A transcript and a draft
Record ninety seconds of two people who know each other well, and transcribe it faithfully. The page will contain overlapping speech, sentences abandoned halfway, a scatter of um and you know, a run of mm and yeah from the listener, and a topic raised and dropped. It is accurate, it is close to unreadable, and it discloses less about the two speakers than a competent novelist discloses in five lines. Novelistic dialogue contains almost none of those features. By any measure of fidelity to the recording it is false, and readers experience it as lifelike.
That asymmetry is where this chapter starts. A description of a room is not expected to be an inventory, and no one defends one on the ground that rooms really look like that. Dialogue alone is habitually defended on the ground that people really do talk this way. They do, and they also produce greetings, weather and scheduling that no reader will sit through. What a reader registers as realism in dialogue is fidelity to a set of social facts about talk: that people want things from each other, that they seldom say so directly, that saying and meaning come apart, and that every turn places its speaker relative to the last. A reader who says an exchange “sounds real” is reporting that the wants under it are legible, not that the words would match a tape.
The common draft fault runs the other way from the transcript. It is fluent, clean and helpful. The household novel used throughout this book supplies one. Elena has driven across town to her brother Dov’s front door, not knowing that their mother changed the locks of the family shop in March. In the first draft Dov opens with “I know you want the shop key. I’ve been dreading this conversation,” and Elena answers that it is humiliating to have to ask for one. Within five turns Dov has explained the locks, his divided loyalty, his mother’s reasons for not explaining and his regret, and he apologises. Nothing in it is false to these people. It is inert, because every speaker says what they feel, nothing is withheld, and the reader has nothing to do. The worked analyses diagnose it and revise it.
The claim
The chapter’s governing claim is the following.
Two characters trading lines are a transcript. A dialogue exchange is a transcript with an engine under it: each speaker wants something from the other, wants it now, and uses the words to get it, so that the balance of information, standing or feeling between them is different at the end.
Three limits on the claim keep it usable. The want need not be large: I want you to admit you were wrong or I want to leave this room without answering will drive an exchange, as The Purposes of a Scene (Chapter 1) argues of scene goals, provided the want is legible and resisted. The want need not be conscious, since the gap between what a character pursues and what she thinks she pursues is one of the form’s resources, provided the reader can infer it. And not all speech in a novel must be an exchange in this sense. Ritual talk and small talk establish register, mark time and supply the norm against which a deviation shows; what they cannot do is carry a scene.
Where the chapter sits
The chapter applies the value-shift test of The Purposes of a Scene (Chapter 1) (the question what changed?) at the scale of a run of turns, and adopts the fifth domain proposed there, the relational, because in talk the standing between two people moves more often than anything else. It proposes a sixth, the silence domain, for what an exchange does through what is not said. Who reports the speech, and how much of the gap between saying and meaning the narration exposes, is a viewpoint matter, handled in The Viewpoint Shift (Chapter 13).
The household facts in play follow the canon kept for this book. The doorway is Dov’s, across town from the shop; the key is the new key to the shop door; and Elena learns there, four months after the event, that the locks were changed. Scene and Sequel (Chapter 2) picks up her drive to the shop afterwards.
See also: the Reference introduction, with the claim’s qualifications in full; the Reference account of transcribed talk, with a table of the features of real conversation that dialogue omits.