On the 1.7.1990 a new law was introduced in Denmark whereby specific consent had to be obtained from the deceased's relatives before autopsy could be performed. Autopsy records from two periods before and after this date, namely from 1.7.1980 to 1.7.1981 and from 1.7.1990 to 30.6.1991, are examined here with respect to autopsy rate and agreement/disagreement between clinical assessment as to cause of death and findings on autopsy. With an autopsy rate of 82.7% before and 10.3% after the new law we found that there was significantly more disagreement between the clinical assessment and the autopsy findings in the 1990/1991 period, presumably because of a selection bias towards difficult cases. The autopsy material from 1990/1991 could only be used for statistics in a limited fashion. A very low autopsy rate of about 10% implies that 25-30% of all death certificates from hospitals must be assumed to be incorrect as concerning serious diseases. The study shows that there is a considerable need for the performance of autopsies for the sake of medical statistics.