A cohort of 21,698 U.S. refinery and chemical plant workers was observed for eight years to determine if there were interplant or other variations in causes of mortality. Plant populations in three geographic locations were combined to develop an internal standard for comparing subgroups within the total cohort. At no one geographic site were consistently different rates for all major causes of death observed. The adjusted mortality rates for potentially exposed workers were slightly greater than those for nonexposed workers for most causes examined. Smokers incurred a higher risk of mortality from many causes of death when compared with nonsmokers, regardless of occupational category. After controlling for smoking, there remained a slight excess in mortality for potentially exposed as compared with nonexposed workers.