The Evidence on Teaching

Subject

Teaching music

Bottom line. Teach music because music is worth teaching. Instruction moves musical skill by about half a standard deviation, which is a normal, respectable instructional return and the only claim about music that needs no hedge. Every spillover promise is weaker than advertised: the two largest randomised trials — 3,004 five-year-olds given daily music for a year, and a 2,914-child lottery for El Sistema — found nothing on reading, maths or cognition, and the disadvantaged-pupil subgroup came in at exactly 0.00. A small effect on laboratory executive function is genuinely contested and worth watching, but it is not a budget argument. Meanwhile the selection you are being sold as an effect is measurable: taking up an instrument is 78% heritable, self-selecting children are already 0.29 SD ahead before the first lesson, and the SAT advantage of music students goes from +37 points to +0.23 the moment prior achievement enters the model.

The program (what a school should actually do)

Music is the subject where the gap between what is claimed and what is known is widest in this entire archive — and, unusually, the honest case is still a good one. The trouble is that almost nobody makes it.

Teach music for musical skill, and expect a normal instructional effect.

  • Music teaching moves musical outcomes at d ≈ 0.56 overall — music literacy and sight-reading 0.54, performance and aural skills 0.51. Singing instruction moves singing accuracy at g = 0.43 across 34 studies.
  • Buy feedback, not a branded method. Feedback (self, teacher, computer) is the strongest teaching moderator in the singing literature, and specialist systems — Kodály, Orff and their relatives — produce lower effects than generalist teaching in the only meta-analysis that has compared them. Technology adds a negligible amount.
  • Start singing early: ages 5–8 is where the singing-accuracy evidence concentrates, and singing costs nothing. (evidence: music instruction)

Do not fund music on the promise that it raises attainment.

  • The EEF's First Thing Music trial randomised 3,004 five-year-olds to fifteen minutes of music every school day for a year: reading g = 0.07 [−0.02, 0.16], and the free-school-meal subgroup 0.00 [−0.20, 0.19].
  • A lottery evaluation of El Sistema across 2,914 Venezuelan children found no cognitive or academic effects at one year.
  • Maths is null in every meta-analysis of music training that has ever been run. (evidence: music and cognitive transfer)

Know which claim is actually still open, and keep it out of the budget meeting.

  • Whether instrumental training moves laboratory executive function is genuinely contested. Sala & Gobet put it at g = 0.056 under active controls, falling to −0.021 after bias correction; a reanalysis of their own script gives 0.234; an instrumental-only meta gives 0.26; and eight active-control RCTs on inhibition control specifically give g = 0.60 with zero heterogeneity.
  • Two independent teams locate the surviving effect in the same place — executive function — which is the best reason to keep watching. Nothing about persistence, generalisation or school consequence has been shown, which is why it is a research question and not a purchase.

Assume selection until a randomiser removes it.

  • Taking up an instrument is 78% heritable with no detectable shared environment, and genetically correlated with verbal IQ (r_g = 0.44–0.80).
  • Children who choose music are 0.29 SD ahead before the first lesson; randomised children start level.
  • The music–IQ association survives controlling family income and parental education, and dies controlling music aptitude — training's unique contribution is 0.63% of variance.

What to refuse to spend money or time on

  • Any music programme justified by test scores. Two randomised trials with 5,900 children between them say no, and the subgroup where the benefit is always promised is exactly zero.
  • The Mozart effect, in any packaging. Mozart d = 0.37 against silence; any other music d = 0.38. There is an arousal effect. There is no Mozart effect (misc-fads).
  • Branded classroom-music systems bought for their pedagogy. Specialist methodologies underperform generalist teaching on musical outcomes, and whole-class music-appreciation programmes are where the randomised transfer nulls concentrate.
  • Aptitude screening. Engagement is 78% heritable and ability is substantially heritable, but heritability describes population variance and licenses no prediction about a child — the same conclusion the archive reaches for athletic talent.
  • "Start before seven or it's too late." The one genetically informative test found the early-start advantage in achievement is accumulated practice hours, and the residual aptitude association was fully explained by familial factors.

The hereditarian bottom line for a founder

Music is the domain where practice explains more of attainment than anywhere else — Platz's corrected r꜀ = 0.61 is the highest practice–achievement correlation in any studied field — and it is also the domain with the single cleanest demonstration that practice is not the cause. In 10,539 Swedish twins, monozygotic pairs discordant for practice by up to 20,228 hours showed no within-pair difference in music discrimination: rhythm p = .64, pitch p = .24, melody p = .43.

Both facts are true and the reconciliation matters. Practice correlates with attainment because the same people who have the aptitude do the practice. Instruction still works — literally nobody plays an instrument without being taught — and per the premise, teaching every child music converts "who was offered lessons" into "who has the aptitude," which is evidence the teaching worked, not that it failed.

One scoping correction the archive insists on: Mosing's null concerns music discrimination, an auditory perceptual test. The authors concede that instrument-specific motor and notation skills may still respond to practice. Do not read it as "practice does not improve playing."

Open questions a founder should watch

  • No randomised trial of instrumental instruction with an independent measure of musical skill exists. It would be cheap and it would settle the base claim. That it has never been run, while the field ran dozens of trials on maths scores, is the most revealing fact in this subject.
  • Whether the instrumental-versus-classroom-music distinction really explains the meta-analytic disagreement is untested by anyone outside the disputing parties.
  • Whether the inhibition-control effect survives a large preregistered active-control trial. One is registered and has not reported.
  • Persistence, everywhere. Not one estimate in either topic has a follow-up after training stops.
  • Whether structured group arts activity does something behavioural for vulnerable children — El Sistema moved self-control +0.20 SD and aggression −0.24 SD for boys exposed to violence — is a better-supported question than the cognitive one, and it is not really a question about music.

Evidence topics

← Home