Over the next few decades we must boost crop productivity if we are to feed a growing world population, which will reach more than 9×109 people by 2050; and we should do it in the frame of a sustainable agriculture, with an increasing scarcity of new arable land and of water for irrigation. For all important crops, average yields are only a fraction-somewhere between 20% and 50%-of record yields; these losses are mostly due to drought and high soil salinity, environmental conditions which will worsen in many regions because of global climate change. Therefore, the simplest way to increase agricultural productivity would be to improve the abiotic stress tolerance of crops. Considering the limitations of traditional plant breeding, the most promising strategy to achieve this goal will rely on the generation of transgenic plants expressing genes conferring tolerance. However, advances using this approach have been slow, since it requires a deep understanding of the mechanisms of plant stress tolerance, which are still largely unknown. Paradoxically, most studies on the responses of plants to abiotic stress have been performed using stress-sensitive species-such as Arabidopsis thaliana-although there are plants (halophytes, gypsophytes, xerophytes) adapted to extremely harsh environmental conditions in their natural habitats. We propose these wild stress-tolerant species as more suitable models to investigate these mechanisms, as well as a possible source of biotechnological tools (‘stress tolerance’ genes, stress-inducible promoters) for the genetic engineering of stress tolerance in crop plants.