This study examined whether the relationship between high poverty and infant mortality rates (IMRs) varied across race- and ethnic-specific populations in large urban areas. Data were drawn from 1990 Census and 1992-1994 Vital Statistics for selected U.S. metropolitan areas. High-poverty areas were defined as neighborhoods in which > or = 40% of the families had incomes below the federal poverty threshold. Bivariate models showed that high poverty was a significant predictor of IMR for each group; however, multivariate analyses demonstrate that maternal health and regional factors explained most of the variance in the group-specific models of IMR. Additional analysis revealed that high poverty was significantly associated with minority-white IMR disparities, and country of origin is an important consideration for ethnic birth outcomes. Findings from this study provide a glimpse into the complexity associated with infant mortality in metropolitan areas because they suggest that the factors associated with infant mortality in urban areas vary by race and ethnicity.