The Enigma of Harmony: Can Music Be Fully Explained by Scientific Theory?

Introduction: Why Does Music Move Us?

Why do certain combinations of notes sound harmonious and pleasing, while others feel harsh and unsettling? This seemingly simple question is actually a profound problem spanning physics, mathematics, physiology, and psychology. The 2012 article Harmony Explained: Progress Towards a Scientific Theory of Music attempts to use a scientific framework to explain this artistic phenomenon that humans have intuitively experienced for thousands of years.

The article sparked heated discussion in the tech community on Hacker News. For engineers and scientists, music harmony theory is both an ancient artistic tradition and a system that can be precisely described through mathematics and physics—this duality is precisely what makes it fascinating.

Starting with Physical Harmonics: The Scientific Nature of Sound

Sound as a Composite Waveform

The scientific foundation of harmony theory begins with understanding the physical nature of sound. When an instrument produces a note, what we hear is not a pure tone of a single frequency, but a composite sound formed by the superposition of a fundamental frequency and a series of harmonics (overtones) at integer multiples of that frequency.

For example, a vibrating string not only vibrates as a whole to produce the fundamental frequency, but simultaneously vibrates in segments of one-half, one-third, one-quarter, and so on, producing overtones at 2x, 3x, and 4x the frequency. It is the relative intensity of these overtones that determines the timbral differences between instruments—middle C sounds completely different on a piano versus a violin. This phenomenon can be precisely described using Fourier analysis: any periodic waveform can be decomposed into a superposition of sine waves, each corresponding to a harmonic component. This mathematical tool is not only the cornerstone of modern signal processing and audio engineering but also provides a unified theoretical language for understanding how instruments produce sound.

The Mathematical Laws of Consonant Intervals

The ancient Greek Pythagoras discovered long ago that when two tones have frequencies in simple integer ratios, they sound more consonant. Octaves (2:1), perfect fifths (3:2), and perfect fourths (4:3) sound pleasing precisely because their frequency ratios are simple and their overtones overlap significantly, reducing auditory conflict.

The Pythagorean school discovered the relationship between intervals and string length ratios through monochord experiments in the 6th century BCE, considered one of the earliest quantitative physics experiments in human history. They observed that a string length ratio of 2:1 produces an octave, 3:2 produces a fifth, and 4:3 produces a fourth. This discovery profoundly influenced Western philosophy—the Pythagorean school proposed the cosmological view that "all is number," even believing that celestial motion follows the proportions of musical harmony (the concept of "musica universalis" or "music of the spheres"). This idea of viewing mathematics as the fundamental explanation of natural beauty remains the intellectual source of scientific music theory to this day.

This discovery forms the first building block of scientific music theory: the sense of harmonic consonance largely stems from the simplicity of frequency ratios and the degree of match between overtone structures.

The Gulf Between Physics and Perception

Why Pure Mathematics Cannot Explain Everything

However, if harmony were merely a mathematical game of frequency ratios, music theory would not have today's complexity and controversy. The real challenge is that human perception of consonance and dissonance is not entirely determined by physics—it is deeply influenced by auditory physiological mechanisms and cultural habits.

One of the article's core contributions is its attempt to bridge the gap between physical acoustics and human perception. The cochlea in the human ear has specific frequency resolution capabilities, and when two frequencies are too close together, they produce "beating" and roughness on the basilar membrane—this is considered an important physiological source of the sensation of dissonance.

Specifically, the cochlea is a spiral, fluid-filled tube whose internal basilar membrane gradually becomes wider and more flexible from base to apex, with different positions sensitive to different frequencies—the base responds to high frequencies, and the apex to low frequencies. This structure is essentially a biological spectrum analyzer. When two tones with close frequencies sound simultaneously, they activate adjacent regions on the basilar membrane, creating mutual interference and producing the so-called "critical bandwidth" effect. Two frequencies within the critical bandwidth produce a rough beating sensation—this is precisely the physiological basis of the dissonance theory proposed by Hermann von Helmholtz in his 1863 book On the Sensations of Tone (Die Lehre von den Tonempfindungen). Helmholtz's work remains a classic reference in music psychoacoustics, and later researchers such as Plomp and Levelt further quantified the relationship between critical bandwidth and consonance through systematic experiments in 1965.

Equal Temperament: A Carefully Designed Mathematical Compromise

A classic example is the twelve-tone equal temperament universally adopted in modern music. To allow instruments to play in any key, the octave is divided into 12 equal semitones, with each semitone having a frequency ratio of the twelfth root of 2 (approximately 1.05946). The cost of this approach: except for the octave, all intervals are no longer perfect integer ratios but approximations. For example, the perfect fifth in equal temperament is 700 cents, while the mathematically perfect fifth (3:2) is approximately 702 cents—a small but real difference.

Equal temperament was not an overnight invention but evolved over hundreds of years. Before it, Western music successively used Pythagorean tuning (based on stacking pure fifths), meantone temperament, and various well temperaments. Bach's famous Das Wohltemperierte Clavier (The Well-Tempered Clavier) likely used some form of well temperament rather than strict equal temperament—different keys had subtle color differences. Chinese Ming Dynasty mathematician Zhu Zaiyu precisely calculated the mathematical formula for twelve-tone equal temperament in 1584, about twenty years before Europe's Simon Stevin. The widespread adoption of equal temperament had to wait until the industrial era of the 19th century, as mass-produced pianos and organs required standardized tuning systems, and the demands of orchestral ensemble playing and free modulation made a unified temperament a practical necessity.

In other words, the "harmonious" music we hear today is essentially a series of carefully designed mathematical compromises. This perfectly illustrates that music is both rooted in physical laws and filled with human conventions and cultural constructions.

The Ambitions and Limitations of Scientific Theory

Where Are the Boundaries of Explainability?

The article's title humbly uses the phrase "Progress Towards," acknowledging that a scientific theory of music remains an unfinished endeavor. It can explain why certain intervals are consonant and why overtone structures affect timbre, but for higher-level musical phenomena—such as the emotional expression of melody, the tension and resolution of tonality, and the evolution of style—science's explanatory power remains insufficient.

This limitation echoes the "explanatory gap" problem in cognitive science. We can precisely measure the physical parameters of sound waves and trace the electrical signals of auditory nerves, but the leap from physical stimulus to subjective musical experience—why minor keys sound "sad," why dissonant chords create "tension" that produces "satisfaction" when resolved to consonant chords—still lacks a fully satisfying reductive explanation. Cross-cultural research also shows that while preferences for basic consonant intervals may have some biological universality, emotional responses to complex harmonic progressions and modal colors exhibit significant cultural differences.

In the Hacker News discussion, the tech community was generally interested in this kind of interdisciplinary attempt. Reducing artistic phenomena to computable, modelable systems inherently appeals to engineering thinking. But some perspectives cautioned that music's charm may lie precisely in those parts that cannot be fully formalized.

Implications for AI Music Generation

Worth extending is that such scientific music theory has direct relevance to today's AI music generation. Whether rule-based algorithmic composition or deep learning-based music generation models, understanding the physical and perceptual foundations of harmony helps build systems that better align with human aesthetics. Only when a model has "learned" the patterns of consonance and dissonance can it generate works that truly move people, rather than mechanical piles of notes.

Current AI music generation has two main technical approaches. The first is symbol-based methods, such as using Transformer architectures to process MIDI sequences (exemplified by Google's Music Transformer and OpenAI's MuseNet)—these models work directly at the note level, making it easier to incorporate harmonic rules. The second is audio waveform-based methods, such as Google DeepMind's WaveNet and later MusicLM, Meta's MusicGen, and others—they directly generate audio signals, capturing timbre and expressive details, but their understanding of harmonic structure is more implicit. Understanding the physical and perceptual basis of consonance can help design better loss functions, evaluation metrics, and training data filtering strategies, making generated results more acoustically aligned with human aesthetic preferences. In recent years, some researchers have begun exploring injecting music theory knowledge as inductive bias into model architectures, or using music-theoretic constraints to guide the generation process—this represents the cutting edge of convergence between scientific music theory and AI technology.

Conclusion: Where Art Meets Science

Harmony Explained represents a valuable effort: using rational tools to understand sensory experience. It tells us that musical harmony is not a purely subjective, mysterious phenomenon, but one that can be partially explained by physical laws, physiological mechanisms, and mathematical structures.

However, it also honestly acknowledges that we still have a long way to go before a complete "science of music." Perhaps this tension—where science and art never fully coincide—is precisely what enables music to continue inspiring human curiosity and creativity. For those of us in the AI era, revisiting such classic questions also provides a solid theoretical starting point for thinking about how machines might understand and create art.