Computational Ethnomusicology: Analyzing Oral Folk Music

This article explores how modern ethnomusicologists utilize pitch-histogram microtonal analysis and rhythmic-meter artificial intelligence to study the transmission of oral folk songs across generations. By adopting these computational methods, researchers successfully bypass the long-standing biases of Western staff notation, which historically forced diverse global musical traditions into a rigid, twelve-tone, metrically uniform framework. Together, these technologies allow for the objective measurement of subtle melodic shifts and complex rhythmic variations, preserving the authentic nuances of non-Western and unwritten musical heritage.

For over a century, the standard tool for documenting oral folk traditions was Western five-line staff notation. However, this system was designed specifically for European art music and relies on a fixed twelve-tone equal temperament scale and rigid, mathematical subdivisions of time. When ethnomusicologists transcribed indigenous or rural folk songs into staff notation, they were forced to round off "in-between" pitches (microtones) and regularize fluid rhythms to fit standard time signatures. This practice obscured the true nature of the music, erasing the very stylistic nuances that define how songs are taught, remembered, and altered from one generation to the next.

To capture melody without the filter of the Western scale, researchers now use pitch-histogram microtonal analysis. This process begins by running audio recordings of folk singers through fundamental frequency (\(F_0\)) extraction algorithms, which measure the exact vibration rate of the voice in Hertz (Hz) and convert those measurements into cents (where 100 cents equals one Western semitone). Instead of snapping these notes to a grid, the software generates a continuous pitch histogram—a visual graph showing the exact frequency distribution of every sound the singer produces.

Through pitch histograms, ethnomusicologists can see the precise tuning systems used by a specific community. When studying generational transmission, researchers compare the pitch histograms of elder singers with those of younger apprentices. This reveals whether specific microtonal inflections—such as neutral thirds or intentional pitch glides—are being faithfully preserved, gradually shifting due to physiological differences, or flattening out under the influence of modern, Westernized media. The data represents the singer's actual tonal center and scale structure without ever referencing a piano keyboard.

Just as folk melodies resist the twelve-tone scale, oral rhythms often defy the strict bar lines and metronomic beats of Western notation. Many folk traditions rely on non-isochronous meters (beats of unequal length), rubato (expressive speeding up and slowing down), or patterns guided by the natural cadence of spoken language. Rhythmic-meter artificial intelligence addresses this by analyzing the raw acoustic onset of notes and syllables directly from the audio waveform.

Machine learning models are trained to recognize recurring patterns of emphasis, duration, and acoustic energy without assuming a standard 4/4 or 3/4 time signature. By clustering these timing intervals, the AI identifies the underlying rhythmic architecture inherent to the performance itself. It maps how a singer stretches a specific beat for dramatic effect or compresses a phrase to match a breath cycle, creating a bespoke rhythmic grid for that specific song and performer.

By combining microtonal pitch histograms and rhythmic AI, ethnomusicologists can conduct large-scale, objective analyses of how oral traditions evolve. Researchers can feed decades of archival field recordings alongside contemporary performances into these systems to track the genealogy of a song.

The technology highlights exactly which elements of a song a culture considers mandatory to preserve and which elements are open to individual improvisation. For example, the data might show that over three generations, the microtonal intervals of a traditional lullaby remained mathematically identical, while the rhythmic pacing adapted to modern speech patterns. Ultimately, these computational tools ensure that oral music is studied on its own terms, honoring the acoustic reality of the performers rather than the limitations of a written page.