-book promise, honored every time

Romance is the hardest genre in which to run a series arc, and the reason is contract rather than craft. The per-book promise is non-negotiable, and romance readers enforce it more strictly than any other readership: this volume's central couple must end settled and happy. Layer one can never be deferred. The Series page covers these same books as a question of entry points into a long-running romance series. The question here is narrower and harder. If every volume is obliged to end itself, how can an arc move at all?

Singh's solution is the genre's standard one, run unusually long. Slave to Sensation (Berkley, September 2006) introduces a world with two populations, one of them conditioned under a protocol named Silence that strips emotion away. Every later volume is a complete romance with its own couple and its own resolution. By Allegiance of Honor (June 2016) there were fifteen books, and that one was designed as an ensemble bridge rather than a single couple's story. The underlying arc is whether Silence can fall, and at what cost. No single love story answers that question. Every love story moves it, because each successful pairing is evidence against the protocol. The romance plot and the arc are made of one material, which is why the result is fifteen books that hold together rather than a world tour with couples attached.

Singh's own account beats any outsider's analysis. In a post on her blog dated June 18, 2014, listed below, she described the Silence arc as the series' defining element so far. Some readers had assumed that Heart of Obsidian (June 2013), the twelfth book, which set the protocol's fall in motion, had ended the arc. Her view was that the volumes dealing with the direct aftermath had to come first, and that which book would close the arc depended on how much the current manuscript could wrap up. She framed the whole thing in television "seasons": a first arc closing, a second opening, the cast carried across.

Several lessons follow. The arc's answer can land before the arc ends, because the consequences belong to the arc too. Jumping straight from the answer to fresh material is one common reason a series seems to have skipped a beat. Thinking in seasons helps. And an arc that closes and hands over to a successor is far more sustainable than an arc required to last the entire series. The successor here is Psy-Changeling Trinity, from Silver Silence (June 2017) onward, still running at the time of writing; check where it stands now. Through all of it, no couple's romance was ever deferred to a later book. The toughest per-book promise in publishing was honored in every single volume while the arc kept moving. Romance proves it can be done, and every other genre loses its excuse.

Isaac Asimov's Foundation (1942 to 1953, and 1982): a retrofitted arc with a checkable record

The Foundation stories ran in the magazine Astounding Science-Fiction from May 1942 through January 1950, were collected by Gnome Press as Foundation (1951), Foundation and Empire (1952) and Second Foundation (1953), and returned with Foundation's Edge (Doubleday, 1982), the first new book since 1953. Asimov wrote up how the series was built in a piece called "The Story Behind the 'Foundation'", which makes this a rare case where a long arc's construction can be checked. The arc was far more improvised than the finished shape implies.

The arc was agreed in rough outline before a word was written, and that is the only part that was planned. On August 1, 1941, travelling to see his editor John W. Campbell after reading Gibbon, Asimov had the idea. Roughly an hour's conversation produced the concept: a vast set of linked stories about a collapsing empire, a long interregnum (a dark age between empires), and a science able to shorten it. That science became Hari Seldon's plan, and the series question is simply will the plan hold? It passes all three tests. Everything else arrived later, story by story, years apart.

Three details from Asimov's account outweigh most of the craft advice on this subject. First, he withheld the ending of the first story on purpose, to force Campbell to order a sequel. He was open about it being a commercial tactic, and it was also the origin of the arc. The next installment promptly ran into a dead end and nearly killed the project, which is the normal outcome of that move and the reason a two-layer contract is needed at all. Second, he tried to end the series and was overruled. Campbell refused the ending, demanded a revised one, and extracted a promise of one more story. Any writer whose editor or publisher has demanded a volume the arc had no room for knows the feeling. Third, when he returned in 1982 after years of pressure from readers and his publisher, he deliberately left one item open so the series could continue. In 1982, with the first story dating from 1942 and the third collection from 1953, he still preferred options to promises.

The craft lesson is not that planning is unnecessary. The lesson is how tiny the upfront need is: one question and one span. Coherence in a discovered arc comes from new variables rather than new scale. Asimov's self-declared favorite story hinges on a factor the Plan could not foresee, and every retrofit can do the same. Look for the variable the premise left out instead of a bigger threat.

Agatha Christie's Curtain (1975): the last book, written first and locked away

Curtain was written around 1940, during the Second World War, and published by Collins Crime Club in September 1975, with a Dodd, Mead edition in the United States the same year. The book is Poirot's final case and Hastings's final appearance, the first time the two had been paired since 1937, and its ending is the series' ending. Sleeping Murder is Marple's equivalent, written in the same wartime period and published in 1976; Christie had died that January. Of every example on this page, this one is the most practical, and its subject is the unfinished arc.

Christie's series are thin-arc by design, and superbly so. Poirot and Marple get a fresh case each time, the world barely moves, and you can begin anywhere. The move worth copying is what she did with the endings. With bombing raids over London and her own survival uncertain, she wrote one final case for each detective and had the two books locked away. The Christie estate's site says one of them was a gift for her daughter, should Christie die in the raids. They stayed under lock and key from around 1940 until 1975, more than thirty years.

If you have not read Curtain, skip the next paragraph. The twist has been in print since 1975, but a twist is still a twist.

Poirot dies, and the manner of his death reverses the moral stance he held across his whole series. His death was reported as news: The New York Times ran his obituary, written by Thomas Lask, on its front page on August 6, 1975. He is the only fictional character ever to be treated that way.

Look at what the banked ending achieved. From roughly 1941 on, both of her great series were immune to being left unfinished. Her own fate no longer mattered to the ending, because the ending existed, it was her own, and the right people held it. Each intervening volume stood alone, so a reader who quit at any point lost nothing. The ending arrived in 1975, thirty-four years on from 1941, and it cost rather than explained, which is why it reached a newspaper's front page.

There is a lighter version that takes one afternoon. Write the ending of the final book now, two pages of plain prose, and put it in the continuity bible. Nobody reads it, and it is not a draft. What it does is make the arc answerable, the first of the three tests, and turn "how will this end?" from a worry into a document you can revise.

Tamsyn Muir's The Locked Tomb (2019 onward): a trilogy that became four books

Of all the examples here, this one is the most recent and the most useful for anyone mid-series now. The Locked Tomb was planned as a trilogy: Gideon the Ninth (Tor, 2019), Harrow the Ninth (2020) and a finale called Alecto the Ninth. In July 2021 the publisher announced it would be four books. Muir told Constance Grady at Vox, in an interview published January 27, 2022, that the finale's planned first act had reached around 140,000 words, a novel's length, before the finale had properly started. That act was published as Nona the Ninth (Tor, September 13, 2022). Alecto the Ninth kept its title and was pushed back.

The case teaches two lessons that point in opposite directions. The first: the number of volumes changed and the planned arc did not. The question is unchanged, the ending is unchanged, only the packaging moved. The alternative would have been cramming a hundred and forty thousand necessary words into an opening act and producing the kind of finale readers call rushed. When the plan says four and the material says five, trust the material. Clinging to the planned number confuses the arc with the schedule. There was a cost attached, and it was a real debt: readers waiting for a conclusion learned, mid-wait, that an extra installment came first. The fair way to handle that debt is the one Muir took. The unplanned volume has to be a whole book and never a holding pattern.

The second lesson is unresolved. The finale had not been published at the time of writing, and its announced dates had moved more than once; check the current position yourself. None of that is criticism. Long books take the time they take. What the case shows is the clearest live example of building for a series that may never finish. All three published volumes close their own stories, startling closes included. They reward the reader now rather than saving everything for later, and an arc of the same length with deferred endings would be far worse off.

An everyday scenario: a mountain rescue series stuck at book three

Picture a writer two books into a planned four-book crime series. Her lead is Rhiannon Kewley, who heads a mountain rescue team in a national park. Book one gave readers a crag and a corpse beneath it. Book two gave them a hiker reported missing who wasn't. Sales were modest and the reviews were good. Book three has stalled at eleven thousand words, and she cannot say why.

She runs the paid-and-moved test on the published books. Book one paid off the crag death, which is fine. And it moved the series question from nothing to a state the reader can name: something is wrong with the account of the callout where Cai Nollan died, because a page in the team's callout log is in a different hand, and the reader saw it in chapter nine. Book two paid off the hiker, also fine. Its movement, stated with no flattery, comes to this: two people who know something were introduced, and nothing is known at the end of book two that was not known at the end of book one.

So the diagnosis is clear. Book two's stillness is why book three is stuck. The question sat static for a whole volume, and she has been trying to write the third act of a middle that has no second act.

The fix is repair one, promoting a sub-answer. The identity of the person who rewrote the log page was reserved for book four. She promotes a piece of it. Book three will settle when the page was rewritten: after the inquest, not before. Small as it looks, that is a real answer. It spoils no ending, it changes what the reader believes, and book three is now about something other than waiting. After that comes the bookkeeping, and the audit she just ran is the real takeaway.

Try it yourself

Two procedures follow, and each one yields a document rather than a feeling. Procedure A, steps 1 to 10, takes about an hour and is for a series still in planning or one volume in. Procedure B, steps 11 to 20, takes an afternoon and is for a series stuck midway.

  1. Write the series question as a question. One sentence, ending in a question mark. If it takes two sentences, you have two questions, which is fine: number them, and flag the one the final book answers.
  2. Write the answer. One sentence, plain, and private. If you cannot write it, the question has failed the answerable test, so narrow it until you can. "The protocol falls, and the person who breaks it loses the thing it was protecting" is an answer. "Magic turns out to be more complicated" is not.
  3. Write the refusal. Describe the no-ending in one sentence. If there is no version of it a reader would accept, the question is decoration and the arc carries nothing.
  4. Write the question as a reader would say it after book one only. If that version is far vaguer than your own, book one is under-planted. If the two are identical, check that no character actually said it out loud.
  5. List the key facts. Which facts from book one does the answer need? Names, dates, ages, who was present, the rule. Write them down, then verify each one on the page. Roughly half the time, one of them exists only in your notes.
  6. List every running clock. An age, a term, a date, an anniversary, a prison sentence, a pregnancy, a lease. Each one is a promise about elapsed time, and each one is a trap waiting in book four. Start clocks deliberately or not at all.
  7. Sort the plants into options and promises. List every item in book one that looks significant. Beside each, write the number of fair payoffs it has. One means a promise: decide now whether you are keeping it. Three means an option: leave it alone, and do not explain it in book two.
  8. Write the ending. Two pages of plain prose covering the last chapter of the last book. Nobody will read it and it is not a draft. Put it in the continuity bible, dated today, and expect to rewrite it.
  9. Name the one rule you will not break. The rule the answer rests on. Post it where you can see it.
  10. Stop. Books three through six stay unoutlined.
  11. Open the published books and start a promise register. Procedure B starts here, and it is the reason this element exists. Work from the books themselves, never from memory, because memory flatters. Build a register with one row per published volume and five columns, headed Volume, Paid off (its own story), Moved the question to, Promised, not yet delivered and Planted, still live.
  12. Fill the first three columns from the books. Each volume's last two chapters plus its first two are enough to tell you what it resolved and where the big question was left. Write the "moved to" entry as a state, never as an event: the reader now believes X.
  13. Read the blanks. A blank in column two or column three is a diagnosed fault. Column two blank means an unresolved volume, which explains the cooler reviews. Column three blank means the arc did not move, which locates the sag. Both blank is the book where you lost readers.
  14. Fill column four with no flattery, then count it. Every question opened and not closed, including the small ones: the unopened letter, the character who is due back, the minor figure's hint nobody followed. The count matters more than the contents. More than about a dozen across four books is an overdraft, whatever the quality of the volumes.
  15. Fill column five. Everything planted and not yet fired. Beside each, write the number of fair payoffs it still has, as in step 7. Zero means the plant is quietly dead: fire it or stop mentioning it. The Foreshadowing page has the test for a single plant.
  16. Add elapsed time to each row. How much story time the volume covers, and the gap between it and the next. Two simultaneous volumes, or a run of books that each span four days, is the commonest cause of a static series. This step usually produces the finding that reshapes the next book.
  17. Choose one promotion from column four. A reserved question that can be answered in the next volume without spoiling the ending. One promotion, and only one. Answering everything at once reads as panic, so three is too many.
  18. Choose one closure from column four. A thread that has sat static for two volumes. Close it or kill it. Not deciding is what has been happening.
  19. Reread your private answer from procedure A. It has drifted. Rewrite it to fit the series you have published rather than the one you planned, and note the cost in column five, which is usually one or two plants that are now dead.
  20. Write the next volume's two sentences before writing the volume. "This book pays off ______." "This book moves the question from ______ to ______." Pin them up. Any scene that serves neither sentence is a scene to interrogate. The sheet below gathers both procedures on one page, so copy it into the continuity bible and fill it in.

Series question: __________ Answer (private): __________ Refusal (the plausible no): __________ Rule I will not break: __________ Running clocks: __________ This volume pays off: __________ This volume moves the question from __________ to __________ Promised and not yet delivered (count): __________

Common mistakes and misunderstandings

Treating the arc as a longer plot. An arc is a question with consequences hanging from it. It sits beneath four or ten complete stories and does not replace them. Thinking of it as one giant plot sliced into volumes produces books that end mid-sentence, with the real structure floating somewhere above them. The Plot page covers the chain of cause and effect inside a volume. The arc is the chain between volumes.

Writing the advance-only volume. The commonest breach of the contract. It feels productive while you write it, because a great deal happens and the arc moves a long way. It reads as half a book. No amount of arc progress compensates for an unresolved volume, and the reviews will be specific about it.

Writing the resolve-only volume. The second commonest breach. It does less damage per book and more over time. Four self-contained volumes with nothing building between them are a format, and a reader who notices treats the next book as optional. The fix is small: one irreversible change per volume is enough.

Confusing the theme with the question. "This series is about how power corrupts" cannot be answered, so it cannot carry an arc, and a series built on it drifts. Keep the theme; the Theme page explains how it unifies the books. Then find the yes-or-no question underneath it.

Writing prophecies as promises. Prophecies, visions and predictions are promises in disguise. Any statement about the future is a published contract you can no longer edit. Write them as categories rather than names, choices rather than acts, shapes rather than sequences. Otherwise book five must obey the literal wording, however much the story would prefer something better.

Planting a detail with one meaning. A detail that can mean only one thing has already spent the ending. Of all the mistakes in series fiction it is the costliest, it happens in book one, and it usually comes from enthusiasm. Plant options instead.

Treating the sagging middle as a phase. The sag is a diagnosis, and persistence will not cure it. Treading water almost always means two things: the arc has not moved for a book, and the timeline has stopped. Both can be fixed in the current volume. Waiting for the next one fixes nothing.

Expecting the plan to prevent a dead middle. A plan prevents aimlessness. It does not prevent a dead middle, and it can cause one. An over-detailed plan produces scenes that exist to satisfy the document. When a chapter is defended by the outline rather than by what someone in it wants, the plan is the problem.

Delivering an answer without a bill. Revealing is the easy half of a series ending. Making somebody pay is the hard half. Most revelations lose to the private theories readers have built. A cost competes with nothing.

Underestimating how series end. Sales, health, interest and life all stop series, far more often than writers plan for. A set of books whose volumes each resolve keeps its value even if the arc stays open forever. A set with every ending deferred becomes worthless the day it stops.

Mistaking a thin arc for laziness. The detective and children's series that run for decades carry very little arc, and they do it deliberately. A thin arc is the only structure that lasts forty years without buckling. Plan direction and drift rather than a question, and stop apologizing for it.

Related elements