Cancer mortality trends from 1953 to 1991 were assessed by means of data supplied by the Department of Vital Statistics of the Ministry of Public Health. The population at risk was obtained from the Bureau of Statistics and Censuses. Age-specific and age-adjusted mortality rates were calculated, using the world standard population, for a number of sites or groups of sites. In order to obtain relative risks of death for each period, Poisson regression models were fitted to the data using the GLIM program. The main model included age and period as explanatory variables. Among males, the principal increase was observed for lung cancer, followed by prostatic cancer. The rates were mainly stable in colon cancer and leukaemias, whereas gastric cancer showed a marked decline. Also, a recent decline was seen for oesophageal cancer. In females a steady decline in mortality was observed for all sites combined. Major decreases were seen for oesophageal, gastric, cervical and total uterine cancers. The only cancers showing significant increases were breast cancer, and lung cancer for the most recent period. Providing that there were no changes in death registration or in survival rates, changes in prevalence of risk factors might be responsible for the observed trends.