What causes a particular meme to take the Internet by storm, dominating image boards and inspiring hundreds of variations, while another one languishes?
It’s a tantalizing question in the nascent field of meme theory, and not just because the answer could shed light on our collective online subconscious. It’s also possible that research into it could eventually explain broader aspects of cultural consumption—why an entire work, perhaps even a novel or a painting, might gain a following, flop, or eventually fade into obscurity.
So far, the bulk of research into why a meme goes viral examines how its current position in a social network can be used to predict how it will continue to spread. The idea is that if you look at how influential the people are who have already shared it, their relationships to other people, and whether they shared it at a time when others are likely to see it, then you can crunch the numbers and make an educated guess as to whether it will continue to spread or peter out.
The problem with that approach, according to Michele Coscia, a postdoctoral fellow who studies complex networks at the Harvard Center for International Development, is that it ignores the content of the meme itself. After all, an influential person can create a new meme and share it at the optimal hour and day of the week, but if it’s boring, it’s still likely to be a dud. It’s not that he sees extrinsic meme theory as wrong-minded, he says; he just wants to look at the other side of the coin.
“There must also be some value in the content itself,” said Coscia during a Skype interview.
Technically, a meme is any single unit of cultural exchange, and it can refer to anything from a chain letter to the figure of speech “jumped the shark.” To start with, though, Coscia opted to study just one type of meme, and one of the most visible today: image macros, in which internet comedians riff on current events or ongoing jokes in bold white text that appear over images known as Scumbag Steve, Forever Alone, and many others. His hypothesis was that the more similar a meme was to existing fare, the less likely it would be to find viral success. If you
think about it, though, either possibility is tempting: fresh ideas are more memorable, but Foul Bachelor Frog spawned Foul Bachelorette Frog, and so forth.
“I think there's support for Michele's idea,” said Eytan Adar, a professor of information and computer science at the University of Michigan who studies the flow of information online. “There's certainly a pressure and incentives in social media to demonstrate that you knew or said something first. Memes that are very similar to something said before are probably treated as ‘old.’”
First, Coscia needed a dataset. He used the Meme Generator API to look at memes shared during the summer of 2013, and analyzed the name of each one to see whether it was related to a preexisting meme, as with Socially Awkward Penguin and Socially Awesome Penguin. Then he used the image analysis tool SURF to detect which images looked like one another, and finally broke down the actual text of each meme variation to figure out which were talking about the same current events.
Then, Coscia used that data to create an enormous map. On it, each node represents a meme. The larger the node, the more variations it spawned; the closer to orange it appears, the more votes it got on Meme Generator. Lines connect memes that Coscia’s analysis showed to be related to each other—so Socially Awkward Penguin is connected to Socially Awesome Penguin, and both are connected to Socially Awkward Awesome Penguin.
The center of the map is a tangle of interrelated jokes and references. Off toward its periphery, though, it develops into longer chains: series of piggybacking memes that are less closely related to the groupthink at the center. And it’s in that perimeter, outside the space where the memes are most similar, where the big orange nodes—the superstar memes that generated lots of popular variations—are most likely to fall.
What that suggests is that Coscia’s hypothesis is correct. Similarity to other memes decreases the probability that a meme will be successful. Coscia, who first became interested in memes after reading Richard Dawkins’ 1976 “The Selfish Gene" and lurking on meme-centric subreddits, believes that his research is the first to suggest that finding.
It demonstrates “that the intrinsic characteristics of memes and their similarity with one another is connected with their likelihood of going viral,” Coscia wrote in a paper on the work that has since published in Scientific Reports, an open access journal published by Nature Publishing Group. “This is a remarkable result: it allows researchers to detect meme characteristics and use them to objectively explain why a meme is popular.”
“If you are very similar to what is already there, I can guarantee that you will not be successful,” Coscia said. But “dissimilarity does not necessarily imply success. There is no magic potion that tells you whether you’ll be successful or not."
Of course, there’s no reason yet to assume Coscia’s results apply to anything other than image macros. Though he’s tightlipped on what he’s working on next, he says he’s now working on a much larger dataset—and that his very long term goal is to study how the structure of very complex works, like art or literature, affect the way they spread.
It’s tempting to frame Coscia’s research as an instruction manual for creating successful memes, but it’s more useful, he says, as a guide for what not to do: if you want to be successful, don’t jump on a bandwagon.
You “will never hear me reciting guru-like advices to reach success like ‘be different,’” he wrote on his blog when the paper was published. “That’s just bollocks.”
reader comments
46