Why the narrator’s reliability is a structural device

A butler describes a dignified career in loyal service. Nothing he says is false, and by chapter nine the reader has assembled a second account from the same sentences — one about a life spent not looking at something. The book never states the second account.

What everyone sees

Unreliability is read as a property of a character: this narrator is dishonest, or confused, or self-deceiving. That locates it in the wrong place. Reliability is a relation between the narrator and the structure surrounding him, and the structure is what supplies the reader with the material to disagree.

What is actually happening

Booth introduced the term by defining reliability against the implied author rather than against the facts — a narrator is unreliable when the work’s own norms diverge from the narrator’s, which means unreliability is constructed by the text rather than discovered in the character. Genette’s analysis of narrative levels supplies the machinery: the distance between who speaks and who sees can be manipulated independently, so a text can grant the reader perceptual access the narrator lacks while withholding any commentary. The reader’s suspicion is therefore engineered. It arrives as the reader’s own achievement, which is why it is far more durable than a stated judgement would have been.

Why it stays hidden

The engineering hides because the reader experiences discovery. Working out that a narrator is unreliable feels like penetrating a disguise, and that feeling is the point — a conclusion reached by the reader is held more firmly than one supplied. Nothing in the text takes credit, because taking credit would dismantle the effect.