A report on the pilot study of the Confidential Enquiry into Perioperative Deaths was published recently in the United Kingdom. The scheme was designed as a specialized form of clinical audit relating to perioperative patient care. Because the report placed little emphasis on autopsy, we have looked at all autopsied perioperative deaths, as defined by the enquiry, over a three-year period (1986-88) in a defined catchment area of Northern Ireland. Class I major discrepancies of diagnosis between clinical and postmortem records were found in 21% of 213 autopsied perioperative deaths. By definition, these represented instances in which an adverse impact on patient survival had resulted from the discrepancy. In an additional 29% of cases, there were Class II major discrepancies, which were discrepancies in the primary diagnosis not relevant to life-saving treatment. In 30% of cases there was a discrepancy in a secondary diagnosis, which might have affected the eventual prognosis, had the patient survived, but which was not related directly to the cause of death; these were termed Class III discrepancies. In 47% of cases, there were Class IV discrepant secondary diagnoses, which were 'incidental' findings and had no bearing on prognosis. To put these findings in context, the autopsy rate in Northern Ireland was 12% in 1987, with a higher rate (23%) in the two main teaching hospitals and a lower rate (8%) in all other hospitals. The rate of coroner's autopsies, 6%, is uniform throughout Northern Ireland. Coroner's autopsies are carried out mostly by the salaried staff of the State Pathologist's department. The observed differences in the rates of hospital autopsies reflect local deficiencies of pathologists in relation to the work load. In a series of consecutive autopsies carried out at one of the main teaching hospitals, the highest autopsy rate (68%) was found for paediatric patients, with rates of 29% for surgical and 23% for medical cases. The proportion of hospital autopsies carried out in perioperative surgical patients was the same as that for medical patients, reflecting the fact that no particular emphasis is placed on use of the autopsy as a form of clinical audit for the perioperative group.