Researchers at the University of California, Berkeley have developed a machine learning algorithm to analyze storytelling in pop music lyrics. The study examined more than 5,000 songs from the Billboard Hot 100 list spanning 1960 to 2024, seeking to understand how narrative elements in popular music have evolved over time.
David Bamman, associate professor in the School of Information at UC Berkeley and first author of the study, explained the motivation behind the research: “We wanted to see if computational methods could measure the stories that are present in songs in order to help us understand how storytelling in music has changed on a larger scale over the past half of a century.”
Contrary to expectations that narrative songwriting peaked during the ballad-driven era of artists like Joan Baez and Bob Dylan in the 1960s, researchers found an increase in narrativity since the 1990s. This trend is largely attributed to hip-hop’s rise, which often features strong storytelling components. For instance, Ice Cube’s “It Was a Good Day” ranked as one of the most narrative songs analyzed, just ahead of Taylor Swift’s “All Too Well.” The lyrics describe a day-to-day experience involving relationships and challenges such as racism and police surveillance.
Tom McEnaney, corresponding author and associate professor at UC Berkeley’s Departments of Comparative Literature and Spanish & Portuguese and director of the Berkeley Center for New Media, commented on broader implications: “Many of us are excited about the computational modeling of cultural data because it grounds some of the claims that can be made from a smaller set of examples. When you have this larger set of data, you can get a picture of literary changes or developments that are clarifying and often go against the prevailing wisdom in that field.”
The study also challenges previous assumptions within literary theory regarding genres. McEnaney noted: “There was this assumption that poetry gets into popular music through hip-hop, and that hip-hop is the most lyric of genres. If literary theory also tells us that lyric is non-narrative, then we wouldn’t expect to find hip-hop as the main driver of narrativity in popular music. But when you get to the 1990s, you just see the narrativity score shoot through the roof. This turns on its head the last 30 years of literary approaches to hip-hop.”
Bamman shared how his interest began after hearing Bruce Springsteen’s “Thunder Road,” realizing its narrative strength: “I finally got it, because the song is a story,” Bamman said. “It was really clear that Bruce Springsteen is a storyteller, and I could see how that would resonate with people.”
For over ten years Bamman has worked on computational tools for analyzing literature and film. In previous research published last year by Bamman’s team using facial recognition technology on film footage confirmed increased diversity among actors appearing onscreen over two decades (source).
The current project involved undergraduate students Sabrina Baur, Mackenzie Hanh Cramer and Anna Ho who helped evaluate more than 1,000 song lyrics based on objective criteria such as presence of characters or agents, sequences of events unfolding over time or creation of vivid worlds.
Using these assessments they trained their algorithm to analyze thousands more songs for narrative content.
Findings also showed storytelling marks prestige within country music; Grammy-nominated Best Country songs featured higher levels of narration compared with other tracks from those albums.
McEnaney added: “People often think of hip-hop as a primarily Black genre and country as a primarily white genre. Beyoncé’s Cowboy Carter is only the most recent challenge to that assumption, and at a moment of absolute racial divisiveness in the country, it’s interesting to think about how those genres are the most popular genres and also have this shared interest in telling stories.”
