Résumé : Context and objectives: Since 1998, researchers in Burkina Faso have enrolled about 760 vulnerable women (sex workers, bar waitresses, pedlars, students, etc.) in an open cohort (the Yerelon cohort) to determine whether adequate management of sexually transmissible infection and exposure to well-designed, well-delivered, and plentiful communication for behaviour change (CBC) can reduce their vulnerability to HIV. This qualitative social anthropology study was conducted to: i) analyse the messages assimilated by the women who follow CBC sessions; ii) determine the impact of these messages on the women's behaviour. Methods: The researchers observed CBC sessions in the clinic, in women's homes and at their places of work for 45 days. They also conducted 37 individual in-depth interviews, 8 group interviews and a focus group, selecting a "reasoned" sample of 80 women from the cohort. The data were entered with Word software ana analyzed with the 'Atlas TI' qualitative data analysis software. Results: This study shows that: i) these vulnerable women have good knowledge of the themes raised during CBC sessions with the Yerelon peer-leaders and are quite able to reproduce the messages. Most can repeat and explain them; ii) the women find the messages presented during CBC, by peers who live similar lives and speak their "language", to be understandable and applicable to them; iii) they learned much about condom use and negotiation, HIV/AIDS, genital herpes, and other STIs. They know for example that condoms have two sides, one lubricated and the other not, and that condoms effectively protect against pregnancy and some STIs; iv) they negotiate condom use better with all types of sexual partners, thus improving their relation to it; v) a large majority report new positive behaviour, selecting clients and partners, and using condoms routinely, taking it off after intercourse themselves. They also now routinely use modern medical care for any STIs; vi) Of 524 women seronegative at inclusion in 2003, only 0.8% became infected in the first three months of the study, and no others in the year that followed. Conclusion: In Bobo-Dioulasso, Burkina Faso, the communication offered in the CBC programme to vulnerable women by peer-advisers trained by a multidisciplinary team, improved to varying degrees the women's knowledge of the various topics covered. This improvement changed their behaviour positively, leading to routine condom use, especially with occasional partners. This has resulted in low HIV incidence. This model can and must be spread to other vulnerable populations. Admittedly, the free services and products provided in the programme make it harder to perpetuate and disseminate. The promotion of health mutuals is one method to provide funding.