OBJECTIVE: To evaluate the hypothesis that breast-feeding is associated with reduced type 2 diabetes among African-American, Hispanic, and non-Hispanic white youth, mediated in part by current weight status. RESEARCH DESIGN AND METHODS: The SEARCH Case-Control Study, an ancillary study to SEARCH for Diabetes in Youth, was conducted in two of six SEARCH clinical sites. Eighty youth with type 2 diabetes aged 10-21 years were included. Nondiabetic control participants were recruited from primary care provider offices (n = 167). Breast-feeding information was recalled by biological mothers. RESULTS: Prevalence (%) of breast-feeding (any duration) was lower among youth with type 2 diabetes than among control subjects (19.5 vs. 27.1 for African Americans, 50.0 vs. 83.8 for Hispanics, and 39.1 vs. 77.6 for non-Hispanic whites). The overall crude odds ratio for the association of breast-feeding (ever versus never) and type 2 diabetes was 0.26 (95% CI 0.15-0.46). Results were similar by race/ethnic group (P value for interaction = 0.17). The odds ratio for the association after adjusting for 12 potential confounders was 0.43 (0.19-0.99). When current BMI z-score was added to the model, the odds ratio was attenuated (0.82 [0.30-2.30]), suggesting possible mediation through current childhood weight status. Analyses that incorporated duration of breast-feeding, adjusted for potential confounders, provided evidence for dose response (test for trend, P value <0.0001), even after inclusion of BMI z-score. CONCLUSIONS: Breast-feeding appears to be protective against development of type 2 diabetes in youth, mediated in part by current weight status in childhood.