Songs that share a common melody belong to the same melody family. Berit Janssen conducted research at the Meertens Institute into families of melodies in the Liederbank. She developed a computer program to detect recurring elements in melodies.
Whoever searches for There was a girl in the Netherlands recently, finds dozens of hits. Not only different sources of the same song, but also many other texts, to the same melody:From here to 't Groenewout, With this sweet Christmas time, Hear from the little Jan, There was a Molenaarszoon recently etcetera. These songs all belong to the same family of melodies. They are therefore linked to each other in the Dutch song bank.
PhD student Berit Janssen explains how the relationships came about:on the one hand by employees of the Liederbank, who discovered similarities on the basis of their own hearing, and on the other by the computer. Colleague Peter Kranenburg developed a so-called melody search engine, which makes comparisons based on musical notation. In this way, similarities that were not immediately detected with the naked ear were also detected.
The melody families formed the starting point for Janssen's research. The main question she wanted to answer was:what motifs can be found in all the songs of a family? A motif is the smallest unit in musical sentence analysis.
Computer vs Human
But that question turned out not to be so easy to answer. Janssen:“For example, I had fifteen different versions of folk songs from one melody, recorded in different parts of the country, and all transcribed with music notation. Usually no two were identical. Even in short motifs of two or three notes there are already small changes.”
The computer technology she initially used could only bring up identical fragments, so that didn't work. That is why she opted for a different approach. She had the computer analyze how often certain melody fragments with slight variations occur in the different variants. To do this, she first had experts assess whether fragments in melody variants resembled each other. Janssen then tried to imitate the choices they made in a computer model. In this way it was possible to conduct large-scale research with the computer, which would have cost human ears too much time.
Featured by the editors
MedicineWhat are the microplastics doing in my sunscreen?!
AstronomySun, sea and science
BiologyExpedition to melting land
Unknown songs
She then had the computer compare 1,700 melodies based on their notation. The songs were all from the collection of folk songs that are part of the Meertens Tune Collections. Most of these songs were collected at the time by Wil Scheepers and Ate Doornbosch, among others for the radio program Onder de Groene Linde. In the 1950s, the researchers made hundreds of so-called "field recordings" of people across the country. In addition, the collection includes versions of the same songs from songbooks from the nineteenth and twentieth centuries.
The PhD student herself tried to listen to the songs as little as possible at this stage of the research, so as not to unconsciously influence the development of the method. “Afterwards I did check the stable melody fragments that the computer picked out. But I only listened to about 50 of those 1700 songs.” There are well-known songs, such as Once upon a time there was a girl, or Kortjakje is always sick, but the majority is unknown. The people who sang these songs to the field researchers did not pass them on to their children, the researcher explains. “They have only survived in these recordings.”
Human cognition
With the help of the computer, Janssen found melody fragments that occur more often than others. She then wanted to know the properties of these fragments. Or in other words:which bits stick better in your head? This is where human cognition comes in. After all, the variation between the songs can be explained by oral transmission. Fragments that were easier to remember are more often found in the song versions.
Janssen came up with five predictors:the position within the melody (because a fragment at the beginning might stick better), whether or not there is repetition within the melody, a long or short note combination (short is easier), a combination that itself has a repeating pattern (as for example in the melody of the sentence There was a girl loos recently ). But the connection with the musical expectation could also predict whether a music fragment is remembered better.
Surprisingly, a combination of all these five factors turned out to be the best predictor. The PhD candidate did not immediately see this coming:“There are so many different factors that influence the oral transmission of songs, apart from the music itself. For example, maybe you're more likely to adopt versions of people you like or hear more often. I couldn't measure all those kinds of factors in my research. It is therefore very interesting that we can predict here which melody fragments will be preserved on the basis of musical properties alone.”