Grammar for Writing: Evaluation report and executive summary

155 schools randomised; 5,182 pupils analysed on the primary outcome (2,362 EverFSM)

Year 6 pupils (aged 10-11) in English primary schools

Independently evaluated two-arm cluster RCT (University of York), commissioned by the EEF as an EFFECTIVENESS trial - the scalable version under everyday conditions, following an earlier EEF efficacy trial. Teachers got three days' training plus a fourth; two units (4 weeks + 2 weeks) delivered whole-class in daily 1-hour sessions in place of existing writing lessons. Primary outcome was a BESPOKE writing test built from previous KS2 papers - i.e. researcher-commissioned, not a published standardised instrument - while the secondary outcomes (KS2 reading, KS2 writing teacher assessment, KS2 GPS) are the national tests and are independent. EEF security rating: moderate to high, reduced because >28% of starting pupils were missing from the final analysis after school withdrawals. Process evaluation found high adaptation: nearly three-quarters of intervention teachers adapted the programme, and fidelity to two of its central tenets (connecting grammar to effect/purpose; discussion to tease out choice-making) was judged compromised in a number of observed schools.

No evidence of benefit anywhere. Primary outcome (bespoke writing test) ES = -0.02 (95% CI -0.08 to 0.03), 0 months' progress, n=5,182. EverFSM pupils ES = 0.05 (-0.03 to 0.13), 1 month, n=2,362, not statistically significant. No effects on KS2 reading or KS2 writing attainment; a SMALL NEGATIVE effect on KS2 GPS (grammar, punctuation and spelling), equivalent to one month less progress - the programme did not even move the outcome it most directly targets. Prior-attainment moderation ran the wrong way for the group most often invoked to justify grammar teaching: pupils at or above the pre-test median showed no benefit, and LOWER-performing pupils showed a potential small negative effect (ES = -0.11, ~2 months less progress, not statistically significant). Teachers' own grammar knowledge, measured by a quiz, did not improve either - though the evaluator flags the quiz as possibly unreliable. More than 90% of surveyed teachers nonetheless found the programme, training and materials useful, which is a clean demonstration that practitioner satisfaction carries no information about effect. Cost approximately GBP 28.70 per pupil in year 1, ~GBP 10 per pupil per year over three years.

Low. Randomisation was at school level with an independent evaluator, so treated and control pupils have the same expected distribution of ability-relevant alleles. The residual threat is differential attrition (>28% missing at post-test after school withdrawals), which the evaluator itself flags as the reason for the downgraded security rating.