How the relationship between species richness and productivity changes along environmental gradients remains poorly understood. We examined the context dependency of complementarity processes underpinning this relationship (biotic feedbacks, resource partitioning and facilitation) using the framework of Grime's (1973) humped-back model. We considered several scenarios of variation in competition and facilitation along environmental gradients, either monotonic with the most common or intense facilitation at the most abiotically severe end of gradients or nonlinear with the strongest facilitation at intermediate positions along gradients. How competition shifts to facilitation along environmental gradients is a key for determining where the effect of species richness on productivity occurs. Based on the literature, the original Stress Gradient Hypothesis would likely predict that complementarity effects should be the greatest, or the most important, in the most abiotically stressful environments. Alternatively, both the ‘collapse of facilitation’ and the ‘shift back to competition’ scenarios predict that the highest overall complementary effects on productivity, not biomass, would most likely occur at intermediate positions along environmental stress gradients, but this might vary depending on the source of stress. This latter prediction is consistent with a great deal of literature on natural gradients of productivity and species richness. Synthesis. Our predictions illustrate the importance of better understanding the context dependency of complementarity processes and the key role of facilitation along environmental gradients to better focus conservation efforts where ecosystem functioning is more likely to be negatively affected by species loss, in particular in species-rich communities. © 2021 British Ecological Society< Réduire