A Testable Theory Of Frisson: Music That's Better Than Expected

26 July, 2015
Musical frisson can be a function of the violation of musical expectations – but I hypothesize that this can be better stated as: frisson is something that happens when music turns out to be better than expected.


Frisson is the tingly feeling that people sometimes get when listening to music.

Thrills, chills, frissons, and skin orgasms: toward an integrative model of transcendent psychophysiological experiences in music (Frontiers in Psychology 2014, LD Harrison & P Louie), is a recent article that summarises the current state of research into this aspect of the human response to music .

Violations of Musical Expectation

One observation that has been made about frisson is that it seems to happen more when expectations are violated.

However, the notion of "violating expectations" has to be stated with a caveat – the music still has to be musical. After all, the easiest way to violate expectations when playing some music is to play some wrong notes. But music listeners don't get frisson every time someone plays a wrong note. For a frisson to occur, the notes have to be initially perceived as "wrong", but actually turn out to be "right".

In other words, music is more likely to provoke a frisson in the listener, if the music turns out to be much better than the listener expected it to be.

This restatement of the hypothesis has two consequences which are worth following up.

Firstly, there may be other ways of manipulating a listener's musical expectations so as to induce a frisson, that don't involve playing "wrong" notes. (In my own personal experience this is indeed the case, and I will expand upon that shortly.)

Secondly, we can ask how it is possible for a music listener to develop incorrect musical expectations about how good a given musical item is going to be. It is as if there is one part of the brain determining how musical each part of the music actually is, and there is some other part of the brain attempting to guess, in advance, how good each part of the music is going to be – and if the difference between the two is large enough, then viola, frisson!

Shopping Mall Talent Shows

When I was a certain age, my children were a certain age, and they belonged to a singing and performance club. As part of their club activities, they took part in school holiday talent shows at the local shopping mall.

Every now and then, when watching these talent shows, I would hear a young singer who was very good – surprisingly good for a local talent show.

And quite often, when listening to one of these very good young singers, I would experience a frisson.

The song that was being sung was usually one that I was familiar with, and in most cases it wasn't even one of those songs that I would consider my favourite songs. Often it was one of those songs for which one might expect a young singer to have difficulty reaching all the notes comfortably.

One possible explanation for these unexpected frissons is the expectation that I had: I didn't expect a young singer at a local talent show to produce music that good, but, contrary to my expectations, the music was very good. It wasn't necessarily any better than the commercial studio-recorded version of the song that I had heard on the radio or television, but it was much better than what I expected to hear, at that moment, sung by a live singer, in the shopping mall.

An Experimental Design

If frisson can be provoked by the expectations of the listener, then it should be possible to reproduce this experimentally.

Psychology has a proud tradition of doing experiments on willing volunteers who are deliberately misled in some way.

I would like to follow this proud tradition and design an experiment where volunteers are fed false information which leads them to have inaccurate low expectations about one or more musical performances they are going to hear.

Such an experiment could be based on my example of shopping mall talent shows. The experimental subjects could be told that they are taking part in a study of how adults listen to musical performances by children, using children from a local school – but secretly a child singing prodigy is inserted into the line-up, and afterwards the subjects are asked if they experienced any frissons.

To provide a proper control group in this experiment, it would be necessary to have half of the subjects be told in advance about the child prodigy being included, and how good that child actually was (for example as compared to professional adult singers), and specifically, which of the child performers was the prodigy.

Also, following the logic of "double-blind" experimental protocol, none of the singers should know which of the listeners are in the control group or not (of course this is easily achieved if the singers aren't told anything in advance about these particulars of the experiment), and the self-reporting of the subjects' responses to the music should not depend on any face-to-face questioning by the experimenters.

With a suitable control group defined, it should be possible to calculate a significance value for the results, following standard statistical methods.

Why Music Expectations Can Be So Wrong

It might seem strange that the human mind can somehow wrongly predict its own response to a stimulus that it is perceiving. In most situations there is no particular separation between what we perceive about something and what we expect to perceive given the detailed perceptions of that thing.

But music seems to be a special case.

In science, theories and prediction go together – a theory is defined by the predictions it makes, and we measure the success of a theory by observing if its predictions are correct.

With music, it's as if the mind of the listener has an internal theory about its own response to music, and this theory is not actually a very good theory – it can be quite wrong, when listening to music, what the perceived musicality of that music will be. (One implication of this analysis is that the determination of the "musicality" of music must occur, at least some of the time, a measurable period of time after the direct perception of the sounds of that same music.)

Our inability to form very good theories about our own responses to music does help to explain, in a round-about fashion, the very difficulty of composing music. For example, when considering scientific theories, if we have a theory that accurately predicts some phenomenon in nature, we can usually use that same theory to design and engineer systems that solve certain problems based on the predicted behaviour of natural systems that the theory explains.

For example, if we have a theory that explains the motions of planetary bodies as they move around the solar system, we can use that theory to design a spaceship, and provision it with sufficient rocket power, and formulate a flight plan for that spaceship to fly past Pluto.

We would be skeptical about someone who claimed to have a theory about how and why objects move through space, and that person organised numerous Pluto missions, but all the missions ended up in completely the wrong place.

Similarly, we might suppose that the great difficulty that most of us have in composing music (including even experienced musicians), is a result of the not very good internal theories that we have inside our own brains about the expected response of our brains to any sequence of sounds not already known to be musical.

The CPAI Hypothesis, and Extra-Neural Circuits

According to my hypothesis of Constant Patterns of Activity and Inactivity, musicality is defined by interleaved patterns of activity and inactivity in cortical maps that are relatively constant over a certain timescale.

This hypothesis supposes some means by which the brain observes the geometric nature of patterns of cortical neuron activity. But, neurons don't really "know" where they are in relation to other neurons, they only "know" which other neurons they are synaptically connected to. On the other hand, there are many other brain cells, including glial cells, which are tasked with managing neurons that are "near by", and which plausibly do have some local awareness of the activity and inactivity of neurons located in their immediate neighbourhood.

So one possible model of how the brain measures the shape of patterns of neural activity is that glial cells are involved in the measurement. A necessary secondary hypothesis of this model is that there has to be some means by which the glial cells making the measurement feed the results back into the neural circuitry, possibly into a part of the brain distinct from where those measurements were made.

If musicality is a function of the positional location of neural activity, and if neurons don't a priori "know" what their positional location is, in relation to other neurons, then we have a plausible model of why the neural circuits in the brain are not immediately aware of which sequences of neural activity are likely to result in maximum musicality, and we can explain why those neural circuits are not very good at predicting the brain's determination of musicality from the immediate auditory (and visual) percepts that cause that musicality.

(These same neurons will gradually "learn" certain rules that approximately predict musicality as a function of activity and inactivity of other neurons that they are connected to, but even after years of such learning, these rules may not efficiently generalise to all the possible items of music that the listener's brain has not yet heard. And each time something new is learned about the "rules" of musicality, well, that's another frisson that might be experienced.)