The role of positive interactions has become widely accepted as a mechanism shaping community dynamics. Most empirical evidence comes from plant communities and sessile marine organisms. However, evidence for the relative role of positive interactions in organizing terrestrial animal communities is more limited, and a general framework that includes positive interactions among animals is lacking. The ‘stress gradient hypothesis’ (SGH) developed by plant ecologists predicts that the balance between positive and negative interactions will vary along gradients of biotic and abiotic stress, with positive interactions being more important in stressful environments. Paralleling the SGH, stress gradients for terrestrial herbivores could be equated to inverse primary productivity gradients, so we would expect positive interactions to prevail in more stressful, low productivity environments. However, this contradicts the typical view of terrestrial animal ecology that low primary productivity systems will foster intense competition for resources among consumers. Here we use alpine herbivores as a case study to test one of the predictions of the SGH in animal communities, namely the prevalence of positive interactions in low productivity environments. We identify potential mechanisms of facilitation and review the limited number of examples of interspecific interactions among alpine herbivores to assess the role of positive and negative interactions in structuring their communities. A meta-analysis showed no clear trend in the strength and direction of interactions among alpine herbivores. Although studies were biased towards reporting significant negative inter actions, we found no evidence of competition dominating in harsh environments. Thus, our results only partially support the SGH, but directly challenge the dominant view among animal ecologists. Clearly, a sound theoretical framework is needed to include competition, positive and neutral interactions as potential mechanisms determining the structure of animal communities under differing environmental conditions, and the stress-gradient hypothesis can provide a solid starting point.