In a study published in Science in 2006, participants went to a website and listened to songs, rating and downloading the ones they liked. The 14,000 participants were randomly assigned to different "worlds." Individuals in the "independent" world simply rated and downloaded songs without any input about what others were doing. In the other seven dependent worlds, raters saw which songs other participants downloaded and how they rated them.
The researchers figured if ratings were based solely on each participant's taste, then the best songs would rise to the top and all the worlds would mirror the independent world. But if, as they suspected, participants were influenced by others' ratings, then different songs would be rated highest in each world.
The researchers found vast variation in rankings between different worlds. Often, which song was rated highly simply depended on who the first raters happened to be. In some circumstances, "if you rewound the world and played it again, you could see a potentially different outcome," says Dr. Salganik.