Groove and microtiming
A Socratic walk-through of groove and microtiming — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does a drum machine playing perfectly on the beat feel less danceable than a drummer who is slightly late?
You have heard the explanation, and it is a lovely one: the machine is too perfect, the human drummer sits a hair behind the beat, and that sliver of imprecision is what makes a body want to move. Before we accept it, let us ask what the claim predicts. It predicts that music locked to an unwavering grid should be hard to dance to. Yet the most reliably floor-filling music ever made — disco, house, techno — is quantised to a machine grid, frequently by an actual machine. If lateness were the engine of groove, that entire catalogue ought not to work. So either the story is wrong, or it is describing something other than what it says.
Reasoning it through
REASONING #The nice thing about this claim is that it is testable, and it has been tested. Take one drum pattern, render it exactly on the grid, then render versions with the events nudged off the grid by various small amounts, and ask listeners which grooves more. When Frühauf, Kopiez and Platz did this in 2013, the quantised version scored highest, and ratings fell as the deviations grew. Davies and colleagues, working the same year with similar manipulations, likewise found no groove benefit from added microtiming. So the folk account is not merely unproven; it was examined and came out backwards.
Is that the end of it? Not quite, and honesty requires the other side. Datseris and colleagues, in 2019, delayed a jazz soloist by small amounts against a metronomically exact rhythm section; professional musicians rated the delayed versions as swinging more, while being unable to say what had been altered. Critics of the null results argue that the earlier experiments shifted notes in ways no player would, so of course they sounded wrong. The fair reading is that microtiming is not the engine of groove, but neither is it nothing, and the size of its contribution is still argued over by people who have listened far more carefully than we have.
If timing deviation is not the driver, what is? The strongest and most replicated predictor is syncopation — notes placed off the strong beats — and it works as an inverted U. Witek and colleagues found that medium syncopation maximised both pleasure and the desire to move; too little is inert, and too much dissolves the beat altogether. Why should the middle win? Because the beat is not in the signal. Nothing in a recording is the pulse; you infer it and your body keeps it. Groove seems to be what happens when there is a gap between the pulse you are predicting and the notes actually arriving — the urge to move is, on the leading account, the motor system filling that gap in. Call that account promising rather than settled; it is a live research programme, not a closed question.
Now look again at what "human timing" turns out to be when measured. Friberg and Sundström timed jazz drummers' ride cymbals and found that the swing ratio — long note against short — is not scattered error but a systematic function of tempo: it falls steadily as tempo rises, while the short offbeat note stays close to a tenth of a second regardless. That is a rule. Drum machines have had a knob implementing it since the 1980s.
So what actually makes a cheap machine pattern lifeless? Mostly that every snare is the identical recording. A snare struck harder is not a louder copy of a softer one — the spectrum changes, the rattle changes. Strip that and you strip the accent structure that tells a listener which pulses are strong, and you also remove every fill and variation. The gap between a machine and a drummer, once you control for the grid, lives in the sound, not the clock — which is why sample libraries now record dozens of velocity layers and alternate between takes on repeated hits.
The analogy
THE ANALOGY #Think of an old speech synthesiser reading a sentence in a flat monotone. If you diagnosed its problem as timing — the syllables come too evenly — and set about jittering their lengths, you would not produce a human voice. You would produce a drunk one. What the monotone is missing is stress: which syllables get force, which get length, which get pitch. The drum machine's flatness is the same complaint, and the same misdiagnosis.
Speech stress points outward — it marks contrast and meaning in the world — whereas a drum accent refers to nothing beyond the pattern; and real speech timing does carry structural information at clause boundaries, so timing is not purely decorative there in the way an unswung sixteenth is here.
Clarifying the model
THE MODEL #The correction worth holding is not "microtiming never matters" but that three separate things get bundled under one word. There is random deviation, which the experiments suggest does not help and can hurt. There is systematic offset — swing ratios, a snare that consistently lags, a bass consistently early — which is rule-like, stylistically specific, and entirely implementable by a machine. And there is sound variation in loudness and timbre, which is where most of the perceived humanity actually sits. The folk account takes the evidence for the second and third and credits it to the first.
One more caution about the word itself. In this literature "groove" is defined operationally as the pleasurable urge to move, captured on rating scales; it is not a synonym for musical quality, and a track can score low and still be excellent.
A picture of it
THE PICTURE #How to readThe curve is derived, not measured: it assumes the finding that a swung drummer's short offbeat note stays near 100 milliseconds while the beat itself shortens with tempo, so the ratio is simply the remainder of the beat divided by that fixed 100. Read left to right as the band counts off faster: at a walking 150 the long note is three times the short, and by 300 the swing has flattened to even eighths. The point of the picture is the smoothness — a quantity this orderly is a rule the player is following, not the human error the folk story needs it to be.
What became clearer
WHAT CLEARED #The premise inverted itself under inspection. Perfectly gridded music dominates dancefloors; experimentally added timing deviation reduces groove more often than it raises it; and the "human" timing that survives measurement turns out to be a systematic, tempo-dependent rule a machine can execute exactly. What a bare drum machine really lacks is accent — variation in force and timbre from hit to hit — and beneath that, what makes any of it move you is syncopation against a pulse you are supplying yourself.
Where to go next
ONWARD #- Why a listener's own movement, once started, changes how strongly the beat is felt.
- How different genres settled on different systematic offsets, and how those became stylistic signatures.
Key terms
TERMS #| Term | What it means |
|---|---|
| Microtiming | small departures of note onsets from the metrical grid, measured in milliseconds. |
| Quantisation | snapping recorded or programmed notes exactly onto that grid. |
| Syncopation | placing notes away from the strong beats, so the played pattern and the felt pulse diverge. |
| Swing ratio | the duration of the first note of a pair divided by the second, in a swung subdivision. |
| Velocity layer | a separately recorded sample of the same drum struck at a different force, used so repeated hits differ in timbre and not merely in volume. |
Every term the collection defines is gathered in the glossary.