Brain imaging research shows a higher correlation between specific brain activity and future success of music than how much subjects said they actually like the songs.
An Emory University research project measured brain responses from a group of adolescents while they were listening to songs by nascent or unknown artists. The group also gave feedback on how they liked the songs subjectively. The researchers tracked the popularity of these songs over the next few years, and saw significant correlation between increased brain activity in certain areas of the brain and later song success, while likeability did not predict sales.
In this study concerned with predicting popularity, researchers Gregory Berns and Sara E. Moore used fMRI (functional magnetic resonance imaging) to measure reward-related regions of the brain. Cultural popularity - "something that is popular in the broadest sense and appeals to a large number of individuals" - was tracked through brain activity to see if the link between activity in reward regions and future purchasing decisions applied to the population at large.
Brain activity over direct questioning lends credibility to the field of neuromarketing, according to Scientific American. This study basically underlines the main explanation for the the field by demonstrating that the raw brain data "might provide better information about future purchases than the consumer."
The difference between raw and processed data brings interesting possibilities to the table. As SA suggests, it might be a shortcoming of the research process, or it might be that people change their answers based on expectations - "downplaying our embarrassing inner rapture over Justin Bieber, or overstating our enjoyment of Miles Davis."
Neuromarketing presupposes some sort of disconnect between what consumers say they like and what they actually buy. Recent topic coverage adds credence to the value in this area of research, including Nielsen's acquisition of neuroscience and ad research company NeuroFocus.