The mortality experienced by a cohort of 36,691 rubber workers during 1946-85 has been investigated. These workers were all male operatives first employed in any one of the 13 participating factories in 1946-60; all had worked continuously in the industry for a minimum period of one year. Compared with the general population, statistically significant excesses relating to cancer mortality were found for cancer of the pharynx (E = 20.2, O = 30, SMR = 149), oesophagus (E = 87.6, O = 107, SMR = 122), stomach (E = 316.5, O = 359, SMR = 113), lung (E = 1219.2, O = 1592, SMR = 131), and all neoplasms (E = 2965.6, O = 3344, SMR = 113). Statistically significant deficits were found for cancer of the prostate. (E = 128.2, O = 91, SMR = 71), testis (E = 11.0, O = 4, SMR = 36), and Hodgkin's disease (E = 26.9, O = 16, SMR = 59). Involvement of occupational exposures was assessed by the method of regression models and life tables (RMLT). This method was used to compare the duration of employment in the industry, the duration in "dust exposed" jobs, and the duration in "fume and/or solvent exposed" jobs of those dying from causes of interest with those of all matching survivors. Positive associations (approaching formal levels of statistical significance) were found only for cancers of the stomach and the lung. The results of the RMLT analysis are independent of those from the SMR analysis, and the study continues to provide limited evidence of a causal association between the risks of stomach cancer and dust exposures, and the risks of lung cancer and fume or solvent exposures in the rubber industry during the period under study.