“If I want to know how the guitar and saxophone became the important instruments throughout classical repertory or how chord progressions have changed, those are questions musicology has been unable to approach,” says Associate Professor of Music Michael Cuthbert. Spotting trends and patterns in a large corpus of music is nearly impossible using traditional methods of study, because it requires the slow process of examining pieces one by one. What his field needed, Cuthbert determined, was a way to “listen faster.”
With that goal in mind, Cuthbert gathered a team of computer programmers at MIT to develop music21, a suite of software that provides many of the same tools for analyzing music that so many of us depend upon when working with text and data. For example, musicologists can use music21 to search for chords or graph pitches.
“In music, we haven’t been able to approach analyzing pieces with anything like the bird’s-eye view that major text repositories give us,” explains Cuthbert, who is also an experienced programmer. “Music21 was the first modern system to lay the groundwork and set libraries for working with music.”
“Once the data is in there,” he says, “we can ask 1,000 questions.”
An open-source, cross-platform toolkit written in Python, music21 employs more than 9,000 pages of code to endow the computer with a basic understanding of music theory, thereby enabling researchers to examine large bodies of works, identify trends, and pinpoint interesting outliers worthy of deeper study.
Developed with funding from the Seaver Institute, which supported the project from 2008 to 2012, music21 now boasts 3,000 downloads a month. The toolkit has been used for a wide range of projects worldwide, from identifying who wrote various segments of collaborative Beatles songs to supporting opposing arguments in a Society of Music Theory debate about the concept of key.
About a year and a half ago, Cuthbert finally put music21 to work on his own research in medieval European music, using the computer for such tasks as determining a composition’s country of origin and reconstructing damaged notation. For example, music21 was able to determine the missing clef and signature information for one 14th-century Italian piece by running through all the approximately 300 possible options in minutes—saving Cuthbert two days of work.
Music21 has also proved a boon to teaching, Cuthbert says, noting that students in his Fundamentals of Music class use it to test their understanding of music theory and to practice composition. The software allows him to spend less time grading papers, which means he can accommodate more than twice the usual number of students in the popular class. Next spring, he plans to pilot the use of music21 for teaching in other MIT classes and at other universities.
While music21 currently requires its users to have some programming capability, Cuthbert says he is planning to develop new applications tailored to those who lack such skills. “Music21 has been career-changing for musicians who have some technical ability,” he says. “We’re hoping to bring it out to the rest of the world.”