Why do some rumors spread like wildfire but burn out quickly, while others seem to smolder for years?
According to recently published physics research, it's all in the different ways people handle information.
Take the Web page of French champagne house Veuve Cliquot, for example. It features quite a cordial warning.
"Dear websurfer, A promotional deal is currently on the Net regarding a free offer of a case of 6 bottles of Veuve Cliquot champagne. This is a hoax, totally beyond our control ... We strongly condemn the author of this hoax and hope that it will end."
The warning refers to an email hoax that promised readers the case of champagne for forwarding the message, ostensibly as a reward for helping expand the company's email database. It's been circulating for more than four years. The disclaimer was added to the site over two years ago.
In recent years, researchers who study networks have suggested that information spreads like a viral epidemic, moving steadily through the social chain with a given incubation period for each infection.
Given the common assumption that, on average, people typically take a day to forward an email, persistent hoaxes like Veuve Cliquot should be impossible. So what's happening?
"Information is different, because it's valued by people," said Esteban Moro, a math professor at the University Carlos III of Madrid.
When an email pops up in our inbox, we make a quick decision whether to respond immediately or file it in a "do later" pile. Unlike a computer virus, which we can spread by simply opening an attachment, a funny video or chain letter requires us to make the decision whether to forward it, and when.
"[Information spread] is going to be affected by the way we schedule our tasks," Moro said.
Although the average delay between receiving an email and forwarding it on is about one day, actual response times tend toward extremes. About half of us answer an email within the first hour, while 20 percent take more than a week.
"Most of us know people in our departments or company who are very, very active, and some who take a long time to respond," Moro said.
In a paper to be published in Physical Review Letters, Moro determined how "fast forwarders" and information slowpokes decide the fate of the latest funny YouTube cat video, chain mail hoax, or juicy bit of Aston Kutcher gossip.
He teamed up with Jose Luis Ibarren, the head of IBM's European e-marketing division, to design a real-world viral marketing experiment, a campaign that awarded a raffle ticket for a laptop to users who referred friends to a newsletter subscription.
Carried out in 11 countries and ultimately reaching more than 30,000 people, the experiment gave Moro the data to create a detailed model of viral information flow.
He discovered that information is caught in a tug-of-war between hyper-connected users and those who let emails languish in their inbox.
A piece of information goes viral if each infected person quickly spreads the email, video, or gossip to at least one other person.
"If you are infected and infect someone else, the number of people this 'someone else' will infect … is the number that tells you whether the campaign or infection will be alive in the second round or not," Moro said. "As long as the fast forwarders who first get wind of the news spread it to more people than originally heard it, it's over the tipping point."
"If you're above the tipping point, the message not only spreads out to a large fraction of people of society, but [does so] in a matter of hours," Moro said. "Not because the message is important, but because the more active people are the ones driving the information diffusion."
This results in a viral phenomenon like the videos of Susan Boyle singing on Britain's Got Talent, which reached 47.7 million people in less than a week.
About 90 percent of viral marketing campaigns fail to reach this tipping point, but they're often cost-effective because they may continue to spread, although at a much more casual pace. Below the tipping point, slowpokes take the reins, as they have in the case of the Veuve Cliquot champagne hoax.
Moro's experimental campaign lasted two months after the original emails were sent; his model predicted it would persist for a year if he maintained the website supporting the campaign.
But companies who want to use viral marketing to keep costs down, Moro said, should decide quickly whether to kill a campaign or keep it alive.
"If you're in a viral marketing department, you want it fast, cheap, and big," Moro said. "What we're telling them is run it for two days."
By that point, the campaign will spread to half the people it will ever reach; the other half will trickle in gradually over months, dragged out by the slow forwarders. If reaping the benefits from these late-comers requires maintaining a website or some other cost, it might not be worth it.