Résumé
Sorghum is the staple crop of Burkina Faso. It is grown in a range of cropping systems and with various production objectives. Until to date little success has been obtained by the national breeding program in diffusing modern varieties. There are various reasons for this, amongst others, that breeding has been too centralised with low representative growing conditions, and that farmers' needs and preferences have not been taken into sufficient consideration. The approach of an in-situ preservation and variety enhancement sorghum project recently initiated in Mali and Burkina Faso is to focus on local landrace-based population improvement through recurrent selection, participatory selection of segregating materials and early decentralised yield testing. The project involves national and international scientists, development organisations and farmer organisations. Farmer organisations and other local relevant actors are involved in all stages of the research activities. Our text will provide preliminary results of on-farm breeding work carried out in 2001. Two on-farm trials were conducted in different agro-ecological zones in Burkina Faso (villages of Some and Zikieme), each comprising 64 sorghum cultivars sown in a 8 x 8 lattice design with two replications. The cultivars were F3/F4 sorghum lines derived from: crosses between guinea and caudatum types, previously released varieties and local landrace varieties. Workshops for 20-40 farmers, 3 scientists (2 breeders, one agronomist) and a local NGO were organised at both experimental sites in order to discuss and to identify farmers' needs, as well as to rank farmers' criteria for selecting a new sorghum variety. In a second working session, the trials were evaluated by groups of 4-6 farmers (male and female, and from different areas). At the Some site, all 64 entries of the trial were scored by two male and two female groups as well as one senior sorghum breeder for the three most important criteria identified in the first working session. The farmer groups and the breeder next selected for the better appearing panicles in the best preferred lines. At the second site of Zikieme, farmer groups from different areas within the zone selected their preferred lines and panicles in both repetitions. Agronomic data was collected by technical staff. Selection, following the habitual procedure of pedigree selection, was also carried out in the same F3/F4 lines on-station by the national sorghum breeder. To assess these evaluation methods and to develop perspectives for participatory pedigree selection for the above mentioned project, the different sets of collected data were analysed and compared by using simple correlation, Kendall coefficient of concordance (multiple correlation) and comparisons of means. Some of these results were discussed and then compared with previous PVS results done in the same areas. Finally, the efficiency and quality of the participatory selection method is discussed and compared to the conventional method. (Texte intégral)