Tapping your foot
Tapping your foot to a song, an orchestra following the conductor's hand, a whole hall clapping in time: people are remarkably good at moving with a rhythm. In the laboratory this skill is usually studied in its simplest form: tapping a finger along with a metronome.
In this experiment a metronome clicks twelve times in each round and you tap the screen on every click. After twelve clicks the sound stops, and you carry on for twelve more taps at the same tempo. This two-part design is the classic task researchers call 'synchronization–continuation'. The first part shows how well you lock on to an external rhythm, the second how well your internal clock holds the tempo on its own.
Why do we tap before the click?
Bruno Repp's 2005 review and Repp and Yi-Huang Su's 2013 review summarise hundreds of studies in this field. One of the most surprising and most robust findings is that when people tap with a simple metronome, their taps land, on average, before the click. This is called the negative mean asynchrony. People do not notice it; it feels as if they are tapping exactly with the click.
This anticipation shows that the brain is not reacting to the click but predicting when the next one will come and starting the movement ahead of time. Why exactly the taps run early is still debated. One explanation is that touch information from the finger reaches the brain later than sound, so the finger has to go a little early to feel 'simultaneous'. Another is that the interval between clicks is perceived as slightly shorter than it is. Adding extra sounds between the clicks reduces the asynchrony, which supports the second idea, but not every finding fits it.
Musicians and tempo
The asynchrony is smaller in musicians. In one study described by Repp and Su, with a metronome clicking once every 800 ms, drummers' asynchrony was about −20 ms, while pianists, singers and non-musicians were around −50 ms. Tempo matters too: in non-musicians the asynchrony grows as the clicks get further apart, whereas in musicians it stays small and fairly constant.
Synchronizing with a visual rhythm is much harder than with sound. With a flashing light, taps become noticeably more variable; Repp and Su report that with a static flashing light it gets especially hard once the clicks are less than 500 ms apart. For people who cannot use sound we also offer the metronome as a light, but we flag those rounds; they should not be read on the same scale as rounds with sound.
When the sound stops
Once the metronome falls silent, your internal clock is the only way to keep the tempo. During synchronization the brain makes small corrections after each tap: if one was early, the next is delayed a little. When the sound stops, that chance to correct disappears; small errors pile up and the tempo can drift.
In 2005 Rasmus Flach focused on the moment of transition. As soon as the sound stopped, taps sped up, and this did not change even when participants knew exactly when the transition would come. According to Repp and Su, Flach also found that people who tapped earlier during synchronization tapped with shorter intervals during continuation, which fits the idea that the interval between clicks is perceived as shorter than it is. In this experiment we compare your average tap interval in continuation with the metronome's at every tempo.
Measuring rhythm in a browser
Rhythm experiments deal in milliseconds, and an ordinary computer or phone is not designed to be that precise. Getting sound out of the speaker, registering a touch on the screen and the browser processing it all add delays of a few tens of milliseconds that vary from device to device. In 2022 Manuel Anglada-Tort, Peter Harrison and Nori Jacoby showed that this latency and jitter are the biggest problem for online tapping experiments, and developed a method that gets around it by recording through a microphone.
We do not use the microphone. We schedule the clicks on the browser's audio clock in advance, use the audio output time the browser reports, and take each tap with the event's own timestamp. Even so, we cannot know the touch latency. That is why we compute the score not from the mean asynchrony but from two measures that a constant delay does not affect: how much your taps wobble from one click to the next, and how far your tempo drifts once the sound stops.
What your score does and does not tell you
Half of each round's score comes from synchronization and half from continuation. The synchronization score is computed from the standard deviation of the asynchronies on the nine clicks after the warm-up: a variability of 40 ms gives 5 points, 20 ms about 7.5. The continuation score adds half the variability of the intervals to the percentage difference between your average tap interval and the metronome's: a combined 6% gives 5 points. The maximum total is 50.
We show your mean asynchrony on the results screen because it is the measure this experiment is famous for, but that number also contains your device's delay. The same person may see a different mean asynchrony on a phone and on a computer. As crowd data comes in, we will show how asynchrony and continuation speed-up are distributed by tempo.