The Filled Duration Illusion: Why Subdivision Drags the Beat
Fill the gap between two beats and the gap feels longer. Then you speed up to fix a problem that was never there.
Short answer: an interval that contains events is perceived as longer than an empty interval of the same physical length. This is the filled duration illusion, and in music it has a specific consequence: when the beat is subdivided, the beat feels slower, so players compensate by speeding up. Repp and Bruttomesso demonstrated this in the laboratory in 2010, and then found the same tendency in commercial recordings by well-known pianists, who played subdivided variations faster than the simple themes that preceded them. Nobody in that data was trying to rush. The perception was wrong first, and the hands followed.
The basic effect
Take two intervals of identical duration. Leave the first empty, marked only by a tone at each end. Fill the second with a few events in between. Ask which lasted longer and the filled one wins. The effect has been studied for a long time under exactly that name, and it generalises across modalities and across a range of durations.
Applied to a musical beat, the arithmetic is uncomfortable. The gap between two quarter-note beats at 100 BPM is 600 ms whether it is empty or packed with sixteenth notes. But it does not feel that way. Filled with subdivisions it feels like a longer gap, and a longer gap between beats means a slower tempo. That is how a purely perceptual bias becomes a performance problem.
What Repp and Bruttomesso measured
Their 2010 paper in Attention, Perception, & Psychophysics attacked the question from three directions, and the design is worth following because each part answers a different objection.
In the first experiment, musically trained participants produced the subdivisions themselves within a synchronization-continuation paradigm, the standard task described in our guide to research timing test batteries. They sped up, which is exactly what you would predict if the subdivided beat felt slower and they were correcting toward the tempo they believed they had started with.
In the second experiment, a reproduction task, the same speeding pattern appeared. So the effect was not an artefact of one task's structure.
In a third part, participants simply judged the tempo of a subdivided sequence relative to a preceding simple beat sequence, with no production involved. They rated the subdivided sequence as slower. That closes the loop: the perception is biased whether or not the hands are involved.
The same tendency in real recordings
What makes this study unusually persuasive is that it did not stop in the laboratory. Repp and Bruttomesso measured tempo in famous pianists' recordings of two variation movements from Beethoven sonatas. In a theme-and-variations structure the theme is written mostly in simple beats and the first variation typically subdivides them. There is a strong tendency in those recordings to play the first variation faster than the theme.
They also had musicians perform a simple theme and variations in the laboratory with an explicit instruction to keep the tempo constant. The same tendency appeared anyway. An instruction is not enough to cancel a perceptual bias, which is a familiar finding to anyone who has read our guide to rushing and dragging.
The authors' own summary of the mechanism is the clean statement of the whole effect: intervals containing events seem longer than empty intervals, and therefore must be shortened to be perceived as equal in duration.
Why filled intervals feel longer
There is no single agreed explanation, and the honest answer is that several accounts survive. One family appeals to the sheer number of events processed within the interval. Another appeals to attention. A third, and the one with the most direct experimental support, appeals to predictability.
Pariyadath and Eagleman, publishing in PLOS One in 2007, showed that an unexpected item in a repeated sequence is judged as lasting longer than the repeated items around it. Two details of their result matter. First, the dilation was tied to the unpredictability of the stimulus rather than to its salience or to the amount of attention deployed, since increasing salience did not increase the dilation. Second, the distortion was specific to duration: other temporal judgments about the same stimulus were not correspondingly slowed, which means subjective time in general was not stretching. A follow-up in the same journal in 2012 linked these subjective duration distortions to neural repetition suppression.
Density and predictability are not the same thing, but they interact in music constantly. A subdivided beat is both fuller and, once the subdivision is established, more predictable, and the net effect on felt duration is what the tempo measurements capture.
Duration is bent by more than density
Filling is one of several acoustic properties that shift perceived duration, which is worth knowing before you attribute a drift to subdivision alone. Dawson, Aalto, Šimko and Vainio, publishing in 2017, designed experiments to disentangle the contributions of fundamental frequency, intensity and physical duration to judgments of loudness and duration. Both fundamental frequency and intensity had an impact on duration judgments, with higher and louder sounds judged as longer.
Put together with the register effects described in our guide to tempo illusions, the picture is that felt time in music is a composite of rate, register, loudness and density, and only the first of those is what a BPM number describes.
What it means at the instrument
| Situation | What feels true | What is true | Typical result |
|---|---|---|---|
| Theme, plain quarters | Tempo as written | Tempo as written | Stable |
| Variation, subdivided | Slower than the theme | Same beat rate | Player speeds up |
| Sparse slow passage | Faster than it is | Same beat rate | Player may drag |
| Dense fill before a return | Longer than it is | Same bar length | Early re-entry |
The remedy is not to stop subdividing. Subdivision remains one of the most effective timing tools available, for the reasons set out in our guide to what subdivision is: it fills the space between beats with reference points, which is precisely what makes placement accurate. The remedy is to distrust the feel of tempo while subdividing and to keep an external reference running, which is what our silent-bar metronome drills are built around.
One more consequence worth naming: this illusion is a good reason to be sceptical of tempo judgments made from memory. If your reference for a piece is a passage that happened to be dense, the tempo you stored is biased. Our guide to tempo memory and recall covers what stored tempo references are and are not good for.
Judging a tempo relative to two others is exactly this problem. In Chrontween you hear two tempos and have to find the one exactly halfway between them. Plain tones, no subdivision, nothing to fill the gap: the clean version of the judgment this article is about.
Frequently asked questions
What is the filled duration illusion?
It is the finding that an interval containing events is perceived as longer than an empty interval of the same physical duration. Applied to music, an interval between two beats that is filled with subdivisions feels longer than the same interval left empty, so the beat itself feels slower.
What did Repp and Bruttomesso find?
In a 2010 paper they replicated and extended earlier evidence that metrical subdivision slows perceived beat tempo. Musically trained participants who produced the subdivisions themselves sped up, compensating for the perceived slowing, both in a synchronization-continuation paradigm and in a reproduction task. In a separate judgment task, participants rated a subdivided sequence as slower than a preceding simple beat sequence.
Is there evidence of this in real performances?
Yes. In the same paper, tempo measurements of famous pianists' recordings of two variation movements from Beethoven sonatas showed a strong tendency to play the first variation, with subdivided beats, faster than the theme, which was mostly simple beats. A similar tendency appeared in laboratory performances by musicians who had been instructed to keep the tempo constant.
Why do filled intervals feel longer?
There is no single agreed explanation. Accounts appeal to the number of events processed, to attention, and to how predictable the contents of the interval are. Pariyadath and Eagleman showed that unpredictable stimuli are judged as lasting longer, and that the dilation is tied to prediction violation rather than to attention or salience, so predictability is at least part of the picture.
Does this mean I should stop subdividing?
No. Subdivision is one of the most useful timing tools there is, because it fills the gap between beats with reference points. The finding is that subdivision carries a systematic bias, so you should check your tempo against an external reference while subdividing rather than trusting the feel.
Does pitch or loudness change perceived duration too?
They do. Dawson, Aalto, Šimko and Vainio reported in 2017 that both fundamental frequency and intensity influence duration judgments, with higher and louder sounds judged as longer. So density is one of several acoustic properties that bend felt duration.
Keep reading
- What is subdivision: the tool this illusion attaches itself to.
- Tempo illusions: pitch, register and context biases on perceived speed.
- Rushing and dragging: the practical symptom, and how to catch it.
Sources: Repp & Bruttomesso, A filled duration illusion in music: effects of metrical subdivision on the perception and production of beat tempo (Attention, Perception, & Psychophysics, 2010); Pariyadath & Eagleman, The effect of predictability on subjective duration (PLOS One, 2007); Pariyadath & Eagleman, Subjective duration distortions mirror neural repetition suppression (PLOS One, 2012); Dawson, Aalto, Šimko & Vainio, The influence of fundamental frequency on perceived duration in spectrally comparable sounds (PeerJ, 2017).