A symposium on retrospective exposure assessment for epidemiologic studies of workplace hazards, jointly sponsored by NIOSH and the National Cancer Institute, was presented. Information ranged from philosophical issues in health effect studies to recommended improvements in quality of exposure data to be collected. Five major topic areas were addressed: selecting an approach to exposure assessment; methods of exposure assessment for industry based studies; methods of exposure assessment for community based studies; evaluating validity and precision of exposure estimates; and measures to facilitate future epidemiologic studies. For each area, the current state of practice was summarized, and important issues and new methods to address them in the future were identified. For the first topic, initial stages of a study were addressed, including disease characteristics, worker populations and exposures themselves. The second area considered use of exposure measurements in epidemiologic studies, role of biological monitoring in exposure assessment, derivation of exposure information from sources other than measured exposures, development of exposure based classification strategies for epidemiologic analysis, and reconstruction of historical exposures. The third area covered exposure determinations for subjects from a geographical area grouped on the basis of a common disease, rather than a common place of employment. Questionnaire based data and use of job exposure matrices were discussed. The fourth area addressed statistical evaluation of exposure assessment strategies, validation of work history based job information and effect of exposure misclassification in epidemiologic studies. Finally, summaries and commentaries were provided on past approaches and future aspects of exposure assessment.