Saturday, August 4, 2012

The Science Of Making Boring Pop Music Is Developing Apace: My Analysis Of "Measuring the Evolution of Contemporary Western Popular Music"


In May of this year a study was released that empiricaly shows western pop music has become more homogeneous over recent decades.  The article is titled Measuring the Evolution of Contemporary Western Popular Music.  The researchers quantitatively analysed a publically available collecton of musical metadata called the Million Song Dataset (MSD) to arrive at their conclusions.

The MSD is interesting in itself.  Available for download from Columbia University, it is a 280 GB file that contains dozens of individual datum on each of one million popular song recordings. Data ranges from what you find in ID3 tags (year, artist, track, album) to sophisticated quantifications of the music itself, developed using The Echo Nest's clinically named software, Analyzer (also called "Analyze").  This analytic data ranges from fairly objective things like key and tempo to algorythmic assessments of more subjective things like danceability.

To me, the most interesting thing about the MSD is how the million songs were chosen.  Like the music analysis tool, Analyze, the universe of possible songs came from "music intelligence" firm, The Echo Nest.  Starting from the many millions of songs The Echo Nest had data on, the MSD developers first limited the set to music that was "familiar." From there they broadened it again to collect similar music to fill out the overall set.

In Measuring The Evolution the researchers specify their data set includes "rock, pop, hip hop, metal, [and] electronic" music. Clearly, however, they aren't studying any of those genres.  The bias toward familiarity in populating the MSD means they studied music in these styles that achieved popular success, plus music with something significant in common with the popular.  A heavy metal song that also becomes a hit on pop radio does not cease to be heavy metal, but it does become pop music. There is also probably something it has in common with pop and that distinguishes it from the millions of heavy metal songs that would never be accepted by pop radio.  These are the commonalities and distinguishing marks this type of analysis can uncover.

That the MSD researchers explicitly used the word "familiar" in describing their method tells us something important about their definition of "popular music" and therefore the subsequent work of the Measuring the Evolution team.  In some musical contexts "popular music" is everything that is neither "traditional music" (AKA, folk) nor "art music" (AKA, classical).  Even jazz is made to fit in these three containers, making them each really, really huge and, more importantly, focused on the music itself and its production.  Popular music, in this case, has nothing to do with popularity or commercial success.

In contrast, by starting with "familiar" both sets of researchers set their sights on the pop music of big record sales, radio saturaton and soft drink endorsements.  This music is defined and shaped by popularity.  It reaches big audiences and has a fundamentally different relationship to the surrounding culture than music that is tied to a relatively small subculture or some other market niche.  This sort of pop music is where eclectic meets predictable and new meets familiar. It is a musical genre organized around recreating its own past hits while also making every song memorably unique. This is Pop Music, the perpetually changing but unmistakable musical genre that has dominated youth-oriented radio for decades, not Popular Music, the huge musicological category for everything low-brow that isn't folk music.

What did they actually say?

Measuring the Evolution of Contemporary Western Popular Music shows, in numbers, that from 1955 to 2010 pop music...

+ used a progressively narrower range of chord and melody transitions.  In other words, it became easier to predict what chords and notes would follow their predecessors through the course of a song.
+ has progressively become "louder."  I won't address this as too much has been said already.
+ became less diverse timbrally.  Like notes and chords, the overall harmonic spectra of songs have become more similar, suggesting that the choices of musicians, arrangers, engineers and producers have narrowed right along with the singers and songwriters.

As I read other's reactions to Measuring The Evolution I saw two main types of reactions to it.  One was exactly as I expected.  People like myslef, who pride themselves either on being musical omnivores or their devotion to obscure and experimental music are cheering; "See! We told you this crap all sounds the same!  And it's getting worse!"  But really, save your breath.  Nobody cares why we prefer the weird stuff.  Most people like their music to go down relatively easily, not be a bunch of work.

The other, I didn't see coming, but should have.  Some people, especially those who are classically trained, object to the exercise in its entirely.  The argument goes something like "Quantizing and quantifying music for statistical analysis does not tell you anything useful and fundamentally offends the spirit of all musical endevor."  It's not so much an argument as a plea for the programmers and statisticians to stay far, far away.

"Stop looking at my violin like that!"

Where I part ways with pretty much everybody who's opinion I have heard or read in response to to Measuring The Evolution is that I don't think there is any cause for alarm.  I don't think it is bad that pop music is getting more homogeneous.  I also don't have any objecton to this type of analysis.

To worry about or criticize pop music for spanning a narrower range of musical variety (in certain ways) fundamentally misses what pop music is.  To create the perfect pop song is to perfectly blend musical familiarity with novelty.  What Measuring The Evolution really tells us is that writers, producers and artists are refining the familiarity side of the equation.  They can't neglect the novelty side because that is what distinguishes their record.

It is a cliche and slightly off target to say pop producers "give the people what they want."  It's more a need than a want.  They give the audience the musical comfort they need to listen through to the end of the song for the first time.  We already knew creating an accessible, memorable, danceable pop song is an art (even if it isn't "art").  Measuring The Evolution opens up some of the science involved.

Somebody needs to figure out how to quantify "heavy."  To quote Skwisgaar Skwigelf, "I can't tune down any farther."

No comments: