In the Sahel region where it is largely grown, pearl millet is a subsistence crop with an important role in food security. In this geographical area, yields are considered low. Breeding programs aim to release high-yielding varieties supposed to meet local environmental constraints. Certified varieties are developed and tested under optimal growth conditions and their performances are usually not as good as expected in farmer's field. In the groundnut basin in Senegal, farmers usually multiply and use their seeds from one year to another. Such local varieties are known to be well adapted to the farm conditions where they have been selected. Based on these observations we aimed to evaluate i) agronomical performances of local varieties under on-farm conditions comparing them to a reference variety from breeding programs, and ii) whether performance depends on pedoclimatic constraints. To this end we carried out on-farm trials in two locations contrasting for rainfall (550mm or 800mm). In each location, 10 farmers sowed both their own selected pearl millet variety and the certified variety recommended by agronomical institutions. Farmers were asked to follow their cultural itinerary for both cultivars. We measured stover and grain yield in the different varieties, and we performed soil analysis to compare field's fertility before sowing and after harvesting. Results showed a significantly higher grain and stover yield for local varieties than for certified varieties in both locations. One location was characterized by a significantly lower rainfall and lower values for 6 of the 7 soil nutrients tested, while achieving overall higher grain and stover yields than in the other location. Reasons for higher performances are still to be explored but we hypothesize that varieties multiplied yearly on-site are more likely to adapt to their environmental constraints than certified varieties developed under optimal controlled growth conditions. Our results raise questions about