This posting delves a little more into the focus of my research: the cognition of music. In particular, we will look at an article recently published in Scientific American called “Brain Scans Predict Pop Hits.” It is a summary article of a paper by Gregory Berns and Sara Moore from Emory University. This study looks at how well brain response via fMRI can predict pop music hits.
This posting delves a little more into the focus of my research: the cognition of music. In particular, we will look at an article recently published in Scientific American called “Brain Scans Predict Pop Hits.” It is a summary article of a paper by Gregory Berns and Sara Moore from Emory University. This study looks at how well brain response via fMRI can predict pop music hits.
Before we all get excited in a Madison-Avenue-attacks-again kind of way, keep in mind that the results here were anything but dead-on conclusive. The only indicator of predictability came from the nucleus accumbens, which as the author of the Scientific American article says was only “loosely predictive of higher sales.” Nevertheless, the smoking gun of the nucleus accumbens is pretty cool because this region of the brain has been implicated in neural circuits involved in reward processing.
We have to be careful, though, when saying that the nucleus accumbens is an indicator because of the statistical significance tests done to indicate it. In particular, along the lines of an earlier posting (see “When can we consider “significance” significant?), imaging studies are done with paired tests usually between two or more conditions. In the case of this study, it was likely a comparison (I haven’t read the actual article, but am generalizing from other fMRI experiments) between brain hemodynamic response for the pop sensations and the pop duds. One of the issues with this approach is that the single-trial nature of hearing that song for the first time is averaged out of such an analysis.
Of course, we will go easy on Berns and Moore because this experiment was started in the heydays of imaging (2006), when it was a bold step just to even ask these questions and answer them by averaging data over many trials. Now though, single-trial analysis that actually tries to understand the trial-to-trial variability and how this impacts perception is the focus. In fact, this is the focus of what I am working on with my colleagues at Columbia.
So far, we have found that we might be able to track learning over the course of a music experiment in which adjustments to the piece are made at repeatable locations. In other words, we use single trial analysis to look at how a given trial changes as a function of the ongoing experiment, and the supposed concurrently happening learning.
We hope to push more into the direction of Berns and Moore though, attempting simultaneous EEG and fMRI. This will allow us not only to have a strong eye on the dynamics but also on the spatial changes happening in the course of the music experiments.
Being part of a blog on career advice, the take away message here is: cool things are happening out there and it’s always important to see how they relate to the project you’re working on. It’s always good to know what the audience has seen recently!
