Data on abortion in sub-Saharan Africa are rare and non-representative. This study presents a new method to collect quantitative data on clandestine abortion, the confidants method, applied in 2001 in Ouagadougou, Burkina Faso. Preliminary qualitative work showed that individuals are aware of their close friends' induced abortions: women usually talk to their peers about the unintended pregnancy and ask them for help in locating illegal abortion providers. In a survey of 963 women of reproductive age representative of the city of Ouagadougou, we asked respondents to list their close relations, and, for each of them, and for each of the 5 years preceding the survey, whether they had an induced abortion. According to these data, there are 40 induced abortions per 1000 women aged 15-49 in Ouagadougou annually, and 60 per 1000 women aged 15-19. Adverse health consequences followed 60% of the reported induced abortions, and 14% of them received treatment in a hospital. Extrapolating these results to the entire city, we estimate that its hospitals treat about 1000 cases of abortion complications a year. Hospital data indicate that these centers admitted 984 induced abortions (adding all "certainly", "probably" and "possibly" induced abortions in the WHO protocol) in 2001; the age distribution of patients admitted for induced abortion also corresponds to the confidants method's projections ("certainly" induced abortions only). At least two biases could affect the abortion rates estimated by the confidants method, pertaining to the selection of the sample of relations and to the varying number of third parties involved in the abortion process. The confidants method, which is similar in its principle to the sisterhood method used to estimate maternal mortality levels, might generate accurate estimates of illegal abortion in certain contexts if these two biases are controlled for. Further testing is necessary.