We conducted a case-control study of mental retardation (MR) in which case children (aged 10 years) were identified from existing records at multiple sources, primarily the public school systems. Control children were drawn from a roster of public school students not receiving special education services. We found that maternal educational level at the time of delivery was strongly and inversely related to a form of MR not accompanied by other serious neurologic conditions. For this isolated form of MR, maternal educational level was by far the most important predictor from among seven sociodemographic variables examined. There was a significant race-education interaction that indicated a steeper gradient in risk among white mothers than among black mothers. Relative to children of white mothers with 12 years of education, all children of black mothers, except those whose mothers had 16 or more years of education, were at increased risk. The results may be useful as a guide for selecting high-risk groups as candidates for early childhood intervention programs.