ACROSS the French countryside, in the summer of 1789, rumours swirled about vengeful aristocrats bent on the destruction of peasants’ property. It was not true. The Great Fear, as it is now known, tipped France into revolution with a flurry of fact-free gossip and rumour.

Two centuries later the methods for spreading nonsense are much improved. In the first paper of its kind, published in Science on March 8th, Soroush Vosoughi and his colleagues at the Massachusetts Institute of Technology present evidence that, on Twitter at least, false stories travel faster and farther than true ones.

The study, carried out at MIT’s Laboratory for Social Machines, showed this by examining every tweet sent between 2006 and 2017. The researchers used statistical models to classify tweets as false or true, by applying data taken from six independent fact-checking organisations. That allowed them to categorise over 4.5m tweets about 126,000 different stories. Those stories were then ranked according to how they spread among Twitter’s users.

The results were stark. False information was retweeted by more people than the true stuff, and faster to boot. True stories took, on average, six times longer than falsehoods to reach at least 1,500 people. Only about 0.1% of true stories were shared by more than 1,000 people, but 1% of false stories managed between 1,000 and 100,000 shares.

The reason false information does better than the true stuff is simple, say the researchers. Things spread through social networks because they are appealing, not because they are true. One way to make news appealing is to make it novel. Sure enough, when the researchers checked how novel a tweet was (by comparing it, statistically, with other tweets) they found false tweets were significantly more novel than the true ones. Untrue stories were also more likely to inspire emotions such as fear, disgust and surprise, whereas genuine ones provoked anticipation, sadness, joy and trust, leading to the rather depressing conclusion that people prefer to share stories that generate strong negative reactions. Perhaps not coincidentally, fake political news was the most likely to go viral.

The paper also sheds some of the first peer-reviewed light on the impact of “bots”—automated accounts posing as real people. The idea that Russian bots in particular helped sway America’s presidential election has lodged itself firmly in the public consciousness. Yet the paper finds that, on Twitter at least, the presence of bots does not seem to boost the spread of falsehoods relative to truth.

The researchers were able to conduct a study of this breadth thanks to the business relationship between one of their number, Deb Roy, and Twitter, which provided its entire historical dataset at a steep discount. But more are likely to come. Technology companies, and particularly social-media firms, are facing a backlash from regulators and consumers worried about the harm from their products. Twitter, for its part, has said it is ready to offer the same dataset to other outside experts.