← Three short discoveries on WatWat
A young woman sitting in a chair listening through large headphones plugged into an old device
Art

A wholly predictable song is boring and a wholly surprising one is exhausting, so plainly there is some middle. To find it, calculated given what came before, and people rated how much they liked what they heard. Where was the peak?

Tested in people

In the middle, as . Across two experiments, with 43 participants and then 27 more, liking did not rise in a straight line with order or with surprise. It peaked between them: enough pattern to build an expectation, and enough change to update it. A shape like that is stronger evidence than a straight line, because it is hard to explain by any simple bias, and it held even after .

A wholly predictable song is boring and a wholly surprising one is exhausting, and that is a feeling you can measure rather than merely have.

The researchers used a model that computes how predictable each note is given what came before, which gave them a number for every moment instead of a subjective judgement.

Across two experiments, with forty-three participants and twenty-seven more, liking did not rise in a straight line with order or with surprise. It peaked in the middle: enough pattern to build an expectation, and enough change to update it.

A shape like that is stronger than a straight relationship, because it is hard to explain by a simple bias, and it held even after seven repeated hearings. So musical pleasure may come out of the play between a successful guess and an error you can learn from. Both samples were small, and the model describes one aspect that leaves out lyrics and personal memories.

wat. Among 70 listeners across two experiments, the music people liked best was neither the most predictable nor the most surprising.

wat wat

Musical pleasure may come out of the play between a successful guess and an error you can learn from.

How do we know?

Two independent experiments combined liking ratings with a mathematical measure of musical expectation.

Tested in peoplePeer-reviewed study

What is still unknown

Both samples were small, and the model describes one aspect of music that leaves out lyrics and personal memories.

Learn something?

Gold et al., Journal of Neuroscience· 2019· DOI 10.1523/jneurosci.0428-19.2019

Where to next

Can an audience in a hall really hear a Stradivarius?musicDoes a Mozart sonata really reduce epileptic seizures?musicWhy does a chill arrive just before the peak of a song, and not at it?music

Want another one worth repeating?

Start with three short discoveries. No sign-up, nothing to pick first.

Surprise me
Browse the topic libraryAbout WatWat, and how sources get checked