Sociologist Matthew Salganik and his colleagues at Columbia University set out to test the theory that music listeners simply like the music they know other people enjoy. They set up a Web site and recruited more than 14,000 participants--mostly American teenagers--by offering free, licensed downloads of 48 songs by different up-and-coming (and therefore unlikely to be known) bands. They randomly assigned new participants to one of two groups. The first group picked songs to listen to just by title or band name, ranked them on a star scale ranging from one (the worst) to five (the best), and then were offered the opportunity to download the song. The researchers argue this gave them a measure of a song's inherent quality.
The second group, however, saw the same song and band names but also the number of times other participants in the group had downloaded that particular song. This second group faced two experimental conditions: one in which songs were randomly presented and one in which they were presented in descending order of popularity. The subjects in this group were also divided into eight subgroups, called worlds by the researchers, in order to assess whether hit songs varied from world to world.
That is exactly what they found. Although popular songs remained relatively popular from world to world, they did not achieve the same level of success. The sociologists also found that a song's overall popularity or disfavor was generally higher or lower in the presence of peer information than without it. In other words, hit songs in the groups that saw ranked lists were even more popular than in the groups working without any knowledge of what their counterparts thought.
The researchers argue that this means would-be impresarios (and sociologists) will continue to struggle to identify surefire hits. "Experts fail to predict success not because they are incompetent judges or misinformed about the preferences of others, but because when individual decisions are subject to social influence, markets do not simply aggregate pre-existing individual preferences," the team writes in the report detailing the findings in today's Science. "In such a world, there are inherent limits on the predictability of outcomes, irrespective of how much skill or information one has."