The unreliable narrator
An unreliable narrator is a storyteller whose account the reader is meant to distrust, because the book quietly supplies evidence that contradicts the narration. The disagreement is engineered. A novel using the device hands the reader details the narrator reports without understanding, conceals, or invents, and lets a second version of events assemble underneath the first. Four kinds of narrator produce that gap: the mistaken, the naive, the evasive and the lying. All four work on the page for one reason, that prose offers no way round the voice. All four are in trouble the moment a film crew arrives, because a camera is a second witness.
Yes, the unreliable narrator is a literary device, and a structural one
Yes, the unreliable narrator is a literary device, and it is structural rather than decorative: it changes how the whole book is read, not how one paragraph sounds. The term was fixed by the American critic Wayne C. Booth in The Rhetoric of Fiction (University of Chicago Press, 1961), which defined the narrator as unreliable when the narrator's values and report diverge from those of the implied author standing behind the text. Booth's formulation is still the one literary criticism argues with, and it locates unreliability not in the narrator's character but in the distance between two levels of the book, the level that speaks and the level that arranges. A rogue is not automatically an unreliable narrator. A saint can be one. The test is structural, and the structure is what an adaptation either rebuilds or loses.
The device is older than the term. Henry James built The Turn of the Screw (1898) so that a governess reporting ghosts and a governess losing her grip fit the same pages, and Agatha Christie turned the technique into a mechanism in The Murder of Roger Ackroyd (1926).
The four types, sorted by what the narrator is doing to the reader
The four types divide by intent: what the narrator is doing to the reader, and whether the narrator knows they are doing it. Two of the four are innocent, and two are not, which is why they behave so differently when a screenwriter gets hold of them.
| Type | What the narrator does | The reader's job | A novel that runs on it |
|---|---|---|---|
| Mistaken | Reports honestly and interprets wrongly, or misremembers | Spot the misreading before the narrator does | Julian Barnes, The Sense of an Ending (2011) |
| Naive | Reports accurately but has no frame to interpret what is reported | Supply the frame the narrator lacks | Emma Donoghue, Room (2010) |
| Evasive | Knows, and steers the account away from it | Notice the shape of the hole | Kazuo Ishiguro, The Remains of the Day (1989) |
| Lying | Constructs a false account on purpose, for an audience | Work out who the account is aimed at | Gillian Flynn, Gone Girl (2012) |
Liz Jensen's The Ninth Life of Louis Drax (2004) runs on the naive type crossed with the mistaken. The boy narrating is 9 years old and unconscious, and he lists his own history of accidents in the flat, boastful register of a child who has settled on bad luck as the explanation. He conceals nothing. He lacks the categories that would let him see what the reader can see forming behind him.
The signals a novel plants so the reader can catch the narrator out
The signals are planted, not accidental, and there are five a reader can look for. A second character corrects the narrator in dialogue the narrator reproduces without comment. The chronology will not close, so dates or ages contradict each other across chapters. The register misfits the event, and something appalling is described in a tone reserved for something small. A physical detail recurs with a different value each time it appears. And the narrator defends a point nobody has attacked, which is the oldest tell of all.
Prose can plant these because prose controls the reader's whole supply of information. There is no channel round the narrator. The reader cannot look past the sentence at the room. This is the single fact that separates the literary device from its screen equivalent, and it is the reason the crossing to film is a rebuild rather than a transfer.
Why the device breaks when the novel becomes a film
The device breaks in adaptation because film adds a witness the novel never had. A camera reports independently of whoever is speaking, so the instant a narrator's claim is dramatised, the image either confirms it or exposes it. Prose has one channel. Film has at least three, image, sound and voice, and the audience trusts the image most. A voiceover that says one thing over a picture that shows another does not read as a character deceiving themselves. It reads as a mistake in the film.
Jensen gives the reader a boy in a coma talking continuously, in his own idiom, for most of the novel, with the doctor treating him supplying the only outside account. The reader is locked inside a 9 year old who cannot be corrected.
Alexandre Aja's 2016 version keeps the boy's voice as voiceover and relocates the story from France to the United States, but the camera spends most of its running time in the room with the adults, watching them from outside anyone's head.
Three specific failures follow. Voiceover is intermittent where prose narration is continuous, so a film narrator is absent for long stretches and cannot maintain a false version across them. Audiences hear voiceover as authorial rather than personal, closer to a caption than a confession, which makes deliberate lying by a voiceover feel like a cheat rather than a characterisation. And film has no equivalent of the unmarked omission: prose can leave a sentence out and nobody notices, while a cut is visible as a cut.
What an adaptation has to rebuild, and how far it usually travels
What an adaptation rebuilds is the gap, not the voice. Measured on the four points this reference uses everywhere, plot, character, voice and theme, an unreliable-narrator novel almost always travels furthest on voice and least on plot. The events survive intact. The person telling them stops being the only way in. Theme follows voice: a novel about the unreliability of a mind becomes a film about the events that mind was wrong about.
The successful rebuilds substitute a technique for the voice: falsifying an image and correcting it later, withholding a point of view and revealing it late, or running two accounts of one event against each other. Each is a director's device rather than a novelist's, and each buys back a portion of what the page did for free. The naive narrator survives best, because a naive voice over honest pictures produces the gap without anyone lying. The lying narrator survives worst, because film cannot lie to an audience without the audience feeling handled.