Scientists have learned to recreate imagined melodies based on brain activity

The experiment involved ten patients with epilepsy, for whom electrodes were implanted on the surface of the brain as part of clinical observation. This allowed the scientists to record neural activity with high precision. The participants first listened quietly to the initial bars of children’s songs, then mentally imagined how the melody continued, and finally hummed what they had imagined.

Data analysis showed that the brain activity contained information about the relative pitches of the sounds in the imagined melody — that is, about the relationships between the notes. Experts built a model that, based on these signals, predicted classes of pitch classes (such as “C,” “D,” “E,” and so on), rather than their absolute pitch. This approach proved to be particularly successful: people often recognize a melody precisely by the interval relationships between notes, even if it sounds in a different key. By combining predictions for individual notes over time, the team was able to recreate the melodic movement that reflected the structure of the songs being mentally imagined.

Brain signals indeed carry information about the pitch of notes in imagined musical fragments. In the long term, this opens up opportunities for decoding other musical characteristics, such as tempo and dynamics. Furthermore, scientists intend to find out whether similar methods can be applied not only to simple, well‑known melodies, but also to more complex music.

The work is of particular importance from the perspective of potential practical applications. Jii Kwon hopes that in the future, such technologies will help recreate imagined melodies not only in people without specialized musical training but also in patients who cannot vocalize their ideas, in particular, in people with amyotrophic lateral sclerosis. In such cases, the ability to read musical images directly from brain activity could become an important communication channel.

Thus, the study not only expands our understanding of how the brain encodes musical images but also paves the way for new neurotechnological solutions — from fundamental tasks in cognitive neuroscience to practical tools to support people with severe motor limitations.