Research
How do minds and brains distinguish music from other sounds?
I'm using behavioral, computational and neuroscientific techniques to investigate how minds and brains distinguish music from other sounds.
Overview
Across every known human society, music is an "absolute" cultural universal (Savage et al., 2015). From lullabies in the Amazon to ritual drumming in West Africa, musical expression appears in every ethnographic record, although its form varies widely within and between cultures (Mehr et al., 2019). Yet despite this universality, there is no single, agreed-upon definition of what music actually is (McLeod, 1971). What one culture considers music—such as overtone chanting, microtonal scales, or rhythmic speech—may be perceived as noise, ritual, or even language by another. This conceptual ambiguity presents a profound challenge for researchers: how do we define music in a way that respects its global diversity while probing the cognitive and perceptual boundaries that shape its recognition?
Three studies which I am conducting, as outlined below, aim to address this challenge from complementary angles. The first study uses behavioral methods to investigate how individuals classify culturally diverse sound samples as “music” or “non-music,” revealing the perceptual and conceptual boundaries that listeners apply. The second study leverages machine learning to model these judgments computationally, training algorithms to classify and generate novel sound samples based on human-labeled data. The third study uses EEG to explore the neural correlates of musicality, asking whether brain responses to traditional and AI-generated music reflect shared cognitive signatures. Together, these studies offer a multi-level investigation—behavioral, computational, and neural—into one of the most elusive questions in cognitive science: what makes music music?
Musical universals [link unavailable]
McLeod, N. (1971). Musical universals. Ethnomusicology 15(3), 379–402.
Universality and diversity in human song
Mehr, S. A., Singh, M., Knox, D., Ketter, D. M., Pickens-Jones, D., Atwood, S., Lucas, C., Jacoby, N., Egner, A. A., Hopkins, E. J., Howard, R. M., Hartshorne, J. K., Jennings, M. V., Simson, J., Bainbridge, C. M., Pinker, S., O’Donnell, T. J., Krasnow, M. M.,Glowacki, L. (2019). Universality and diversity in human song. Science, 366(6468), eaax0868. https://doi.org/10.1126/science.aax0868
Statistical universals reveal the structures and functions of human music
Savage, P. E., Brown, S., Sakai, E., & Currie, T. E. (2015). Statistical universals reveal the structures and functions of human music. Proceedings of the National Academy of Sciences, 112(29), 8987–8992. https://doi.org/10.1073/pnas.1414495112
Exploring three approaches to music cognition by asking, "What is music?"
My research aims to build foundational knowledge in music cognition through initial engagement with three core approaches to the field: psychology, neuroscience, and computation. Specifically, I explore how these approaches illuminate different facets of music cognition by examining how each sheds light on the question: What is music?
What is music?
From handwritten folk songs to global sound archives, the story of ethnomusicology is a journey through culture, technology, and human connection. Beginning in the 19th century with Oskar Kolberg’s documentation of Polish village music, scholars have sought to preserve and understand the world’s musical traditions. Innovations like Alexander Ellis’s pitch measurement system and Carl Stumpf’s early sound recordings laid the groundwork for modern research.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
How do minds and brains distinguish music from other sounds?
Subject participating in online citizen-science "game" study.
Study 1: Behavioral Research
Conceptual Boundaries of Music: A Behavioral Study of Cross-Cultural Sound Classification
Question
Do individuals share a common conceptual boundary for what constitutes “music” and "non-music" across culturally diverse sound samples, and do certain acoustic features predict classification as music?
Approach
The experimental design for Study 1 centers on a behavioral classification task using a curated set of sound samples drawn from traditional societies around the world. These samples vary widely in acoustic and structural features—such as rhythm, melody, repetition, and instrumentation—some aligning with proposed musical universals and others challenging them. Participants will listen to each sample and make a binary judgment: “Do you consider this to be music?” This forced-choice format ensures clarity and comparability across responses. To enrich the data, participants will also rate their confidence on a 7-point Likert scale and optionally provide open-ended explanations for their decisions. The stimuli will be randomized and counterbalanced to control for order effects.
Participants will be recruited from diverse cultural and musical backgrounds to explore potential influences of enculturation and training. The study will be conducted online on one or more platforms allowing for secure and scalable data collection. Analysis will focus on consensus rates, feature-based correlations and cross-group comparisons. Acoustic features of each sample will be extracted and statistically linked to classification outcomes, enabling exploration of which properties most reliably predict musicality judgments. This design allows for both quantitative and qualitative insights into the conceptual boundary between music and non-music, and it lays the foundation for computational modeling in Study 2.
How do minds and brains distinguish music from other sounds?
Simple neural network machine-learning architecture.
Study 2: Computational Research
Learning Musicality: Machine Classification and Generation of Music vs. Non-Music from Human-Labeled Sound Samples
Question
Which machine learning algorithm best classifies culturally diverse sound samples as “music” or “non-music” (K-neighbors, decision tree, random forest, logistic regression, CatBoost or gradient boost)?
Approach
Study 2 builds directly on the labeled dataset from Study 1 to train a machine learning model capable of classifying sound samples as “music” or “non-music.” The first phase involves preprocessing the audio data to extract relevant acoustic features, including MFCCs, spectral centroid, rhythmic patterns, and pitch contours. These features will be used to train and validate classification models K-neighbors, decision tree, random forest, logistic regression, CatBoost or gradient boost. The models' performances will be evaluated using standard metrics (accuracy, precision, recall, F1 score) on held-out test data. Misclassifications will be analyzed to refine feature selection and assess model limitations.
In the second phase, generative AI models—such as MusicVAE, Jukebox, or diffusion-based audio synthesis—will be trained on subsets of the labeled data to produce novel sound samples conditioned on the “music” or “non-music” label. These generated samples will be evaluated in a follow-up behavioral study, where human participants classify them using the same binary judgment and confidence scale as in Study 1. The goal is to assess whether the model has internalized perceptual boundaries of musicality and can reproduce them in novel outputs. This two-part design enables both predictive modeling and creative synthesis, offering a computational lens on human musical perception.
How do minds and brains distinguish music from other sounds?
Subject undergoing electroencephalography (EEG) monitoring.
Study 3: Neuroscientific Research
Neural Signatures of Musicality: EEG Responses to Human-Recognized and AI-Generated Sound Stimuli
Question
Do individuals show similar neural responses to “music” and “non-music”, as classified in the Behavioral Study as they do to AI-generated samples from the Computational Study?
Approach
Study 3 uses EEG to investigate the neural correlates of musicality by comparing brain responses to sound samples classified as “music” or “non-music” in Studies 1 and 2. Participants will be fitted with a 64-channel EEG cap and seated in a sound-attenuated room. They will passively listen to randomized sequences of audio clips from both the traditional dataset and the AI-generated samples. Each clip will be 10–20 seconds long, and inter-trial intervals will be jittered to reduce anticipatory effects. Optional post-trial ratings will be collected to link subjective perception with neural data. The experiment will be divided into blocks to minimize fatigue and maintain attention.
EEG data will be analyzed for event-related potentials (ERPs), focusing on components such as N1, P2, and late positive complex (LPC), which are associated with auditory categorization and attentional engagement. Time-frequency analysis will assess oscillatory activity, particularly frontal theta and beta bands, which may reflect cognitive processing of musical structure. Inter-trial phase coherence (ITPC) will be used to measure neural entrainment to rhythmic stimuli. Comparisons will be made across stimulus types (traditional vs. AI-generated), perceptual categories (music vs. non-music), and participant groups (musically trained vs. untrained). This design aims to uncover whether musicality judgments are reflected in consistent neural patterns and whether AI-generated music evokes similar brain responses to culturally grounded musical forms.
Behavioral Research
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed euismod, nunc at facilisis tincidunt, neque urna fermentum magna, nec ullamcorper justo nulla nec sapien. Integer vitae purus ac lorem tincidunt tincidunt. Suspendisse potenti. Curabitur vel sem sit amet eros malesuada aliquet.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Distinguishing Music from Non-Musical Sounds:
A Review of Psychological Studies
Imagine someone walking from town to town with a notebook, asking people to sing their favorite songs so he could write them down and save them forever. Well, that's exactly what Oskar Kolberg did starting in the 1830s. Kolberg traveled around Poland "recording" songs people sang in villages.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Distinguising Music from Non-Musical Sounds:
Psychological Evidence From Online Experiments Using Methods of Gamification and Citizen Science
Think of it like a giant music museum where you can listen to songs from all over the world — even ones recorded over 100 years ago! In the early 20th century, Carl Stumpf stood at the forefront of a new way to study music — not just as art, but as a window into human culture.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Neuroscientific Research
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed euismod, nunc at facilisis tincidunt, neque urna fermentum magna, nec ullamcorper justo nulla nec sapien. Integer vitae purus ac lorem tincidunt tincidunt. Suspendisse potenti. Curabitur vel sem sit amet eros malesuada aliquet.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Distinguishing Music from Non-Musical Sounds:
A Review of Psychological Studies
Imagine someone walking from town to town with a notebook, asking people to sing their favorite songs so he could write them down and save them forever. Well, that's exactly what Oskar Kolberg did starting in the 1830s. Kolberg traveled around Poland "recording" songs people sang in villages.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Distinguising Music from Non-Musical Sounds:
Psychological Evidence From Online Experiments Using Methods of Gamification and Citizen Science
Think of it like a giant music museum where you can listen to songs from all over the world — even ones recorded over 100 years ago! In the early 20th century, Carl Stumpf stood at the forefront of a new way to study music — not just as art, but as a window into human culture.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Computational Research
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed euismod, nunc at facilisis tincidunt, neque urna fermentum magna, nec ullamcorper justo nulla nec sapien. Integer vitae purus ac lorem tincidunt tincidunt. Suspendisse potenti. Curabitur vel sem sit amet eros malesuada aliquet.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Distinguishing Music from Non-Musical Sounds:
A Review of Psychological Studies
Imagine someone walking from town to town with a notebook, asking people to sing their favorite songs so he could write them down and save them forever. Well, that's exactly what Oskar Kolberg did starting in the 1830s. Kolberg traveled around Poland "recording" songs people sang in villages.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Distinguising Music from Non-Musical Sounds:
Psychological Evidence From Online Experiments Using Methods of Gamification and Citizen Science
Think of it like a giant music museum where you can listen to songs from all over the world — even ones recorded over 100 years ago! In the early 20th century, Carl Stumpf stood at the forefront of a new way to study music — not just as art, but as a window into human culture.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.
Fedorenko, E. 2021. Current Opinion in Behavioral Sciences, 40, 105-112. DOI: 10.1016/j.cobeha.2021.02.023.