Narrative craft
Unreliable narrator: doubt the voice, trust the story
Give readers a stable event, a motivated distortion and enough evidence to reach their own judgment.
An unreliable narrator gives readers reasons to question an account without making the entire story arbitrary. The useful tension is not simply whether someone lies. It is the distance between what a speaker says, what the scene lets us notice and what we can reasonably infer. Design that distance deliberately, and readers can revise their understanding while continuing to trust your storytelling.
This guide uses an invented glasshouse incident. Ada says she secured the door before a storm; her colleague finds it open the next morning. We will build a selective account around ordinary professional pride, not a diagnosis or a supernatural explanation. You can complete the exercise with a notebook and a draft. The aim is a readable human conflict, not a contest to defeat the reader.
An unreliable narrator can report accurately and still mislead
A narrator may distort an event, misinterpret a person or judge behavior by a self-serving standard. Those are different problems. Ada might invent locking the door, accurately report seeing her colleague near it but wrongly blame him, or admit leaving it open while insisting that nobody should complain. Decide which gap your story needs before making every sentence suspicious.
A limited viewpoint is not automatically unreliable. A speaker who says she could not see the latch is acknowledging a limit. A speaker who claims the door was secure despite never checking it is presenting an inference as a fact. Store norske leksikon describes an unreliable narrator as one who gives a false or misleading impression of events. That definition leaves room for more than an outright liar.
Nor must a flawed narrator be wrong about everything. Let Ada identify plants correctly, remember a colleague’s birthday and recognize cold damage. Her competence gives the account texture and makes the disputed claim specific. If every detail is equally false, readers have no ground from which to interpret the story. Uncertainty becomes noise rather than a meaningful difference between perspectives.
Write the incident once without Ada’s explanation
Start with a plain private record: Ada carried trays outside, propped the glasshouse door with a stone, removed the stone at closing and failed to test the latch. Wind opened the door overnight. Some seedlings were damaged. Her colleague Ben discovered them. These are the working facts of this exercise, not sentences that must all appear in the finished story.
Now write what Ada actually knows. She remembers moving the stone and hearing the door touch its frame. She does not witness the night’s weather or Ben’s arrival. Keep those limits distinct from what you, the author, know. Otherwise her account may reveal impossible knowledge in one paragraph and conveniently lose ordinary knowledge in another. That is a continuity fault, not an intentional narrative gap.
Finally mark what can remain undecided. Perhaps neither person knows exactly when the door opened. You can preserve that uncertainty while still establishing that Ada overstated her certainty. A complete solution to every physical detail is unnecessary if the emotional story concerns responsibility. Readers need enough evidence to judge the claim, not an engineering report on the latch.
Give the distortion something to protect
Ada has just asked to manage the glasshouse permanently. Admitting she skipped a check threatens the image of competence she wants Ben to support. That gives her selective account a purpose. She emphasizes how late she worked, how carefully she moved the trays and how casually Ben sometimes leaves tools around. None of that proves the door was locked, but it helps her feel deserving of trust.
The motive should affect the telling, not merely sit in a character biography. When Ben asks about the latch, Ada answers a different question: she was the last person doing any useful work. Her complaint may be sincerely felt. The reader can understand the grievance while noticing that it avoids responsibility. Give the evasive sentence a place in the conversation where it costs something.
Avoid making mental illness, disability or trauma a shorthand for dishonesty. Those subjects require their own careful research when relevant; this exercise does not need them. Shame, loyalty, ambition and an ordinary desire to be liked can all produce a slanted account. Choose a particular pressure and show its operation instead of assigning a label that supposedly explains every contradiction.
Let the scene resist the narrator’s verdict
Write a sentence that contains both Ada’s judgment and an observable action. She tells us Ben is eager to blame her, while he quietly puts a cloth around the damaged plants and asks whether the latch caught. The action does not establish that Ben is perfect. It simply gives readers material that does not fully match Ada’s verdict. They can notice the difference without an author announcing it.
Physical evidence can do similar work. Ada says the doorway was clear when she left, then mentions scraping mud from her shoes because she carried the last tray through it. Do not make that mud magically identify a culprit. Let it establish access or interrupt a too-simple claim. Each clue needs a modest, plausible meaning before it becomes part of a larger interpretation.
Use more than one kind of resistance: a remembered phrase, an object’s position, a consequence and another person’s response. If every clue is Ben saying Ada is wrong, the story depends entirely on choosing which speaker to believe. Independent details let readers test both accounts. Ben can be mistaken about a smaller matter without erasing the important evidence against Ada’s certainty.
Distinguish a withheld fact from an implausible thought
In a retrospective account, ask whom Ada is addressing and why now. She might be writing an explanation for a supervisor. Such a document can openly select information; its purpose shapes what she says. An immediate stream of thought creates different expectations. If she knowingly opened the door, avoiding even a passing thought about it solely to preserve a twist may feel like the author hiding behind her.
You can make omission visible without revealing everything. Ada counts the tasks she completed but stops before describing the final check. Ben asks a concrete question she refuses to repeat. The gap is part of the behavior, not a missing paragraph that appears only when the ending needs it. Let readers know there is something to examine, even if they do not yet know its extent.
Skrivekunst’s discussion of narratology separates reporting, interpretation and judgment, offering a useful conceptual distinction for this revision. Apply it to your own sentences: is Ada misstating an action, misreading a response or defending a value? You need not import theoretical terminology into the story. The practical benefit is knowing which part of the account your scene is challenging.
Make the changed understanding alter a decision
The revelation need not be a confession followed by a lecture. Ben can ask Ada to test the latch in front of him. It does not catch on the first attempt. Ada must decide whether to repeat her claim or demonstrate the check she omitted. The new understanding becomes an immediate action: she tests again, then agrees to change the closing routine with him.
Do not solve a relationship merely by solving the mechanism. Ben may accept the new routine but decline to endorse her promotion yet. Ada’s effort matters without cancelling the damaged plants or her evasiveness. A consequence keeps the ending from rewarding honesty with instant absolution. Alternatively, she may double down. In that version the reader’s knowledge changes while her behavior exposes its cost.
Look backward after choosing the ending. Does the early account contain enough material for the revised interpretation? Remove evidence that promises a different culprit unless that alternative has a real purpose. Keep the voice recognizably Ada’s after she changes her position. A narrator does not need to become a neutral encyclopedia the moment she accepts one uncomfortable fact.
Test the reader’s inference before polishing the voice
Give a fresh reader the scene without explaining which sentences are clues. Ask what they believed about the door at the beginning, where that belief changed and which detail caused the change. Then ask what they believe Ada wanted. Record the answers before explaining your plan. A correct guess about the door means little if the reader cannot understand why Ada shaped the account.
If the reader never questions her, strengthen the connection between a claim and contradictory evidence. If the reader doubts every sentence immediately, restore a few dependable observations and reduce repetitive defensiveness. If they feel cheated, inspect the omitted information and the chosen narrative distance. Adding more ominous adjectives will not repair a passage that deliberately conceals an inevitable thought.
You can keep this evidence record in ordinary notes. If you already draft in Fyrlo, place accepted events and Ada’s beliefs separately in the plot and character context, then compare one passage with a stronger or weaker evasion. Review the result yourself: Fyrlo is a planning and drafting aid, not a guarantee of continuity, originality, factual accuracy or publication readiness. A generated narrator can contradict the plan accidentally.
For the final pass, underline every claim of certainty and check what supports it. Read the dialogue aloud for evasions that sound like a person rather than a puzzle setter. Preserve the mundane observations that make the setting real. The reader should finish with a changed understanding of Ada, not the suspicion that any detail could have been replaced whenever the author wanted a surprise.