Back to the table

46 Intermediate Ag Agency Character

Agency

The character makes choices that change the plot.

What it is

Agency means the character causes what happens. She makes a choice, and the story changes because of it. That's the whole idea, and it's narrower than it sounds. Courage doesn't give a character agency. Neither does skill, or power, or a defiant speech. She has agency only where a decision of hers is the reason something happened next.

You can't see it by reading a chapter at a time. Agency lives in the joins between scenes, in what causes what. Read one chapter and you see a character feeling things and doing things. Read the chapter list and you see who moved the plot. That's why the test for it is done on paper.

The Protagonist page uses agency as the test that separates a lead from a passenger. This page goes further: how to measure it, where it can fairly thin out, and what to do when it's gone.

Here's the test. Write a plain summary of your plot in ten to fifteen sentences, with no adjectives. Then delete the protagonist from the summary, and only from the summary. Rewrite each sentence as if she had never existed, and let the other characters do what they were going to do anyway.

Now look at what survives. The conspiracy still comes out, because the journalist was already digging. The brother is still saved, because the doctor was already on the way. The war still ends, because the treaty was already drafted. If most of your summary reads like that, you've found the problem. Your lead was there for everything and reacted to everything. She caused none of it.

The test is blunt on purpose. It says nothing about how good the writing is or how likable she is. It measures one thing: how much of the plot rests on her.

Not every choice counts. A decision carries weight when it passes three checks:

Run those three checks on your turning points and you get a number instead of a feeling. In revision, a number is what you need.

A common piece of bad advice says the protagonist must always be proactive. Follow it and you get a lead who is exhausting to read, in a book where nothing ever lands on her from outside. Reacting is fine. Reacting is the raw material. Something happens to the character, she feels it, and then comes the part that matters: she decides what to do about it. Cause and effect is the chain. Agency is the character's hand on it. Drafts go wrong when the character reacts and then stops. The feeling arrives, beautifully written, and leads to no decision, so someone else has to deliver the next event.

A reactive opening is normal. A reactive climax is fatal. Almost every novel starts with something happening to the lead: she's fired, bereaved, drafted, jilted, robbed, or chosen. The inciting incident is, by definition, something she didn't arrange. The first act exists to convert her. Somewhere in it, being acted on turns into acting, and she commits to a course she could have refused. The climax needs the opposite. It's the one scene where the reader has to watch the outcome hang on a decision of hers. A climax she only reacts to is the most expensive structural fault a draft can have, because it drains the meaning out of everything before it.

You measure agency twice, and the two results can disagree. At scene level it works like a scene goal. The viewpoint character wants something in this room, goes after it, and the scene ends with a result she partly caused. A book can be full of scenes like that and still fail the removal test, because none of the small wins add up to a decision about the plot. That's the pleasant, busy draft that critique partners call well written, and nothing more. At book level you count the major turns that trace back to her. A novel has eight to twelve of them. How many exist because she decided something? There's no correct ratio, and anyone who gives you one is guessing. The direction is clear, though. The further you are from the ending, the more room you have for events she didn't cause. The last two turns should be almost entirely hers.

The reverse problem is rarer, but it happens. A lead can own every big turn and drift through the scenes in between, deciding nothing for pages at a time. Book-level agency with no scene-level agency reads like a summary. Scene-level agency with no book-level agency reads like a diary.

One more thing agency is not: competence. A character can choose badly, again and again, and still have total agency. A bad choice is often the strongest beat you have, because it gives you the consequence and the flaw in one move. The try-fail cycle is made of a character acting and being wrong. Writers who mix the two up water down their leads. They want her to be admirable, so they make her careful, and a careful character holds back and lets other people decide. Don't try to make her good at things. Make her the reason things happen, including the bad things. Agency isn't power, either. A queen with an army who does whatever her council advises has less of it than a prisoner who decides what to say. That case comes up often enough to get its own example below.

Why it matters

Readers never name this fault. They report its symptoms, and the symptoms point somewhere else. When a lead has no agency, beta readers say the book is slow, the middle drags, and they didn't connect with her. Each note sends you to the wrong fix. Slow gets treated by cutting. Dragging gets a subplot. Not connecting gets more of her feelings, more of what she senses, more backstory. That last one makes it worse, because now the reader spends even longer inside a person nothing depends on. Learn to test for agency directly. A wrong diagnosis costs you a whole revision pass spent on prose.

The fix is cheap, which is rare for a fault this serious. Most structural problems cost you scenes. This one usually doesn't. You keep the same events, in the same order, at the same length, and rewire them so the character causes them. She works out where the meeting is instead of being taken there. She turns down the help and goes alone. She's the one who opens the envelope. You aren't writing new chapters. You're changing whose choices the existing chapters come from. If you write to a schedule, that should matter a lot, because it's one of the few serious faults you can fix in two weeks.

Agency holds the middle up. A sagging second act is almost always a stretch where the protagonist stopped deciding and started just being there: traveling, learning, being briefed, being trained, waiting for the antagonist's next move. Pacing looks like the culprit. Agency is the cause.

It's what readers pass on. When the lead drove the book, a reader can tell a friend what she did. When the lead was carried, the reader can only describe what happened around her, and a description of a situation doesn't get passed on. Across a series this adds up. Readers come back for a lead whose decisions they enjoy watching.

Real-world examples

The carried protagonist, and the century-old name for it

In an unpublished manuscript, this fault rarely comes from a lazy character. Far more often the lead is surrounded by people more decisive than she is.

It's easy to build by accident, because each of those choices makes its scene easier to write. The mentor knows what to do next, so he says so. The best friend has the contact. The love interest owns the car. The antagonist has a plan, which neatly fills the next three chapters. Each of those is a sensible piece of supporting cast design. Together they leave the protagonist nothing to do but agree. A mentor is the usual culprit, and a capable sidekick comes a close second.

You can spot it in the dialogue, where the lead's lines are mostly questions and the answers always turn out right. You can spot it in the chapter list, where the verbs next to her name are is told, is taken, is warned, is shown, is saved.

Genre fiction noticed this long ago. Critics of mystery novels have a name for one version of it: the Had-I-But-Known school. It's credited to Mary Roberts Rinehart and The Circular Staircase (1908), and it was named after a satirical Ogden Nash poem of 1940. Its trademark is a narrator who looks back and tells you what she might have done to prevent the disaster. The crime drags on because she failed to act, and regret takes the place of a decision. People were calling it a weakness within about thirty years of the form appearing.

The fix is almost always to reverse the direction of a request. Instead of the mentor telling her where to go, she works it out and tells him. Instead of the friend producing the contact, she pushes him into giving up a contact he didn't want to share. Same information, same scene, with the cause running the other way.

J.R.R. Tolkien's The Hobbit (1937): a carried opening, a converted middle, and a climax he sleeps through

George Allen & Unwin published The Hobbit on September 21, 1937. Its first act shows how much passivity a reader will forgive at the start of a book.

Bilbo Baggins doesn't really decide to go on the quest. Gandalf arranges it. He marks Bilbo's door, fills his house with dwarves, and maneuvers him into a contract. For several chapters Bilbo is luggage: carried along, then rescued from trolls, from goblins, and from wolves. If the book stopped before Mirkwood, he'd fail the removal test completely.

Then he converts, and he converts through decisions. In the spiders' clearing, he chooses to fight for the dwarves instead of slipping away while invisible. He plans the barrel escape himself. He chooses to go down the tunnel to the dragon alone. And in the book's most interesting act of agency, he steals the Arkenstone and slips out of the Mountain to hand it to the men and elves laying siege to it. He does it against the interests of his own party, knowing it will look like treachery. Thorin sees it exactly that way and banishes him. You won't find a purer form of agency: a costly decision, made alone, that changes the political situation of the whole novel.

Then comes the part to study hardest, because it's a warning more than a model. During the Battle of Five Armies, Bilbo puts on the ring, a stone strikes his helm, and he's knocked out. He lies there, invisible and unconscious, through the whole climax, and he wakes up to be told who won. The battle is decided when the eagles and Beorn arrive. That's help from outside, arriving in time, and it owes nothing to any choice of his.

Tolkien knew about this shape. In his 1947 essay "On Fairy-Stories", based on a lecture he gave in 1939, he coined the word eucatastrophe for the sudden joyous turn, the good catastrophe. He saw it as the mark of the fairy story, the way tragedy is the mark of drama. He defended it as a matter of form, and the eagles are the example people quote most.

So a reactive opening can survive at almost any length, as long as the conversion is real and the reader can see it. A rescue at the climax can survive too, in a story that has told you it's a fairy tale and isn't making a modern novel's promise. In a contemporary thriller, romance or mystery, it reads as the author solving the problem instead of the character, and readers put that in the one-star reviews.

Three novels that break the rule on purpose: Wells (1898), Vonnegut (1969), Williams (1965)

The rule isn't absolute, and pretending it is just makes it easier to argue with. Here are three widely read novels whose protagonist mostly doesn't drive the story, and none of them is a failure.

H. G. Wells's The War of the Worlds ran as a serial in Pearson's Magazine in 1897, and William Heinemann published it in hardcover in 1898. The narrator has no name, and over the whole invasion he causes almost nothing. He runs, hides, watches, nearly gets killed, gets trapped in a ruined house, and survives. The ending is the most famous thing about the book, so here it is: the Martians are killed by Earth's bacteria, which they have no immunity to. Nobody defeats them. That's the argument the novel is making. On this scale, human agency doesn't matter, and a narrator who saved the world would ruin the book.

Kurt Vonnegut's Slaughterhouse-Five, published in 1969, goes further and says it out loud. Billy Pilgrim is unstuck in time, shuffled between Dresden, Ilium and Tralfamadore without choosing any of it. In chapter eight, the narrator says the story has almost no dramatic confrontations, because its people are "listless playthings of enormous forces". One effect of war, he says, is that people are discouraged from being characters. The missing agency is the book's whole point. A Billy Pilgrim who made decisions would be a different book, and a much smaller one.

John Williams's Stoner is the subtler case, and the most useful. Viking Press published it in 1965. NYRB Classics reissued it on June 20, 2006, and after that, decades after its author's death, it became an international bestseller. William Stoner gives in to most of his life: a marriage that turns cold, a career that stalls, a daughter he slowly loses, a love affair he doesn't fight for. Almost all the book's heat comes from the one thing he refuses to do. He blocks a favored graduate student he believes is a fraud, and he keeps blocking him, at a cost he pays for the rest of his career. A novel in which he accepts almost everything turns on one long, stubborn act of will.

That shows where the line really falls. These books aren't exceptions to the rule so much as proof of what agency buys you, because they show what a book has to do without it. Take it away and you can still write something powerful. But the power has to come from somewhere else, usually an argument the book is making about powerlessness, and you have to do it on purpose. Jane Alison's Meander, Spiral, Explode (Catapult, 2019) is the serious study of those other engines. It looks at how pattern, repetition, spirals and cell structure can hold a narrative together where decisions don't. Read it before you decide your quiet novel is exempt. The question to ask isn't "does my lead have to drive?" It's "if she doesn't, what is holding this up instead?" If the answer is nothing, you don't have a counter-example. You have a passive draft.

Herman Melville gave the shortest version of the difference in 1853, when "Bartleby, the Scrivener" came out in two parts in Putnam's Monthly Magazine that November and December. Its title character does almost nothing except refuse, in five words, again and again. Refusing is a choice. Making no choice at all is something else.

Agency in a narrow box: Solzhenitsyn (1962), Whitehead (2016), Gaiman (2002)

The excuse writers reach for most is that the character is boxed in. It's almost always wrong. Prisoners, children, the powerless, the very ill: the narrower the choice, the heavier each choice weighs. These aren't exceptions. They're the strongest agency stories there are, because the reader can see the whole cost of every decision.

Aleksandr Solzhenitsyn's One Day in the Life of Ivan Denisovich appeared in the November 1962 issue of Novy Mir, and it covers a single day in a Soviet labor camp. Ivan Denisovich Shukhov controls nothing: not where he works, not what he eats, not whether he's punished. What he has is a string of tiny decisions, and every one of them is real. He decides whether to report sick and how hard to push it. He decides to sneak a second ration at lunch. He decides to smuggle in a scrap of metal to make into a tool, knowing what a search would cost him. He decides to do the bricklaying well, which nobody asks of him, and it's the one thing in his day that's truly his own. He does small favors for another prisoner in exchange for a share of a parcel. At the end he counts the day as almost a good one, and the reader can add up why, because every item on the list is something he chose. His whole room to act is the width of a bunk, and the novella is still one of the books most packed with agency on the shelf.

Colson Whitehead's The Underground Railroad was published by Doubleday in 2016. It won the National Book Award that year and the Pulitzer Prize for Fiction in 2017. Its protagonist, Cora, is enslaved on a Georgia plantation. When Caesar suggests escaping, she refuses. Later she chooses to go. That order is deliberate. The refusal shows what the decision costs, so when the decision comes it's the biggest act in the novel, and everything after it grows out of it.

Neil Gaiman's Coraline does the same job for a child protagonist. It was published in 2002, won the Hugo and Nebula awards for best novella in 2003, and won the 2002 Bram Stoker Award for best work for young readers. Coraline has no power at all, in either world, so her agency is one decision made twice: she goes back through the door, frightened, to get her parents. The whole book is built so you can see that choice as a choice. That's why it's frightening and not just spooky.

The very ill and the very old deserve the same treatment, with no discount. A character who can't leave her bed can still decide what to tell, whom to forgive, what to sign, what to refuse, and whom to protect by lying. Those decisions change the plot, and they're usually more interesting than a chase. If your ill, imprisoned or child protagonist has no agency, don't blame the constraint. You haven't worked out yet what she still controls.

John Kennedy Toole's A Confederacy of Dunces (1980): every decision wrong, every event his

Toole died in 1969 with the manuscript unpublished. His mother, Thelma, spent years pushing it on people until she reached the novelist Walker Percy. Louisiana State University Press published it in 1980, and it won the Pulitzer Prize for Fiction in 1981.

Ignatius J. Reilly is the best proof you'll find that agency and competence have nothing to do with each other. He's lazy, grandiose, self-deluding and wrong about everything, and he's also the cause of the entire novel. His mother's car crash creates the debt that forces him to work, and that's the one thing done to him. After that, every disaster in the book is his. He takes the job at Levy Pants and wrecks it. He throws away whole drawers of files. He sends an insulting letter to a complaining customer on the company's stationery, over the owner's signature. He tries and fails to stir the factory workers into a protest for better pay. He takes the hot dog cart and eats the stock. He launches a political scheme to impress a woman in New York, and he's thrown out of the party where he starts it. Not one of these decisions is good, and every one of them makes trouble the rest of the cast has to deal with.

Fair warning: the next part gives away the ending. Ignatius guesses that his mother is arranging to have him committed. When Myrna, the woman he's been writing to, arrives from New York, he urges her to drive him away at once, and they pass the ambulance on the road. Somebody else supplies the car. The decision to get into it, and the timing, are his. That's the difference between a rescue and an escape. Hold on to it, because a great many climaxes can be fixed with it. Help may arrive. What help must not do is decide.

An everyday scenario: the protagonist three other people were driving

Say you have a finished draft of eighty-eight thousand words, and two beta readers have given you the same unhelpful note in different words: strong writing, lost interest around the middle.

Your protagonist is Verity Kehoe. You know her very well, and you can hear her voice. So you run the removal test on your own synopsis. You delete her name and let everyone else do what they were going to do.

Most of it survives. The investigation still happens, because Ambrose Tay, the retired inspector, was already reopening the case and has the files. The safe house is still found, because Freda Okonkwo makes the call. The confession still comes, because the suspect's brother had decided to talk and needed someone to talk to. Verity is there for all three. She causes none of them.

The synopsis is rough, so next you build a decision ledger, one line for each of your twenty-six chapters: what did she decide, and who decided it? Ambrose's name is in the second column eleven times. Freda's is there five times. Verity's own name appears four times, and three of those are in the first act.

None of this is a prose problem, and going deeper into her head won't fix it. The fixes are conversions, and they cost no new chapters. Ambrose doesn't reopen the case. He refuses to, and she spends two chapters getting the files out of him, which gives her something to want in a room where somebody is saying no. Freda doesn't find the safe house. Verity works out the address from something she noticed in chapter six and goes without telling anyone, which is why she's followed. The brother doesn't volunteer the confession. She has to offer him something she can't afford to give, and the reader watches her pay.

The word count hardly changes, and the events are the same events. It isn't the same book anymore. The ledger now reads eleven to four the other way.

Try it yourself

Do this with your chapter list open. Allow an hour for a full novel, and write everything down. The exercise fails if you do it in your head, because you know what you meant, and the page only has what she caused.

  1. Run the removal test on paper. Write your plot as ten to fifteen flat sentences. Then rewrite the whole thing with the protagonist deleted, and let the rest of the cast do what they were going to do. Count the sentences that survive unchanged, and start from that number, not from your gut feeling.
  2. Build the decision ledger. Make one row per scene or chapter, with three columns: what changed in this scene, who decided it, what it cost the decider. Fill the second column with names. The worst thing you can find there is somebody else's name, over and over, in a long unbroken run.
  3. Trace the information. Go through every important thing the reader learns and mark it sought or delivered. Delivered means it just arrived: overheard, mailed, confessed, explained by an expert, found in a handy diary. Sought means she went after it and gave something up to get it. If most of the marks say delivered, you have an agency problem, even if the lead is busy on every page.
  4. Count the noes. Read your lead's dialogue looking only for refusals, conditions and counter-offers. A character who only ever says yes is a tool for other people's plans. Rudeness doesn't count here. Look for the moments where she says no and the plot has to bend.
  5. Test the climax on its own. Write one sentence: the ending comes out this way because [Name] decided to ______. If you can't finish it, or if the truthful answer is "to be there", stop the rest of the exercise and fix that first. Nothing else on this list matters as much. See climax.
  6. Convert, don't add. Take the three weakest rows in your ledger and use one of three moves on each. Reverse the request: she asks for the thing and is refused, instead of being given it. Move the discovery: she finds out instead of being told. Attach a price: she gets it, and it costs her something specific, named on the page. You're changing what causes what. The word count can stay where it is.
  7. Do the constraint inventory. If your protagonist is in prison, ill, a child, employed, poor, or boxed in some other way, list twenty things she still controls. What she says and doesn't say. What she notices and tells no one. Who she protects, what she refuses, what she does well when nobody is watching. Then pick three and give each one a consequence.
  8. Check the want is live. Agency with nothing behind it reads as random. If the ledger is thin, the cause is often further upstream, in motivation or in the gap between want and need. She isn't deciding because you haven't yet made her want anything enough to risk something for it.

Here's a block to paste into your project notes:

Sentences surviving the removal test: __ of __ Turns she caused / total turns: __ / __ Information sought / delivered: __ / __ Refusals in the whole book: __ The climax happens this way because she decided to: __________ Three conversions lined up: __________

Common mistakes and misunderstandings

The plot that would happen anyway. The basic case, and the reason the removal test exists. Everything in the book is caused by the antagonist's schedule, the mentor's knowledge or a friend's push, and the protagonist supplies reactions. The draft often reads well scene by scene, so it gets through several revisions before anyone spots the fault.

A lead who is busy but not deciding. Being active doesn't give her agency. She can run, fight, travel, investigate and argue for four hundred pages while every one of those actions was somebody else's idea. The ledger catches this. The page never will.

Agency mistaken for aggression. When writers are told their protagonist is passive, they often make her louder: more defiance, more sarcasm, more refusing to be handled. The chain of cause and effect stays exactly as it was. A quiet character who makes one decision that sends the plot a new way has more agency than a combative one who is still being steered.

The rescue at the climax. Cavalry, eagles, a sniper on a roof, a confession from a surprise witness, a villain who overreaches just when it suits the plot. You can keep any of these if the protagonist's choice is what puts the help in place or what makes it usable. Help arriving is fine. Help deciding isn't. The climax page covers the endings this ruins.

Answers that arrive in the mail. A quiet cousin of the rescue, and much harder to see. The reader learns the key fact from a letter, a dying confession, a handy file or a character who explains. Nothing is technically wrong, and the middle goes soft anyway, because the protagonist's whole job has become receiving things. Make her hunt down one of those facts and pay for it.

She never once says no. Agreeable protagonists are easy to write, because a yes always lets the next scene start. A character who never turns anything down has no preferences the reader can see, and readers take that as having no self. One well-placed refusal does more for a lead than a whole chapter of her thoughts.

Waiting for the reactive opening to fix itself. Starting with something done to the character is right and normal. Letting it go on past the first act isn't. If you can't name the scene where she stops being pushed around and starts choosing, the reader can't either, and the sag they report in the middle starts right there.

"But she's a prisoner, she can't do anything." The exemption writers claim most, and the least valid. A narrow choice is still a choice, and a tight box concentrates agency instead of removing it. If you really can't find a decision open to your captive, ill or very young protagonist, research her situation more. You haven't found a structural exemption.

Treating agency as a character trait. Agency belongs to your plot's wiring. It isn't an adjective. You can't fix it in the character sheet or by rewriting her introduction, because it only changes when you change which events come from which decisions. It's a structural repair, and it belongs at the start of a revision, not at the end.

The decision made off the page. She does choose. You know she does, and the next chapter follows from it. But the choosing happened in the white space between scenes, and readers only give credit for what they see. If the plot hangs on a decision, the reader should be there when she makes it, ideally with someone in the room arguing the other way. That's where internal conflict and this element meet.

Faith in a strong voice. A strong voice will cover this for a while. A distinctive narrator can carry a reader through a whole first act of drifting. That's why drafts led by voice get the warmest early feedback and the most complaints about the middle. The voice hasn't stopped working. The reader has figured out that nothing depends on the person talking.

Related elements