AIMS: The aim of this study was to examine whether early retirement is gained by persons with poor health or whether retirement leads to negative changes in health. Mortality for persons in Denmark using one of two publicly financed retirement schemes was investigated. Disability benefit (førtidspension) was granted for health reasons mainly, while early retirement benefit (efterløn) was earned through long-term membership of an unemployment benefit scheme. METHODS: Data from Danish population-based registers were used for each year in the period 1986-96. The study setting was the population of Denmark, born between 1926 and 1936. RESULTS: the standardized mortality ratio in employed persons was low (0.59 and 0.51 for men and women, respectively) and high in disability benefit recipients (2.31 and 1.66). The mortality in the early retirement benefit recipients (0.88 and 0.72) was in between the mortality for the disability benefit recipients and the employed persons. Disability benefit recipients had a high relative risk of death immediately after retirement. In early retirement recipients the relative risk of death increased with time since retirement. CONCLUSIONS: Disability benefit was mainly gained by persons with poor health. In contrast, the increasing mortality of the early retirement recipients is consistent with an adverse effect on health of retirement itself, but may also be due to the cessation of health selection in the group of disability benefit recipients after retirement. It is difficult to disentangle the effects of a change in activity following retirement from the effects of an earlier health-associated selection into the two retirement schemes.