Coroners' files have been examined to ascertain the numbers of deaths involving self-poisoning with analgesic drugs with specific reference to the paracetamol/dextropropoxyphene combination. The period of study was 1976-1980. This report concentrates on cases in England, although reference is made to similar deaths occurring in Scotland and Northern Ireland. Data have been collected extensively on a wide range of issues concerning fatal self-poisonings mainly by visiting coroners' offices in England to make direct investigation of records. The total number of cases where the paracetamol/dextropropoxyphene combination can be considered as ingested in the self-poisoning episode is underestimated. The number of cases involving alcohol and/or other drugs taken together with the combination product is particularly underestimated. Involvement of people aged 30 years and below comprises 32% of all cases. 'Gesture' overdoses comprise an estimated 14% of all cases in England. An analysis of fatalities from cases where quantities in post-mortem blood of dextropropoxyphene less than 1 microgram/ml and paracetamol less than 50 micrograms/ml are found and of cases where death occurs within 2 h of ingestion of the overdose has been carried out. The results are inconclusive. In most cases alcohol and/or other drugs are found to be involved. Critical inspection of coroners' files shows relatively few of these cases where the combination product is ingested on its own. There is wide variability in the data available in coroners' files. In many cases data of value to this research are not recorded. Medical history and quantitative levels of drugs suspected (particularly dextropropoxyphene) are particular examples of factors which may not be recorded. Office of Population, Censuses and Surveys (OPCS) mortality data are based on certified causes of death. Because of the underestimate of the involvement of this combination product and the under-reporting of other drug and/or alcohol ingestion with the combination, care must be exercised in quoting or drawing conclusions from OPCS statistics.