Background: Despite a great deal of studies on blended language learning, synthesized empirical evidence of its effects on language performance is still lacking. Objectives: To bridge the gap, this study aimed to meta-analyse the effects of blended language learning on EFL learners' language performance. Methods: Drawing on the framework of AT (Activity Theory), this paper reported results based on a meta-analysis of the effects of blended language learning on EFL (English-as-a-foreign-language) learners' language performance with 43 valid effect sizes from 39 experimental and quasi-experimental studies published during 2000-2021. Results and Conclusions: The results showed that the overall effect size was significantly moderate for language performance, suggesting that blended language learning is more effective than traditional methods. The moderating effects of AT-related moderators were analysed. The educational levels, intervention durations, geographic regions and software types were found to be significant moderators, while others did not find a significant moderating effect. Implications: The results also provided some practical insights into the pedagogical implications for teachers, practitioners, and researchers in blended language learning.