Sociometric questionnaires were administered to all pupils in the fourth, fifth, and sixth grades in an elementary school in February 1970 (soon after forced integration had occurred), again in June 1970, and again to fifth and sixth grade pupils in February 1971. Both blacks and whites at all three grade levels chose significantly more members of their own race than members of the other race; blacks rejected whites significantly less than would be expected by chance, an effect which increased over time for older pupils; and, initially, whites rejected blacks less frequently than expected by chance, but shifted to approximately chance rejection of blacks. In January and February 1971, two observers counted interactions among pupils in each grade for ten minutes once per week for a period of five weeks. As expected, interactions increased significantly with grade level; however, subsequent analysis revealed that the primary variable determining interracial interactions is the proportion of blacks to whites in the classroom. A relatively low proportion of minority members (black or white) appears to be most effective in improving interracial relations.