Music and Surprise: Which Causes Which?

24 May, 2013
Is music musical because it's surprising? Or is music surprising because it's musical?

Expectational Theories of Music

There are many different theories about what music is, and why we like listening to music.

There is one group of theories about music, which don't really have an official name, but they have to do with expectations, so I like to call them "expectational theories".

When we listen to music, we have expectations about what notes will come next, and when they will come, and what they will sound like. We also have expectations about how the music might affect us, emotionally, or otherwise.

So far I haven't really said anything that implies that the role of expectations in the perception of music is particularly different from role that expectations play in the perception of anything else.

We have expectations about things that are going to happen, sometimes what happens is what we expected, sometimes what happens is not what we expected. Whatever.

What the expectational theories claim is that as music "unfolds", the expectations that we have about music play a critical role in whatever perceptual process it is that makes us perceive that music is musical.

One fan of expectational theories is David Huron, who wrote a whole book about music and expectation.

Surprise Implies Music?

If you are interested in music science, like I am, you might have seen the recent paper Interactions Between the Nucleus Accumbens and Auditory Cortices Predict Music Reward Value by Valerie Salimpoor et al. Or you might have read one of the popularized accounts of this research such as Brain's music pleasure zone identified at the Guardian, or Why Your Brain Loves That New Song at Scientific American.

This research seems to "prove" that our appreciation of music has a strong relation to expectations.

To quote Scientific American's quote of David Huron: "It's a lovely, lovely piece of research."

Salimpoor discovered that the nucleus accumbens is more active when music listeners listen to new music that they like more. She discovered this by getting her subjects to listen to new music while being brain-scanned in an MRI machine, and afterwards measuring how much the subjects liked the new music by asking them to say how much they would pay for the music they just heard if it was on iTunes – and actually requiring them to pay that much money.

What Salimpoor found was that appreciation of the music which the subjects liked the most was correlated with higher activity in the nucleus accumbens.

And to quote Salimpoor (as quoted in the Guardian article): "Activity in the nucleus accumbens normally would indicate that expectations are being met or surpassed."

The implication is, apparently, that the musicality of music is somehow determined by the brain's calculations of its expectations about music.

This sounds plausible. But I remain unconvinced that the perception of musicality has anything at all to do with expectations.

Just because the nucleus accumbens is active when you hear new music that you like, doesn't mean that the nucleus accumbens has anything to do with how much you like music.

In fact, to quote the Scientific American article: "The nucleus accumbens is believed to be responsible for pleasant surprises, or 'positive prediction error,'"

Which leads to my alternative hypothesis ...

An Alternative Hypothesis

My alternative hypothesis is that the determinant of musicality is not in the nucleus accumbens at all, and the reason why new good music strongly violates our expectations is that our brains are hopeless at predicting the musicality of new music. The human brain is hopeless at predicting musicality, because the brain does not have any understanding of how or why it appreciates music.

And because the brain is hopeless at predicting musicality, whenever new good music is heard, the brain will always be surprised at how musical it is.

When listening to new music, at first the brain perceives the notes of the new music, and based on its limited knowledge and understanding of musicality, it has no reason to suppose that the new music will be any good. But then, after hearing enough of the new music, the mechanism that responds to musicality is activated, at which point the brain "feels" how good the new music really is. Which it finds quite surprising.

So the nucleus accumbens does not cause the perception of musical quality, rather it is responding to the perception of musical quality, which has already happened somewhere else. And the response of the nucleus accumbens is to be "surprised", because no other part of the brain succeeded in predicting the musical quality that actually happened.

Why is the brain so terrible at predicting its own appreciation of music?

Possibly the brain has no understanding of its own appreciation of music because the appreciation of music involves mechanisms that are not entirely contained within the brain's neural circuitry. For example, as I have suggested elsewhere, musicality may be computed in glial cells, which respond to patterns of activity in neurons, and then (somehow, and somewhere) feed information about the computed level of musicality back into neural circuits.

The possible role of "non-neuronal" circuits in the perception of musicality may explain why scientists have failed to detect a "music centre" in the brain via fMRI, which specifically images neural activity. (Additionally, for my non-neuronal theory to be plausible, I have to assume that the relevant "patterns" of neural activity have a structure that is too fine to be detected by any existing brain-scanning technology. Which is entirely possible, because there are quite a few neurons inside each "voxel".)

In any case, regardless of any specific explanation of why the brain cannot predict musicality in advance, it is unclear from the description of Salimpoor's research that she has any reason to suppose, as commentator David Huron wants us to believe, that how we process our expectations about music has anything to do with why we like music in the first place.

To fully test Huron's "expectational" hypothesis it would be necessary to include some type of control in the research. A good example of a control would be something else that was unexpectedly good, like unexpectedly good food, or unexpectedly good sex. And one would have to observe that the nucleus accumbens' response to new good music was qualitatively different to the nucleus accumbens' response to anything else that is better than expected.

Until such a comparison is done, we should be skeptical of claims that Salimpoor's research constitutes proof that music is all about expectation.

The Other Problem With Expectational Theories of Music

Salimpoor's experimental protocol specifically observed brain response to new music, thus avoiding the question of how the brain responds to old music, and how we can still like music, even after we know exactly what notes are in that music and when they occur.

It is true that we eventually find music more boring if we hear it too often. But compared to other types of entertainment, this boredom effect is quite weak, and we can happily listen to the same item of music many times, on different separate occasions, and still enjoy listening to it every time.

We can even enjoy music repeatedly when we play it. There cannot be much "surprise" about what notes are in a musical item if the listener is the same person as the person playing the notes. So a theory that claims that our appreciation of music somehow depends on "violation of expectations" is not very convincing.

To me this seems like a fatal flaw in any "expectational" theory. If the appreciation of music depends on the relationship between the music and our expectations about that music, then any music that we are familiar with should be boring and un-musical, just like we laugh at new jokes, but we find old jokes boring.

But that does not match up with most people's experience of music.

Which leaves us looking for a better theory.