Epidemiological studies of nuclear workers are an important source of direct information on the health effects of exposure to radiation at low doses and low dose rates. These studies have the important advantage of doses that have been measured objectively through the use of personal dosimeters. However, to make valid comparisons of worker-based estimates with those obtained from data on A-bomb survivors or persons exposed for medical reasons, attention must be given to potential biases and uncertainties in dose estimates. This paper discusses sources of error in worker dose estimates and describes efforts that have been made to quantify these errors. Of particular importance is the extensive study of errors in dosimetry that was conducted as part of a large collaborative study of nuclear workers in 15 countries being coordinated by the International Agency for Research on Cancer. The study, which focused on workers whose dose was primarily from penetrating gamma radiation in the range 100 keV to 3 MeV, included (1) obtaining information on dosimetry practices and radiation characteristics through the use of questionnaires; (2) two detailed studies of exposure conditions, one of nuclear power plants and the other of mixed activity facilities; and (3) a study of dosimeter response characteristics that included laboratory testing of 10 dosimeter designs commonly used historically. Based on these efforts, facility- and calendar year-specific adjustment factors have been developed, which will allow risks to be expressed as functions of organ doses with reasonable confidence.