The purpose of this study was to check the reliability of the classification of death in cases regarded primarily as accidents or undetermined. The material used for the analysis consisted of the documents of autopsies performed at our department in 1981. There were altogether 283 cases of suicides, accidents or undetermined deaths, accounting for 8.4% of total mortality, and of these 180 cases were of the two last categories. In 125 cases primarily regarded as non-suicides (accidents or undetermined) there were some features of suicide leading to reclassification as probable suicide in 24 cases. Fifteen of these were primarily poisoning "accidents", mainly alcohol intoxications. The result indicates that the suicide rate would be 3.7% instead of 3.0% of the total mortality. It may thus be concluded that the official figure for suicides could be as much as 18.9% smaller than the "actual" figure.