This study was designed to estimate the risk of developing diabetes in relation to adult height components, namely leg length and leg length/height ratio. Data on 12,800 individuals without diabetes were obtained at the baseline examination from the ARIC cohort. Cox proportional hazard models were used to estimate hazard rate ratios of diabetes for each 5-cm difference in leg length and 1 SD difference in the leg length/height ratio. During a mean follow-up period of 7.6 years, the age-adjusted incidence per 1,000 person years of follow-up was 25.8, 24.2, 10.4, and 16.2 in African American (AA) women, AA men, white women, and white men, respectively. The hazard ratio for diabetes (95% CI) per 5-cm difference in leg length was 0.85 (0.75–0.95) in white men, 0.79 (0.69–0.90) in white women, 0.90 (0.75–1.07) in AA women, and 0.99 (0.77–1.27) in AA men, after adjusting for age, parental history of diabetes, parental socioeconomic status, and weight at age 25. The hazard ratio for diabetes per 1 SD difference in leg length/height ratio followed the same trend. Leg length is inversely and independently related to an increased risk of diabetes in middle-age white men and women but not in African-Americans. This sex–race heterogeneity suggests that nutritional and environmental factors in childhood may modify this risk through different pathways.