A retrospective chart review was conducted to determine the types and frequency of errors, other than those of diagnostic accuracy, made in recording information on death certificates and to assess factors that might affect those rates. The sample (n = 426) consisted of 50% of all deaths in a London, Ontario teaching hospital over one year. For each certificate reviewed, 6 questions were asked based on W.H.O. guidelines: 1) Was there an acceptable cause of death? 2) If mechanisms of death were recorded, were they adequately explained by an underlying cause of death? 3) Were there any sequencing errors? 4) Were there 2 competing causes of death listed? 5) Was there recorded any time interval between onset of the condition and death? 6) Was there any other inappropriate information recorded? The death certificates were filled out in an acceptable fashion 68.1% of the time. Comparing the 6 major departments in the hospital, there was significant difference in the error rates of the different departments (p = .0035). Error rates were not significantly better for certificates that had been signed by a coroner nor in those that had an autopsy performed. The majority of the death certificates (89.4%) were completed by house staff. More attention has to be devoted to raising physicians' awareness of the types of errors made in completing death certificates. Recurring educational sessions and feedback, if provided in teaching hospitals, could be helpful to increase the accuracy of these important documents.