Does Music Help You Focus? Two Meta-Analyses Disagree, and the Lyrics Explain Why

Most articles on this question pick a side. The honest position is that the field has not picked one either - two meta-analyses in the same journal, twelve years apart, report effects of essentially opposite sign.

Working out why they disagree turns out to be more useful than either result on its own, and it produces advice you can actually apply.

The two meta-analyses

Kämpfe, Sedlmeier and Renkewitz (2011) included 97 studies, of which 66 allowed a direct comparison of background music against no music, covering 4,501 participants.

Overall effect: r = .03. Effectively zero.

The domain breakdown is where the interest lies:

Domain	Effect (r)	Studies
Emotion	+.11	6
Mundane behaviour	+.08	22
Cognition - judgement	+.03	13
Cognition - achievement	-.01	43
Memory	-.09	8 (n = 274)
Reading	-.11	8 (n = 680)

For study performance specifically - the row with 43 studies behind it - the pooled effect is nil. For reading, it is negative, and Kämpfe notes that all eight reading studies pointed the same way.

de la Mora Velasco, Chen, Hirumi and Bai (2023) pooled 71 effect sizes from 47 studies on learners specifically and found d = 0.314 - small but positive, favouring background music. We could not retrieve the confidence interval or total sample size and are not printing them.

The reconciliation

The later meta-analysis identifies a moderator that resolves most of the conflict: the effect was larger when music played before the assessment rather than during it.

That fits a mechanism that has been in the literature for decades. Music before a task acts on arousal and mood - it wakes you up, it puts you in a better frame of mind, and those carry into the task. Music during a task competes for the same processing resources the task needs.

Which is why the domains split the way they do in Kämpfe’s data. Mundane behaviour and emotion, where arousal helps and there is nothing to interfere with, come out positive. Reading and memory, where verbal material is being held and manipulated, come out negative.

The most defensible summary: music before you start plausibly helps. Music while you read plausibly hurts. Both effects are small.

Lyrics are the active ingredient - and liking the song does not save you

Perham and Currie (2014) tested 30 undergraduates on reading comprehension under four conditions: quiet, liked lyrical music of the participant’s own choosing, disliked lyrical music (thrash metal), and instrumental music (instrumental thrash metal).

Main effect F(3, 87) = 8.05, MSE = 0.47, η² = 0.22, p < .05. Two comparisons matter:

• Quiet beat liked lyrical music, F(1, 29) = 10.43, p < .05
• Quiet beat disliked lyrical music, F(1, 29) = 16.96, p < .001

And the two nulls that make the finding useful:

• No significant difference between quiet and instrumental music
• No significant difference between liked and disliked lyrical music

So it is not about enjoyment, and it is not about music as such. It is about words. Choosing music you love does not protect your reading comprehension; removing the lyrics does.

Thirty participants is a small sample and the effect estimate is imprecise. But the direction is consistent with a much older and larger literature on irrelevant sound.

Why words specifically

Kattner and Meinhardt (2020) tested 50 participants on serial recall of tonal and verbal sequences under white noise, speech and instrumental music. The disruption turned out to be domain-matched.

For verbal recall, only speech was reliably disruptive (d_z = 0.19); instrumental music was equivalent to white noise. For tonal recall, instrumental music disrupted more than noise (d_z = 0.45) and more than speech (d_z = 0.24).

That is a cleaner result than the blanket claim usually made. Pitch interferes with pitch memory; speech interferes with verbal memory. Which means:

• Reading or writing - lyrics are the problem, instrumental is roughly free
• Learning music, or anything with a pitch or rhythmic structure - instrumental music is the problem

It also means “music wrecks your memory” is an overstatement of what has been shown.

The Mozart effect is dead, and how it died is instructive

Pietschnig, Voracek and Formann (2010) meta-analysed 39 studies, 104 independent samples, over 3,100 participants:

Comparison	Effect (d)	95% CI	Effects
Mozart vs control	0.37	[0.23, 0.52]	38
Any music vs control	0.38	[0.13, 0.63]	15
Mozart vs other music	0.15	[0.02, 0.28]	11

The killing comparison is the middle row against the top. Any music performed the same as Mozart. There is a small general arousal effect and essentially nothing specific to the composer.

The authors also found effects more than three times larger in laboratories affiliated with the original proponents than in independent ones, along with detectable publication bias. Their verdict: “little evidence left for a specific, performance-enhancing Mozart effect.”

That lab-affiliation pattern is worth remembering the next time a study of a study aid comes from the people who invented it.

What this does not establish

Nothing about lo-fi playlists or “focus music” products specifically. No controlled trial of those exists that we could find. The general findings apply, but the products have not been tested.

Nothing over long periods. These are single-session laboratory tasks. There is no evidence about habitual listening while studying over weeks.

The confidence interval for the positive meta-analysis. We could not retrieve it, so we cannot tell you how precise d = 0.314 is.

Whether individual differences matter as much as they do for noise. In the white noise literature, about a third of one clinical sample performed worse with sound, and the predictor was a stable personal characteristic. Nothing comparable has been mapped for music.

What follows

Four things you can act on, in descending order of how well evidenced they are.

If you are reading or writing, take the lyrics out. This is the best-supported result on this page - verified across a within-subject experiment and consistent with the irrelevant-speech literature. Instrumental music was statistically indistinguishable from silence.

Do not assume your favourite song is exempt. Liked and disliked lyrical music produced the same impairment. The intuition that music you love fades into the background did not survive testing.

Consider putting the music before the work rather than during it. This is the moderator that reconciles the two meta-analyses, and it costs nothing to try.

Expect a small effect either way. The pooled effect on cognitive achievement across 43 studies is r = -.01. Whatever music is doing to your studying, it is not the main thing determining how it goes.

Music is one of several study conditions where the felt benefit and the measured benefit come apart - see the cognitive offloading guide for the wider picture, does white noise help you focus for the unstructured-sound version, and spaced repetition for a technique that does survive testing.