Rhizobial bacteria, endomycorrhizal fungi (also known as arbuscular mycorrhizas), and pseudomonad bacteria associated with plant roots can provide substantial benefits to the plants by fixing nitrogen, supplying phosphorus, or controlling root pathogens, respectively. A significant fraction of plant photosynthetic carbon may be used by these associated microorganisms, both to support their beneficial activities and for microbial growth and reproduction. Because many microbial individuals are associated with each individual plant, the individual benefit to a microbe that allocates more resources to its own reproduction (thereby allocating less to fixing N-2, supplying P, or producing antifungal metabolites) would exceed its individual loss from any resulting reduction in collective benefits (mainly plant carbon substrates). An initially rare "free rider" mutant strain might therefore be expected to displace its more cooperative parental strain. Yet, the mycorrhizal and legume-rhizobium mutualisms have persisted (often coexisting with "cheating") for millions of years. This paper discusses the importance of microbial cooperation (with plants and with other microbes) and possible reasons for its evolutionary persistence in the rhizosphere. In undisturbed soils, spatial structure can favor kin selection, but this may be counterbalanced by the increased likelihood that future competitors will be among the beneficiaries of current cooperation. In loose associations, direct fitness benefits to microorganisms may explain the evolutionary persistence of activities (e.g., production of antifungal compounds) that can benefit plants as a side effect. In closer, more symbiotic, relationships, host sanctions against individuals or clones that fail to perform their symbiotic function may be more important. New molecular methods and other research tools are facilitating research on this topic, and some of these conclusions soon may be revised.