Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz
Modeling the Missing Fundamental: Effects of Musical Exposure in Humans and Neural Networks
Poster Session B - Sunday, March 8, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom
Lukas Grasse1 (), Matthew S. Tata; 1University of Lethbridge
Pitch is a fundamental aspect of human sound perception. Tones are composed of a fundamental frequency (F0) and harmonics, and the pitch of a note usually corresponds to F0. Humans sometimes experience the "missing fundamental" (MF) illusion, in which the perceived pitch corresponds to an F0 that has been removed and only some harmonics remain. This illusion is more prevalent among those with musical training. We evaluated human participants and auditory artificial neural networks (ANNs) on a task designed to measure their perception of the missing fundamental. We hypothesized that musical exposure and experience would lead to improved pitch perception and increase missing fundamental responses in the ANNs, as it does in humans. The ANNs were trained on speech, music, and other audio using a variety of self-supervised and supervised learning tasks. Models trained on music and other audio were susceptible to the missing fundamental illusion, whereas self-supervised models trained solely on speech were not. In undergraduate students, we found a significant correlation with perception of the missing fundamental and musical experience. By fine-tuning models that did not initially exhibit strong missing fundamental representation, Wav2Vec 2.0 and Data2Vec, we found that this musical fine-tuning improved the models' pitch representation; however, it failed to induce a strong representation of the missing fundamental. This work suggests that while certain types of sound exposure, like music, might be a key component to perceive the missing fundamental illusion, it is not sufficient on its own. Specific musical experiences may be critical.
Topic Area: PERCEPTION & ACTION: Audition
CNS Account Login
March 7 – 10, 2026