First Thing Music: Evaluation Report

3,004 pupils randomised (1,492 treatment, 1,512 control) across 121 classes in 64 schools; 2,150 analysed (1,116 / 1,034), i.e. 28% attrition

English Year 1 pupils, aged 5-6, in state primary schools

The largest randomised trial of music instruction in existence and, structurally, the best test anyone has run of whether classroom music raises reading. Class-level cluster randomisation, a full school year of Kodály-inspired music for 15 minutes EVERY school day, independent evaluators, and a published standardised primary outcome (Progress in Reading Assessment). Grade B rather than A for two reasons the report states itself: the control was business-as-usual rather than an active alternative, and 28% attrition with free-school-meal status significantly predicting missingness means the estimate "may be biased." EEF security rating 2 of 5 padlocks. The trial also sits inside the wider Learning About Culture programme, so do not count it twice with that source.

Reading attainment Hedges g = 0.07 (95% CI -0.02 to 0.16), p = 0.13 - equivalent to one month's additional progress and not statistically distinguishable from zero. In the free-school-meal subgroup, which is where arts advocacy usually locates its effects, the estimate is exactly 0.00 (95% CI -0.20 to 0.19), p = 0.97, n = 451: zero months. Secondary outcomes were null-to-negative: social skills -0.09 and creative self-efficacy -0.03. The dose is the point - this is not a token exposure but daily music for a whole school year in a programme designed by specialists, delivered to 3,004 five- and six-year-olds, and the reading effect is a month of progress that the evaluators decline to call real.

Low. Class-level randomisation within schools removes selection into music. The live threat is differential attrition by free-school-meal status, which the report flags.

No replication. Its null is consistent with Sala & Gobet (2020) - whom the evaluators cite in the report as the prior reason to doubt music-to-cognition transfer - and with the EEF Act, Sing, Play trial, where music workshops did not outperform drama workshops on maths or literacy. It is in tension with Román-Caballero et al. (2022) only under their own distinction: this is classroom vocal music, not instrumental tuition, and that meta-analysis argues the null belongs to exactly this kind of programme.