Meta-Analysis of the Contribution of Cooperative Learning Models to Science Cognitive Outcomes: Evidence Toward SDG 4 in Indonesian Elementary Education

Abstract

This study evaluated the cognitive learning outcomes of elementary and junior high school students in science subjects after implementing cooperative learning models, using a mean-score meta-analytic approach. A systematic search of Google Scholar, DOAJ, ERIC, Springer, Elsevier, and ResearchGate identified 40 relevant studies, of which 24 met the inclusion criteria and yielded 34 usable effect-size artifacts. Data were analyzed in JASP 0.8.4.0 using a random-effects model (Restricted Maximum Likelihood). For elementary school students, the pooled effect size was 71.86 (95% CI: 66.43–77.29); tested against the minimum completion criterion (KKM) of 75, this value was statistically significantly lower than the criterion (z = -2.12, p < .05), so H0 was rejected. For junior high school students, the pooled effect size was 76.74 (95% CI: 73.75–79.73), which did not differ significantly from the KKM (z = 1.41, p > .05), so H0 was retained. Egger's regression test showed no evidence of significant funnel-plot asymmetry at either level (elementary: z = -0.01, p = .989; junior high: z = -0.95, p = .345). These findings indicate that, on average, elementary students' science achievement after cooperative learning remained significantly below the minimum completion criterion, while junior high students' achievement was statistically indistinguishable from it. This pattern suggests that implementation fidelity and students' developmental readiness for group-based learning likely moderate the effectiveness of cooperative learning across school levels.