MRIs of teenage brain waves can predict pop hits.
bizjournals.com
bizjournals.com
Correlation does not imply causation.
As it happens, there was a correlation. But we don't know what the songs were actually like. (Perhaps the poor songs were incredibly bad, so much so that any musician would have been able to flag them up as likely failures). We don't know much about the 27 participants. We don't know much about the experimental method.
This is not well reported science.
You can reconstruct the source of neural activity using EEG (it is an example of the inverse problem, http://en.wikipedia.org/wiki/Inverse_problem), but you need at least 64 channels if you want any semblance of precision. 128 or better, 256 channels give better results (although, at least a few years ago, the 256 channels machines often were an overkill (diminishing returns)).
Every major label has a deal with them, nearly every hit on the top 50 for the last 5+ years was pre-screened by management. The last numbers I heard were $1m/year for the service with a $20k/song screening charge.
Color me amazed.
I think the fact that the fMRI data predicted success, but that the explicit ratings given did not predict success is the interesting part of this. It would be totally uninteresting if you could predict the song's success just from asking people what they like, or how good they think something is. That an fMRI can tell you more about the goodness of the song than the people in the fMRI listening to the music, is pretty cool.