Does the Student Grouping Study really support a call to back setting?

Posted on 06/05/2026

Rachel Marks looks closely at a recent and much-reported EEF study. She finds it falls far short of the headline claims.

Imagine that for a known medical condition there are two treatments, offered by different hospitals. The treatments are evaluated through a naturalistic quasi‑experiment. Treatment A is used by 27 hospitals; treatment B by 62. Hospitals are imperfectly matched, differing by inspection rating. There is ‘local variation’ in how treatment B is administered, with different doses and regimes. Forty per cent of hospitals nominally offering treatment A switch the sickest patients to treatment B. Overall, 29 per cent of patients drop out of the study. Outcomes are measured largely in terms of survival, rather than a wider range of quality‑of‑life indicators. The headline result comparing treatment A to treatment B is an effect size of −0.05 (CI −0.14, 0.03). Would you push for hospitals to be advised to adopt treatment B on this basis?

In healthcare, the answer would almost certainly be ‘no’. In education, however, we appear far more willing to act on such an evidence base. The recent Education Endowment Foundation (EEF) Student Grouping Study illustrates this vividly. Despite a non‑significant headline effect and a low‑to‑moderate evidence security rating, the study has been widely reported as showing that mixed‑attainment grouping in secondary mathematics (call it treatment A) has ‘failed’, prompting calls to support setting. This reading is understandable: the report foregrounds a headline result that students in mixed‑attainment classes made ‘one month’s less progress’. Yet this framing risks asking the study to answer a question it was never equipped to address; that is, whether it is grouping structures themselves, rather than the beliefs and practices enacted within them, that shape students’ mathematical experiences.

Deeply permeated

The study compares schools that organise students differently. It cannot compare classrooms genuinely free from fixed‑ability thinking with those structured around it. That distinction matters. Research repeatedly shows that fixed‑ability thinking survives organisational change. Even in ‘mixed attainment’ classrooms, teachers and students often behave as if ability were intrinsic, measurable, and largely immutable. Read in this light, the findings do not undermine mixed attainment grouping so much as reveal how deeply the discourse of ability permeates schooling.

While the report mostly uses the language of ‘attainment’, the conflation of ability and attainment is never far below the surface. The study is framed as exploring differences between ‘grouping by subject ability (or setting)’ and mixed attainment grouping (p8). The head of mathematics survey refers to ability or abilities 11 times (Appendix C.7). This language matters. It frames ability as a natural property of students rather than an outcome shaped by opportunity (within and beyond schooling). If the goal is to challenge fixed‑ability thinking, can we do so while giving practitioners a vocabulary that normalises it? One head of mathematics, reflecting on mixed-attainment teaching, remarks that ‘the best students are held back and the confidence of the less able students is shattered’ (p55, emphasis added). These labels are not attached to attainment at a moment in time, but to students themselves. The antonyms of ‘best’ here speak volumes.

The study also acknowledges that only a small proportion of English secondary schools teach mathematics in mixed-attainment classes throughout both Years 7 and 8. This matters profoundly. These schools operate within a system including league tables, curriculum frameworks and assessment structures that is structurally aligned to setting. Schools adopting mixed-attainment grouping under such conditions tend to differ from the mainstream not just in observable ways, but in professional learning capacity, curriculum vision and willingness to tolerate short‑term uncertainty. As a result, the study does not compare setting and mixed attainment as educational ideals. It compares established setting systems with a small number of schools attempting an alternative within a system built against it. To treat this comparison as evidence that setting is preferable is to confuse systemic inertia with pedagogical effectiveness. The headline finding of slightly lower progress in mixed-attainment groups is therefore better understood as a consequence of structural constraint than of inherent weakness. The lesson observation data support this. Under external accountability pressures, pedagogies looked strikingly similar regardless of grouping:

Teaching practices are broadly similar across schools that used mixed attainment grouping and schools that set by attainment … driven by national policy around mastery. (p69)

Further:

The overwhelming majority of the lessons observed in both mixed attainment and setting schools were dominated by teacher-led whole-class teaching with all students tackling the same sets of tasks and problems. Often this was interspersed with brief opportunities for paired discussion of less than one minute duration … We observed very few instances of groupwork. Typically, the focus was on procedures and methods with little focus on developing understandings of the underlying concepts. (p59)

In short, changing how students were grouped did not meaningfully change how mathematics was taught.

The real question

The data also reveal persistent assumptions tied to fixed ability, including the belief that pace and early exposure to examination content signal success:

… top sets provided more challenge through a faster pace and a greater emphasis on GCSE (even though these students were in Year 7 and Year 8, three or four years before the GCSE examination). (p70)

Taken together, the findings do not show that setting produces better outcomes, nor that mixed-attainment grouping is ineffective when well enacted. They show that structural reform without ideological challenge reproduces existing hierarchies. In a context of intense accountability, limited professional development and policy borrowing, pedagogical practices converge.

Over‑interpreting the study results risks reinforcing practices with well‑documented equity costs, while obscuring the conditions under which mixed-attainment teaching might succeed. The real question is not whether setting outperforms poorly implemented mixed-attainment teaching, but whether this study provides credible evidence that setting is a superior policy. A close reading of the report suggests it does not. Rather than justifying an expansion of setting, the findings invite a harder question: what conditions are necessary for equitable, high‑quality mathematics teaching, whatever the grouping, and why are these conditions so rarely met? Reframing the debate in these terms would better serve both research integrity and educational justice.