7 min read

How your brain keeps time

Sight has eyes. Hearing has ears. Timing has nothing: no organ, no receptor, no nerve running from a clock to the brain. And yet you can catch a ball, take a conversational turn without stepping on the other person, and know roughly when a kettle is about to boil.

Interval timing is one of the strangest abilities we have, because it works well enough to run your whole day while having no obvious hardware behind it. Psychologists have spent sixty years arguing about what is actually doing the counting. The argument isn't settled, but the parts that are settled explain almost everything you notice when you play a timing game: why you drift, why the errors grow with the interval, and why some days you're half a second fast on everything.

Three clocks, not one

The first thing to get straight is that “timing” is at least three different jobs, and the brain seems to do them with different machinery.

• Circadian timing, on a roughly 24-hour cycle, run by the suprachiasmatic nucleus in the hypothalamus. This one really is a dedicated clock, and it's the reason you get sleepy at a consistent hour.
• Interval timing, from about a second to several minutes. This is the range a timing game lives in, and it's the contested one.
• Millisecond timing, below a few hundred milliseconds, which underpins speech perception, motor control, and rhythm. It behaves differently enough from interval timing that most researchers treat it as a separate system.

The evidence for the split is that they come apart. Patients and experimental manipulations can damage or distort one range while leaving another intact, which is hard to explain if a single master clock feeds everything.

The classic model: a pacemaker and a bucket

The dominant account for decades has been the pacemaker-accumulator model, formalised in the 1970s and 80s as Scalar Expectancy Theory. The idea is deliberately mechanical. Something in the brain emits pulses at a roughly steady rate. When an interval starts, a switch closes and pulses start dropping into an accumulator, like water into a bucket. When the interval ends, the switch opens, and the level in the bucket is compared against a remembered level for the target duration. Close enough, and you call it a match.

It sounds crude, and it is, but it predicts real behaviour with uncomfortable accuracy. In particular it predicts the single most important fact about human timing:

This is why a well-built timing game can't score you on a fixed margin. Landing within 0.1 seconds of a 1-second target is an ordinary result; landing within 0.1 seconds of a 10-second target is remarkable. Temvo scores relative error rather than absolute, which is not a design preference so much as an admission that the alternative would be measuring the wrong thing.

The competing accounts

Striatal beat frequency

One influential alternative drops the pacemaker entirely. Instead of one oscillator, it proposes many: populations of cortical neurons firing at different frequencies, all reset at the start of an interval. As they drift out of phase with each other, they produce a pattern that is unique at every moment. Neurons in the striatum learn to recognise the pattern that coincided with reward. Timing becomes pattern recognition rather than counting.

Intrinsic network states

A more radical position says there is no clock at all, dedicated or otherwise. Any neural network that is still settling from a disturbance carries information about how long ago the disturbance happened, in the way ripples carry information about when a stone landed. On this view timing is a free side effect of neural dynamics, computed locally by whichever circuit needs it.

These aren't settled debates and you don't need to pick a side. What matters for a player is that all three accounts agree the mechanism is noisy, that the noise scales with duration, and that it can be recalibrated by feedback.

Why your clock speed changes

If there is something like a pacemaker, then anything that changes its rate should change how long an interval feels, and this is where the research gets satisfyingly weird.

Raising core body temperature tends to make time feel slower, as though the clock were running fast and racking up pulses more quickly than usual. Dopaminergic drugs shift timing in the direction their pharmacology predicts: raise dopamine and intervals are typically reproduced short, as if the clock sped up. Arousal, stimulants, and even loud repetitive stimulation nudge the same dial. Effect sizes vary between studies and none of this is a party trick you can reliably perform on yourself, but the direction of the effect keeps showing up.

Practically, this is why timing is a skill with weather. You will have days where every round comes in long and days where every round comes in short, and it isn't only concentration. Your clock has a rate, and the rate moves.

Knowing you're being timed changes the answer

There is a hard split between prospective timing, where you know in advance that you'll be asked how long something took, and retrospective timing, where the question arrives afterwards. They rely on different information. Prospective judgements draw on attention paid to time as it passed. Retrospective ones are reconstructions, built from how much you remember happening.

Every mode in a timing game is prospective, which is why it feels so different from estimating how long you were stuck in a queue. It also means the skill you build here is narrower than “a better sense of time”. It's specifically a better ability to attend to elapsed duration while it elapses. The attention story is where that gets interesting.

Vierordt's law, or why you regress to the middle

One last effect, and it's the one players notice fastest. Across a set of intervals, people tend to overestimate the short ones and underestimate the long ones, pulling every answer toward the middle of the range they've been exposed to. It was described in the 1860s and it has been rediscovered constantly since, in part because it is remarkably robust.

The modern reading is that it's not a flaw but an inference. When your measurement is noisy, the statistically optimal move is to blend it with your prior expectation, and your prior is roughly the average of what you've seen lately. That produces exactly the observed pull toward the mean.

For a player, the useful consequence is that your errors are not random. If your long rounds are short and your short rounds are long, you aren't inconsistent, you're being pulled toward your own internal average, and that is a bias you can correct for deliberately once you know it's there.