From songs on the radio to birds chirping outside, humans are constantly surrounded by music. However, while making music appears to be a universal phenomenon, the vast diversity of music across different cultures also seems to point toward variation. Furthermore, human song’s global similarities have never been proven through research.
In a November 2019 study conducted over five years, an international team of researchers composed of political and data scientists, musicians, and linguists tried to answer questions related to commonality in human song.
According to Timothy O’Donnell, an assistant professor in the McGill Department of Linguistics and a lead member of the research team, the inspiration for the study stemmed from the lack of a cross-cultural music database.
“We wanted to study music in various societies in great depth, but no one had […] built a database that allowed this kind of breadth,” O’Donnell said in an interview with The McGill Tribune. “What kind of questions could we ask [once we had these data]?”
The project was essentially a large study of songs across different cultures around the world. For this purpose, two databases were built, the first of which was called Ethnology. Human Relation Area Files (HRAF), which has records annotated by anthropologists, was a major resource for this database and for cross-cultural study in general.
“We extracted all the references to human song from [HRAF] and built our own database of references to songs and further analyzed and annotated those,” O’Donnell said. “Basically any time any ethnographer that had studied [a] society in the world had mentioned a song, we were able to pull that out and see what they had mentioned, whether it was a war event or a childbirth event, and analyze the basic components surrounding the event.”
Although the second database was a relatively small collection of 120 song recordings, it required an enormous amount of work to gather them.
“These were very old, sometimes obscure recordings that we had to go to archives to find and [that were] in some cases from societies that no longer exist,” O’Donnell said. “So, we built that, and we called it discography. [It was a] tremendous amount of leg work. Sam Mehr [the lead author] spent many months tracking down these recordings, and then the team spent many years transcribing and translating them.”
O’Donnell contributed his skills as a data analyst to the project, applying the artificial intelligence techniques that he usually uses on language to music.
“Most sophisticated analysis was in Ethnography, which was analyzing text,” O’Donnell said. “We took huge amounts of data and did dimensionality reduction such as histograms on [them], using techniques from [natural language processing] that are widely used in political science.”
The study revealed that there was more variability in music events within cultures than between cultures.
“Most cultures seemed to display most kinds of music,” O’Donnell said. “It wasn’t the case that one culture only used music for healing and another for war, but [rather] that all tended to use music for similar functions.”
The research team also found some preliminary evidence for music features that were universally shared, such as rhythm and tonal structure. Another important conclusion was that music is not a fixed biological response with a single function. It is used worldwide in many contexts that vary in formality, arousal, and religiosity.
The team is currently building a larger dataset with 1,600 recordings from all over the world. Using more sophisticated and thorough analyses, they hope to answer questions such as whether tonality is universal, a cultural construction, or the product of some auditory effect.