费曼学习法对学生学习效果的影响：一项包含41项实证研究的元分析

The Impacts of the Feynman Learning Method on Students’ Learning Outcomes: A Meta-Analysis of 41 Empirical Studies

摘要: 近年来，费曼学习法作为以输出驱动为核心的“以教代学”学习策略，被广泛应用于各阶段教学实践，但现有实证研究对其学习增益效果尚未形成统一结论。为系统厘清费曼学习法对学生学习效果的整体作用及适用边界，本研究整合国内外41项对照实验开展元分析，以学段、学习组织形式、学科领域、知识类型四大维度构建亚组分析框架，探究该方法效应的调节机制。研究共纳入44个独立效应量，总样本量4071人；发表偏倚检验结果显示文献无显著发表偏差。随机效应模型分析表明，费曼学习法对学生学习效果存在中等程度正向促进效应(g = 0.468, 95% CI [0.379, 0.556], P < 0.001)。亚组分析结果显示：费曼学习法在K12与在职教育阶段、自主学习模式、非STEM学科、陈述性知识教学场景下，效应更显著且稳定；大学阶段、合作学习模式、STEM学科、程序性知识场景下虽仍存在正向效果，但研究间异质性更高。研究证实费曼学习法具备可靠的教学推广价值，但不存在普适性，其实施效果受多重情境因素约束。据此建议教师结合学生认知发展阶段、学习组织形式、学科与知识类型差异化设计教学方案，精准发挥费曼学习法的育人优势。

Abstract: In recent years, as an output-driven learning strategy centered on “learning by teaching”, the Feynman Learning Method has been widely adopted in teaching practices across various educational stages. However, existing empirical studies have not reached a consistent conclusion regarding its promoting effect on learning performance. To systematically clarify the overall effect of the Feynman Learning Method on students’ learning outcomes as well as its applicable boundary conditions, this paper conducts a meta-analysis synthesizing 41 controlled experiments worldwide. A subgroup analysis framework covering four dimensions (educational stage, learning organization form, disciplinary field and knowledge type) is constructed to explore the moderating mechanism of the method’s effect. A total of 44 independent effect sizes involving 4071 participants are included in this research. Publication bias tests via funnel plot and Egger’s regression indicate no significant publication bias among the selected literature. Results from the random-effects model reveal that the Feynman Learning Method exerts a moderate positive effect on students’ learning outcomes (g = 0.468, 95% CI [0.379, 0.556], P < 0.001). Subgroup analyses demonstrate that the method yields more significant and stable effects in K12 and on-the-job education, self-directed learning settings, non-STEM disciplines, and declarative knowledge teaching. Although positive effects still exist for university education, collaborative learning contexts, STEM disciplines and procedural knowledge learning, substantial heterogeneity is observed among relevant studies. This study verifies that the Feynman Learning Method possesses reliable promotion value in teaching, yet it is not universally applicable, and its effectiveness is constrained by multiple contextual factors. Accordingly, teachers should design differentiated teaching plans by taking students’ cognitive development levels, learning organization form, disciplines and knowledge types into full consideration, so as to maximize the educational advantages of the Feynman Learning Method.