Beyond the average: a systematic review and meta-analysis of differentiated instruction for diverse English learners

Across 45 studies contributing 56 effect sizes in 18 countries, differentiated instruction helps diverse English learners, on average. But the same lesson can transform one classroom and barely move the next. This systematic review and meta-analysis looks past that average to ask how much the benefit of differentiated instruction depends on where, how, and with whom it is used. Following PRISMA 2020 guidelines, we searched six databases, identified additional studies through citation tracking and expert advice, and analyzed 56 effect sizes from controlled studies conducted between 2011 and 2024 and involving 16,773 learners. For each study, we calculated a standardized effect size (Hedges’ g ) from group means and standard deviations, and combined the estimates using a random-effects model in Jamovi. Differentiated instruction produced a medium-to-large average improvement in cognitive academic language outcomes [Hedges’ g = 0.78, 95% CI (0.557, 1.003)]. This pooled mean, however, masked substantial between-study heterogeneity ( I ² = 97.75%): true effects varied so widely that the 95% prediction interval (−0.84 to 2.40) included the possibility of no benefit, indicating that the effect is conditional rather than uniform. This pooled mean should therefore not be read as a reliable expected effect for any new classroom, nor used on its own as a policy benchmark; the conditional pattern of effects, not the single average, is the primary result. Three conditions were each associated with a three- to fourfold difference in the effect: when the study was conducted, whether English was taught as a foreign or a second language, and the learners’ educational level. A close look at the 13 exemplar studies with very high effects ( g ≥ 1.20) revealed eight shared practices, which we organize into a simple, usable evidence-derived framework (SAM-DI). The key point is not that differentiation works or fails, but that its benefits are conditional: strong where it is used strategically and for the learners who need it most, and weak or absent elsewhere. Interpreted cautiously about the direction of causation, reading past the average gives teachers and policymakers a clearer, fairer guide to where differentiation is worth the effort for diverse English learners.