Rumour sharpening

Why does a story grow sharper and more dramatic each time it is passed along?

A story that reaches you fourth-hand is rarely a faded copy of the original. It is usually a better story: shorter, cleaner, with a sharper point and a more obvious villain. Fading is what imperfect memory should produce — blur, loss, uncertainty. Instead what comes back is more definite than what went out.

So the ordinary account, that retelling wears a story down, cannot be the whole of it. What kind of process subtracts and intensifies at the same time?

Start with what a teller actually has to work with. Not a recording, but a memory of having been told — and memory does not hold a message, it holds whatever the listener made of the message. Every retelling is therefore a reconstruction from a partial trace. The question becomes: which parts survive reconstruction, and why do the survivors grow?

Allport and Postman put this on a bench in the 1940s with what they called serial reproduction. One person describes a picture or hears a passage, tells a second, who tells a third, down a chain of six or seven. They named three regularities in what came out the far end.

Levelling: the account gets shorter, fast at first and then flattening out. What goes is detail, qualification, hedging, names, numbers — the material that makes a story hard to hold and hard to tell.

Sharpening: a small number of details survive and become more prominent than they were, doing narrative work the original never assigned them.

Assimilation: the whole is reshaped toward the teller's expectations. Gaps close with the expected thing, motives get supplied, sequence gets tidied into cause. Their most-quoted case is a subway scene containing a razor held by a white man; along many chains the razor migrated to the hand of a Black man standing nearby. Nothing drifted at random. It drifted toward what the tellers already believed.

Now notice something about the first two. They are not separate processes. If most of the detail is dropped and the story still has to work as a story, whatever remains must carry more of the load. Sharpening is what levelling looks like from the point of view of the surviving detail. You do not need an extra mechanism for emphasis; you need only loss plus the requirement that the thing still be tellable.

And the survivors are not a random sample. The teller keeps what they can retell — what is vivid, what fits, what earns a reaction — so the loss is biased toward the memorable.

Here is where honesty is owed about the method. In a serial-reproduction chain, nobody may ask a question, nobody hears the same thing twice, nobody has a second source, and nobody has any reason to care whether they are accurate. Real rumour is nothing like that. It moves through conversation, with challenge, correction, cross-checking, and many overlapping lines of transmission rather than one chain. Shibutani's treatment of rumour as improvised news — collective problem-solving when official information is inadequate — describes groups that sometimes converge on accuracy rather than away from it. So the experiments demonstrate that reconstruction distorts in a direction. They do not establish that this is what happens to a rumour in the world, and that gap is a live criticism of the work.

Credit is owed backwards too. The deeper source is Bartlett, whose "War of the Ghosts" studies a decade earlier introduced both the serial method and the idea doing the real work here: the schema, a prior structure that reconstruction leans on. Bartlett's own experiments are loose by modern standards — small numbers, informal procedure — though the core finding has held up.

The word "sharpening" invites a misreading: that tellers are embellishing, exaggerating for effect. Mostly they are not. The teller is reporting what they remember, in good faith, and their memory has already done the editing. That is why the process is so hard to catch from the inside — no one in the chain experiences themselves as changing the story.

Two refinements worth keeping. First, the sharper version is better by narrative criteria, not by truth criteria: single-cause, agentive, resolved. Those are the properties that make a thing easy to hold and rewarding to pass on, and they are largely unrelated to whether it happened.

Second, none of this requires a distorting individual. Suppose every teller is scrupulous and forgets at random. Transmission still selects, because a version easy to remember and satisfying to repeat gets passed on more often than one that is not. Sharpening can be a property of the population of retellings rather than of any teller in it — which is why hunting for whoever "started the exaggeration" so often finds nobody.

• How rumour behaves when transmission is a network with feedback rather than a chain.
• Whether the same selection pressure explains which urban legends survive centuries.

Every term the collection defines is gathered in the glossary.