Résumé
The effectiveness of on-farm and/or on-station early generation yield testing was examined to maximize the genetic gains for sorghum yield under smallholder famer production conditions in West Africa. On-farm first-stage yield trials (augmented design, 150 genotypes with subsets of 50 genotypes tested per farmer) and second-stage yield trials (replicated a-lattice design, 21 test genotypes) were conducted, as well as on-station a-lattice first-and second-stage trials under contrasting phosphorous conditions. On-farm testing was effective, with yield showing significant genetic variance and acceptable heritabilities (0.56 in first-and 0.61 to 0.83 in second-stage trials). Predicted genetic gains from on-station yield trials were always less than from direct testing on-farm, although on-station trials under low-phosphorus and combined over multiple environments improved selection efficiencies. Modeling alternative designs for on-farm yield testing (augmented, farmer-as-incomplete-block, multiple lattice, and augmented p-rep) indicated that acceptable heritabilities (0.57 to 0.65) could be obtained with all designs for testing 150 progenies in 20 trials and 75 plots per farmer. Ease of implementation and risk of errors would thus be key criteria for choice of design. Integrating results from on-station and on-farm yield testing appeared beneficial as progenies selected both by on-farm and on-station first-stage trials showed higher on-farm yields in second-stage testing.