OBJECTIVE: To examine prospectively the association between age, BMI, and subsequent incidence of type 2 diabetes in Australian aboriginal people. RESEARCH DESIGN AND METHODS: We performed a stratified analysis of incidence data from a community-based longitudinal study. Measures included fasting and 2-h postload glucose concentrations, and BMI, stratified into four categories. Subjects were 882 male and female participants in diabetes screening initiatives in two remote Australian aboriginal communities, free from diabetes at baseline, ages 15-77 years. RESULTS: There were 46 incident cases of diabetes over 2,808 person-years of follow-up. BMI modified strongly the sex- and community-adjusted association between age and diabetes incidence (P < 0.001). Adjusted for age, sex, and community, the population diabetes incidence rate was 20.3 cases/1,000 person-years, with BMI-specific rates of 10.7-47.2 cases/1,000 person-years, and relative risks (95% CI) for BMI strata beyond the reference category (< 25 kg/m2) of 3.3 (1.5-7.0), 2.7 (1.1-6.8), and 4.4 (1.7-11.6), respectively. The population's attributable risk (95% CI) associated with BMI beyond the reference category was 70.1% (58.1-82.4). CONCLUSIONS: BMI-specific diabetes incidence rates in Australian aboriginal people are among the highest in the world. Diabetes incidence in the lowest BMI category (10.7 cases/1,000 person-years) is two to five times greater than corresponding rates for non-aboriginal populations. An urgent need exists to prevent weight gain associated with diabetes. Further study is required to determine for aboriginal people an optimal range of BMI, likely lower than that suggested for non-aboriginal populations.