The Effect of Drama-based Pedagogies on K-12 Literacy-Related Outcomes: A Meta-Analysis of 30 Years of Research

32 studies, 49 independent samples, 209 effect estimates; total N not reported

K-12 students in drama-based pedagogy interventions, 1985-2015

The most useful drama meta-analysis for this archive, not because of its headline number but because it publishes the moderator that dissolves it. Random-effects models in Comprehensive Meta-Analysis, trim-and-fill applied (no publication bias detected within any outcome grouping), five samples winsorized. Grade C: the pooled designs are quasi-experimental, and the paper reports d and p but NO confidence intervals anywhere in its tables, which is a real reporting gap for anyone trying to recompute.

Headline effects are large - achievement d = 0.64 (adjusted 0.68), 21st-century skills 0.52 (adjusted 0.70), motivation 0.60, attitudes 0.40 (adjusted 0.40 but non-significant), arts 0.34. Then the measure-type moderator (Q_b = 8.88, p < .05): NONSTANDARDIZED researcher-made tests give d = 0.89 (adjusted 1.27) while STANDARDIZED tests give d = 0.29 (adjusted 0.35). Observer ratings give 0.59, surveys 0.28. Proximity does the same work: direct outcomes 0.68 versus indirect 0.24. So does measurement volume: studies using a single assessment give d = 0.87, studies using more than ten give d = 0.04 (Q_b = 26.65). And so does dose, in the wrong direction: 3-10 hours gives 0.87 while more than 20 hours gives -0.01. By domain, reading 0.66, writing 0.71, mathematics 0.24, and self-concept 0.12 (ns). Leader matters: classroom teacher 0.89 > researcher 0.54 > teaching artist 0.31. The authors are candid that the field oversells itself - "Many advocates and researchers of drama-based pedagogies over-generalize or over-characterize its effects" - while defending the researcher-made measures as more relevant than "the austere, irrelevant nature of standardized tests," a defence the archive does not accept.

Unmanaged. The corpus is quasi-experimental; the operative biases here are measure proximity, single-outcome selection and unblinded observer rating rather than heredity.

Reproduces the direction of Podlozny (2000) and Lee et al. (2015) on an overlapping literature, and its standardized-test estimate of 0.29-0.35 is the number those earlier syntheses should be read down to. Contradicted where it matters most by the two randomised tests with standardised outcomes: the EEF Speech Bubbles trial (n = 821, null on reading and oral narrative) and Hautala et al. (2023), where an active oral-reading control BEAT the drama arm.