A method for retrospectively assessing exposures in epidemiological studies where little or no data on past exposures were available was developed and was illustrated by applying it to a case/referent study of the cancer risk associated with spraying herbicides and manufacturing phenoxy herbicides and chlorophenols possibly contaminated with dioxins. The study group consisted of a cohort of 21,183 workers from 11 countries assembled by the International Agency for Research on Cancer. The exposure assessment consisted of estimating cumulative exposures by a team of industrial hygienists unaware of the case/referent status of the subjects. The first step in the assessment process involved determining in a qualitative manner whether a given subject had been exposed or not using work histories and company supplied information on processes and environmental conditions. If these analyses indicated that the subject was not exposed, a cumulative exposure score of zero was assigned. If exposure was indicated to have occurred, durations of exposure were estimated from the work histories and levels of exposure (LEs) were estimated by determining the product of weighting factors such as emission factors, contact factors, and personal protective factors. In estimating durations of exposure, the last 5 years before entry into the study were excluded from the analysis. The various weighting factors were estimated from industrial hygiene and monitoring data, if available, and company supplied data on production processes, products manufactured or used, and the availability and use of local ventilation and personal protective equipment as control measures. Cumulative exposure scores were calculated as the product of the duration of exposure and LE. Representative data indicated that 23% of the referents had been exposed to 2,4,5-trichlorophenoxyacetic-acid (93765) and 2,4,5- trichlorophenol (95954) for an average of 8.0 years. The LE was 1.6. The estimated cumulative exposure of these subjects was 13.0 LE years. The authors conclude that this approach can be applied to retrospectively evaluating exposures in industry wide studies where few exposure data are available. Although the results are likely to be less valid than assessments based on a comprehensive set of measurements, the approach is more valid and reliable than subjective assessments made by an expert.