Tap along with the crowd
Tap to a candombe recording exactly as GlobalTap participants did, judge the crowd's beats against the expert's, then see where you fall among them.
Where is the beat? For most music, people agree without thinking. For some music they agree with each other, but not with the expert who wrote the beats down.
In GlobalTap, 746 people in ten countries tapped along with 30-second excerpts from 59 countries. Their pooled taps became a consensus beat track for each excerpt. On this recording of Afro-Uruguayan candombe drumming, the crowd keeps the expert’s tempo but lands consistently later in every beat, and listeners preferred the crowd’s version. It is one of the phase-offset cases in GlobalTap’s validation set.
Tap along with the same 30 seconds the participants heard, judge the two beat tracks yourself, then see where you land among the 36 people who tapped this excerpt.
Practice mode Nothing you do here leaves this browser. We plan to open this as a citizen-science study once our ethics amendment is approved. How taking part will work
1 Check your volume
Headphones work best. Play a few seconds and set a comfortable level that lets you hear every drum.
2 Practice with a metronome
Tap on every click, using the space bar or the pad below. This also measures how much your keyboard or screen delays each tap.
3 Tap along with the music
Keep your taps simple, steady and evenly spaced. Adjust if the tempo changes.
That is the instruction participants saw, word for word. The excerpt lasts 30 seconds. We score the taps between 10 and 25 seconds, as in the paper.
4 Rate two beat tracks
Before you see any results: here are two versions of the same 15 seconds with clicks on top. One follows the crowd's beat and one the expert's, in random order. Each plays once. Rate how well the clicks match the beat of the music.
5 Where you fall among the tappers
- crowd tap density and consensus beats
- expert annotation
- each online tapper, sorted by timing
- you
One expert beat drawn as a clock face. Each red dot is one online tapper: the angle is where in the beat they tapped on average, and the distance from the centre is how steady they were. The blue dot is you.
6 Listen back
The same window with your own taps, the crowd's or the expert's as clicks.
What you’re seeing
The red curve is where all 36 participants tapped, pooled and smoothed, and the red ticks are the consensus beats fitted to it. They land about 80 ms after the expert’s beats, at the same tempo.
Neither answer is a mistake. The paper notes that offsets like this are consistent with P-center effects or with attending to a different instrumental or rhythmic layer, though the study can’t tell these apart. The clock face shows how tightly most people agree with each other, even where they disagree with the expert.
That is why GlobalTap treats the crowd’s beat as data rather than noise. Disagreements like this one show where listeners hear the beat differently from the annotation.
Where this comes from
Related work
- Beat and downbeat tracking based on rhythmic patterns applied to the Uruguayan candombe drumming Introduced the candombe recordings and the expert beat annotations used here.
- Uruguayan candombe drumming: beat and downbeat tracking data set The recording you tap to, released under CC BY 4.0.
- REPP: A robust cross-platform solution for online sensorimotor synchronization experiments How GlobalTap measured taps precisely through participants' own microphones.
- Rhythmic prototypes across cultures: A comparative study of tapping synchronization Earlier evidence that where people tap depends on the music traditions they know.
The taps, tap density, consensus beats and rating scores on this page are from the GlobalTap study. Taps are measured from your keyboard or screen, and the practice step corrects for their delay. Your offset is accurate to a few tens of milliseconds.
Audio: an excerpt of csic.1995_ansina1_02 from the Uruguayan candombe drumming beat and downbeat tracking data set by Luis Jure and Martín Rocamora (Zenodo, doi:10.5281/zenodo.6533068), introduced in Nunes, Rocamora, Jure and Biscainho (ISMIR 2015). Licensed CC BY 4.0. We cut it to the 30-second study excerpt, mixed it to mono, and added click tracks for the rating step.