The Green Revolution boosted crop yields in developing nations by introducing dwarf genotypes of wheat and rice capable of responding to fertilisation without lodging. We now need a second Green Revolution, to improve the yield of crops grown in infertile soils by farmers with little access to fertiliser, who represent the majority of third-world farmers. Just as the Green Revolution was based on crops responsive to high soil fertility, the second Green Revolution will be based on crops tolerant of low soil fertility. Substantial genetic variation in the productivity of crops in infertile soil has been known for over a century. In recent years we have developed a better understanding of the traits responsible for this variation. Root architecture is critically important by determining soil exploration and therefore nutrient acquisition. Architectural traits under genetic control include basal-root gravitropism, adventitious-root formation and lateral branching. Architectural traits that enhance topsoil foraging are important for acquisition of phosphorus from infertile soils. Genetic variation in the length and density of root hairs is important for the acquisition of immobile nutrients such as phosphorus and potassium. Genetic variation in root cortical aerenchyma formation and secondary development (‘root etiolation’) are important in reducing the metabolic costs of root growth and soil exploration. Genetic variation in rhizosphere modification through the efflux of protons, organic acids and enzymes is important for the mobilisation of nutrients such as phosphorus and transition metals, and the avoidance of aluminum toxicity. Manipulation of ion transporters may be useful for improving the acquisition of nitrate and for enhancing salt tolerance. With the noteworthy exceptions of rhizosphere modification and ion transporters, most of these traits are under complex genetic control. Genetic variation in these traits is associated with substantial yield gains in low-fertility soils, as illustrated by the case of phosphorus efficiency in bean and soybean. In breeding crops for low-fertility soils, selection for specific root traits through direct phenotypic evaluation or molecular markers is likely to be more productive than conventional field screening. Crop genotypes with greater yield in infertile soils will substantially improve the productivity and sustainability of low-input agroecosystems, and in high-input agroecosystems will reduce the environmental impacts of intensive fertilisation. Although the development of crops with reduced fertiliser requirements has been successful in the few cases it has been attempted, the global scientific effort devoted to this enterprise is small, especially considering the magnitude of the humanitarian, environmental and economic benefits being forgone. Population growth, ongoing soil degradation and increasing costs of chemical fertiliser will make the second Green Revolution a priority for plant biology in the 21st century.