The association between body mass index (BMI) and mortality was investigated in 2453 black male (aged 30-79 years) and 2731 black female (aged 40-79 years) members of the Kaiser Foundation Health Plan. During a 15-year follow-up 393 male and 283 female deaths were identified. Analyses were conducted separately in a lower and an upper range of BMI (as well as over the entire range), to isolate separate effects of low weight and high weight on mortality. Particular attention was also paid to potential bias from cigarette smoking and antecedent illness. Cox regression analyses showed that over the entire range of BMI the adjusted BMI-mortality association was significantly J-shaped for the men and essentially flat for the women. The inverse association between BMI and mortality in the lower range of BMI was statistically significant for the men; the adjusted relative hazard increasing from the 10th to the 50th percentile of BMI was 0.76 (95% confidence interval [CI] 0.59-0.98). The positive association between BMI and mortality in the upper range of BMI was highly statistically significant for the men; the adjusted relative hazard increasing from the 50th to the 90th percentile of BMI was 1.37 (95% CI 1.14-1.63). Whether controlled by multivariate analysis, by excluding the first 5 years of follow-up from the analyses, or by analyzing the BMI-mortality association in smoking-specific and/or illness-specific subgroups, smoking and antecedent illness did not have much impact on the BMI-mortality association, in either sex. The general observations on the BMI-mortality association are similar to findings in some white cohorts.