Summary Physiology and genetics of non-symbiotic N2-fixing bacteria have made much progress in recent years, especially in the case of a few reference strains. Nevertheless, understanding the ecology of diazotrophs cannot be achieved by studying only laboratory microorganisms. It is necessary to study naturally-occurring populations, to characterize their densities, size, composition, variability and variations in order to understand how a plant can select a rhizosphere population from a soil population. Very few comparisons of phenotypic diversity and dominant phenotypes in these two habitats have been made up to now. More studies of this type would allow a better knowledge of the selective pressures which actually drive the shift of population and they would permit investigation of the underlying mechanisms. These can vary from mere metabolic adaptation to selection of pre-adapted genotypes. A third mechanism is possible in which 'pre-adapted' genes are maintained in soil populations at very low frequencies and energy costs, and whose transfer is triggered by the selective factor to which they constitute an adaptation.