Grain yield in any given environment is determined by yield components developed at different phenophases. Yield components are influenced by the environmental conditions the plant experiences during the respective phases. The final yield of a given cultivar depends on the interaction between genotype and its responses to environmental conditions. Hence, it is necessary to evaluate the plasticity in yield components formation while selecting genotype for a given environment. For this, we conducted field trials comprising 10 upland rice genotypes representing a large share of genetic variation, with two sowing dates in two consecutive years in three altitudinal locations creating 12 environments in Madagascar. Crop duration, grain yield and yield components (tillers per hill, panicles per tiller, grains per panicle, sterility, grain weight) were strongly affected by sowing dates, location, year and genotypes. Sowing date and years resulted in comparatively more variable environments in high and low altitudes than in mid altitude. Yield stability across environments reflected the target environments the genotypes were originally selected for. Variation in grain yield among planting dates within altitudes was not mainly due to temperature but rather to the combinations of abiotic factors the genotypes experienced during the different phenological stages during which the different yield components were formed. Yield components and their contribution to environmentally induced yield penalties were analyzed in detail. The contribution of individual yield components to final yield changed with the environmental conditions the rice experienced during the development stages. This effect may have a stronger influence on final yield than the genetic control of the individual yield components. New combinations of traits are required to better exploit the environmental potential which may only be possible via advanced crop models simulating the environmental effects on yield com