There’s a rough consensus on which cities are the fashion capitals of the Western world: London, Milan, New York and Paris. Fashion trends start there and slowly diffuse out to the rest of the globe. But what about music? Which cities are trend-setters for our listening habits?

A new paper by two Irish network scientists, Conrad Lee and Pádraig Cunningham of the Clique Research Cluster, tries to answer the question by analyzing listening habits on, a popular music Web site that pinpoints users by geography. The researchers found that, in North America, Atlanta tends to propel trends in hip-hop music, while Montreal leads listening habits for indie music. Over in Europe, meanwhile, Oslo is the big trend-setter. Here’s one of their graphs:

Leader-follower network for the 20 most active cities in Canada and U.S. The height of the nodes corresponds to influence. Gray lines signal a lag in spreading tastes of 1-3 weeks, blue edges a lag time of 4- 5 weeks.

What’s surprising, the authors note, is that the largest cities aren’t always the most influential adopters of new music (or snubbers of stale music). New York and Los Angeles don’t appear to have nearly as much influence over listening trends as, say, Montreal, even though those areas are presumably home to many more local bands and musical groups. What’s more, some cities you’d expect to be fairly similar — like Portland and San Francisco — don’t appear to influence each other very much in the network.

The researchers, by the way, analyzed the music data by “adapt[ing] a method previously used to detect the leadership networks present in flocks of birds.”

Now, granted, there are real limitations to this study. The authors focus on because it offers up a wealth of geographic data. But obviously not all music listeners use (plenty of people use Spotify, or YouTube, or iTunes), and it’s not clear how well users represent of the broader public. What’s more, a few smaller cities have relatively few listeners, leading to a lot of noise in the data. (The paper itself spends the bulk of its time discussing techniques to filter out this noise.)

The authors themselves stress that their paper is “a work in progress.” Still, it’s a neat way to model what they call a “leader-follower relationship among cities.” Perhaps one day we’ll be able to predict musical trends before they get big, simply by following the flow from city to city. But, the authors say, that’s a subject for a future paper.

(The link comes via MIT’s Physics arXiv blog.)