What makes a story last

The oldest story most people can name is written on clay, and it survived because a library burned down.

The Epic of Gilgamesh was copied and recopied across Mesopotamia for well over a thousand years, and the version usually read today was pulled together by a scribe named Sîn-lēqi-unninni. Then it vanished for roughly two millennia. Nobody read it, nobody mentioned it, and no tradition carried it forward. It came back in the 1850s when excavators at Nineveh dug into the ruins of Ashurbanipal's library, and it came back legible because the sack of the city in 612 BC set the place alight and fired the tablets like pottery. Papyrus would have gone up. Clay got baked.

In 1872 George Smith, a former banknote engraver working at the British Museum, was reading through the fragments and found an account of a flood, a boat, and a bird sent out to look for land. He is said to have reacted by leaping about the room and starting to undress. He'd found a version of a story his entire culture assumed had one source.

Which is a useful thing to hold on to before asking why some stories last. Quite a lot of the answer is luck.

Most of what's gone was lost by accident

Before printing, a text stayed alive only if someone kept copying it, and copying was slow, expensive and boring. Papyrus rots in damp climates within a few centuries. Parchment lasts far longer but cost a herd of animals per book, so old manuscripts were regularly scraped clean and written over — the reason the best surviving copy of some of Archimedes' mathematics was found underneath a thirteenth-century prayer book.

The scale of the loss is hard to take in. Sophocles wrote something like 120 plays. Seven survive complete. Aeschylus wrote perhaps 80 or 90, and again seven come down to us. Sappho was collected in antiquity in nine books and was famous throughout the ancient world; what we have now is one complete poem and a heap of fragments, many of them recovered from an Egyptian rubbish dump at Oxyrhynchus.

The clearest case is Euripides. Ten of his plays survive because they were school texts, taught with commentaries, copied constantly. Another nine survive for a much stupider reason: a single manuscript happened to contain a stretch of his plays whose Greek titles fall in one run of the alphabet, and that manuscript made it through. We don't have those nine because anyone judged them his best. We have them because they were filed next to each other.

So the first filter on lasting isn't merit at all. It's whether a text was on a syllabus, or in a monastery that didn't burn, or written on something that survives being wet.

Printing changed the terms completely. Once a book exists in a few hundred identical copies scattered across a continent, losing it becomes very hard. After about 1500, survival stops being the scarce thing. Attention becomes the scarce thing instead, and the question shifts from whether a story can be preserved to whether anyone still wants it.

Oral stories are built to be rebuilt

Stories that lived in performance rather than on a page were shaped by a different pressure, and the shape gives the pressure away.

Milman Parry went to Yugoslavia in the 1930s, with Albert Lord, to record illiterate singers of long heroic poems. The question behind the fieldwork was Homeric: how could anyone hold the Iliad in their head? The answer they found was that the singers weren't holding anything in their heads, not verbatim. They were rebuilding the poem each time out of a large stock of ready-made phrases that fitted the metre, swapping them in as the line required.

Which explains the bits of Homer that read oddly to a modern eye. Dawn is rosy-fingered every single time. Achilles is swift-footed even when he's sitting down sulking. Those aren't lazy decorations — they're components of the right metrical length, kept on the shelf and pulled down as needed. The same logic produces the long catalogues, the repeated scenes of arming and feasting, and the habit of returning to an earlier point in reverse order to close a section off.

All of it is memory engineering. A story that has to survive without writing gets pushed towards strong rhythm, repetition, concrete visual scenes, small casts with sharply distinct characters, and episodes that can be reordered without wrecking the whole. Anything intricate, subtle or dependent on exact wording gets sanded off, because it can't be reconstructed on the fly.

You can still see the residue in stories that reached print centuries later. The threes, the rule that the youngest child wins, the repeated formula that signals a scene is starting again — those are oral load-bearing structures, kept out of habit long after nobody needed them.

What memory actually holds on to

Some material is simply stickier than other material, and the reasons are unglamorous.

Concrete beats abstract. A locked room, a severed hand, a bargain sealed with a name — these lodge in a way that a summary of a theme never will. Causal chains beat sequences. E.M. Forster made the point neatly in 1927: "the king died and then the queen died" is a sequence, and you'll forget it; "the king died, and then the queen died of grief" is a plot, and you won't, because the second half now depends on the first. Memory is much better at holding things that entail one another.

Then there's the counterintuitive, in careful doses. Work in the cognitive study of religion, particularly by Pascal Boyer and by Justin Barrett, suggests that concepts which break one expectation while keeping the rest — a statue that hears you, a wolf that talks, a person who casts no shadow — are recalled and transmitted better than either wholly ordinary concepts or wildly bizarre ones. One violation is memorable. Six is noise.

Social content sticks too. Who betrayed whom, who owes what, who is secretly related to whom. That's most of Greek tragedy, most of the Bible's narrative books, and roughly all of soap opera.

None of this makes a story good. It makes it transmissible, which is a different property and easy to confuse with the first one.

The universal story shapes are oversold

Somebody will have told you that all stories are one story. Be careful with that.

Vladimir Propp's Morphology of the Folktale, published in 1928, is the serious version. He worked through a corpus of Russian wonder-tales and extracted 31 recurring functions — the hero is sent away, an interdiction is broken, the villain is punished — appearing in a consistent order. It holds up remarkably well, and Propp was explicit about the limits: this is a description of one genre in one tradition, not a law about narrative.

Joseph Campbell's monomyth, from 1949, is the version that escaped. It's enormously influential, especially on screenwriting, and it has two problems that specialists have raised since it appeared. The categories are loose enough that almost any narrative can be pushed into them with a bit of shoving, which makes the claim very hard to disprove. And the supporting examples were gathered from a huge range of cultures with the fitting parts kept and the awkward parts trimmed. Christopher Booker's seven basic plots run into the same objection.

There's also quantitative work. A 2016 study by Reagan and colleagues ran sentiment analysis over thousands of texts and reported that emotional trajectories cluster into a small number of shapes — rise, fall, fall-then-rise and so on. It's real and interesting, and it measures the emotional temperature of the prose, which isn't the same thing as plot. Six sentiment curves don't give you six stories.

And where the same story genuinely does turn up in distant places, the explanation is often duller than universality. Folklorists catalogue tale types in the Aarne-Thompson-Uther index, and the Cinderella type appears from Europe through the Middle East to East Asia, with a Chinese version written down in the ninth century. Stories travel along trade routes, with soldiers and merchants and slaves, the way goods do. Two cultures sharing a tale is often evidence of contact rather than evidence of a shared unconscious.

The honest position: recurrence is real, it's weaker and vaguer than the popular versions claim, and much of it comes down to the memory constraints in the previous section rather than anything deep about the human soul.

Characters get out of their stories

Here's something you can test on yourself. You almost certainly know who Sherlock Holmes is, and Dracula, and Scheherazade. Now try to summarise the plot of anything they appear in.

Characters detach. Plots don't. A character strong enough to be described in a sentence can be lifted out and put in new situations indefinitely, and every time that happens the original gets another few decades of life. Conan Doyle learned this the hard way — he killed Holmes off in 1893 to get on with what he considered his serious work, and public demand dragged him back within a decade.

The mechanism is simple enough. Adaptation is cheap when the thing being adapted is a person rather than a structure, so characters accumulate versions, and versions keep the name in circulation. Frankenstein's creature is the standard example of how far this can drift: the figure most people picture, silent and lumbering with bolts in his neck, comes from a 1931 film. Mary Shelley's creature is eloquent, self-taught and reads Paradise Lost. The novel survived partly by being replaced in the public mind by something else with the same name.

Ambiguity is a rereading engine

Stories that only work once tend to get used once.

The texts that hold on across centuries usually have a gap in them that readers have to fill, and the filling changes with the century. Why does Hamlet delay? The nineteenth century said he was too sensitive for action; the twentieth had theories involving his mother; the current fashion is to treat the delay as a problem of political evidence in a surveillance state. The play doesn't say. That's the point — a text that answers its own central question has nothing left to argue about, and gets shelved.

Genre matters here too. A mystery that spells out its solution is finished on the last page. What keeps rereaders coming back is the stuff around the answer: the voice, the world, the character.

Ambiguity isn't sufficient on its own, though. Plenty of obscure and richly ambiguous books have vanished without trace. Ambiguity keeps a story alive once it already has an audience; it doesn't create one.

Translation is life support

A story confined to one language lives exactly as long as the readership of that language, which is a shorter time than people assume.

What keeps foreign texts alive is retranslation. The Odyssey has been put into English dozens of times over four centuries, each version pitched at its own moment, and it's the steady supply of new translations rather than any single one that keeps the poem in circulation. When retranslation stops, a work quietly drops out, still respected, still unread.

Translation also rewrites the thing it carries. Wordplay, metre and puns don't cross, which is why narrative travels far better than lyric poetry. And sometimes translators add. The Thousand and One Nights reached Europe through Antoine Galland's French version early in the eighteenth century, and the two tales everyone knows — Aladdin and Ali Baba — have no known Arabic manuscript behind them predating him. Galland got them orally from Hanna Diyab, a young Syrian traveller he met in Paris in 1709. The most famous stories in the collection entered it through the translation.

Somebody always decides

Canons aren't discovered. They're assembled by people with jobs.

Alexandrian scholars drew up select lists of the poets and playwrights worth copying, and being off the list was effectively a death sentence. Medieval monasteries chose what to spend parchment on. School and university syllabuses have been the single strongest preservation mechanism of the last few hundred years, because a text that's taught is a text that's printed, and a text that's printed is one that survives being ignored for fifty years.

Shakespeare is the case that should frighten anyone comfortable about all this. Roughly half his plays had never been printed when he died in 1616. Macbeth, The Tempest, Twelfth Night, Julius Caesar and about fourteen others exist only because two colleagues, John Heminges and Henry Condell, gathered the scripts and published the First Folio in 1623. Had they not bothered — had one of them died first, or the money not been found — half the canon would be a list of titles.

That's the shape of the whole answer, really. Stories last when they're easy to carry in the head, easy to retell in new forms, loose enough to be argued about, attached to a character somebody wants to use again, and lucky enough to be near someone who kept copying them.

The rest is fire, damp and rubbish dumps.