Recent reports have indicated a substantial increase in the prevalence of pediatric obesity. This increase is of significant concern given evidence that childhood obesity tends to track into adulthood and that adults who were obese as children experience increased morbidity and mortality independent of adult weight. The purpose of this study was to determine which variable tested, either alone or in combination, best predicts body mass index (BMI) in children of different races. The study population consisted of 612 students drawn from schools in a large Midwestern city and suburb. Students were given the FitnessgramR fitness test and completed a 46-item questionnaire designed to elicit responses regarding knowledge, attitudes, and beliefs about physical activity. 432 students (White, N=228; African American, N=107; Hispanic, N=97) completed all aspects of the study and were used in the final analysis. Separate Stepwise Multiple Regression Analyses were conducted for each race. The dependent variable was BMI and the independent variables were: activity levels, perceptions of fitness, self-efficacy, knowledge, barriers and social norms regarding exercise, and seven measures of health related fitness. Predictors that contributed no unique variance to the model were excluded. For African American students, knowledge, barriers, and the mile run made a significant contribution to the variance. The amount of the contribution varied from 16-31%, accounting for 69% of the total variance. For Hispanic students, social norms contributed 10% of the variance and barriers contributed an additional 29% while for White students, the mile run contributed 11% and knowledge contributed 13%. Health promotion programs that target increases in students’ knowledge of physical fitness, potential barriers to physical activity, and increases in actual physical fitness levels may help to reduce the increasing prevalence of childhood obesity. Keyword(s): diversity, elementary education, health promotion