Grammar for writing? An investigation of the effects of contextualised grammar teaching on students’ writing

744 Year 8 pupils in 31 schools

Year 8 pupils (aged 12-13) in English secondary schools

Developer-led mixed-methods trial (Reading and Writing 26(8), 1241-1263) with a randomised controlled component plus a qualitative study - the study that generated the Grammar for Writing programme later tested twice by the EEF. The theoretical claim is that grammar taught as a meaning-making resource EMBEDDED in writing instruction is a different intervention from the isolated formal grammar teaching that earlier meta-analyses found ineffective. This is the strongest positive evidence in the topic and it is also the most criticised: Wyse & Torgerson (2017) show that intention-to-treat analysis was not used and that analysis was conducted at pupil level despite randomisation and delivery being at school/class level, both of which inflate precision and can inflate the point estimate.

Positive overall effect on writing performance for the intervention group, e = 0.21, p < 0.001. But the moderation is the finding that matters and it runs against the usual justification for grammar teaching: "the study also indicates that the intervention impact differentially on different sub-groups, benefiting able writers more than weaker writers." A companion paper (Myhill et al. 2012/2013) reports teachers' own grammatical subject knowledge as a mediator of student outcomes. The authors describe the study as "the first to supply rigorous, theorised evidence for the potential benefits of teaching grammar to support development in writing" - a claim that two subsequent independent EEF trials of the same programme did not reproduce.

Medium. Randomisation should neutralise heritable differences between arms, but the analysis was conducted at pupil level rather than at the level of randomisation and did not use intention-to-treat, so the reported estimate is not protected by the design in the way the headline implies. The subgroup finding (able writers benefit more) is a within-arm comparison and is therefore fully exposed to the archive's premise - "able writers" is substantially a heritable category, so the moderation is an association, not a demonstrated interaction with instruction.