Why do rhizobia expend resources on fixing N(2) for the benefit of their host plant, when they could use those resources for their own reproduction? We present a series of theoretical models which counter the hypotheses that N(2) fixation is favoured because it (i) increases the exudation of useful resources to related rhizobia in the nearby soil, or (ii) increases plant growth and therefore the resources available for rhizobia growth. Instead, we suggest that appreciable levels of N(2) fixation are only favoured when plants preferentially supply more resources to (or are less likely to senesce) nodules that are fixing more N(2) (termed plant sanctions). The implications for different agricultural practices and mutualism stability in general are discussed.