This study concerns the relation between mortality and the short-term outcome of inpatient treatment for alcoholism. A total of 121 patients (87 men, 34 women) of whom 89 were voluntary and 32 compulsorily committed. They had a mean age of 41 ±7 (sd) years and attended a 5-week programme at Runnagården, Örebro, Sweden. Most patients were socially unstable and severely alcohol dependent. Ten months (mean) after discharge, 96% of the patients and their referring social workers were contacted with mail questionnaires. Of these patients, 13% had been totally abstinent and a further 42% were improved but had had relapses. After a mean of 8.5 ±0.27 (sd) years, 27 patients (24%) had died. All abstainers survived, but non-abstainers had a 9 times higher mortality than expected. Non-abstinent improved women tended to survive longer than non-improved women, but among non-abstinent improved men no such tendency was found. In conclusion, a reduction in the frequency and quantity of abusive drinking was not enough to reduce the higher risk of death. Only abstinence seemed to be preventive.