The model rests on four strands of argument. Each opens here with the practical question it answers, and the last says how far the model binds.
Why does a caused event land harder? Aristotle and Forster
A writer who asks why one sequence of events grips and another, equally eventful, does not is asking the oldest question in the field. In chapter 9 of the Poetics Aristotle ranks lowest the episodic plot, whose episodes follow one another without probability or necessity, and highest the plot whose events happen against expectation and yet di’ allēla, “through one another”. His standard is plausibility rather than proof: events must arise from what this kind of person would do in this situation. The Latin pair often attached to the idea, post hoc and propter hoc (“after this” and “because of this”), comes from logic, where inferring a cause from a succession is a fallacy. In narrative the same inference is a competence. Readers make it constantly and on little evidence, and the craft of the chain consists in supplying enough evidence for the inference to feel earned. E. M. Forster’s Aspects of the Novel (1927) gives the formulation most writers half-remember: “The king died and then the queen died” is a story, and “The king died, and then the queen died of grief” is a plot. Three words turn sequence into consequence. Forster does not dramatise the grief or motivate it; he asserts a cause in a prepositional phrase and the reader accepts it, which shows how cheaply the conversion can be bought.
Why must a scene end worse? Desire and delay
The rule that a scene should end in a setback can look like a taste for melodrama. Peter Brooks’s Reading for the Plot (1984) supplies the reason for it. Brooks describes reading as driven by narrative desire, the reader’s want, produced by the text, for the end that will make sense of everything before it. The middle of a narrative is a dilatory space (from the Latin dilatio, delay), a stretch of detour and postponement whose function is to stop the story reaching its end too soon. A disaster is the device that keeps the end away: each time a goal is nearly reached and turned into a worse problem, the reader’s desire is renewed instead of spent. A clean win spends it, and its damage shows in the next chapter, whose goal must then be invented rather than inherited. Kenneth Burke’s Counter-Statement (1931) states the same idea most briefly: form is “the creation of an appetite in the mind of the auditor, and the adequate satisfying of that appetite”. On that account a scene creates and sharpens an appetite, and a sequel turns an appetite left unsatisfied into a new one.
Why must a goal be stated? The reader’s search for causes
Writers are told to imply rather than state, and for goals the advice is wrong. Roland Barthes, in S/Z (1970), called the reader’s running ability to recognise an action in progress and give it a name the proairetic code, from the Greek proairesis, the choosing of a course of action; his examples are names such as Stroll and Murder. A stated goal licenses the reader to name the sequence. Once Elena wants the key, the reader opens a file called getting the key and measures every sentence against it; without a goal there is no title to file events under, and they are read as a list.
Experimental work supports a narrower point. Tom Trabasso and colleagues model a story as a causal network, and find that events on the chain connecting opening to outcome are recalled better and judged more important than events off it, and that the more causal links an event has, the better it is remembered (Trabasso and van den Broek 1985). Graesser, Singer and Trabasso (1994) add that readers routinely infer why something is happening now but infer what will happen next only weakly. The asymmetry has two consequences for practice. A chain that is present need not be underlined, because the reader is already looking backwards for the cause. A goal must be stated, because the reader will not reliably project it. These findings concern comprehension and memory, not literary value: the chain makes a narrative cheap to process, and that cheapness is a resource a writer may spend or decline.
How far does the model bind? Where it comes from
A writer should know where a tool was made before relying on it. Swain (1915–1992) wrote for the American all-fiction magazines and later taught professional writing at the University of Oklahoma, and his book codifies a market that paid by the word and bought reliably shaped stories in volume. The model’s features follow from that market. It is generative, telling a writer what to write next, which criticism never does; its units fit a reading session; and every unit ends in a state that motivates the next. Jack M. Bickham’s Scene & Structure (1993) carried it to a wider readership, and from there it passed, simplified and often unattributed, into online craft advice. It appeared when, in Mark McGurl’s account in The Program Era (2009), the university workshop was teaching nearly the opposite lessons: the image and the epiphany rather than the chain and the unresolved close. That is the model’s period convention, the set of assumptions it carries from its time and market. Agency, escalation and external goals are the virtues of one tradition, and a book that does not share them is not broken for that reason.
Barthes made the deeper qualification in an essay of 1966, arguing that narrative works by confusing what comes after with what comes because of. Causality is conferred by readers rather than contained in texts, and three things follow. Adjacency does much of the work the model credits to structure, so causal signposting can usually be lighter than a writer fears. The chain can be faked in connecting sentences, so the diagnostic tests are run on events with the connectives removed. And refusing the chain is a legitimate choice: Don Quixote makes the arbitrariness of sequence part of its joke, and Catch-22 makes the absence of causal logic its argument. The model is therefore most useful as an instrument that tells a writer which kind of book a draft currently is.
See also: the full theoretical account in Reference, including Propp’s functions, Todorov’s equilibrium schema, Burke’s pentad and situation models; The Purposes of a Scene, “How It Works” (Chapter 1), on kernels and satellites and on why a scene without change leaves no trace.