The Beauty of Musical Sound Combinations as Coincidence in the Upper Spectrum: Partial Alignment, Envelope Periodicity and Overtone Spectral Entropy
We propose that the aesthetic quality of a simultaneous combination of musical tones (its consonance) is decided in the upper, overtone part of the combined spectrum. When the overtones of the tones coincide in frequency, the combined sound has a strictly periodic waveform with a short period. Its envelope does not fluctuate in the roughness range, and its overtone spectrum contains fewer distinct components. We formalise this as an overtone-coincidence hypothesis. Aesthetically preferred combinations show greater partial alignment in the overtone region. All else being equal, they also have a lower spectral entropy there, measured at the time resolution of hearing. We derive three exact properties of dyads of harmonic tones with frequency ratio p/q. A fraction 2/(p+q) of all partials coincide with a partial of the other tone. Every coincidence lies at or above the p-th harmonic of the lower tone. The combined waveform repeats every q/F seconds, where F is the frequency of the lower tone. Under small mistuning, the beat rates of near-coincident pairs grow in proportion to harmonic number, so tuning errors show up mainly in the upper spectrum. We then test the hypothesis numerically with an auditory filterbank model and the classical consonance order of the twelve chromatic intervals. All rank correlations are averaged over 20 random draws of the partial phases. The overtone spectral entropy, computed in 100 ms frames, predicts the order with mean Spearman ρ = 0.88 in equal temperament (range 0.85–0.90) and ρ = 0.84 in just intonation (range 0.52–0.97; exact coincidences make the measure phase-sensitive). In equal temperament it outperforms the same measure computed over the full spectrum and the classic Plomp–Levelt roughness model (ρ = 0.73); in just intonation it is on a par with total envelope fluctuation and with the analytic index p+q. The prediction for tempered intervals is best for analysis windows of 50–100 ms and degrades for longer windows, while just intervals are unaffected. When partials are stretched so that the "octave" becomes 2.1:1, the entropy minimum moves from 1200 to 1284 cents, the stretched pseudo-octave. A second, deliberately simple spectral model without an auditory filterbank gives the same direction of effect. Two parts of the intuitive formulation are not supported. The energy of the spectral tail does not track consonance. Slow amplitude fluctuations (below 20 Hz) in the overtones mark the imperfect tuning of a consonance, not dissonance: in equal temperament they are largest for the fifth and the fourth. The relevant quantities are therefore entropy and roughness-range fluctuation, not energy and slow fluctuation. The measure is close to the spectral entropy that Hinrichsen introduced as a criterion for tuning pianos. We restrict it to the overtone region, give it the time resolution of hearing, and state it as a falsifiable hypothesis about preference. Finally, we discuss structural analogies with physics. Commas in tuning behave like the holonomy of a lattice gauge field. Level matching in closed-string theory is a coincidence condition between mode numbers. We state the limits of these analogies.
Authors
- Maxim Chakhovsky
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-14
- DOI
- https://doi.org/10.5281/zenodo.24220571
- Primary Topic
- Neuroscience and Music Perception
- Type
- preprint