A Million Songs Later
Serra et al (2012) report that music is a "human universal, involving perceptually discrete elements displaying organization." They note, if you had seven years with nothing to do (and presuming you didn’t sleep) you might choose to listen to one million songs recorded from 1955 to 2010. And, if you did, you could make some observations and draw some conclusions.
In fact, the "Million Song Dataset" does exist (includes rock, pop, hip hop, metal and more) and permits open access (to data and content analysis) and interesting opportunities and challenges. Serra et al (2012) report patterns and metrics (in music) have been relatively stable for the last 50 years. Further, assuming the past is a reasonable predictor of the future, they predict less variety with regard to pitch transitions, greater homogenization with respect to timbre and louder music-based on the one million songs studied. Indeed, they suggest, old tunes might be easily modernized by simply changing the harmonic progressions, the instrumentation and the loudness to those that are common and fashionable.
McFee et al (2012) have authored "The Million Song Dataset Challenge" to see if someone can create an advanced music information retrieval technology, algorithm (or other functional analysis protocol) based on the same (million song) dataset to predict which songs that will be listened to by an individual listener, with due respect and knowledge as to the listener’s musical preferences. McFee et al report organizations such as Netflix, iTunes and Amazon use collaborative filtering, more so than content analysis, and the commercial implications of content analysis with regard to musical preference is unknown.
For More Information, References, and Recommendations
Beck DL., Bhatara A.(2012) Musicians, Hearing Care Professionals, and Neuroscientists. Hearing Review, February.
Sera J, Corral A, Boguna M, Haro M, Arcos JLI. (2012) Measuring the Evolution of Contemporary Western Popular Music. Nature.