Itzhak Perlman recently spoke with the L.A. Times about an aspect of frisson that has perplexed me from the start of our development process: how a moment that works for me may not work for you at all (and vice versa). When asked what moments in music move him, Perlman replied:
I’m actually capable of being moved by a particular phrase, or a particular sound of either a singer or instrumentalist. It’s incredible; it really affects me. Sometimes I listen to things on YouTube that I like, and if it’s amazing I wind up in tears. It doesn’t have to be so much the way it’s performed, but the actual music itself…Sometimes I question my students: ‘What gives you goosebumps?’ And I love the fact that the same goosebump piece to me, for you is, ‘Nah, that one doesn’t do it for me but this one does.’ I always find that interesting.
I couldn’t have said it better myself. Over the past year, there have been multiple times I’ve discovered a great moment that gave me chills and, after eagerly sharing it with my girlfriend, been met with a response of her shrugging and dismissively saying “It’s good, I see why you like it, but it doesn’t give me goosebumps.” Bah!
A more concrete example of this dilemma emerged for our team with the song Touch Me I’m Going to Scream by My Morning Jacket. Several people submitted the song to our dataset and flagged the moment at 4:53 as the most chills-inducing passage. When our annotators reviewed the song, however, each annotator flagged a different moment, none of which were 4:53. Wild! To give just a few examples, one annotator flagged 2:24, another flagged 4:27-4:35, and another flagged 5:41 (there were of course other moments flagged beyond the four above, and plenty of listeners that said no parts of the song worked for them). Have a listen and see if any of these four moments work for you:
So, what the heck is going on? How can it be that, even within the same song, no one on our team got goosebumps during the same moment, much less during the main moment flagged in the crowdsourcing process? Below are three hypotheses (note if our annotators were each listening to different recordings, or using different volume settings, or using different headphones, or in rooms with different ambient temperatures, or if some had consumed a lot of caffeine prior to listening, or if some knew My Morning Jacket and disliked the band for some reason, these could all have affected the results too, but we largely controlled for them in our listening process):
1) Our annotators are non-representative of the broader population
It’s possible, but unlikely. We used a large group of listeners purposefully selected to vary by their age, gender, country of origin, genre preferences, and whether they played music or not. And, in an early analysis of our first five annotators, their results were not highly correlated with each other (making it improbable that all five would deviate from the trends in the broader crowdsourced data).
Further analysis suggested there was a wide range of sensitivity to frisson among our annotators. Certain individuals typically got chills during multiple moments in many songs, while others at best got chills during only one moment in a song, and even then it was rare to find songs that worked for them. As a result, I do not think this hypothesis explains the divergent results with Touch Me I’m Going to Scream.
2) Our annotators’ minds are wandering during parts of the song
While we used carefully controlled listening conditions with our annotators, it’s still a reality that any of us can easily find our minds wandering during longer (8+ minutes in Touch Me I’m Going to Scream) songs with contemplative lyrics. And the frisson research is clear that active, focused attention is required for a passage to work. So, each of our annotators may have been differently engaged with the music at various points, contributing to their divergent frisson tags. In everyday listening this problem is magnified and really makes a mess of things (e.g., the same moment that works for one person does not work for another simply because the second person got a pop-up Slack message right when they were listening to that particular passage, etc.).
I think this hypothesis has some validity, but doubt its the whole story given that our annotators listened to each song multiple times. Its unlikely they could have all become distracted during the same parts of the song every time.
3) Our annotators have different responsiveness to each of the eight patterns
Our leading hypothesis is that either due to innate differences or varying exposure to certain sounds and types of music growing up, we are all differentially sensitive to the eight frisson patterns; we each have a unique Frisson Profile. Let’s take a closer look at the four moments flagged in the My Morning Jacket song:
- 2:23 is primarily a Startle pattern (with a supporting Surround pattern)
- 4:27-4:35 is a vintage Scream pattern
- 4:52-4:53 is a combined Danger pattern and Resolution pattern
- 5:41 is identical to the Startle pattern at 2:23 but has that high-pitched synth layered on top that produces an additional, momentary Harmonicity pattern
If people are differentially sensitive to frisson patterns, this would indeed help explain the divergent results here (and the broader low correlation numbers among our annotators in the table above). It would also be consistent with my personal experiences arguing with my girlfriend over frisson moments. I am highly sensitive to the Surround pattern, likely resulting from the fact that I played percussion growing up and spent so much time sitting behind the high-tonal volume brass sections in orchestras. And yet, this pattern usually does not work on my girlfriend. In contrast, what does appear to reliably work for her is the Scream pattern, to which I am less sensitive. When I am listening to the crowdsourced data and an acute Scream pattern comes on, she often shouts across the house “that one actually worked for me!” This has held true across various genres.
So, is it all subjective? I don’t think so. To use film preferences as an analogy, some people generally like horror movies, some action movies, some dramas, but occasionally there are great films in each genre that execute the particulars of that genre so well that they move broad audiences beyond niche enthusiasts. Most people, for example, don’t usually watch period dramas about classical or jazz musicians. And yet the superlative acting and music in Amadeus (1984), Shine (1997), and La La Land (2017) resulted in Academy Awards and broad commercial success. This makes me think exceptionally talented directors and actors periodically find ways to re-invent and execute certain genres so effectivelly that they can move people who generally don’t watch, of even dislike, that genre.
A recent Harvard study on universal human responses to music affirms my belief that the same is true of music in general and frisson in particular. While we each have a unique Frisson Profile, there are certain rare compositions and performances that deploy frisson patterns in such a novel and exceptional way that they create moments that “work” for broad audiences. The great Hollywood composers have figured this out, from John Williams’ soaring climax in Jurassic Park when they first see the dinosaurs and the brontosaurus rears up on its hind legs, to Howard Shore’s swelling full ensemblewhen the entire city bows to the hobbits at the end of The Return of the King, to Hans Zimmer’s triumphant introduction of the Elgar variation when the small boats appear during the climax of Dunkirk. Those moments helped create blockbusters and, in my view, are tapping into something real.
Perlman himself has endorsed the idea that frisson is not all subjective, stating that the one song that “works” on his audiences no matter what country he performs it in (and Perlman has performed in almost every country in the world) is John Williams’ theme from Shindler’s List. Make sure you are in the right mood and have a listen:
Incredibly moving stuff. And, consistent with Perlman’s experiences with his audiences, nearly every one of our annotators reported getting chills during this piece, particularly during the ending section of 3:34-3:54 (especially those final two notes at 3:52). The exceptional Grief patterns in Williams’ composition (and Perlman’s performance) are what we are all about here at Qbrio: stunning moments in music that move many of us to tears and goosebumps.