Predicting the future of tundra plant communities is a major intellectual and practical challenge and it can only be successful if underpinned by an understanding of the evolutionary history and genetics of tundra plant species, their ecophysiology, and their responsiveness (both individually and as component parts of communities) to multiple environmental change drivers. This paper considers the types of experimental approaches that have been used to understand and to predict the future of tundra plant communities and ecosystems. In particular, the use of 'environmental manipulation' experiments in the field is described, and the merits and limitations of this type of approach are considered with specific reference to the International Tundra Experiment (ITEX) as an example to indicate the key principles. The approach is compared with palaeo-environmental investigations (using archives-or proxies-of past change) and the study of environmental gradients (so-called 'space-for-time substitution') to understand potential future change. Environmental manipulation experiments have limitations associated with, for example, short timescales, treatment artefacts, and trade-offs between technical sophistication and breadth of deployment in heterogeneous landscapes/regions. They do, however, provide valuable information on seasonal through decadal phenological, growth, reproductive, and ecosystem responses which have a direct bearing on ecosystem-atmosphere coupling, species interactions and, potentially, trophic cascades. Designed appropriately, they enable researchers to test specific hypotheses and to record the dynamics of ecosystem responses to change directly, thus providing a robust complement to palaeo-environmental investigations, gradient studies and ecosystem modelling.