This study explores the relationships between patterns of grouping students for mathematics instruction and students' achievement and perception of learning environment. One hundred ninth-grade minority students were involved in the study. These students' scores on the Numerical Ability Subtest of the Differential Aptitude Test were used to group students into high ability, low ability, and heterogeneous groups. Scores on the Fundamental Interpersonal Relations Orientation Inventory were used to divide each of these groups into compatible and incompatible classes; class sizes did not exceed 17. The Metropolitan Achievement Test was given as a pretest, and four months later as a posttest. The Classroom Learning Environment Inventory was also used as a posttest. Using class group as the unit of analysis, data were submitted to analyses of variance and covariance. Findings indicated that grouping by ability and compatibility may be necessary but not sufficient for maximizing achievement. Groups differed in their perception of the learning environment. (SD)