Mutualisms are of central importance in biological systems. Despite growing attention in recent years, however, few conceptual themes have yet to be identified that span mutualisms differing in natural history. Here I examine the idea that the ecology and evolution of mutualisms are shaped by diverse costs, not only by the benefits they confer. This concept helps link mutualism to antagonisms such as herbivory, predation, and parasitism, interactions defined largely by the existence of costs. I first briefly review the range of costs associated with mutualisms, then describe how one cost, the consumption of seeds by pollinator offspring, was quantified for one fig/pollinator mutualism. I compare this cost to published values for other fig/pollinator mutualisms and for other kinds of pollinating seed parasite mutualisms, notably the yucca/yucca moth interaction. I then discuss four issues that fundamentally complicate comparative studies of the cost of mutualism: problems of knowing how to measure the magnitude of any one cost accurately; problems associated with using average estimates in the absence of data on sources of variation; complications arising from the complex correlates of costs, such as functional linkages between costs and benefits; and problems that arise from considering the cost of mutualism as a unilateral issue in what is fundamentally a reciprocal interaction. The rich diversity of as-yet unaddressed questions surrounding the costs of mutualism may best be investigated via detailed studies of individual interactions.