How Has Popular Music Changed Since the 1950s?
All I know since yesterday / is everything has changed. (Taylor Swift, 2012.)
So don't fear if you hear / a foreign sound to your ear. (Bob Dylan, 1965.)
How has popular music changed since the 1950s?
I mean, apart from obvious changes like the rise of new genres (rock, hip-hop, etc), new instruments (synthesizers, drum machines, etc.), and new forms of distribution (CDs, streaming, etc.), how have the songs themselves changed, and what do these changes tell us about the state of our culture?
In this newsletter I'll be sharing some new data on the topic.
Why is this important?
Some of the new findings seem to hint at cultural decline. For instance, over the past seven decades, melodies have become simpler, while lyrics have grown more redundant and negative. From data like this, it's a short hop to the conclusion that our culture is becoming dumber and more gloomy. In a 2022 New York Times essay, David Brooks actually used the song lyric data in support of such a conclusion. (See my response here). Others have stuck to the narrower conclusion that it's merely the music itself that's getting worse. (Try Googling "Is popular music getting dumber?").
As you'll see, these are terribly simplistic readings of the evidence. We need to be more careful with the data – and with broader inferences about cultural change.
A quick note on scope
My focus here is on popular music, which can be defined in many ways but refers, in essence, to music that lots of people purchase and/or listen to. ("Pop" is just one genre of popular music, along with rock, hip-hop, R&B, dance/electronic, country, etc.)
More specifically, I'll be focusing on the most popular songs in America, as the data I review is connects with Billboard's U.S. charts. (The popularity of these songs is similar among western countries but not identical. This week, for instance, the top 5 songs on Billboard's Hot 100 are Not Like Us, A Bar Song, I Had Some Help, Million Dollar Baby, and Espresso. On Billboard's Global 200, these songs are ranked #1, #4, #6, #8, and #2, respectively.)
A new study
The new study, published in Scientific Reports this July 4, was authored by Madeline Hamilton and Marcus Pearce in the Music Cognition Lab at Queen Mary University of London.
Hamilton and Pearce analyzed changes in popular music melodies from 1950 through 2022. Their sample consisted of the Billboard year-end top 5 songs for each year.
Billboard's rankings are based on physical and digital sales, radio airplay, and streaming. Thus the top 5 songs are probably the ones that the largest number of people have heard at least once, as well as the ones people most often hear more than once.
Hamilton and Pearce analyzed MIDI files available from Billboard for all songs. MIDI files are suitable given the researchers' interest in the abstract features of melody that make it more or less complex.
To illustrate those features, compare the Star-Spangled Banner, Mary Had a Little Lamb, and Papa's Got a Brand New Bag. (I'll bet you've never seen those songs mentioned in the same sentence before.)
The melody of Star-Spangled Banner includes extensive changes in pitch. By the time you've sung "Oh-oh say can you see", you've already spanned an entire octave. That's more range than we find across the entire melodies of the other two songs. We can say then that the Star-Spangled Banner is relatively complex with respect to variability in pitch, as well as in the average distance in pitch between one note and the next.
At the other end of the spectrum, Mary Had a Little Lamb is quite limited in pitch. The entire melody consists of four notes, without much distance between the highest and lowest ones. The tempo is limited too: It's basically quarter notes, and the pauses after the word "lamb" are brief and comparable in duration. We can say then that Mary is relatively simple with respect to variability in pitch, the distances between one pitch and the next, and the duration of notes as well as the silences between them. It's also simple owing to its redundancy (which makes it predictable) and to the fact that it tends to be sung slowly.
Papa's Got a Brand New Bag has a broader range in pitch than Mary, though not as much as the Star-Spangled Banner. But there are huge differences in the complexity of the tempo. In James Brown's original recording, much more time passes between lines (almost 3 seconds between one line and the next). Then the notes come fast. "Ain't no drag" drops nearly an octave in about a second. As the third verse opens, there's an unexpected reduction in the amount of time between lines. So, we can say that it's a very complex song with respect to tempo, and perhaps moderately complex in pitch.
These are crude illustrations of most of the pitch and tempo features of melodic complexity that Hamilton and Pearce analyzed.
The researchers' main findings include an overall decrease in the complexity of popular music melodies since 1950, along with an increase in the density of notes. During this seven-decade period, melodies have become less variable in pitch, more predictable in rhythm, and otherwise less complex, but at the same time we're hearing more notes per second. In short, simpler, quicker melodies. I'll discuss what that might mean in a moment.
Some evaluative remarks
This is a great study, in my view. It's also a great example of the kind of lab study that can get picked on for the wrong reasons.
First, the sample is obviously limited. Hamilton and Pearce acknowledge this in their article, and Madeline Hamilton reiterated the point in an email to me, noting that the top 5 songs per year is an "infinitesimally small fraction of the music listeners enjoy".
I may view the sample more positively than Hamilton does. Yes, it's extremely small, but sample representativeness can be more important than size. The top 5 songs in any given year are probably fairly representative of, say, the top 10 or 20 songs. A lot of people hear those songs, including the artists who'll be creating the popular music that listeners encounter in the years to come.
Second, the measures are obviously limited, since MIDI files reduce melodies to single notes. As Hamilton pointed out in her email to me, "when we listen to melodies, we always listen to them in the context of the song, i.e., with all the other instruments playing at the same time."
In songs like In Da Club (#1, 2003) and Yeah (#1, 2004), for instance, the opening riff is repeated throughout the song (six notes in In Da Club, four notes in Yeah) and form a sort of counterpoint to the singer's words. This creates a melodic experience that the MIDI files couldn't reflect. In addition, MIDIs fail to capture that slight, poignant crack in a singer's voice on certain notes, or the timbre of their voice, or the interplay between the notes and the lyrics.
All the same, MIDI files do allow researchers to isolate melodic features that shape the listening experience. As Hamilton noted in her email to me, objective complexity (the kind she and Dr. Pearce explored) is correlated with listeners' subjective ratings of complexity. This, the study reveals meaningful historical trends that exist alongside whatever else affects the perceived complexity of popular songs.
What do the findings mean?
Hamilton and Pearce are admirably cautious. Rather than drawing firm conclusions, they lay out two kinds of interpretations consistent with their data.
One possibility is that because people tend to like moderately complex music, popular songs achieve a sort of equilibrium: If some aspects of a song are more complex, others tend to be simpler. Thus, the increasing simplicity of melodies since 1950 may be balanced by greater complexity in other attributes. A likely candidate is the increase in note density that Hamilton and Pearce found. Melodies are simpler now, but they contain more notes per second. Regardless of the overall tempo of the songs, the melodies are moving more quickly.
From a singer's perspective, this makes sense. The simpler the melody, the more rapidly it can be sung. Eminem's "supersonic" rap in the midst of Rap God – just over 6 words per second – is only possible because there's almost no variation in pitch. Nobody could sing the Star-Spangled Banner, on key, at anywhere near that speed.
This possibility also makes sense given other data suggesting that the overall tempo of popular music has declined since the 1950s. Slower tempos allow more notes to be articulated per second.
A second possibility is that decreasing melodic complexity is reflective of cultural decline. That's a disturbing notion. Referring to the work of director and composer Yuval Shrem, Hamilton and Pearce note
"The digital era increasingly demands the compression of language, so that what we have to say will stay under character limits and fit into headlines. According to Shrem, this decreases the complexity of our messages and perhaps even our ability to digest complex ideas, and this manifests in the popular music we enjoy. Philosopher and music critic Mark Fisher argues that, in addition to this fragmentation of language, we are overwhelmed with the rapid pace of modern culture and technological progress and therefore do not have the mental capacity to enjoy or create truly complex or novel art."
I'll revisit these possibilities in a moment. In my view, the first one seems more consistent with the data. In particular, the emergence of hip-hop and rap, which are just as complex musically as earlier genres, may account for Hamilton and Pearce's findings as well as others.
A new analysis
On July 10, six days after publication of Hamilton and Pearce's paper, data scientist Daniel Parris posted an article to his newsletter Stat Significant entitled "How has music changed since the 1950s? A statistical analysis".
Although Parris analyzed different variables, his findings are both consistent with and extend Hamilton and Pearce's.
Parris explored Billboard Hot 100 songs over the past seven decades, making use of Spotify's database on five attributes of each song.
1. Duration.
Parris found that song length increased up through the late 1990s and then began to decline. From a listener's perspective, this is a big effect – song lengths averaged around 4.5 minutes in the 1990s but only about 3 minutes now.
Parris links the increase in length to technological changes, such as transitions from vinyl to cassettes and then CDs. As for the decline, he notes that the "laziest" explanation is a reduction in consumer attention span. More plausible, in his view, are changes in the music industry. iTunes and streaming led musicians to focus on individual songs rather than albums, and streaming services shifted to pay-per-play revenue models. These changes favored shorter individual tracks. (Why create three 4-minute songs when you can earn more money from four 3-minute songs?)
2. Danceability.
Spotify defines this as "how suitable a track is for dancing". So, presumably, Bridge Over Troubled Water (#1, 1970) gets a much lower score than Levitating (#1, 2021).
Spotify doesn't appear to publicize how danceability is calculated, other than noting that it's based on "tempo, rhythm stability, beat strength, and overall regularity", and that scores range from zero to 1.
Parris found that the average danceability of popular music has increased since the 1950s. It's a small effect (from 0.5-ish to 0.65-ish on that 0 to 1 scale) but meaningful, given that it's based on a variety of music, not all of which is intended to be danceable.
Parris offered a number of theories for this change. Most resonant for me is a theme that cross-cuts some of these theories and connects with Hamilton and Pearce's data: The ascendance of hip-hop and rap. As you might guess, Parris found that Spotify tends to view songs from these genres as most danceable, with traditional pop and soft rock scoring lowest on this dimension. Hey Jude (#1, 1968) may be a deeply moving song, but it doesn't move your feet.
With the growth of hip-hop and rap, it's understandable that analyses like Parris's would show increasing danceability since the 1950s. It's also understandable that Hamilton and Pearce would find less melodic complexity during the same time period. This is not to say that hip-hop and rap are less complex forms of music. Rather, when focusing on melodic elements such as variability in pitch, that's not where the complexity in these genres resides. 50 Cent's rapping In Da Club (#1, 2003) includes some subtle shifts, but the majority of the words represent the same pitch.
3. Instrumentalness.
Parris noted a decline in the instrumental elements of popular music – e.g. lengthy guitar solos – a shift that goes hand-in-hand with the next major trend:
4. Speechiness.
With the decline in instrumentalness, lyric-driven music has become more prevalent. Here again, I think we're seeing the influence of hip-hop and rap, not only as separate genres but as forms of vocal delivery that influence other genres. In Last Night, the #1 hit from 2023, you can hear Morgan Wallen shift into a rap at the 27 second mark for about 10 seconds ("No way it was the last night..."). It's a subtle change – not much of a departure from the way he'd been singing up to that point – but it adds words to the song at a high rate (while reducing melodic complexity, since most of this part is sung at the same pitch).
5. Valence.
"Valence" refers to the emotional content of a song, and Parris noted a steady decline in positivity since the 1950s.
Spotify doesn't explain exactly how the company measures valence, but a number of peer-reviewed studies, using different approaches to measurement, show that over the past seven decades, popular song lyrics have become gloomier, angrier, and more pessimistic – in a word, more negative. This is consistent with a proportional increase in the number of songs written in minor keys.
However, few of the experts who've identified these trends, including Parris, claim to be certain that more negative lyrics mean that people themselves are becoming more negative. (A notable exception, perhaps, is the critic Ted Gioia, who does link lyrical valence to cultural mood, albeit in a nuanced way.) Maybe we're gloomier than we used to be, maybe not. Song lyrics don't necessarily reflect public state of mind.
For instance, I believe that the shift in lyrical valence over the past 70 years reflects, at least to some extent, greater receptivity to public expression of the kinds of personal struggles, sensitive topics, and dark themes that are now routinely aired in popular music.
People have always experienced profound gloominess, and artists often find ways to convey it (think of Strange Fruit, Billie Holiday's stunning 1939 response to the lynching of Black Americans). But selling a lot of records in the 1950s and early 1960s was easier for songs like Surfin' U.S.A. (#1, 1963), I Wanna Hold your Hand (#1, 1964), or Wooly Bully (#1, 1965), fluffy tunes that would both appeal to and be deemed suitable for the burgeoning mass audience of young people. Over time, as the youth market became more established and cultural norms shifted, there was more room in mainstream music for anger, excessive sadness, despair, and responses to everything else that flew out of Pandora's box.
Conclusions
How has popular music changed since the 1950s? Among other things, the melodies are less complex but move more quickly. Overall tempos are slower. The songs are more danceable yet contain more lyrics. The lyrics themselves are darker.
In my view, these changes reflect the influence of hip-hop and rap, along with greater public openness about difficult themes, although other factors surely played a role, and there were changes I haven't discussed, such as shifts in standardization that Daniel Parris teased out in an article this February.
To the extent that hip-hop and rap had such an influence, the evolution of popular music since the 1950s does reflect an important cultural change: The increasing presence and influence of Black voices in mainstream culture, coupled with greater acceptance of whatever those voices choose to say, even when the message is violent, or sad, or otherwise darker than I Wanna Hold Your Hand.
This is not to say that musical culture is free of racism, or that Black artists are the only musicians who record hip-hop and rap tracks, or that these are the only genres that feature relatively simple, fast-paced melodies and negative lyrics. Rather, I'm just suggesting that a general historical trend (for which there are many exceptions) may be attributable in part to the rise of these genres. Historical change doesn't mean that everything changes. It simply means that some things change, in the same way, and the change is extensive enough to impact our perception of the whole.
In their Introduction, Hamilton and Pearce note that during the 20th century, writers in a variety of fields produced sophisticated narratives about popular music, but it wasn't until the recent digitization of music and the development of suitable quantitative methods that more sophisticated analyses could be run.
Statistics do add to our understanding of popular music. If you approach the topic informally, without the stats, you risk getting caught up in the music you're most familiar with or find most meaningful. For instance, when I first saw the data on increasingly negative lyrics, I felt skeptical, because I immediately thought of Robert Johnson, Woody Guthrie (check out the verses of This Land is Your Land that people don't sing), Billie Holiday, and countless others who conveyed bleakness and despair and anger with their words even before the 1950s. But the data tells me that no matter how many counter-examples I might cite, the lyrics of the most popular songs have, on the whole, become more negative over time. We'll see what the future holds, since, as always, the times will continue to be a-changing.
Thanks for reading!