If you think popular music has grown angrier and sadder over the years, it’s not just your imagination. It’s science.
A scientific analysis of popular music lyrics from the 1950s through 2016, conducted by a student and professor at Lawrence Technological University, showed that song lyrics have become both angrier and sadder since the 1950s.
The analysis, conducted by Kathleen Napier of Novi, Mich., who graduated with a master’s degree in computer science in May 2018, and computer science professor Lior Shamir, also showed that sentiments such as disgust and fear have increased over the years, while the expression of joy has declined.
Napier and Shamir’s analysis was accepted for publication in the Journal of Popular Music Studies, an academic journal published by the University of California. It is included in the December 2018 issue at http://jpms.ucpress.edu/content/30/4/161.
“Music fans in the more recent years prefer songs with sadder and angrier lyrics, which were much less popular in the 1950s or 1960s,” Shamir said of the study.
Napier and Shamir analyzed the lyrics of more than 6,000 songs of the Billboard Hot 100 in each year from 1951 through 2016. They note that the Billboard Hot 100 has been traditionally ranked by popularity measurements such as record sales and radio broadcasting of a song, with indicators such as streaming and social media mentions added in more recent years.
The emotions expressed in each song were analyzed by applying automatic sentiment analysis., a process of analyzing the association of each word or phrase with a set of sentiments and tones that the word or phrase expresses. The combination of sentiments expressed in all words and phrases of the song determine the tones and sentiments expressed in it. The sentiments expressed in the songs of each year were averaged, and the average of each year determines whether the expression of that sentiment increases, decreases, or remains constant over time.
The analysis showed that the expression of anger in popular music lyrics increased over time. Anger was least expressed during the mid-1950s, and increased gradually until 2015 and 2016, which were the last two years tested in the study.
The expression of sadness, disgust and fear also increased over the years, although the increase was less significant than the increase in anger. The analysis also showed that joy was a dominant sentiment of popular music lyrics in the late 1950s, but had become much less dominant by 2016.
Researchers pointed out that since the study covered the most popular songs of each year, the study didn’t show that music has changed – but rather, that the public’s tastes in music changed. While music fans preferred joyful songs during the 1950s, modern music consumers are more interested in songs that express sadness or anger. The paper was submitted in early 2018 and has been under review since.
Shamir said the idea for the study came up in discussions with the student during the spring 2017 semester. Napier, whose degree concentration was data science, wrote the code to automate the analysis of thousands of files of lyrics. Shamir said automatic sentiment analysis is most commonly used by businesses to analyze social media posts and customer satisfaction reviews.
Napier said she had worked earlier on text analysis in political tweets and classical literature, and was working with Shamir on a course manual introducing data science concepts to computer science students when the idea for the study came up. She said she used a “tone analyzer” developed for IBM’s Watson question-answering computer system to analyze the sentiment of the lyrics.
Shamir has long had an interest in using computer science techniques to analyze music and sound, earlier publishing a computer-aided analysis of the evolution of the music of the Beatles, and publishing another paper on using computer techniques to analyze whale song. He’s also applied computer analytic techniques to everything from determining the spin of galaxies to differentiating between the abstract art of professional artists vs. children’s doodles.