The learning results obtained by 4.292 Elementary School students, 32% in multigrade and 68% in unigrade, are compared, considering subjects, classroom size and cycles. It is a fi eld, correlational, explanatory and cross-sectional study that uses ANOVA and Multiple Linear Regression Analysis for the analysis of variance. The results indicate that multigrade groupings, in contexts similar to unigrade ones, can enhance learning outcomes in the second cycle of Elementary School and especially in mathematics. The size of the classrooms seems to have some infl uence; but it does not affect all groups or subjects in a similar way. [ABSTRACT FROM AUTHOR]