Why Stories Work: Narrative and Language Acquisition

Ask anyone who has learned a language to fluency what they did, and sooner or later a book comes into the story. Rarely a textbook. Usually a novel they were too impatient to put down, or a series of detective stories, or the same three volumes read four times. This is not sentimentality about literature. Narrative has properties that make it unusually good input, and they can be named.

The comparison that matters

The alternative to a story is not silence. It is the material most courses are made of: dialogues written to display a structure, passages assembled to introduce a set of words, short texts about ordering coffee and describing your flat.

Those texts are written backwards. Somebody chose the grammar point, then invented a situation in which two people would plausibly use it eleven times. The result is comprehensible, level-appropriate and completely inert: nothing in it needs to happen, so nothing about it is worth predicting, and once the structure has been displayed the text has no further business. Nobody rereads a textbook dialogue for pleasure, and the reason is not that the topic is dull. It is that there is no reason to turn the page.

A story is written forwards. Something is set up, something goes wrong, and the language exists to deliver it. Whatever else follows from that, one thing does immediately: you keep reading. Since acquisition is mostly a function of how much text goes past your eyes, an input source that makes you continue is doing the largest part of the job before any of its other properties come into play.

Repetition you did not ask for

The most practical advantage of narrative is arithmetic.

A word is not acquired on meeting; it is acquired over a series of encounters, and the usual estimates run to a dozen or so. The problem for a learner is where those encounters are supposed to come from, because most words are rare, and a reader who moves from one unrelated text to another meets each mid-frequency word once and then loses it.

Narrative solves this without anybody planning it. A story stays in one world for as long as it lasts. A novel set on a farm keeps needing the words for barn, fence, harvest, mud and cattle, not because the author is teaching them but because that is where the characters are. A murder mystery repeats its dozen investigative verbs on every third page. The vocabulary of a single long text is far narrower and far more repetitive than the vocabulary of the same number of words drawn from a magazine, and the repetition falls exactly on the words that matter to the story you are following.

This is the strongest argument for reading whole books rather than collections of unrelated passages, and for the old advice to read several books by the same author. The fourth Maupassant is easier than the first, and not only because you have got used to him: it is largely the same vocabulary again.

Prediction, and why it makes words stick

Stories are the one kind of text where you know roughly what is coming. Not the events — the shape. Somebody wants something, something obstructs them, it resolves. Children have this structure before they can read, and it works across cultures with only local variation.

For a learner that structure is a second source of comprehensibility, running alongside the words. When you meet an unknown verb in a sentence where a character has just drawn a knife, the range of things it can mean is already narrow. The story has told you what must happen; the word only has to confirm it. That is precisely the condition in which meaning gets inferred from context, and it is the condition a textbook passage cannot create, because a text with no momentum makes no predictions.

It also changes what happens after the inference. A word met in a scene is stored with the scene attached — who said it, what it cost, how it ended. Memory for events is strong and cheap; memory for items on a list is weak and expensive. Readers routinely find that they can recall the sentence a word came from years later, and this is not a quirk of unusually visual people. It is the ordinary behaviour of memory when the material arrives inside something that happened.

Caring is a mechanism, not a mood

Krashen’s later formulation of his own hypothesis added a word: input should not merely be comprehensible but compelling — interesting enough that you forget it is in a foreign language. The addition is often read as encouragement. It is better read as engineering.

Two things follow from wanting to know what happens. The first is volume, which is the whole ballgame: a reader who cannot stop will get through twenty times more text in a year than a reader who is being conscientious. The second is that attention goes to the meaning rather than to the form, which is the state in which language is acquired rather than merely studied. A learner scanning a paragraph for the subjunctive is doing something useful, but they are not doing this.

There is a corollary that people resist: a badly written thriller you cannot put down is better input than a masterpiece you keep abandoning. The measure here is pages, not prestige.

What kind of story

Not all narrative is good input, and the failures are predictable.

• Dialogue-heavy beats description-heavy. Dialogue is short, present tense, high-frequency, and supported by the situation. Long descriptive passages are where the rare vocabulary lives — trees, fabrics, weather, furniture — and they are the usual place where a reader stalls.
• Recent beats old. A nineteenth-century classic asks you to learn a vocabulary nobody has used for a century, on top of everything else. It may still be worth it, but the cost is real, and it is one of the reasons those texts are adapted.
• Plot beats interiority. A novel in which the events are external gives you the second channel of comprehensibility described above. A novel that takes place inside somebody’s head withdraws it.
• Beware the voice. Dialect, heavy slang, an unreliable narrator or an experimental style will defeat a reader who could handle the same content plainly told.
• Series and single authors. The repeated vocabulary is the point, as above.

This is also why the stories in a good graded collection are chosen rather than invented. An adapted Chekhov or Pirandello keeps a plot that was built to work; a story written from scratch to use the past tense of two hundred verbs generally does not.

What stories do not give you

A reader who has only ever read fiction ends up with a characteristic shape: fluent, wide in the concrete and emotional vocabulary, at ease with dialogue and narrative tenses — and helpless in front of a contract, a scientific abstract or a leading article.

Fiction does not teach the academic and formal registers, the connectives of argument, the vocabulary of institutions and technology, or the impersonal constructions those texts live on. None of that is a reason to start with non-fiction. It is a reason to add it once reading is comfortable, and it is the point at which a subject you already know in your own language becomes the easiest text in the language.

How many encounters a word actually takes is the subject of its own article; why reading the same story twice is not wasted time is another; and the theory underneath all of it is comprehensible input.