Description of the subject. For the ecological zone of the mid-western highlands of Madagascar, six upland rice varieties were selected by a research program. Objectives. The aim six as to determine their potential for extension in this area. Method. The six varieties were co-assessed for two years across a network of 20 reference farms. Results. The multilocation trials showed a diverse range of environments indicated by the average grain yield in rice, ranging between 1.04 and 4.50 t.ha(-1). FOFIFA 182 was significantly superior (2.90 t.ha(-1)) and SCRID 195 largely inferior (1.65 t.ha(-1)). In addition, there was a group of four undistinguished varieties (around 2.40 t.ha(-1)). The study of the genetic-environment interaction made it possible to differentiate a variety that produced a lower yield under low-fertility conditions (SCRID 90) and a less stable one, SCRID 91. SCRID 195, the lowest yielding variety, was also the most infected by Striga asiatica, a hemiparasitic plant. The farmers' assessment at the end of the 2-year period was added to the results of the research study on rice yield and Striga constraints, which were their two main criteria for variety performance. The farmers' assessment also reported on six other secondary criteria: threshing, earliness, drought resistance, grain quality, taste and grain weight. Conclusions. This research-farmer co-assessment made it possible to define the best-performing varieties in terms of yield and Striga control, namely NERICA 4 (already in extension) and FOFIFA 182 (proposed for extension). It also identified new varieties of interest, based on other criteria selected by the farmers and highlighted the need for future studies focused on a cropping system approach related to upland rice performance.