You have no sense organ for time. Light has eyes and sound has ears, but duration is constructed by the brain from other signals: how many events happened, how much attention was spent, how often you looked at a clock. That is the reason why a wait in a queue seems endless and a good conversation seems to pass in a flash. It is also why estimating a few seconds is a good test of intuition.
A clock with noise
Many models describe an internal clock with a pacemaker that produces pulses, a gate that opens when attention is on time and an accumulator that counts. The best-known version is scalar expectancy theory, developed by John Gibbon in the 1970s from animal experiments. Its key prediction is the same as Weber's law for the other senses: the error in a time estimate is proportional to the duration. A person who is off by 100 ms for one second will be off by about a second for ten seconds. This relation has been found in humans, rats and pigeons, and is called the scalar property of timing.
It matches what the duration game shows. You are accurate to perhaps ten or fifteen percent for durations of a few seconds, and a bit worse for very short and very long intervals.
Biases in duration
- Vierordt's law. Karl von Vierordt noted in 1868 that short intervals tend to be overestimated and long ones underestimated, with an indifference point somewhere around a second or two. Your estimates drift toward the middle of the range you have been seeing, which is called central tendency.
- Attention. The more attention you give to the passage of time, the longer it seems. Watching a pot makes it longer. A demanding task, which takes attention away, makes it shorter.
- Novelty and the oddball effect. A new or unexpected stimulus in a series of repeated ones is judged to last longer. Tse and colleagues in 2004 showed this in a visual experiment with an "oddball" that lasted the same as the others.
- Emotion. Threatening images appear longer, perhaps because arousal speeds up the clock.
- Body state. Fever and stimulants speed up the clock, and cold and sedatives are claimed to slow it. Results are mixed.
Prospective and retrospective time
There are two different questions. A prospective judgement is made when you know you will be asked: you watch a dot and then say how long it was visible. A retrospective judgement is made after the fact, from memory: how long was that meeting? Prospective judgements depend on attention to time. Retrospective judgements depend on how many events you remember, which is why a holiday full of new experiences feels long in memory, and an uneventful week feels short. Our games use prospective judgements, as that is the only one that can be measured cleanly.
Counting, tapping and music
Subdividing helps. Counting "one-Mississippi, two-Mississippi" works because it replaces a fuzzy internal clock with a rhythmic action of nearly constant length, close to one second. Counting is not a bad strategy; it just measures your counting more than your intuition. Tapping along with a pulse and tempo judgements profit from the same effect. Musicians are on average better at timing tasks, and trained drummers can keep a tempo with an error of just a few percent.
People also compare better than they name. Deciding which of two pauses was longer, as in our Gap game, is easier than saying how long one pause was, since you do not need a unit, only a comparison. Similarly, judging a beats-per-minute value is easier if you know the tempo of a few familiar songs, because they serve as reference points.
Practising
If you want to improve, play the duration game without counting for a few rounds and check the sign of your error. If you are mostly early, your clock runs fast; if you are late, it runs slow. Knowing this is half the correction. The next day, do it again. Time is the quantity where feedback matters most, because there is no external scale to check against.