Swedish twins have been followed for mortality since 1961, when the Swedish Twin Registry was formed. During the years 1961-73 there were 1290 deaths among twins born in 1901-25. In 1156 cases the cause of death could be established from collected records and classified according to the 1965 revision of ICD. Using the review of records as the standard, rates of detection and confirmation relating to the death certificate diagnoses were calculated. It is concluded that Swedish death certificate data are fairly valid for use in epidemiological studies and mortality statistics with regard to most cancer forms, cerebrovascular disease, ischemic heart disease, bronchitis, asthma and emphysema, accidents and suicides, but not for diabetes mellitus, alcoholism, mental diseases, rheumatic heart diseases and other heart diseases. However, in selected clinical-epidemiological studies it is often necessary to collect all available documents prior to judging the cause of death.