Vibrato as Rhythmic Subdivision
A free metronome that plays the beat and the vibrato subdivision together, so you can hear the relationship this research describes. Runs in any browser, works offline, and is also available as an iOS app. The page below explains where its tempo zones come from.
Open Vibrato Trainer →Vibrato rate is usually described as a stable trait of the performer — roughly 5–8 Hz, independent of what the music is doing. This project asks a different question: not whether rate is relatively stable, but what holds it there.
Vibrato Trainer picks a subdivision for any tempo you give it. Those boundaries are not guesses. They are posterior estimates from a Bayesian model fitted to 100 commercial recordings.
| Crossover | Estimate | 95% interval |
|---|---|---|
| Triplet 16ths → 16ths | 63 | 58–69 |
| 16ths → Triplet 8ths | 99 | 91–108 |
| Triplet 8ths → 8ths | 148 | 135–162 |
| 8ths → Triplet quarters | 227 | 210–246 |
| Triplet quarters → Quarters | 310 | 284–343 |
Each recording was timed against a metronome and its vibrato assigned to the subdivision it tracks. Implied rate is then tempo × multiplier ÷ 60. Across all six categories, every category mean falls between 5.5 and 6.7 Hz — the rate stays put while the subdivision underneath it changes five times across a sevenfold tempo range.
The corpus spans 15 genres and 10 instruments, 61 vocal and 43 instrumental observations. Four recordings use two different subdivisions in different sections at one unchanging tempo; three of those four sit within 16 BPM of an estimated crossover.
Every recording is commercially released and publicly available, and the dataset carries a timestamp for each one. Any coding decision here can be checked against the source in about a minute. That is deliberate: this is a single-rater corpus, and the answer to that limitation is to make the evidence easy to audit.
Dataset — 100 recordings (CSV) Song, artist, tempo, coded subdivision, instrument, genre, timestamp and recording link. Conference poster (PDF, A0) Presented at ICCCM 2026, 4th International Conference on Computational and Cognitive Musicology, Würzburg, Germany. Full method, figures and limitations.Van Bebber, M. (2026). Vibrato as rhythmic subdivision: cross-genre evidence for vibrato–tempo coupling in 100 commercial recordings. Poster presented at the 4th International Conference on Computational and Cognitive Musicology, Würzburg, Germany.
Questions, corrections, or a recording you think breaks the pattern — I would like to hear about it: mvanbebber@gmail.com