A multilevel Bayesian meta-analysis of the effects of translanguaging pedagogy on L2 achievement

Abstract Globalisation and the multilingual turn in applied linguistics have exposed a persistent tension between learners’ linguistic repertoires and traditional monolingual pedagogies. It is out of this tension that translanguaging pedagogy (TP) emerged as an instructional framework grounded in translanguaging theory, witnessing a burgeoning empirical literature on its classroom value. However, the empirical evidence has failed to produce a coherent account of TP’s effectiveness in L2 education, as findings have exhibited considerable heterogeneity across instructional contexts, learner populations, and outcome measures. The present study sought to reconcile these inconsistent findings through a multilevel Bayesian meta-analysis of 108 effect sizes drawn from 40 independent studies among 3,145 learners. The pooled effect size of Hedges’ g = 1.014 (standard error = 0.138, 95% credible interval [0.748, 1.291]) establishes a positive and statistically credible effect of TP on L2 achievement. Moderator analyses identified educational level and target language skills as credible sources of heterogeneity, with tertiary-level learners and speaking outcomes yielding the largest effect sizes. Target language, linguistic distance, and treatment duration did not reach the posterior probability threshold for credible inference. These findings advance the theoretical understanding of TP and furnish an empirically grounded basis for instructional decision-making in L2 education.