- 18/08/2021
- Category: Digital Intelligence Lab ID
Ever heard the elderlies saying how they don’t dig the popular songs nowadays for the reason that it’s not concrete enough? Or how you feel like the old songs don’t capture enough of the essence that you’re looking for, but you can’t quite put your finger on what it is? It’s clear that certain aspects of songs have changed over the past decades, and in this research, we’re going to tell you what they are. Our research is based on the following questions:
- What have changed in song aspects, over the decades?
- What do these changes suggest?
Music and emotion have always been heavily linked. Sometimes, you’re having a normal day, but a shuffled sad, melancholy song suddenly makes you feel blue. Or how you would try to vibe to uplifting songs after a bad day. Whether you’re feeling something after listening to a certain song, or you’re listening to specific songs because of your current mood, it’s safe to assume that they’re interlinked.
A circumplex model of affect from Russell (1980) proposed an explanation of emotion theory based on 2 dimensions: valence and arousal. These two dimensions then create a quadrant, in which valence goes from left to right (unpleasant to pleasant), and arousal goes bottom to top (deactivated to activated). In short, valence would mean negative or positive emotions, while arousal is the intensity. Each quadrant would give different feelings, few examples would be:
- low-arousal positive emotions: calm, relaxed, content
- high-arousal positive emotions: happy, excited
- low-arousal negative emotions: bored, depressed
- high-arousal negative emotions: anger, tense, upset
As you read these examples, your brain might remind you of a few songs. Like how the valence and arousal are expressed and translated into music composition. This study would only focus on exploring the pattern in each decade’s top songs audio features and their linkage to emotions.
Research Design
The data was retrieved from Spotify’s playlists: All Out 50s, All Out 60s, All Out 70s, All Out 80s, All Out 90s, All Out 00s and All Out 10s. Each of these playlists are created by the official Spotify, described as the biggest songs in each decade. Thus, this study is exploring the pattern changes in the most popular songs—or in other words, the global public’s favorites—rather than the representation of the whole sets of music released during each decade. Data consists of the song’s energy, valence, instrumentalness, speechiness, mode, and tempo.
Results
Using an Independent-sample Kruskal-Wallis non-parametric test, it was confirmed that distribution of the songs’ energy, valence, instrumentalness, speechiness and mode are indeed different for each decade’s top songs, except for the tempo. Thus, there was no significant difference in tempo over the decades. Let’s take a closer look on how each aspect differs decade to decade:
- Energy
Figure 1. Energy scatter plot
A song’s energy is “a measure from 0.0 to 1.0 of perceptual measure of intensity and activity”, thus translates to a song arousal, going more intense up until 1. In the 50s, the top songs’ energy is more varied, having rather equal songs with low and high energy. As the decades go on, we can notice that the range and median go to the right and even less varied, being more intense over time. However, a decrease happens in the 2010s, going a little bit to the left than the songs in the 2000s.
- Valence
Figure 2. Valence scatter plot
A song’s valence is also a measure from 0.0 to 1.0, with 0 being negative and 1 being positive. In rough visualization, there seems to be not much of a difference in each decade. However, we can notice that in the 2010s, it ranges even wider, with a lot of frequent songs in middle valence (around 0.3-0.6). Few songs within this area are “thank u, next” by Ariana Grande, “Castle on the Hill” by Ed Sheeran, “Starving” by Hailee Steinfeld, in which illustrate longing for love. This pattern is slightly different than other decades with high frequencies on around 0.7-0.8 valence.
- Instrumentalness
Figure 3. Instrumentalness scatter plot
Instrumentalness is used to measure the composition—vocal-dominated or relying on sounds made by music instruments. The closer the score is to 1.0, the more likely it is having no vocal sound. The data might be hard to look at since most of the data swarms around 0.01 of instrumentalness, but here’s a certain highlight: as the decades go on, the top songs tend to have less instrumentalness. In the 1950s up until the 1990s, some songs vary in instrumentalness. However, it got fewer in the 2000s with “The Way I Are” by Timbaland with 0.75, while the highest scoring top song in the 2010s is “Titanium” by David Guetta ft. Sia with 0.15.
- Speechiness
Figure 4 Speechiness scatter plot
Speechiness, defined as “the presence of spoken words in a track”, explains that the closer a score is to 1, the more spoken words are identified in the song. As we can see, the 1950s’ top songs have rather varied speechiness, with “Little Bitty Pretty One” by Thurston Harris having the highest speechiness score with 0.24. From the 1960s, it seemed to be more swarming on the less speechy side, getting wider range over the decades. It peaked in the 2000s, with more diverse speechiness in songs. Few songs with higher scores of speechiness would be “Irreplaceable” by Beyoncé, “Empire State of Mind”, by Alicia Keys, and “Survivor” by Destiny’s Child. In the 2010s, there seems to be less songs with high speechiness than the 2000s.
- Mode
Figure 5 Mode bar chart
A song mode (whether in major or minor key) is also often associated with emotions, as a major key is usually often used in happy songs while a minor key for sad ones. Starting from the 50s, songs in a major key seemed to dominate the popular songs. This trend kept going, but became more balanced between major and minor keys over the decades. In the 2000s, it got even more balanced with 87 songs in major and 63 songs in minor. However, the gap got wider again in the 2010s with 103 songs in major.
After recognizing the patterns from each decade, a linear regression was run to see the model in general from all the songs as a whole. The decade of the release was the dependent variable, while valence, energy, instrumentalness, and speechiness were made as the independent variables. The result is as follows.
Table 1 Linear regression results
Each of the independent variables is found to be significantly linked to the decade of release, with the exception of instrumentalness (p>0.05). As the decades go by, the energy and speechiness go higher, with energy (0.527) having more increase than speechiness (0.132). Valence is also significantly found, but rather negatively linked, meaning that as the decades go by, the music released tends to have lower valence, also acknowledged as “sadder” music.
Changing of Trends
It seems like from the 1950s onwards, people seem to favor more intense, more speechy, and less instrumental music as the decades go by. There seems to be not much of a difference in valence range from the 1950s up until the 2000s. In the 2010s, though, a few of these patterns somehow turned around compared to the 2000s, such as how the energy and valence are more varied, a slight decrease in speechiness, and more songs in major.
As speechiness kept getting higher and more varied, the rise of hip hop and rap might have to do something with the trend. The history of hip hop can be traced back to around the 1970s from the African American communities in NYC, or Bronx specifically. The MCs would talk and rhyme while introducing the DJs and thus, over the years, became more prominent with some calling the 1980s and 1990s as the “golden age of hip hop” (Dye, 2007; PQ, 2019). However, as the charts showed, the top songs in the 2000s are more varied in speechiness and closer to the score of 1, implying that a lot of rap songs were favored by the public in the 2000s. Thus, the legacy of rap and hip hop itself still dominated pop culture in the 2000s.
Certain aspects can be seen getting prominent and more favored as time goes by. However, when analyzed specifically for each aspect, it can be observed that some of these aspects seem to change in the 2010s. The pattern in the 2010s is more diverse and harder to predict compared to the previous decades that seemed to create a consistent shift pattern. A few phenomena that happened in the 2010s might considerably relate to this unpredictability of top songs pattern, namely the rise of social media. Indirectly, the change in people communication also takes part in how the top songs type have shifted. Exposure and topics in social media have made it easier for foreign songs to top the charts globally. K-Pop or Korean Pop for instance, is one of the subcultures that benefited a lot in 2010s, with social media exposure taking part in its success of expansion to the West and globally (Ahn, Oh, & Kim, 2013). If people used to rely on charts from TV programs, critic reviews, and radio charts to discover new songs, meaning that they are only exposed to selected sources from certain figures, social media then enables lots of songs to get their buzz from the general public. Thus, it has become more decentralized as people can easily interact with everyone and vary their sources of music recommendation. In addition to social media, the presence of mass music streaming platforms in the 2010s, namely Spotify, Tidal, Deezer, Joox, and so on contributes to giving exposure to more artists. Accessibility is reaching wider both artists and audiences more than ever.
Electronic dance music (EDM), also named itself as one of the popular genres in early to mid 2010s, with live raves such as Tomorrowland having big crowds (Lombardo, 2019). Not only is the genres more varying, but also the vocal technique seems to create itself a new trend, whisperpop—the songs characterized with whisper-like vocalization rose to fame (Robinson, 2017). It’s safe to say that the top songs in the 2010s became more diverse and unpredictable than the previous decades.
Songs and Their Emotional Arousal
A song modality has always been assumed to be correlated with emotions, like how minor mode is usually used in sad songs (negative emotions), and major mode in happy songs (positive emotions). In determining whether a song is happy or sad, aside from analyzing the valence and the use of the modality, further analysis and interpretation of the lyrics are still needed. However, the use of major and minor keys are heavily related to the nuance intended—“happy-sounding” ones or blue, sorrowful-sounding ones. There are cases of songs being happy-sounding but are actually expressing sadness, regret, or bitterness through the lyrics (and thus, is counted as a sad song). Take, for example, “One Last Time” by Ariana Grande which expresses regret towards a lost partner, and having low valence, despite being in major key. But for most of the cases, the use of the modality and the nuance in the songs have been quite consistent. It was also proven that major chords tend to be used in songs with higher valence (Kolchinsky, Dhande, Park & Ahn, 2017). Parncutt (2014) offered an explanation how modality and emotion are linked, namely for reasons that minor keys are more dissonant, ambiguous, less common, and minor chords contain lower pitches which are more associated with sad speeches. So from the data, it seems like more “happy-sounding” songs are more balanced with the “sad-sounding” songs over the decades.
Human emotions exist not in one dimension, thus can be understood by comprehending each dimension. Both arousal (proxied by energy) and valence work in deciphering emotions. For example, the song with the lowest valence in the 1950s is “Love Me Tender” by Elvis Presley, while the lowest in the 2010s is “All of Me” by John Legend. Though both have similar valence (0.3), both differ in energy. “All of Me” has the energy score of 0.264, while “Love Me Tender” has the energy score of 0.037, meaning “All of Me” evokes more intense emotions rather than “Love Me Tender”. Another example would be despite having rather similar energy (0.36), “Eternal Flame” by The Bangles and “Happier” by Ed Sheeran evoke emotions differently as they differ in valence, with “Eternal Flame” having 0.499 valence, in contrast to “Happier” which has 0.236 valence. Both arouse rather equivalent intensity, but with different emotions—which is also backed up by the lyrics: “Eternal Flame” reassures the significant other, while “Happier” tries to let go.
Hence, both music and emotions are intertwined in reflecting each other. What humans are feeling at a specific time might trigger someone to listen to a specific song, and vice versa, a song might also arouse certain emotions. Each decade has a unique pattern of the public’s preference of music in that time setting, which can be seen on how the patterns from each aspect can differ from time to time.
In conclusion, other than the trend change over the decades, these changes also reflect how human feelings are translated into an artwork that is music. Take a look at the scatter plot of valence and energy change from each decade as shown in Figure 7. It’s worth noting how human behavior, what they’re going through—socio-culturally, historically, across media from electronic to cyber ones—can reflect in popular music. As the phenomena explained above in the 2010s were assumed to be related to the decade’s top songs pattern, future studies might dive deeper into these phenomena and their linkage to the pattern. It’s also interesting to see whether the upcoming decades, starting from the 2020s, will follow the same unpredictability or rather go back creating another pattern starting from the 2010s.
Figure 7. Valence and energy of each decade scatter plot
Author: Nuhida Kinansa Husainy
Reviewer and Editor: Josia Paska Darmawan
Data collecting: Vidiskiu Fortino K.
REFERENCES
Ahn, J., Oh, S., & Kim, H. (2013). Korean pop takes off! Social media strategy of Korean entertainment industry. 2013 10th International Conference on Service Systems and Service Management, 774-777.
Dye, D. (2007). The Birth of Rap: A Look Back. NPR Music. Retrieved from https://www.npr.org/templates/story/story.php?storyId=7550286
Kolchinsky, A., Dhande, N., Park, K., & Ahn Y. (2017). The Minor fall, the Major lift: inferring emotional valence of musical chords through lyrics. R. Soc. open sci, 4, 1-11.
Lombardo, S. (2019). The Evolution of Dance Music Genres in the 2010s. EDMTunes. Retrieved from https://www.edmtunes.com/2019/12/evolution-dance-music-2010s/
Parncutt, R. (2014). The emotional connotations of major versus minor tonality: One or more origins? Musicae Scientiae, 18(3), 324-353.
PQ, R. (2019). Hip Hop History: From the Streets to the Mainstream. The Icon Collective. Retrieved from https://iconcollective.edu/hip-hop-history/
Robinson, P. (2019). ‘Whisperpop’: why stars are choosing breathy intensity over vocal paint-stripping. The Guardian. Retrieved from https://www.theguardian.com/music/2017/nov/11/whisperpop-why-stars-choosing-breathy-intensity-over-vocal-paint-stripping
Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161–1178.
Spotify for Developers. (n.d.) Reference Index. Retrieved from developer.spotify.com/documentation/web-api/reference/#category-tracks
