Pursuing the Psychological Building Blocks of Music

Five years ago, cognitive scientist Samuel Mehr asked a question that has, no doubt, played in the heads of many of us who have dipped into Spotify’s vast international song database: Is music universally present in different societies across the world? If so, what aspects of music do different societies have in common? Music is often discussed as being universal, even by scientists, but these claims are rarely backed by data.

A graduate student at the time, Mehr partnered with fellow student Manvir Singh and got to work. Eventually building a large team of researchers, including anthropologists, psychologists, and political scientists, they now have the data to answer Mehr’s original question. They recently published their findings in Science.

The first step in their research: decide what counts as music. For this project, they defined it as vocal song, because of its fundamental nature across the world. Next, they analyzed an extensive ethnography database comprising 315 societies, developed and maintained by the Human Relation Area Files (HRAF) organization at Yale University. In every society’s ethnography, they found mention of music.

Given the diversity of the 315 societies analyzed—from the Highland Scots to the Libyan Bedouins to the Maasai of eastern Africa—the presence of musical behavior in all of them is noteworthy. For the first time, there is data to support the claim of the universality of music.

For the first time, there is data to support the claim of the universality of music.

But some music scholars aren’t necessarily impressed. In their article, Mehr and colleagues quote ethnomusicologist George List, who they say sums up the view among many global music scholars: “The only universal aspect of music seems to be that most people make it.… I could provide pages of examples of the non-universality of music. This is hardly worth the trouble.”

The more meaningful question, perhaps, is whether the music from all of these different societies has something in common. Is it just that everywhere people produce music, or are there similarities to that music that run deeper? Is music purely a by-product of culture, or do humans possess a sort of “musical grammar,” akin to language, that allows us to produce and understand music?

This question is where Mehr and his colleagues turned next. To answer it, they needed to build a database of their own. They selected 60 societies representative of the cultural diversity in the Yale database. Since they already possessed the ethnographic information, what they needed was systematic data on the contexts in which music occurred. So the team coded the ethnographic data for these 60 societies, noting information like the number of people present, any lyrics, and descriptions of performances.

Analyzing this database, Mehr and his colleagues found that yes, the similarities do run deeper.

For one, they found that songs from societies around the world vary predictably in three ways—how formal, exciting, and religious the song is (or isn’t). The research team also found that on these three dimensions, music was more similar across societies than within them.

In other words, people around the world tend to pull the same three levers when they create music. That is not to say that these three levers are the only ones a society might use. Each society draws on a much broader set. And it’s how they pull all these levers differently that creates the diversity of music we observe.

“Our paper makes the claim that there are some universal building blocks, and…that the diversity in human music turns up as a function of these universal building blocks.”

Mehr and colleagues also explored if song occurs alongside common behaviors—think rituals, entertainment, play, or mourning. After reviewing previous research and interviewing music scholars, they catalogued 20 different behaviors that are commonly associated with song. They found evidence across societies that song does tend to accompany these situations.

But the team didn’t just look at the relationship between songs and the behaviors that go with them. They also built a discography with actual song recordings, in order to understand if the music itself shared common features. The discography consists of one example of four types of songs—love, healing, lullaby, and dance—from 30 regions around the world. In total, the discography consisted of 118 songs (they couldn’t find recordings of healing songs from Scandinavia or the British Isles).

They put these recordings to use in several ways. First they played the recordings to about 30,000 people and asked them to guess which type of song it was, using their website The Music Lab. For all four song types, people were able to identify songs at better than chance, conceptually replicating previous research and suggesting that song types have common features across societies. All types weren’t equally easy to identify though—dance songs were the easiest and love songs were the hardest.

Take a listen
See if you can correctly identify the four song-types below. You’ll find an answer key at the end of the article with more information about each song.

Song #1

Song #2

Song #3

Song #4

Mehr also tested whether machine learning techniques would be able to reliably identify the four types of songs at better than chance, like the human listeners. It was.

“One of our most surprising results is that machines, which have no knowledge of human psychology or music theory, can be trained to recognize lullabies, healing songs, love songs and dance songs—even in cultures that they’ve never seen before,” explained Dean Knox, a computational social scientist and co-author, on Princeton’s website.

Showing that humans and computer models can identify songs by type is interesting but leaves something important out. It doesn’t tell you what musical features make a dance song a dance song, or a lullaby a lullaby. For that, the authors combined expert and computer annotations of the musical features of the songs into a model. This part of their work crosses into complex music territory. For illustration, they found that, perhaps unsurprisingly, tempo can distinguish between dance and lullaby. The researchers uncovered subtler relationships too, such as the fact that dance songs tend to have a more variable melody than healing songs. (If you’re a music buff, I recommend diving into Table 2 of the study for more on these features.)

Mehr and colleagues have put together an interactive version of their ethnography and discography. In the discography, you can easily listen to songs from different societies around the world.

Taking these findings together, Mehr and his colleagues suggest that humans are wired to produce and understand music, or at least song, and the propensity toward music is not simply the result of cultural influence. That’s how music is both universal and exceptionally diverse, they say.

“Our paper makes the claim that there are some universal building blocks, and that those get elaborated, that the diversity in human music turns up as a function of these universal building blocks,” Mehr told the Canadian Broadcasting Corporation.

Anthropologist Luke Glowacki, a co-author of the study, put it this way: “Dance songs sound a certain way around the world because they have a specific function. Lullabies around the world sound a certain way because they have a specific function. If music were entirely shaped by culture and not human psychology you wouldn’t expect these deep similarities to emerge in extremely diverse cultures.”

There are caveats. For one, it was only Western listeners who participated in identifying the four song types. This is a problem because they’re interpreting the music through a similar cultural lens. It’s unclear if individuals from non-Western societies would judge the different types of songs similarly. For instance, previous research found that unlike Western listeners, the Tsimane’—a native Amazonian society with limited exposure to Western music—did not rate examples of consonant chords and vocals (notes sounding like they belong together) as being more pleasurable than dissonant ones (notes sounding like they clash).

Another issue is that the music was transcribed using Western music notation, a method that could omit certain aspects of music. Shannon Dudley, an ethnomusicologist not involved in the study, told Scientific American, “Subtleties of rhythm, subtleties of pitch differentiation, articulation and timbre—there are a lot of things that have a huge impact on the way people hear music that aren’t there in [Western] notation.”

If the divided reaction to Mehr’s earlier research—upon which the current paper builds—is any indication, the debate on universality isn’t likely to end soon.

If the divided reaction to Mehr’s earlier research—upon which the current paper builds—is any indication, the debate on universality isn’t likely to end soon.

“The assumption that we can recognize and name the intention of the expressive culture of small-scale societies without, ourselves, participating in the same kinds of activities seems extraordinarily imperialist, and essentialist,” ethnomusicologist Anne Rasmussen told The Atlantic at the time.

In the current study, Mehr and his colleagues used multiple music analysis methods, an attempt to mitigate the bias inherent in any one method. They also attempted to control for bias in the ethnographic reports they used to build the database. And they’re open to other scientists’ having a go—they’ve made their data, along with the database and discography, available for others to analyze.

Like musicians riffing on an old song, perhaps another group of scientists will pick up the notes of this research and improve it, asking questions that unlock another dimension of the debate. We’re keeping our ears open.

Answer Key:

Song #1: A lullaby from the Greek society in Southeastern Europe.
“Lullaby” [Disc 1, Track 6]. On Tēs Leros Ta Tragoudia [CD]. Leros, Greece: Music Folklore Archive (1998).

Song #2: A love song from the Rwandan society in Central Africa.
“Lama” [Side A, Track 2]. On TR-181: Sons d’ Afrique (Rwanda) [LP]. Johannesburg: International Library of African Music (1952).

Song #3: A healing song from the Tunisian society in Northern Africa.
“Sīdī bū ra’s el-‘ajmī” [Track 4]. On Stambeli: Music, Trance, and Alterity in Tunisia [CD]. Chicago: University of Chicago Press (2010).

Song #4: A dance song from the Saramaka society in the Amazon and Orinoco.
“Alesingo” [Side 1, Track 3]. On From Slavery to Freedom: Music of the Saramaka Maroons, Suriname [LP]. New York: Lyrichord Discs (1981).