The association of obesity with mortality was investigated in population-based samples of Micronesian Nauruans (n = 1,400), Melanesian Fijians (n = 1,279), and Indian Fijians (n = 1,182), over 10 years from 1982 in Nauru, and 11 years from 1980 in Fiji. At the end of follow-up, vital status was known for all Nauruans and all but 3.5% of Fijians. Mortality rates were higher in Nauru than Fiji, and in Melanesians than Indians. The mean body mass index of decedents was similar to or less than (Nauruan men, p < 0.001) that of survivors in each sex-ethnic group. Crude mortality rates showed an inverse relation with body mass index in Nauruan men, with inconsistent relations in other sex-ethnic groups. After stratification by diabetes status, there was no relation between mortality and obesity in nondiabetic subjects, but an inverse relation was observed among diabetic subjects in each population. These findings persisted even after the exclusion of subjects who died within the first 2 years of follow-up. After controlling for age, smoking, and diabetes status in Cox proportional hazard models, body mass index (as a continuous variable) was not related to mortality in any sex/ethnic group and tended to be negatively associated with mortality risk. Interactions of body mass index with age, smoking, and diabetes status were not significant. Mortality risk was significantly increased in older subjects and in diabetic subjects, and cigarette smoking also increased risk in some groups. Stratification of analyses according to cigarette smoking did not alter the nature of the results. The association of mortality and body mass index categorized by quartiles was also investigated. After adjusting for age alone, or age, smoking, and diabetes status, the lower quartiles of body mass index were consistently associated with the highest relative risk for mortality. Quadratic terms for body mass index did not improve Cox models in subjects with normal glucose tolerance. Relations with cardiovascular disease mortality were also assessed and results were inconsistent, although positive trends were observed in Nauruan women (p = 0.02) and Melanesian men (p = 0.06). Overall, there was little evidence to suggest that obesity was a risk factor for total or cardiovascular mortality in these populations. However, obesity is clearly associated with a high risk of diabetes and other morbid conditions and at least on this basis it would seem desirable to prevent obesity in these and other Pacific populations.