Do Songs of the Summer Sound the Same?

Elena Georgieva
5 min readAug 16, 2018

--

This week, the New York Times published an article that immediately grabbed my attention:

Why Songs of the Summer Sound the Same

Why Songs of the Summer Sound the Same

We’re all familiar with the concept of a ‘song of the summer.’ You know — that song that plays all summer, every single time you turn on the radio, sometimes on multiple stations at the same time? You know you shouldn’t like it. You should be tired of it by now. Yet you let it play again and again… and when you’re at a bar 6 months later and the song comes on, you can’t help but smile as the song brings back memories of warm weather and carefree living.

Fast-forward to spring and the song is completely irrelevant. “Oh, a throwback,” you think to yourself as it plays in the grocery store.

A quick Google search reminds me of last summer, when “Despacito,” “That’s What I Like,” and “Shape of You” dominated the charts. What are the 2018 songs of the summer? It’s too soon to call, but New York Times writers Sahil Chinoy and Jessica Ma chose to look at three top contenders: “Psycho (Post Malone)”, “I Like It (Cardi B)”, and “Nice For What (Drake).”

The New York Times used Spotify’s API to gather information on these three songs, as well as three songs from 1988 and three songs from 2010. They take each track’s rating from 5 audio features:

  1. Loudness: “Volume of the song”

All songs have around the same max volume (producers use compression so we don’t have to constantly adjust the volume knob), so this feature likely measures the average volume rather than the peaks.

2. Energy: “How fast and noisy the song sounds”

This includes factors like dynamic range (loudest-quietest part of song), perceived loudness, timbre (character of the musical sound), and onset rate (think attack of each drum hit).

3. Danceability: “Strength and regularity of the beat”

Tempo, strength of each beat, rhythmic consistency. Who knew data could quantify what I like to dance to?

4. Acousticness: “Likelihood that the song uses acoustic instruments”

Think “Hey There Delilah” circa 2005.

5. Valence: “How Cheerful the song sounds”

The only way I can think to do this is with machine learning. Ah, technology…

As someone who’s used Spitfy’s API in the past, I am surprised at how the authors chose to omit several factors of available information in their analyses:

6. Speechiness: “How much spoken words are in a track”

Highly relevant as rap is starting to appear more and more on the top charts.

7. Instrumentalness: “Detects whether a track contains no vocals” — most pop tracks score 0 or near 0.

8. Liveness: “Detects whether the track was performed live” — again, most score near 0. Ironically, live is dead.

9. Tempo: “The beats per minute of a track”

Highly relevant in analyzing song similarity.

10. Duration: “Duration of the track in minutes.”

I’m confident with my prediction on this one: pop songs have been getting shorter, as have our attention spans…

11. Mode (Major/Minor)

12. Key or Tonality

13. Time Signature

So… do songs of the summer all sound the same?

I present to you an analysis, incorporating all 13 of these variables for the same 9 songs from the New York Times:

‘Songs of the Summer’ 1988: 1–2–3 (Gloria Estefan), Hold On To the Nights (Richard Marx), Pour Some Sugar on Me (Def Leppard)

‘Songs of the Summer’ 2010: California Gurls (Katy Perry), Ke$ha — Your Love Is My Drug (Official Video), Alejandro (Lady Gaga)

‘Songs of the Summer’ today! Psycho (Post Malone), I Like It (Cardi B), Nice For What (Drake)

Sonic fingerprins for `songs of the summer’ from 1988, 2010, and 2018 looking at seven audio features. Each is scaled [0, 1] where 0.5 is the average amount of that quality across all tracks on Spotify.
Song Tempo, Loudness, and Duration sorted by year. Audio features extracted from Spotify.
A bit more information about each song, formatted into a nice colorful table.

Looking purely high-level, it looks like the “songs of the summer” were somewhat varied in 1988 (ah, the good ‘ol days), and grew to be shockingly similar in 2010. This is especially obvious in the sonic fingerprint, where the fingerprints of the three 2010 tracks overlap almost entirely. Similarly, the loudness and tempo values of the three 2010 tracks are much more similar in value.

Song duration is decreasing (no surprises there), and the values of mode, key, and time signature are interesting but give little information on variation.

I was overjoyed to see in this data that “songs of the summer” are starting to show more variation this summer! We see a spread in the sonic footprint, as well as in the values of loudness and tempo.

Everyone, variety is coming back to pop music! We’re seeing more rap, notably more females rapping, more electronic music, lyrics in Spanish, perhaps a return of R&B, as well as crossovers between genres. Things are looking up, my friends! Things are looking up, and I’m here for it.

I’ve decided to put together a poster on this topic for this year’s ISMIR conference in Paris, France. Thanks for reading, and stay tuned for updates! :)

--

--

Elena Georgieva

Lecturer, Researcher @ Stanford University | music & technology | elenatheodora.com