Friday Weird Science: They totally knew Lady Gaga was coming.

Jul 15 2011 Published by under Friday Weird Science

One of my favorite types of posts is the type where I get to throw in a ton of music links. This one is no exception. Put on your headphones!

I saw this paper covered in an article on Jezebel before I left for vacation, and there just wasn't time to cover it then. The basic idea is that the author of today's study used fMRI on a bunch of teenagers (for a study on peer pressure). He used some random music from undiscovered bands for a random part of the study. Imagine his surprise when, a few years later, he hears a song, and it sounds oddly familiar. It's a song from his study! Specifically, this song:

And so the author went back to his old data. Rather than looking for brain activation correlated with peer pressure, however, he looked at the reaction to the music clips themselves. He found a correlation between brain activation and music that went on to become popular. And the authors concluded that this means brain scans could be used to identify new hit songs.

And I'll be honest that I don't think he's wrong, I think we can identify new hit songs before they ever go to market. But I don't think we need a frakkin' expensive fMRI to do it.

Berns and Moore. "A neural predictor of cultural popularity" Journal of Consumer Psychology, 2011.

The authors, after a long introduction on how great it is to be people or things that are popular, the authors took 27 teenagers and stuck them in an fMRI. They got 15 second clips of music that had been downloaded from MySpace, specifically from unsigned and unknown musicians (this was back in 2006), to minimize the chance that the teenagers would recognize them. The teenagers rates the songs as they heard the clips by how much they liked them and how familiar the song was. They then tracked how popular the songs ended up becoming, and looked to see which areas of the brain showed increased activity during listening.

As it happened, the vast majority of the songs they picked didn't end up getting much of anything. Most of the bands on MySpace, alas, will not produce hit singles. But there was one aspect of the study that correlated heavily with the three songs that became hits (or even close to hits, only one song actually became a hit, the rest only achieved gold or nothing).


(Click to embiggen)

This is the correlation between activity observed in the nucleus accumbens during song listening, with songs ranked by popularity. As the popularity of the song increased, the activity in the nucleus accumbens increased. The thing is, the teenagers had NO IDEA which songs were popular (and in fact, none of them were popular YET) because they had never heard them before. So this correlation could be interpreted (and the authors did interpret it this way) as a prediction of future song popularity.

And the activity in this particular brain area makes sense. The nucleus accumbens is a brain area that is known to be associated with the rewarding and reinforcing properties of things, like cocaine, or...music. So activity in this area when you hear something you like is actually pretty predictable. The interesting thing about this study is that the teenagers tested didn't rank themselves as liking the songs MORE. It's possible that this is because they just didn't get enough exposure to the song to really tell what they liked (they only got 15 seconds).

My main issues with the study are that the study size is small (27 for a music study? You can SO do better), and that I think they might have seen something with likability of songs if they'd given the participants a longer exposure to the music. In addition, they only had three songs that went on to become anything like hits, and in fact they only had ONE song that became a real hit. I think I need to see some more hits in there before I say anything.

If this is reproducible, I guess you could use fMRI to determine which songs are going to be popular. I mean, sure you can do that. If you want to waste your money. fMRIs are expensive, and while this study shows that you may be able to use neural activity in the nucleus accumbens to predict popular songs (notice I said "may", I don't know how "real" this effect is, after all they only had three songs become hits), are you gonna run clips of all your songs through a bunch of teenagers sitting in an fMRI?

And do you need to?

Here's the thing: I think that brain activity may indeed correlate with the popularity of certain songs. But I don't think this has ANYTHING to do with the actual "goodness" of the song. I think it may have a lot more to do with how RECOGNIZABLE the song is. That doesn't mean you've necessarily heard it before (remember these songs are not familiar), rather it means there are themes in there that sound familiar, similar to something you've heard before.

For example.

These are pop songs that sound...exactly like other pop songs. They have the same chord structure, the same rhythms. And they sound FAMILIAR. And human beings, we do like the familiar. We like the feeling of something nice that we've had before. We may respond better to songs that have familiar chord structures that we recognize, ones that our brains can identify as having pleased us before and likely to please us again.

And I think this might be a good hypothesis. After all, look at this:

And this:

I mean, they're awesome, and they are saying the same thing. A lot of pop songs, really popular pop songs, sound the same.

Now you might rant and rail about how pop singers are all plastic and uncreative and everything sounds the SAME. And you might be right. But that's not because the pop singers are inherently horrible. Nope. It's because they aren't stupid. And neither are their handlers, managers, producers, etc. They know the sounds that SELL, even if they don't know it due to having degrees in psychology. They know that songs with specific chord progressions SELL, and so they are going to use them.

This is not even limited to pop music, it's also found in classical. The classical pieces that tend to be popular (usually among non-classically minded people, but often among classical musicians as well), often sound SIMILAR. Handel is tough to hate because everything he ever did sounded the same. Bach sounds the same. Pachelbel...yeah.

This all comes down, to me, to an idea for a study. I want to see a study (not necessarily fMRI, probably just a questionnaire would do it) looking at how much people like songs they don't know. The songs would be either highly unknown songs, or songs that are entirely fake for the purposes of the study. It's my personal hypothesis that the more familiar chords (like the four chords above or the ones in Pachelbel) are present in the song, the more people will find it likable. But then the question would become one of whether they just like the song because they find parts of it recognizable, or whether there's something about those specific chord structures that we have come to associate positive feelings with.

And then I wonder if we could BEND it a little. After all, listen to this:

There's very little about that song that is normal. It has some normal chord structures (though they are separated out and not actually in chords), but not in a normal pattern. But I wonder if we could take songs like these and play them over and over to a group of people, and then test them again on what songs, with which elements, they liked. Would their preferences change as their familiarity increased? I think that it would. After all, we tend to think of Western music, but think of this:

That's WAY outside the Western chord structure norm. But mostly because we're not familiar with it. Perhaps if you tested different cultural groups with different exposures to music types, they would associate most positively with the music structures with which they are most familiar. Obviously, without testing, we can't find out. But do we need an fMRI machine to do it? I think it's a little superfluous.

Berns, G., & Moore, S. (2011). A neural predictor of cultural popularity Journal of Consumer Psychology DOI: 10.1016/j.jcps.2011.05.001

11 responses so far

  • Snarkyxanf says:

    This brings up two of my pet hypothesis:

    1) Over the era of recorded music, people gradually came to like progressively more "discordant" music as they learned to appreciate it (consider that "heavy metal" was used early on to describe Jimi Hendrix's sound). Likewise, "classical" music evolved over the same time period to become much more "difficult" than its predecessors.

    2) fMRI machines are like penises. The people who have them can't seem to help themselves from playing with them.

    • psychotic psychiatrist says:

      Here, Here! well thought out truth; does that make it a law, how many tweets days it take these day?

  • Scicurious says:

    Author's Note: I take no responsibility for LIKING any of the songs featured today. Especially that One Republic song. It kind of makes me cringe.

  • Kevin says:

    Do you listen to planet money? The last episode was about how you manufacture a hit song.

    Maybe all those music studios need to invest in an fMRI machine.

  • Sam McNerney says:

    Awesome post. I have heard this study a few things here and there, but I think this does a nice job of putting it into context and flushing out its consequences.

    Of course, this really worries me. Some of the best art in history has broken rules and made the audience uncomfortable i.e., Stravinski's Rite of Spring. So I hope that producers, artists, and managers don't JUST setting for what the masses want. It is equally important for art to be challenging, rule breaking, and unconventional. That how we get our Picasso's and Steve Reich's.

  • There's also an aspect that I don't think they considered, which is current trends in music. There are certain sounds/patterns in music that are very "now", for any given time. If you'd played people in the early 80s Duran Duran before anyone knew who they were, they'd have been, like, "Wow!"; whereas play it to people who haven't heard that specific band before in 2011, and they'll be like "whatever", or at best, "Retro!". Ditto Beach Boys in the early 60s, and Beethoven in the early 1800s. I'd expand my hypothesis to include an effect of the age of the listener - teenagers are more likely to be in tune with the new stuff.

    • malia says:

      There still seems to be music that defies this "current trends" hypothesis though; if you listen to a lot of the music that's been released in the last year, it's very strongly reminiscent of past decades' trendy music. For examples "born this way" by lady gaga = 80's madonna; "Next Girl" by The Black Keys = 70's classic rock; "big wave" by Jenny and Johnny = early 60's beach boys. My mom is actually loving a lot of this new music bc it reminds her of music when she was a kid (note to my mom...I did NOT mention which decade that was!).

  • Greg Laden says:

    "With a couple of Margaritas in the right company, this could be commercial!"

  • whizbang says:

    It's also important to remember what a mess regression/correlation can be. It's especially vulnerable to outliers, like the poor souls in the lower left who sold less than 100 albums. Take that point away, and I suspect the r value drops quite a bit.

    The overall idea that popular stuff stimulates some part of the damn brain is interesting, but regression is a bitch that can suggest a hypothesis, but never confirm it.

  • Tinny Earsome says:

    It seems to me I've read of similar "can't miss" hit predicting schemes/experiments every few years. I'd certainly think the big labels, at their peaks of wealth, would have explored every such avenue.

    One thing that's interesting to me is how each upcoming group of what is now called "tweens" all through the 20th century came automatically bored with their parent's music. Ragtime, dixieland, big band, crooner pop, Bill Haley/Elvis rockabilly, Beatles and beyond. I myself experienced this. Whatever THAT phenomenon is, it surely isn't based on safe familiarity.

    One other comment that seems pretty obvious to me ix - you're talking more about "recordings" than "songs". Or maybe that's a different issue. But in the pre-rock era it was the song that was the star and very often 3, 4, 5 covers of the same new song would chart. We could test singer/band likability. Or we could have the same artist(s) play & sing different songs with the goal of fixing one big variable of such an experiment.

  • me says:

    I don't know what this says but I found the last video to be the most enjoyable and familiar, although I have never listened to traditional japanese music before. It might be that I prefer the pentatonic scale when I poke at a piano (just stay on the black keys and you can't hit a wrong note).
    It might be that my taste in music is warped by the fact that Loituma's cover of the Ieven polkka is my favorite song.

    Does anybody else feel that the fade on Imogen Heap's song is a little too rapid?

Leave a Reply


nine − = 5