harmonic surprise
The team's metric for how unexpected a chord is, based on the rarity of that chord across a normalized corpus of songs.
As defined in 104 | David Rosen and Scott Miles on the Neuroscience of Music and Creativity
Where it comes up · 1
Most important claims · 3 of 3Search all claims for “harmonic surprise” →
- When sections were aligned, the whole-song harmonic surprise advantage of hit songs was accounted for almost entirely by the verses and bridges leading into choruses; the choruses themselves showed no harmonic surprise difference whatsoever between bottom-quartile and number-one songs—meaning the high-tension pre-chorus, not the famous hook, carries the surprise effect.104 | David Rosen and Scott Miles on the Neuroscience of Music and Creativity
- The most popular (top-quartile) Billboard songs contain absolute harmonic surprise that is consistently about half a bit higher than the least popular (bottom-quartile) songs across the whole song, meaning hit songs are slightly more likely to feature less-common chords.104 | David Rosen and Scott Miles on the Neuroscience of Music and Creativity
- Harmonic surprise can be quantified by treating each chord as an event, computing its frequency across a corpus (e.g., 30,000 chords from 600 major-key Billboard songs, all normalized to one key and truncated at sevenths), then taking the negative log base two of its probability to yield a value in bits—where common chords like C major score around one or two bits and a rare chord like an F-sharp minor ninth scores around seven bits.104 | David Rosen and Scott Miles on the Neuroscience of Music and Creativity