In 1994, a national hospital-based study was undertaken of cases of incomplete abortion presenting to public hospitals in South Africa. Data were collected for all women admitted to a random sample of hospitals with incomplete abortion during a two-week period. The WHO protocol for such studies was used as a basis for developing the methods to describe the epidemiology of incomplete abortion and hospital management of cases. Attempts were made to estimate the proportion of cases that might have been induced. This report focuses on methodological issues arising from the study that have implications for future research. The findings demonstrate that only a small proportion of the women acknowledged having had an induced abortion and that only a few of those who did showed evidence of interference with pregnancy. Clinical opinion of sepsis and the likelihood of induction were found to be highly unreliable. These findings considerably reduce the usefulness of the WHO-protocol method of estimating the likely origin of incomplete abortions. Results presented in terms of three partially overlapping descriptive categories are judged to better reflect the limitations of the data collected.