Understanding “Feeling” Music
If you’ve ever asked yourself how music can make us literally feel, you’re not alone. There’s at least two of us now.
I’m sure you’ve listened to (or seek out) songs that make you feel some sort of actual, embodied physical sensation. In discussing music, we hear people bring up things like frisson and phrases that suggest a song “plucks the heartstrings.” One song I find that really does this for me is the final movement of Camille Saint-Saëns’ The Carnival of the Animals. It’s almost like Saint-Saëns set out to write a song that would do both of the things I just mentioned — that cello is too good.
So, music is pretty weird. On the one hand, it’s something that we hear and think about, and maybe that is capable of inspiring a particular emotional response. But on the other hand, we can think about it being proprioceptive as well, which involves our sense of where we exist three-dimensionally. Have you ever listened to a song that’s so sad it gives you a feeling that your chest is almost depressed? On top of that, we can think about it being something that’s interoceptive, or something we feel internally. Frisson would fall under this category.
So what’s an angle that we can use to begin to understand what elements of music may lead to these feelings? Daikoku et. al consider this question in their recently published study Bodily maps of uncertainty and surprise in musical chord progression and the underlying emotional response.
In doing so, they hone in on a lens centered around the concept of predictive processing. Your brain is consistently generating predictions about incoming sensory information, and then comparing the actual input to what it had anticipated. If there’s a disconnect between the two, this interestingly enough can lead to activation of reward systems of the brain. Think about the feeling of having your expectations tastefully subverted in a song, maybe by proxy of something like a sudden key change.
Daikoku et. al take the idea of predictive processing and split it into two constituent parts: uncertainty and surprise, relating both to the bodily sensations that music can invoke. Here, there is more of a significant gap in the literature.
Generally, 549 participants were then asked to listen to different chord progressions. Then, they completed body mapping tests and rated the emotional valence of each. The chord progressions were generated by a model trained on 890 Billboard pop songs in order to establish a feel for what chord changes were more or less likely to be seen statistically.
This essentially gives us a syntax with which we can start to think about creating chord changes that either follow or intentionally deviate from what is sort of statistically likely to happen in a western pop music canon — and the model does exactly that. So let’s return to the idea of uncertainty and surprise. Chord progressions generated by the model did so playing with these ideas. To be clear, here we’re thinking about uncertainty as being our capacity to anticipate what should happen next in the progression, and surprise as what we feel when a particular chord change really defies our expectations.
It’s a kind of a hairy distinction, to help clarify, I wrote out some of the chord progressions they used in the study to illustrate what’s going on. Each chord progression consists of four chords. Within the progression we consider two groups: in the first, chords 1–3, and then in the second, our final 4th chord. For each of the two groups, we consider different combinations of the progression being high or low certainty/and then high/low surprise.
Chord Progression 1
Group 1: Low surprise, low uncertainty
Group 2: Low surprise, low uncertainty
You can see how it’s kind of easy to see where the progression is going. It’s just 2 chords going back and forth, where the first resolves to the second in a fairly standard way. You probably don’t listen to it and think, “Wow that was unexpected!” or “Wow I had no idea where that was going!” We’d say that both groups (meaning chords 1–3 and the final chord) were fairly unsurprising and fairly easy to anticipate.
Chord Progression 2
Group 1: Low surprise, low uncertainty
Group 2: High surprise, low uncertainty
Here, group 1 is actually identical to the progression featured in group 1. The primary difference lies in our final chord (group 2). You may have found that you went into the last chord relatively certain of what was going to happen, but were then thrown off by the choice to use the suspended version of the chord from chord progression 1. This would be more of an unusual chord choice (at least based on the model’s training data).
Chord Progression 3
Group 1: Low surprise, low uncertainty
Group 2: Low surprise, high uncertainty
Here, the first three chords are relatively standard, but in going to the 4th chord, there is a little bit more uncertainty over where we’ll go. Compare this to the first chord progression. There, we went from G major, to C major, back to G. Based on that, you might anticipate that we’ll go back to G because the pattern seems to be that we’re just alternating between 2 chords. Now in group 1 of progression 3 we go from C major to A minor to Bb Major 7. This is all fairly standard, but the Bb major 7 might cause you to question what the last chord could be (again, based on what the training data suggests is more or less likely), but then we simply return to our A minor chord for the 4th and final chord, which you may find is anticlimactic if anything (and therefore perhaps less surprising in the same way the final chord in chord progression 2 was).
So what did they find?
Hopefully, the above diagram should make a little more sense now. This shows that based on the category of chord progression played, participants experienced bodily sensations in different areas based on body tracking data. For instance, analysis revealed that the first chord progression (which was universally sort of unsurprising and predictable) was associated with sensations of the abdomen. Additionally, in the second chord progression, where we considered making our final chord surprising, this interestingly had the effect of localizing bodily sensations around the heart.
In terms of considering the valence of the chord progressions, I found it fascinating that generally the progressions enjoyed the most were those that kept uncertainty low universally and only introduced surprise for the final chord, if at all — in other words, our first two. And with respect to our vanilla chord progression 1, participants uniquely described that as being associated with calmness, relief, satisfaction, and nostalgia interestingly enough. Generally, emotions were found to have different relationships to bodily sensations depending on the chord progression. For chord progressions 1 and 2 discussed above, the emotional valence was found to elicit bodily sensations. But for other chord progressions, such as those centered around prioritizing high uncertainty, the emotional valence actually caused a downtick in bodily sensations.
So what does this all mean? On the one hand, it’s interesting as someone interested in the creation of music to see the trends we see with participants’ reception of the first two chord progressions. Good music often times will set up very deliberate expectations, and then meet them in a satisfying way, or subvert them in an interesting or unexpected way. It’s interesting that the establishment of clear expectation is consistent in both chord progressions, and that these two are the most preferred across a decently large sample size by a fair margin. Moreover, it’s interesting to see discretely differing chord progressions be reliably associated with different patterns of bodily sensation that are often not influenced by one another (i.e. cardiac sensations occur completely unrelated to those of the abdomen).
While it is undeniably helpful to consider chord progressions in this way, I can’t help but wonder (as the authors do) how the introduction of other elements, such as greater musical context, might influence the generation of proprioceptive or interoceptive sensations. And if you have more questions now than before you read this article, I’ve done my job.
iScience, Daikoku et al. “Bodily Maps of Uncertainty and Surprise in Musical Chord Progression and the Underlying Emotional Response” https://cell.com/iscience/fulltext/S2589-0042(24)00719-3
