It seems probable that Rhizobium was originally a free-living soil organism with the ability to produce nitrogenase, and this enzyme was the source of the nodular nitrogenase. Where the genetic information for leghaemoglobin formation came from is much more a matter for speculation. In the course of natural selection for competence to survive and multiply in soil as a saprophyte between legume-sheltered phases, nitrogen fixation probably became an ecological disadvantage to Rhizobium in its earlier form. Complete repression of the nitrogenase by a system which today is only derepressed by particular compounds in the legume is unlikely, mainly because constitutive mutants would be expected and have not apparently been found. It appears quite likely that part of the genetic information for the nitrogenase system became “banked”, as it were, with the plant. As a direct result of this transfer, the two partners became dependent on their symbiosis for the production of nitrogenase. The mechanism of the return to the bacteroid of the necessary information or its products can only be guessed at in the light of very meagre information. Mechanisms dependent on DNA transfer, either as a chromosome modification or as an episomal addition, are complicated because the added information must be eliminated from each cell which acquires it. On the other hand, transmission of mRNA to the bacterial ribosomes, or of a ready-made protein or part thereof, appears far more probable. Complementation between protein parts synthesized by the plant and by the bacterium could explain the degrees of effectiveness found in symbioses of one plant with several bacterial strains, or of one bacterial strain with several different plants. Many of the possible mechanisms considered here can be examined experimentally. The existence of episomal DNA should be readily detected, protein differences between Rhizobium from laboratory culture and its corresponding bacteroid forms can be explored, and the use of two bacterial strains with two legume hosts may indicate the controlling system for the various enzymatic components of nitrogenases.