Blended instruction integrating off-line and on-line teaching has become an important instrument for promoting educational reform and innovation. However, the results of current empirical studies diverge on the effect of blended instruction on student performance, which necessitates further research on the effectiveness of blended instruction and related factors. This study, using an evidence-based meta-analytical approach, conducts a quantitative analysis of 106 experimental and quasi- experimental studies published from January 2000 to September 2021 in China and abroad, and systematically examines the effectiveness of blended instruction. The research finds that: i) The summary effect size (ES) of the included sample is 0.669 (n=142), indicating that blended instruction has above-moderate positive effects on student performance, especially on student learning motivation and academic emotions and attitude; ii) In terms of education levels, experimental periods and class sizes, blended instruction has the most significant positive effect on junior and senior secondary school students, on a teaching period from one to three months, and on a class size of 51 to 100 students; iii) Regarding the proportion and interactive patterns of online teaching, 50% composition of online teaching and synchronous or synchronous + asynchronous interaction exert the most significant positive effects on student learning. iv) Teaching methods including task-driven learning, role-playing, inquiry-based teaching, and case-based teaching have greater positive effects on student performance than other methods. Group study yields a greater effect on promoting student learning compared to individual study. Based on the findings, the present study also makes suggestions for the effective practice of blended instruction.