Earworms: Why Songs Get Stuck in Your Head (and How They Leave)

This is an unusually well-studied phenomenon for something so trivial, largely because it is easy to collect data on: thousands of people will happily tell you what is stuck in their head right now.

What an earworm actually is

The technical term is involuntary musical imagery, or INMI. The defining features are that it arrives unbidden, repeats, and involves a fragment rather than a whole song — typically a section of chorus or hook lasting fifteen to thirty seconds, looping rather than progressing.

It belongs to the broader category of involuntary memory, alongside spontaneous autobiographical recollections. Music appears to be disproportionately represented in that category, which is itself a clue: musical memory has properties that make it unusually prone to arriving without being asked for.

One counterintuitive finding is worth flagging early. Earworms are not degraded or distorted versions of the music. When researchers compared involuntary and deliberate recall of the same tune, involuntary recall was no less accurate in tempo — people's earworms run at close to the correct speed. What differs is emotional impact, with earworms appearing to affect mood more directly than deliberate recall.

So it is not a corruption of a memory. It is an accurate retrieval that nobody requested, which puts it in the same family as knowing a word arrives when you did not ask for it — involuntary access to a memory system that usually waits to be queried.

Almost everyone, almost every week

Survey work consistently finds involuntary musical imagery to be near-universal, with the large majority of people reporting it at least weekly and a substantial minority daily. It is more frequent in people who engage more with music — musicians and those who listen more report more — which is unsurprising given the role of exposure.

Personality associations have been reported with openness to experience and with neuroticism, and imaging work has reported correlations between earworm frequency and cortical thickness in regions handling auditory perception and imagery. These are modest correlational findings in small samples and should be read as suggestive.

One line of research is particularly informative about mechanism: people with congenital amusia, who have lifelong difficulty perceiving musical pitch, have been surveyed about earworms too. Work in that population helps separate which aspects of the experience depend on accurate pitch processing and which do not — a natural experiment that ordinary samples cannot provide.

Which songs, and why

The most substantial work on this comes from Kelly Jakubowski and colleagues, who took tunes named as earworms by around 3,000 survey participants, matched 100 of them against 100 tunes never named as earworms — controlling for popularity and style — and compared the two sets across 83 melodic features.

Three differences emerged. Earworm tunes had faster average tempos. They had more common global melodic contours — the overall shape of rises and falls followed patterns widespread in pop music. And they had less common average gradients between melodic turning points: unusual movement in the intervals themselves.

That combination is the interesting part. A wholly conventional melody is easy to process and forgettable; a wholly unconventional one is hard to encode. The tunes that stick are conventional at the level of overall shape and unusual at the level of detail.

Recent chart success and general popularity also predicted earworm status independently, which is unsurprising — exposure matters — but the melodic features held after matching for popularity, so they are not simply proxies for how often you have heard something.

The unfinished-loop claim

The explanation you have most likely encountered is that earworms persist because the tune is incomplete, invoking the Zeigarnik effect: unfinished things stay mentally accessible, so an unresolved musical phrase keeps demanding closure until you play the song to the end.

It is an appealing story with much thinner support than its ubiquity suggests. The Zeigarnik effect itself has an inconsistent replication record. The specific extension to musical imagery has not been established, and it does not fit the observation that earworms are frequently fragments of a chorus that resolve perfectly well within themselves and loop anyway.

The connection is worth reporting because it is everywhere, and worth marking as an assumed rather than demonstrated mechanism. The same caveat appears from the other direction in our article on why unfinished tasks intrude: the parallel is intuitive, the shared mechanism is asserted rather than shown.

What starts one

Diary and experience-sampling studies find several reliable triggers. Recent exposure is the most obvious. Memory associations are common — a word, a place, a person connected to the song. And low-attention states matter: earworms are more frequent during undemanding activities like walking, showering or routine work.

The relationship with cognitive load appears to be non-linear. Very low load leaves mental space for intrusions, while very high load also seems to increase them, with the fewest reported at moderate levels of engagement. That pattern has been reported more than once but the underlying explanation is not settled.

Getting rid of one

The most interesting experimental result here involves chewing gum. Beaman and colleagues found that chewing gum reduced both voluntary and involuntary musical imagery, and the proposed reason is articulatory suppression — the jaw movement interferes with the subvocal rehearsal loop that musical imagery seems to depend on.

That is a small finding but a mechanistically informative one: if occupying the articulatory system disrupts an earworm, the earworm is partly running on that system rather than existing purely as abstract auditory imagery.

Other approaches reported by participants include engaging with a moderately demanding task, deliberately listening to the song in full, or listening to a different one. Survey work finds people report all of these working sometimes, which is roughly what you would expect from a phenomenon that resolves on its own within minutes to hours regardless.