We used cross-sectional data on 2,660 black and 2,611 Mexican-American adult participants in the National Health and Nutrition Examination Survey (1999-2006) to investigate the association between metropolitan-level racial/ethnic residential segregation and obesity and to determine whether it was mediated by the neighborhood socioeconomic environment. Residential segregation was measured using the black and Hispanic isolation indices. Neighborhood poverty and negative income incongruity were assessed as mediators. Multilevel Poisson regression with robust variance estimates was used to estimate prevalence ratios. There was no relationship between segregation and obesity among men. Among black women, in age-, nativity-, and metropolitan demographic-adjusted models, high segregation was associated with a 1.29 (95% confidence interval (CI): 1.00, 1.65) times higher obesity prevalence than was low segregation; medium segregation was associated with a 1.35 (95% CI: 1.07, 1.70) times higher obesity prevalence. Mexican-American women living in high versus low segregation areas had a significantly lower obesity prevalence (prevalence ratio, 0.54; 95% CI: 0.33, 0.90), but there was no difference between those living in medium versus low segregation areas. These associations were not mediated by neighborhood poverty or negative income incongruity. These findings suggest variability in the interrelationships between residential segregation and obesity for black and Mexican-American women.