Teaching visual art and drama

The program (what a school should actually do)

Arts education is in a strange evidentiary position. Its transfer claim is one of the best-tested and most clearly failed propositions in education research, and its core claim — that teaching art produces artists — has never been tested at all.

Keep the arts. Just stop justifying them by test scores.

• The EEF's Learning About Culture programme ran five cluster-randomised trials across 400 English schools and ~8,500 children with harmonised standardised outcomes. Every primary attainment estimate is null; the pooled Key Stage 2 writing estimate is −0.04 [−0.14, 0.07] on 5,255 pupils.
• Houston randomised 42 schools and doubled their arts provision for $14.67 per pupil per year: reading −0.019, maths −0.001. The one positive is writing (+0.128 SD) on the smallest sample and the outcome nearest the treatment.
• REAP said this in 2000 from 188 reports and 275 effect sizes, and warned the field: "those who live by instrumental claims risk dying by such claims." (evidence: arts and academic achievement)

Buy trips, and buy the live version.

• Lottery-assigned guided museum tours improved blind-rated critical analysis of an unfamiliar painting three weeks later by 0.10 SD (n = 3,610) and 0.08–0.13 SD (n = 3,958), with effects roughly twice as large in higher-poverty and rural schools.
• Live theatre lotteries moved tolerance +0.142 and plot and vocabulary knowledge +0.101. The decisive result is the null movie arm: students randomised to watch a film of the same play gained nothing on any outcome, so the mechanism is the encounter, not the day out. (evidence: cultural exposure)

Use drama to make a text stick, not as a literacy programme.

• Enacting a story reliably improves recall and comprehension of that story, and REAP found the gains carried to non-enacted texts. That is the use to make of it.
• But the published effects are a measurement artefact at scale: researcher-made tests d = 0.89, standardized tests d = 0.29; one assessment 0.87 versus more than ten 0.04; and the dose-response runs backwards (0.87 at 3–10 hours, −0.01 beyond 20). The EEF's drama trial (n = 821) is null on reading and oral narrative, and in a randomised trial with an active control, plain oral reading beat readers' theatre. Have the class teacher run it — teacher-led 0.89 versus visiting artist 0.31. (evidence: drama and verbal skills)

Teach art on judgement, and say so.

• A systematic review of every (quasi-)experimental arts-education study found one visual-arts study among 26, and it measured self-concept. The strongest study in existence is a 42-student pre-post with no control group.
• The one robust empirical result about drawing is that error comes from misperceiving the object, not from the hand — so instruction should train looking. Nobody has trialled it. (evidence: does teaching art teach art)

What to refuse to spend money or time on

• Any arts-integration package sold on achievement gains. Field average g = 0.11, maths null, and 34 of 44 catalogued interventions have a rationale and no measured effect.
• Turnaround-style pre-post evidence. Eight self-selected schools, no control group, proficiency- rate changes reported as if they were effect sizes, confounded with concurrent federal turnaround funding.
• Screening a film instead of taking the trip. The movie arm of the theatre lotteries was null on every outcome.
• Visiting teaching artists bought to raise literacy. Teacher-led drama gives d = 0.89 against 0.31 for teaching artists across 32 studies.
• "Right-brain" drawing methods. No controlled evaluation of Drawing on the Right Side of the Brain exists, and its R-mode framing is metaphorical rather than neuropsychological — the same mechanism-first pattern as misc-fads.
• Teacher perception as an outcome measure. In the EEF trials, 85–97% of staff perceived improved creativity and 88–98% improved engagement, against measured effects of essentially zero. That contrast is the single most useful line in the report.

The hereditarian bottom line for a founder

Arts participation has no twin study, so the selection story here is inferred rather than measured — but the inference is well grounded and it runs in two directions.

On participation. The music side of this archive supplies the number: taking up an instrument is 78% heritable with no detectable shared environment, and genetically correlated with verbal IQ. Propensity-weighted analysis of national panel data concludes that "selection into arts education is at least as strong as any direct effect on outcomes." Every observational arts-and-achievement finding is estimating who takes part.

On artistic ability itself. Drawing at age 4 is 29% heritable — as heritable as general ability at the same age — with a genetic correlation of 0.52 with g at 14. The entire artists-outperform-non-artists literature is comparing self-selected experts on a heritable trait that shares causes with general ability, and none of it can separate training from selection.

Two things this does not license. Arden's shared-environment component for drawing is non-trivial (≈.26), which is exactly where family and school input would sit — heritability at 29% leaves ample room for teaching. And the randomised nulls on achievement establish something stronger than "the correlation is inflated": they establish that the correlation is entirely attributable to who takes part.

Open questions a founder should watch

• Does school art instruction improve art skill? Genuinely unknown. A school could settle it for the price of two ten-minute still-life tasks and a blind rater.
• Whether the Houston writing effect (+0.128 SD) is real near transfer from composition-adjacent arts work or a chance finding on the smallest sample in the trial.
• Whether REAP's claim that drama gains carry to non-enacted texts survives a randomised test.
• Whether cultural-exposure effects persist beyond the eight-week measurement window, or compound with repeat visits. Both programmes come from the same research team and neither has been independently replicated.
• Whether arts provision does anything for behaviour. Houston's discipline effect (−3.6pp) fails randomisation inference at p = 0.14; the music side's El Sistema lottery moved self-control and aggression for boys exposed to violence. Something may be there, and it is about structured group activity rather than about art.