Buzz Blog

A Scientific Analysis of Meme Success

Monday, April 08, 2013
When you have a joke, anecdote, or quip to share with the internet, attaching your message to a well-known meme — a cultural idea or style spread among people — will likely help it take off. Have a joke about an embarrassing social situation? Then use your joke as a caption for an image of "Socially Awkward Penguin," and there's a better chance thousands of redditors will appreciate your insight.

Ensuring your caption fits the general idea behind the meme will better your chances of upvotes, likes, or shares within an audience familiar with the meme. But what makes one meme more popular than another, and how might meme success depend on the popularity of other memes?

Does "Confession Bear's" latest secret impact the success of "Chemistry Cat's" latest nerdy joke? To investigate the relationships among Internet memes, postdoctoral computer scientist Michele Coscia from Harvard University has adapted ecology research methods and applied them to a popular website closely related to the social media website reddit: QuickMeme.com.

An implementation of "Chemistry Cat," one of many memes found on quickmeme.com


E-cology


Through the course of his research, Coscia saw numerous parallels between the study of species and memes, perhaps leading him to adapt methods from ecology research. For example, many memes evolve and branch over time, such as "Socially Awkward Penguin," and some of its offshoots: "Socially Awesome Awkward Penguin," and "Socially Awksome Penguin."

There's also an atmosphere of competition and collaboration online akin to those found in natural ecosystems. With the plethora of data readily available from QuickMeme and these similarities in mind, Coscia had all of the ingredients needed for some legitimate Internet research.

On QuickMeme, users can rate a meme as either "awesome" (garnering two points for the meme's rating), "Meh" (1 point) or "just bad" (-1 point). Coscia looked at this data for all 499 memes that had at least one featured implementation (a popular instance of the meme's usage) between October 2011 and October 2012. In total, Coscia downloaded over 175,000 implementations of the almost 500 memes.

Coscia found that the vast majority of meme implementations have lower ratings, and the number of implementations decrease as the rating threshold increases, as one might expect. Only the most popular handful of meme implementations reached ratings that exceeded 10,000 points.

As shown in the graph below, the statistics also revealed a "front page effect" that briefly slows the observed decline in the number of memes reaching increasing levels of popularity. Threshold memes that reach the front page often receive an immediate boost solely because they're featured on a more prominent webpage.

The number of meme implementations vs. popularity rating. Notice the small hump around the  1000 point mark, indicating the "front page effect." Image Credit: Michele Coscia via the arXiv

Meme Dependency


After collecting the data, Coscia adapted an ecological null model to tease out any relationships among memes and the effects, if any, those relationships had on a meme's popularity. The null model served to eliminate popularity spikes caused by unusually high traffic on the entire QuickMeme website and account for the average popularity of a meme over time. Comparing data from the null model to the observed data, Coscia could see which observed popularity spikes and dips were truly caused by just the meme in question.

If two memes became popular during the same timeframes, Coscia labeled them as "in collaboration." If one meme succeeded while another failed, then the two memes were "in competition."

Coscia also settled on the following definition for a "successful" meme: Successful memes (among the 499 studied) are those that have a higher-than-average number of implementations (or uses).

Meme success often correlated — and perhaps depended upon — the success of closely related memes. As Coscia points out in his paper, "Chemistry Cat" had a 20.75 percent chance of being "successful" on any given week, but that likelihood rose to 60 percent on weeks that the "Dwight" meme was also successful. This symbiotic relationship is an example of "collaborative" memes.

"College Freshman" — a meme that often ridicules the naivete of a stereotypical college freshman — showed the highest levels of collaboration with other memes. On the other hand, "Hipster Dog" — I'll let you guess what this meme's about — was the most competitive, meaning its success was anti-correlated with the success of other memes. "Hipster Dog" must be ahead of the curve,discovering memes before all the other memes popularize them.

Overall, Coscia found that successful memes correlated with collaborative memes. On QuickMeme, it seems that certain memes may ride on the coattails of others during a surge of popularity. Coscia points out in his paper that many memes may "ride the karma train," by referring to other popular memes.

Collaboration was not the sole indicator of success, however. Memes with small numbers of competitors were more successful, but competitive memes were less likely to be successful if they had popularity peaks in the past.

While these correlations don't necessarily imply causation, Coscia's research has revealed several interesting parallels between meme ecosystem and natural ecosystems. You can read his full paper on the arXiv, where he indicates that he'll present this meme research at the International Conference of Weblogs and Social Media in July.
Posted by Hyperspace