CoChat

Song lyrics have become simpler and more repetitive over the last five decades

Open access
Parada-Cabaleiro, Emilia, Mayerl, Maximilian, Brandl, Stefan, Skowron, Marcin, Schedl, Markus, Lex, Elisabeth, Zangerle, Eva · 2024 · Scientific Reports

The DOI matches its Crossref record; 54 of the 54 references it cites were found in a registry or an index; none of the 36 sources its claims cite is checked for support yet.

Abstract

Music is ubiquitous in our everyday lives, and lyrics play an integral role when we listen to music. The complex relationships between lyrical content, its temporal evolution over the last decades, and genre-specific variations, however, are yet to be fully understood. In this work, we investigate the dynamics of English lyrics of Western, popular music over five decades and five genres, using a wide set of lyrics descriptors, including lyrical complexity, structure, emotion, and popularity. We find that pop music lyrics have become simpler and easier to comprehend over time: not only does the lexical complexity of lyrics decrease (for instance, captured by vocabulary richness or readability of lyrics), but we also observe that the structural complexity (for instance, the repetitiveness of lyrics) has decreased. In addition, we confirm previous analyses showing that the emotion described by lyrics has become more negative and that lyrics have become more personal over the last five decades. Finally, a comparison of lyrics view counts and listening counts shows that when it comes to the listeners’ interest in lyrics, for instance, rock fans mostly enjoy lyrics from older songs; country fans are more interested in new songs’ lyrics.

Overview

CoChat Summary

TL;DR

Analyzing 353,320 English-language Western popular songs from 1970–2020 across five genres, the paper finds that lyrics have become simpler, more repetitive, more personal, and more negative over time. It also shows that lyrics view counts, unlike listening counts, vary by genre and release year, with rock fans favoring older lyrics and country fans favoring newer lyrics.

Key findings

  • Lyrics became more repetitive across all five genres: repeated line ratio increased over time, strongest for rap (m=0.002516) and weakest for country (m=0.000640); ratio of chorus to sections also increased for all genres, strongest for rap (m=0.008703) and weakest for R&B (m=0.000325).
  • Lyrics became more personal and more negative: pronoun frequency increased for all genres except country (m=-0.000145), strongest for rap (m=0.000926); anger increased across all genres, most for rap (m=0.015996); negative emotion increased across all genres, highest for rap (m=0.021701) and lowest for country (m=0.000606). Positive emotion decreased for pop, rock, country, and R&B, while rap showed a moderate increase (m=0.000129).
  • Release-year regression models explained the most variance for rap (adjusted R²=0.32) and R&B (adjusted R²=0.19), and less for country (0.11), pop (0.09), and rock (0.09). Repeated line ratio was the only descriptor with a meaningful positive effect for all genres.
  • Lyrics view count had a positive relationship with release year for R&B (β=0.32, p=.003) and a strong negative relationship for rock (β=-1.47, p<.000); multinomial logistic regression showed country-vs-rap odds increased with lyrics view count and year (β=0.07, SE=0.02, z=3.29, p=.0009), while rock-vs-rap odds decreased (β=-0.05, SE=0.01, z=-5.89, p<.0001). Listening counts showed no significant effects.

Methods

  • Data: LFM-2b last.fm listening events enriched with Genius release year, genre, lyrics, and popularity data; 353,320 English songs spanning 1970–2020 across rap, country, pop, R&B, and rock.
  • Analysis 1: per-genre Huber M robust regression predicting release year from lyrical descriptors after z-score normalization and VIF-based removal of multicollinear descriptors; descriptor importance assessed via regression coefficients.
  • Analysis 2: balanced downsampled dataset of 12,000 songs (2,400 per genre) restricted to 1990–2020; multinomial logistic regression with genre as outcome and popularity × release year predictors, then per-genre multiple linear regression predicting release year from lyrics descriptors and lyrics view count using backward stepwise selection.

Limitations

  • The paper states that last.fm and Genius introduce community bias and popularity bias: last.fm users are not representative, skew male, ages 20–30, and are dominated by the US, Russia, Germany, and UK; pre-platform listening and lyrics-view counts underestimate true frequencies, and physical releases may be missing.
  • The paper states further limitations: genre assignments on Genius may be malformed; the study is restricted to English lyrics; readability metrics are simple and sentence-based; models assume linear change over time; downsampling may introduce sampling bias, especially for pop and country.
  • The text itself makes evident that explanatory power is often low: adjusted R² values are small for several genres, the multinomial model has McFadden pseudo R²=0.01, and the authors note that significant p-values may partly result from the huge sample size and random noise.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41598-024-55742-x

Cites

54 of 54 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Overview of data and analyses performed. Based on a wide variety of descriptors capturing the lyrical characteristics from listening data and lyrics content, we perform two analyses. Analyses 1 identifies descriptors that are characteristic of the release year and genre. Analyses 2 investigates the relationship between the identified lyrics descriptors, popularity (listening counts and lyrics view counts), and release year.
Figure 1 Overview of data and analyses performed. Based on a wide variety of descriptors capturing the lyrical characteristics from listening data and lyrics content, we perform two analyses. Analyses 1 identifies descriptors that are characteristic of the release year and genre. Analyses 2 investigates the relationship between the identified lyrics descriptors, popularity (listening counts and lyrics view counts), and release year.CC BY · PubMed Central
Figure 2 Evolution over time for the descriptors repeated line ratio and ratio chorus to sections for each genre. The linear regression lines (in red) show the evolution of descriptor values over time for each descriptor and genre (Huber’s M robust regression models are trained individually for each descriptor and genre combination). Blue colors denote the density of data points in a given region. R2 and p-values are provided in the yellow boxes.
Figure 2 Evolution over time for the descriptors repeated line ratio and ratio chorus to sections for each genre. The linear regression lines (in red) show the evolution of descriptor values over time for each descriptor and genre (Huber’s M robust regression models are trained individually for each descriptor and genre combination). Blue colors denote the density of data points in a given region. R2 and p-values are provided in the yellow boxes.CC BY · PubMed Central
Figure 3 Forest plot displaying the estimated multinomial logistic regression coefficients (standardized beta) for the prediction of musical genre. As reference class, rap i. e., the genre with the highest average lyrics view count, is considered.
Figure 3 Forest plot displaying the estimated multinomial logistic regression coefficients (standardized beta) for the prediction of musical genre. As reference class, rap i. e., the genre with the highest average lyrics view count, is considered.CC BY · PubMed Central

Verification

Claims the paper makes

0 Valid · 3 Not valid · 26 Not checked · 87 Own results and interpretations

42 more claims

Create a free account to see all 116 claims and to verify the 35 references still open.

See all claims — free account
  • It may state two separate claims as one. 50
  • Probably not a substantial claim. 32

Create a free account to see these claims.

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Not confirmedz = 3.29reported p = .0009 · recomputed p = .001

    “z=3.29, p = .0009”

  • Consistentz = -5.89reported p < .0001 · recomputed p < .001

    “z=-5.89, p < .0001”

2 of 2

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

3 of 3

Citations

Cites

Show all 54 references

Research works printed but not linked: 9

  1. Fell, M. Lyrics classification. Master’s Thesis, Saarland University (2014).
  2. Huber, P. J. Robust Statistics Vol. 523 (John Wiley & Sons, 2004).
  3. Pinheiro, J. et al. Package ‘nlme’. Linear Nonlinear Mixed Effects Models Version 3, 274 (2017).
  4. Toop, D., Cheney, C. & Kajikawa, L. Rap. In New Grove Music online (Oxford music online).
  5. Rye, H. Rhythm-and-blues. In New Grove Music online (Oxford music online).
  6. Middleton, R. Rock. In New Grove Music online (Oxford music online).
  7. Mass, H.-D. Über den Zusammenhang zwischen Wortschatzumfang und Länge eines Textes. Zeitschrift fur Literaturwissenschaft und Linguistik 2, 73 (1972).
  8. Go, A., Bhayani, R. & Huang, L. Twitter sentiment classification using distant supervision. CS224N project report Stanford 1, 2009 (2009).
  9. Mohammad, S. M. & Turney, P. D. NRC emotion lexicon. Natl. Res. Counc. Can. 2, 234 (2013).

Cited by

See all 78

Metadata

Authors
Parada-Cabaleiro, E., Mayerl, M., Brandl, S., Skowron, M., Schedl, M., Lex, E., & Zangerle, E.
Affiliations
  • Universität Innsbruck (AT)
  • Austrian Research Institute for Artificial Intelligence (AT)
  • Johannes Kepler University of Linz (AT)
  • Graz University of Technology (AT)
Corresponding author
Zangerle, Eva
Venue
Scientific Reports
Published
2024-03-28
Volume
14
Issue
1
Pages
5531-5531
ISSN
2045-2322 (ISSN-L)
OpenAlex
W4393255711
Cited by
73
References
53
Citations by year
2026: 36 · 2025: 28 · 2024: 9
FWCI
108.37
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Digital Library of the University of Innsbruck (University of Innsbruck) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Austrian Science Fund (DFH-23, P33526)
Topics
  • Social Sciences › Arts and Humanities › Music › Music History and Culture
  • Social Sciences › Arts and Humanities › Music › Musicology and Musical Analysis
  • Social Sciences › Arts and Humanities › Literature and Literary Theory › Digital Humanities and Scholarship
Keywords
song lyrics, lexical complexity, readability, repetition, music genres, popular music, rock music, country music
MeSH terms
Emotions, Music, Vocabulary

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-20, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-20 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Song lyrics have become simpler and more repetitive over the last five decades. https://app.cochat.ai/papers/10.1038/s41598-024-55742-x
The paper
Parada-Cabaleiro, E., Mayerl, M., Brandl, S., Skowron, M., Schedl, M., Lex, E., & Zangerle, E. (2024). Song lyrics have become simpler and more repetitive over the last five decades. Scientific Reports. https://doi.org/10.1038/s41598-024-55742-x

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.