The purpose of this study was to investigate the effectiveness of the differentiated instruction in mathematics by a meta-analysis. Among the studies conducted for last twenty three years, the relevant 49 research articles were selected, 80 effect sizes were calculated for the cognitive domain and 70 effect sizes for the affective domain. Effect sizes were analyzed with school levels, student level, group organization method such as homogeneous vs. heterogeneous, class transfer and adjusting class sizes for each cognitive domain and affective domain. The results are as the followings: In the cognitive domain, the overall effect size of the differentiated instruction produced a medium effect(effect size=0.68, U3=75.17%). The differentiated instruction showed the highest effect size on elementary school and middle school, mid level students, heterogeneous group, class transfer and not adjusted class size. And in the affective domain, the overall effect size of the differentiated instruction produced a low effect(effect size=0.36, U3=65.36%). The differentiated instruction showed the highest effect size on elementary school, mid level students, and heterogeneous group. Thus the differentiated instruction was proved to be effective in mathematics classes.