effect size

Content and Language Integrated Learning (CLIL) has recently been the focus of numerous studies in language education since it aims to overcome the pitfalls of form-focused and meaning-focused instruction by systematically integrating content and language. This meta-analysis aims to synthesize the findings of 22 primary studies that tested the effect of CLIL on language skills and components. Guiding the analysis are three questions: What is the overall combined effect of CLIL on language skills and components? How do moderators condition the effect of CLIL? To what extent the overall combined effect is conditioned by publication bias? The overall effect size was found to be g=0.81, which represents a medium effect size with respect to Plonsky and Oswald’s (2014) scale. The results of moderator analysis show that CLIL has the highest effect on students’ grammar and listening proficiency and in lower levels of education, especially in elementary schools. It also has the highest effect when combined with hotel management as the subject matter. Fail-safe N test of publication bias shows that the significant positive outcome of CLIL cannot be accounted for by publication bias. The findings have clear implications for practitioners, researchers and curriculum developers.

This meta-analysis was conducted to synthesize the effect of 22 primary studies which have been conducted to test the effect of the integrated skills approach (ISA) on language skills and components. Three questions guide this analyses: What is the overall effect of ISA on language skills and sub-skills? To what extent moderator variables such as learners level of education and proficiency modify the effect of the ISA? What is the magnitude of publication bias in this analysis? The overall effect size was found to be 1.18, which represents a large effect size based on Cohen, Manion and Morrison’ (2007) scale. The results of moderator analysis revealed that the ISA has the largest effect for advanced learners and at a tertiary level of education. The symmetrical funnel plot together with fail-safe N test shows that publication bias does not have any significant effect on the effect size reported in this study. The findings of this meta-analysis have clear implications for practitioners, policymakers, and curriculum developers.