To determine whether the carcinogenicity of dinitrotoluene (DNT) in rodent bioassays was predictive for humans, we examined the mortality experience of exposed workers at two ammunition plants. Cohorts of 156 and 301 men who had worked a month or more during the 1940s and 1950s at jobs with opportunity for substantial DNT exposure were followed through the end of 1980. Numbers of expected deaths and standardized mortality ratios (SMRs) were computed, using mortality rates of US white males as the standard. No evidence of a carcinogenic effect was found, but unsuspected excesses of mortality from ischemic heart disease were noted at both plants (SMRs 131 and 143; 95% confidence limits 65 to 234 and 107 to 187, respectively). Deaths from ischemic heart disease remained high even when compared with expected numbers derived using mortality rates of the counties in which the plants were located. Additional analyses revealed evidence of a 15-year latent period and suggested a relationship with duration and intensity of exposure. Epidemiologic investigations of other heavily exposed populations are needed to confirm the etiologic significance of the association between DNT and heart disease described here.