7 min read

The science of keeping a beat

Tapping along to music is so easy that it looks like nothing. It isn't: it's a prediction problem that most animals can't solve, and the moment you take the music away it becomes a much harder one.

Tapping along is prediction, not reaction

Start with a fact that rules out the obvious explanation. If synchronised tapping were a matter of hearing each beat and responding to it, every tap would land at least a fifth of a second late, because that is roughly what it costs to react to a sound. Tap along to a metronome and your taps land on the beat, not behind it.

So you aren't responding to beats. You're predicting them, running an internal period that you correct against what you hear. The correction is fast, largely automatic, and mostly invisible to you: shift one beat of a metronome slightly and people compensate within a tap or two, including when the shift is too small to consciously notice.

You tap early and you don't know it

One of the field's most reliable findings is that people tapping in time with a metronome place their taps slightly before each beat, commonly a few tens of milliseconds early, while experiencing themselves as dead on. This is called the negative mean asynchrony, and it is remarkably robust.

The classic explanation is that the brain aligns sensory feedback rather than external events: the sound of the beat and the feel of your finger landing travel different distances at different speeds, and lining up the two sensations requires the finger to move early. It shrinks with musical training, which fits, and it isn't fully explained by any single account, which is normal for this literature.

The practical point is that your subjective sense of “on the beat” is not a neutral measurement. It is a construction, with a known offset built into it.

Everyone has a preferred tempo

Ask people to tap at whatever rate feels natural and the answers cluster near two taps a second, roughly 100 to 120 beats per minute, in the neighbourhood of a brisk walk. Accuracy is best near that rate and degrades as you move away from it in either direction.

Very slow beats are the harder side. Once intervals stretch past a couple of seconds, the beat stops feeling like a rhythm and starts feeling like a series of separate events you have to time individually, which throws you back on ordinary interval timing and its proportional noise. Musicians handle slow tempos by subdividing, feeling twice or four times as many pulses as they play, precisely to keep the underlying rate inside the range where rhythm still works.

What happens when the music stops

Now the interesting part, and the one a timing game is built on. Synchronising with a sound you can hear is a corrected process: every beat is a fresh reference, and errors get wiped out as fast as they appear. Continuing after the sound stops removes the reference entirely.

This paradigm, synchronise and then continue in silence, has been studied since the 1970s, and the influential decomposition of it splits your timing error into two independent sources: a central timekeeper that sets the intervals, and the motor system that executes each tap. The two have different signatures, and it is the timekeeper that determines whether you hold a tempo or wander off it.

Wander is the right word. Without an external reference, each interval's error carries into where the next one starts. Small errors don't cancel, they accumulate, and unaccompanied tapping tends to drift like a random walk rather than oscillating around the correct tempo. Many people drift systematically faster as well, which is why an unaccompanied group of musicians usually finishes a piece ahead of where it started.

Playing it better

• Spend the audible phase encoding, not tapping along. Your entire continuation is only as good as the period you locked in. Listen to the gap between beats as a shape you could reproduce, not as a number.
• Move something. Synchronisation is a motor skill more than a perceptual one, and most people hold tempo better with a body movement attached: a foot, a nod, a breath. Just keep it identical across attempts.
• Subdivide slow beats. If the tempo is slow enough that each gap feels like a wait rather than a pulse, feel two beats inside each one. It keeps the rhythm mechanism engaged instead of handing the job to interval estimation.
• Don't chase your own errors.With no reference, you can't actually tell whether you're drifting, so “corrections” mid-sequence are guesses that add variance. Set the tempo, then protect it.

And it's worth remembering how strange this ability is at all. Spontaneously moving in time with a beat is close to universal in humans and rare elsewhere in the animal kingdom, showing up mainly in species that learn their vocalisations. Whatever the brain is doing when you nod along to a song, it isn't a general-purpose trick.