Play It Again

There is a song you have heard four hundred times. You know where the drums drop out. You know the breath before the second verse. Nothing in it can surprise you any more, and you put it on anyway—while you read, while you paint, while you drag yourself through the last set. Why does that work? And is it doing anything to you, other than the obvious?

This turns out to be a question with a substantial and rather badly-reported literature behind it. I went looking for the standard answers and found that most of them are either wrong, unreplicated, or the result of a study everyone cites and nobody has read. What is actually there is more interesting than the folklore, and in one case it changes what you would sensibly do.

The Song You Are Wearing Out

Start with repetition, because that is the part everyone thinks they know. The textbook answer runs: familiarity breeds liking up to a point, then breeds contempt. Psychologists draw it as an inverted U—liking rises with exposure, peaks, and falls away. It is called the Berlyne curve, it is in every popular treatment, and it feels obviously true from experience.

The trouble is the evidence. The largest meta-analysis of the mere-exposure effect—268 separate curve estimates drawn from 81 published articles—found the inverted U reliably for visual stimuli and not for auditory ones. The classic curve everyone imports into music was measured on things people looked at. Nobody had checked whether the import was legitimate, and when somebody did, it did not carry across.

What does hold up is a finding that is more specific and, to me, much more useful. In a set of experiments from 2004, later extended, listeners heard the same music repeatedly under two different conditions. Some were attending to it. Others had it on while doing something else. The two groups came out with opposite curves.

⤢ enlarge

Sit with that for a moment, because it inverts the usual advice. You do not wear out the music you love by playing it too often. You wear it out by listening to it. The album you have had on quietly for three years while you work is in far better condition than the one you played twice a week with your full attention. If you have a record you cannot bear to spoil, the way to protect it is not to ration it—it is to stop paying attention to it.

Two honest caveats. These are laboratory exposure counts, somewhere between two and thirty-two plays, which is nothing next to a real playlist heard daily for a year; nobody knows where the turning point sits in the wild. And the effect was strongest for real, ecologically valid music and weaker for artificial stimuli, which is at least the right direction for once.

What Is Actually Happening in There

The neuroscience is worth handling carefully, because it is where the gap between what was found and what was reported is widest.

The famous result is a 2011 study using PET imaging to detect endogenous dopamine release while people listened to music that reliably gave them chills. It found dopamine release in the striatum, and—the striking part—a functional split: the caudate more involved during the build-up, the nucleus accumbens more involved at the peak itself. Anticipation and consummation, in different places. It is a beautiful finding and it has been repeated in every popular account since.

It is also a very small study, on a highly unusual population—volunteers screened down to the few who get chills reliably—using a technique that infers dopamine indirectly, and it has never been directly replicated. That does not make it wrong. It makes it a single result carrying an enormous amount of weight.

The stronger evidence is elsewhere and less discussed. In a double-blind pharmacological study, listeners took levodopa (which raises dopamine), risperidone (which blocks it), and a placebo, on separate occasions. Levodopa increased their pleasure in the music and their motivation to hear more; risperidone reduced both. Move the chemistry in either direction and the enjoyment moves with it. That is a causal claim, not a correlational one, and it is what the whole dopamine story ought to rest on.

Alongside it sits musical anhedonia: people who enjoy food, money and sex normally, and get nothing at all from music. Their brains show a specifically reduced accumbens response to music and a normal one to monetary reward, with weakened connectivity between auditory cortex and the reward system. A dissociation that clean is worth a great many correlations.

And then some deflation, which the field deserves. A meta-analysis of brain-imaging studies of familiar music—eleven studies pooled—found no peak activations consistent across them at all. The authors went in expecting limbic and reward structures and got, at a permissive threshold, something that looked motor. So the sentence “familiar music lights up your reward centres” is not currently supported by the imaging literature, however often it appears.

Nor is the tidy story that musical pleasure is prediction error. One 2019 study found accumbens activity tracking modelled prediction errors trial by trial. Another, published months later, quantified surprise and uncertainty across eighty thousand chords of chart pop and found pleasure came from an interaction of the two—and that the accumbens tracked only uncertainty, not the interaction. The literature is genuinely split, and anyone who tells you otherwise has read one half of it.

One last gap, which I think is the most interesting hole in the whole field. Nobody has measured whether heavy repetition blunts the neural reward response to a personal favourite. The behavioural decline is documented; a decline in EEG engagement across repeats is documented; the actual dopaminergic question—does your favourite song still pay out like it used to—has not been asked in a scanner. It is the obvious experiment and it has not been run.

Why It Is Always the Songs From When You Were Fourteen

This part is solid, and it is the part that matters most to actual people. When researchers played 111 chart songs spanning 1950 to 2015 to 470 listeners aged 18 to 82, autobiographical salience and familiarity peaked for music from around age fourteen. Not adulthood, not childhood. The early teens.

Music-evoked autobiographical memory has a neural signature too: the dorsal medial prefrontal cortex responds in proportion to how autobiographically loaded a given thirty seconds is, and people with damage there produce memories that are markedly less episodically rich for music while remaining normal for faces. The music is not merely reminding you. It is engaging the machinery by which you are a continuous person.

There is a lovely wrinkle: the bump cascades. Young adults show a second peak for their parents’ era—the music that was playing in the house before they had any say in it. Some of what you love, you inherited.

Repetition Does Not Just Tolerate Music. It Manufactures It.

Here is the finding I would put in front of anyone who thinks repetition is a weakness in music rather than its engine. Take an ordinary spoken phrase, recorded once. Play it ten times, identically. By the tenth, listeners do not hear speech any more—they hear singing. Asked to repeat it back, they sing it, and the pitches they produce are closer to a simple tonal melody than to the original speech.

The effect breaks if the repetitions are slightly transposed, or if the syllables are jumbled. It requires exact repetition. Which means repetition is not a container that music happens to come in. It is one of the operations that turns sound into music in the first place. In a related study, listeners with no taste for difficult contemporary music rated excerpts of Berio and Carter as more enjoyable, more interesting and more artistic when segments had been artificially made to repeat—changes the composers never wrote.

Play a spoken sentence ten times without altering it and people stop hearing speech. They hear a tune. Repetition is not how music is delivered. It is part of what makes it music.

Now: Music While You Are Doing Something Else

The second half of the question, and the half where the answer is genuinely actionable. The mechanism to know is the irrelevant sound effect, and it is one of the most robust findings in cognitive psychology.

Background sound damages your ability to hold things in order. Not memory in general—serial order specifically. And the thing that does the damage is not meaning: sequences of pure tones disrupt serial recall about as much as speech does. What matters is acoustic change. A sound that keeps varying costs you; a sound that stays the same costs you far less. (A 2024 high-powered replication complicated the theoretical account—the effect showed up for verbal but not spatial order memory—but the core phenomenon is not in doubt.)

From that one mechanism, most of the practical findings follow.

Reading. A Bayesian meta-analysis of 65 studies found background noise, speech and music all produce a small but reliable decrement in reading performance—with intelligible speech and lyrics by far the worst. A clean within-subjects experiment put numbers on it: lyrics versus silence cost about a third of a standard deviation on verbal memory and on visual memory, and about a fifth on reading comprehension. Instrumental music versus silence ranged from a small cost to a small benefit, and none of it was credible—statistically indistinguishable from nothing.

And then the detail that makes the whole essay worth writing. In that same experiment, participants correctly noticed that the lyrics were costing them. They also retrospectively rated the instrumental music as having helped, when it had done nothing measurable at all. We can feel the interference we are getting. We cannot feel the absence of a benefit, so we invent one.

⤢ enlarge

One refinement: the language matters. Lyrics in the same language as the text you are reading hurt most, which is exactly what the interference account predicts. If you must have words, have them in a language you do not speak.

A Word About Mozart, Because He Poisoned the Well

It is worth being precise about the study that produced thirty years of nonsense, because almost nobody knows what it actually was.

In 1993, a one-page letter in Nature reported that 36 college students did better on paper-folding-and-cutting and related spatial subtests after ten minutes of Mozart than after silence or relaxation instructions, and that the advantage lasted ten to fifteen minutes. It was never about IQ. It was never about children. It was never about anything lasting. Within six years, a state governor had put Mozart CDs in the hands of new mothers.

The demolition was thorough. Independent labs following the originators’ own procedural checklist found no effect of any practical significance. One study showed the advantage tracked whatever the listener preferred—Mozart or a Stephen King audiobook, it made no difference. And a meta-analysis of about forty studies and more than three thousand subjects delivered the verdict: Mozart against no stimulus, d = 0.37; any other music against no stimulus, d = 0.38; Mozart against other music, d = 0.15, which is nothing. The largest single moderator of effect size was which laboratory ran the study.

What survived is much duller and quite useful: arousal and mood. Faster tempo raises arousal; major mode lifts mood; both feed into performance on the sorts of tasks that benefit from being a bit more awake and a bit happier. That is the entire mechanism. It has nothing to do with Mozart, and it works with whatever you happen to like.

What Music Genuinely Does Help With

Boring things. This is the honest positive finding and it is well demonstrated. On low-demand sustained-attention tasks—the kind where the enemy is not difficulty but drift—self-selected music increases task-focus states and reduces mind-wandering. In a later study run in people’s own homes, where the competing distractions are real, it also sped up reaction times, with mood and arousal formally mediating the effect.

You should also know that two recent syntheses of this field point in opposite directions. One systematic review of 154 experiments found a general detrimental effect on memory and language tasks, worse for lyrics, worse for difficult tasks. One meta-analysis of 47 studies found a small positive effect on learning—and its largest effects came from studies where the music was played before the task rather than during it. That distinction may be the whole reconciliation: music to get into the right state, silence to do the work.

The Gym, and the Death Metal Question

Here the evidence is unusually good. A multilevel meta-analysis pooled 139 studies, 598 effect sizes and 3,599 participants, and the pattern is clear once you look at which numbers are biggest.

Affective valence: g = 0.48. How the exercise feels. Physical performance: g = 0.31. Perceived exertion: g = 0.22. Oxygen consumption: g = 0.15, with a confidence interval that nearly touches zero. Heart rate: not significant at all. The largest effect by some distance is on your experience of the work, not on the work.

⤢ enlarge

Which brings us to the specific question. Does aggressive music make you stronger? There is no evidence that aggression is the active ingredient. What there is evidence for is preference. A meta-analysis comparing preferred music, non-preferred music and silence found preferred music beating silence on motivation by a large margin, on strength endurance and maximal strength moderately, and lowering perceived exertion—with no advantage at all for speed or aerobic endurance.

And a 2026 re-analysis using more robust statistics deflates even that. Pooling 113 effect sizes across anaerobic exercise with proper handling of dependent effects and risk of bias, most outcomes became uncertain. Only repetitions completed and how good people felt survived, and the certainty of evidence was rated low or very low throughout.

So the honest answer to “should I lift to death metal” is: if you love death metal, yes, and that is the entire reason. The genre is not doing anything. Your relationship to it is. Which also means the person next to you with the disco playlist is getting the identical benefit, and neither of you has grounds to be smug.

One more piece of received wisdom that did not survive contact with evidence: matching tempo to intensity. A 2011 study found preference for faster music rising with exercise intensity, and the fitness world turned it into BPM charts. When the same group revisited it in 2024 using unfamiliar, non-lyrical music, there was no discernible relationship between intensity and tempo preference at all. Fast music was best at every intensity; slow music was worst at every intensity. The original effect may well have been carried by familiarity and lyrics rather than by tempo. Play fast things. Do not do arithmetic.

Trance for Reading: The Honest Answer

You asked—or rather, a great many people ask—whether repetitive, minimal, high-tempo music is good for reading and painting and long stretches of concentration. I went looking properly, and the answer deserves to be given straight.

Nobody has tested it. I could find no study comparing repetitive or minimal or trance-like music against other music for sustained attention or flow. Not one. The lo-fi study-music literature amounts to a couple of tiny papers, one of which found the lo-fi condition performed comparably to silence and concluded it offered emotional stabilisation rather than any cognitive gain.

But there is a real mechanism that predicts it should work, and it is the one from earlier. The damage done by background sound is driven by acoustic change, not by sound as such: steady, repeating sequences disrupt far less than varying ones. A four-bar loop that runs for eight minutes without developing is, to the part of your mind that gets derailed, close to a steady state. The intuition has a mechanism. It does not have a study. That is a more respectable position than most things in this field, and I would rather say so than pretend otherwise.

What I would not spend money on is engineered focus audio. Binaural beats have two meta-analyses reporting moderate benefits and a well-powered direct test reporting, in its own words, rather strong evidence against—and the proposed mechanism fails independently: a systematic review of whether binaural beats actually entrain brainwaves found eight of fourteen studies contradicting the hypothesis. The one serious study of a commercial focus-music product found its benefit was largest early in the session, conditional on attentional difficulty, and it was published by authors affiliated with the company. No independent group has replicated it.

What Follows

Reading or writing anything new: no words. Especially not words in your own language. And do not trust your sense that the instrumental track is helping—that feeling is the best-documented illusion in this whole literature.

Painting, or any practised physical work your hands already know: play whatever you like. This is the quadrant where music genuinely helps, and lyrics do no harm, because nothing in the task is competing for the same channel.

The gym: your favourites, loud. Fast rather than slow. Stop calculating beats per minute. Expect it to change how the work feels rather than what you lift, and take that seriously—feeling better about hard effort is how people keep turning up, which is the only variable that has ever mattered.

And the record you are frightened of ruining: put it on in the background. Do not sit down and give it your attention. That, of everything here, is the finding I did not expect and have started acting on.

One closing thought about the limits of all this. Every number in this essay is an average over strangers. Not one of these studies has a column for the fact that a particular song, heard at a particular age, is now permanently fused to a particular July—and that when it comes on unexpectedly in a supermarket you are, for about four seconds, somewhere else and someone younger. The meta-analyses cannot see that. It is, for most of us, the entire reason we keep pressing play.