Traditionally, a high rate of post mortem (PM) examinations has been advocated, primarily for the quality control of clinical diagnosis. In days of increasing scrutiny of medical costs, the role of the 'routine' requesting of any investigation must be questioned, and we report here the results of a policy to request post mortem examinations selectively when there was a reasonable expectation that it would yield new information. Between 21/3/82 and 12/2/86 there were 1356 deaths in patients registered with the Department of Radiation Oncology, Westmead Hospital. During this period, 100 PM examinations were performed. The overall post mortem rate is thus 6.7%, but for patients dying whilst under our direct care it is 23%. The cause of death was changed in 9% of cases following PM examination. The ante mortem assessment of the remission status of irradiated volumes was confirmed in 69% but was not recorded in the post mortem report in 22%, in spite of the medical record being available to the pathologist. Minor modifications were made by PM examination to the establishment of primary tumor site, sites of metastatic disease, histological diagnosis and other significant pathological states. The autopsy is an expensive investigation: we conclude that the low yield of relevant new information in this selective series is a powerful argument against the traditional routine request. Indeed the yield of new information in cases when the histological tumour type was known was even lower. The necessity for a problem-orientated approach is apparent from the data on the remission status of treated sites.