Trust No One: A Guide to the Unreliable Narrator

Spoiler Notice

This guide discusses how several famous narratives disclose withheld information. It therefore contains major spoilers for The Murder of Roger Ackroyd, The Remains of the Day, Lolita, Fight Club, and Memento. Read those works first if discovering their structures unaided matters to you.

Reliability Is a Relationship

An unreliable narrator is not simply a narrator who lies. The useful question is whether the account gives its audience sufficient reason to doubt some part of the narrator’s reporting, interpretation, or judgement.

Wayne C. Booth introduced “unreliable narrator” as a critical term in The Rhetoric of Fiction. In his influential formulation, a narrator is reliable when speaking or acting in accordance with the work’s norms—what Booth calls the implied author’s norms—and unreliable when not (Booth 158–59). The definition remains foundational, but it also opens a difficult question: readers must infer those norms from the work, and readers may not agree about them.

Later narratologists have therefore refined the concept. Dan Shen summarizes a rhetorical model in which narrators can fail along three axes: they can misreport events, misinterpret knowledge or perception, or misevaluate what matters ethically. They can also underreport, underinterpret, or underevaluate—giving an account that is insufficient rather than directly false (Shen 896–910).

This makes reliability a relationship among the narrator’s discourse, other evidence in the work, and the reader’s inferences. It is rarely an all-or-nothing property. A narrator may describe events accurately while misunderstanding their significance, or judge conduct persuasively while withholding a decisive fact.

Fallible Is Not the Same as Deceptive

Greta Olson distinguishes fallible narrators, whose limitations invite correction or supplementation, from untrustworthy narrators, whose values or purposes create greater distance and suspicion (Olson 93–109). The terms are not diagnoses. They describe how an account functions.

That distinction is especially useful for writers. A child who lacks political knowledge, a witness who sees only one room, and a criminal who edits a confession may all provide incomplete accounts, but the ethical and dramatic effects differ. Grouping them under labels such as “madman” or “naïf” can flatten those differences and encourage stereotypes about age, disability, trauma, or mental illness.

Unreliability also need not remain constant. A narrator can be trustworthy about physical action and evasive about motive. Their account may become more candid, less coherent, or differently biased as circumstances change. Design the pattern at the level of individual acts of narration rather than assigning the character one permanent setting marked unreliable.

An AuthZ Craft Framework

The following four patterns are a practical AuthZ framework, not a canonical taxonomy. They translate the theoretical distinctions into questions a writer can use while drafting.

Strategic Misreporting or Omission

The narrator knows more than the account initially reveals and manages information for a purpose. That purpose might be acquittal, admiration, control, privacy, or self-preservation.

Agatha Christie’s The Murder of Roger Ackroyd is the classic spoiler-heavy example. Dr Sheppard records Poirot’s investigation while using selective phrasing and omission to conceal his own responsibility for Ackroyd’s death (Christie). The solution works because the narration does not become random after the reveal: earlier sentences can be reread as carefully limited statements.

For writers, the standard is not whether the narrator has told the whole truth. Few narrators do. The question is whether the eventual disclosure produces a fair and intelligible second reading. Plant evidence that was available but easy to classify incorrectly.

Limited or Contested Perception

The narrator reports an experience whose cause or accuracy remains uncertain. Other details may conflict with that perception, but the work need not supply a single clinical explanation.

In Poe’s “The Tell-Tale Heart,” the narrator opens by asking why the listener calls them mad and insists that heightened hearing proves otherwise. The later account of a dead man’s beating heart can be read against that insistence (Poe). The story creates distance through the narrator’s own language; a craft analysis does not need to diagnose either narrator or author.

When using this pattern, establish what the character perceives, what the surrounding narrative can independently confirm, and what remains genuinely undecidable. A late label for a condition is not a substitute for a coherent point of view.

Self-Protective Interpretation

The narrator may report many events accurately while interpreting them in a way that protects an identity or life story. The gap appears between evidence and explanation.

Stevens in The Remains of the Day recalls his service to Lord Darlington through ideals of dignity and professional loyalty. His own memories nevertheless disclose the costs of emotional restraint and his employer’s political errors (Ishiguro). Readers may disagree about the degree of Stevens’s awareness; the important craft feature is that the counter-account is assembled from material he supplies.

This pattern benefits from repetition with variation. Let the narrator return to a governing value—duty, innocence, success, loyalty—while incidents gradually make that value harder to apply without qualification.

Distorted Evaluation

A narrator can accurately recount conduct while using euphemism, aesthetic display, or moral reasoning to solicit the audience’s approval. The unreliability lies primarily in evaluation rather than event.

In Lolita, Humbert Humbert’s ornate first-person account attempts to redescribe his abuse of Dolores Haze as romance. The novel also preserves evidence of her fear, grief, and constrained choices within the account he controls (Nabokov). It is safer to say the prose invites readers to assess that conflict than to claim it inevitably seduces every reader or makes every reader complicit.

For writers handling abuse or coercion, counterevidence must be legible. Formal brilliance does not automatically create ethical distance. Consider whose pain is made visible, what the narrator normalizes, and whether the work supplies enough resistance to the narrator’s preferred vocabulary.

Signals Readers May Use

No single signal proves unreliability. Several together can make a counter-account available:

• a contradiction between two versions of an event;

• a gap between the narrator’s judgement and the conduct described;

• testimony, documents, or images that resist the narrator’s account;

• conspicuous omissions or abrupt transitions;

• verbal habits that repeatedly minimize, inflate, or rename harm;

• knowledge the narrator could not possess, or information they should possess but avoid;

• a pattern in which predictions and explanations repeatedly fail.

These cues do not force one response. Cultural assumptions, genre knowledge, rereading, and tolerance for ambiguity affect how readers weigh them. Workshop feedback is therefore evidence about particular readers, not proof of a universal effect.

What Unreliability Can Do

Unreliable narration can support suspense, retrospective reclassification, characterization, comedy, ethical conflict, or uncertainty. It does not guarantee any of them.

A mystery may use omission to make earlier scenes change meaning. A character study may allow self-justification to reveal values the narrator cannot examine. A satire may rely on the audience recognizing a gap between praise and the object praised. Another work may refuse to confirm which of several accounts is correct.

Avoid claiming that the device necessarily makes readers “active,” produces a deeper emotional bond, or trains them to detect deception outside fiction. Those are possible responses that would require reader evidence to generalize. As craft goals, they are hypotheses to test.

Designing the Pattern

Choose the Axis

Decide whether the narrator’s account is unreliable about facts, knowledge, or values—and whether it is wrong or merely insufficient. Write the distinction in one sentence before drafting.

Preserve a Usable Baseline

If nothing in the account can be provisionally trusted, the audience has little material from which to infer an alternative. Decide what remains stable: chronology, sensory detail, another character’s testimony, a document, or a repeated action.

Build Counterevidence

Place details that support a second account before the disclosure. Counterevidence can be quiet, but it should not depend entirely on information introduced at the last moment.

Separate Character from Diagnosis

Do not use a mental-health label as shorthand for arbitrary perception, violence, or deception. Research any condition represented, distinguish symptoms from plot convenience, and remember that a narrator can be unreliable without being ill—and mentally ill without being deceptive or dangerous.

Test More Than Surprise

Ask early readers what they believed, when that belief changed, which evidence they used, and what remained ambiguous. Also ask whether the narration obscures harm the work intends them to recognize. A successful first-read surprise can still fail on rereading or produce an unintended ethical emphasis.

Unreliability on Screen

Film and television do not simply replace a literary narrator with an objective camera. Voice-over can conflict with images; editing can restrict chronology; sound and point of view can align an audience with one character’s perception; later scenes can reclassify images that first appeared neutral.

Fight Club uses voice-over, subjective imagery, and editing to conceal that Tyler Durden is a projection of the unnamed narrator. Memento uses two intercut temporal movements: colour sequences proceed in reverse order while monochrome sequences move forward, limiting the audience’s context alongside Leonard’s impaired memory. The structure also exposes Leonard’s capacity to manipulate the evidence on which his future self will rely; the BFI aptly describes him as a narrator unknowable even to himself (Khaldi).

For screen craft, identify which channel carries the unreliable information. Is the character’s speech false? Is the image restricted to perception? Has editing withheld context? Treating “the camera” as the narrator can blur these separate decisions.

Trust as a Designed Question

The most useful unreliable narrators are not machines for delivering twists. They make trust divisible. Audiences may believe an event occurred but reject the narrator’s explanation, or accept an emotion while resisting the moral claim built around it.

Designing that relationship requires more precision than deciding that a narrator is a liar, innocent, unstable, or self-deceived. Determine what the account gets wrong, what it leaves unsaid, what evidence survives against it, and how much uncertainty the work intends to preserve. The result is not a guaranteed lesson in scepticism. It is a carefully built invitation to read one story against another.