Although disease-resistance polymorphisms are common in natural plant populations, the mechanisms responsible for this variation are not well understood. Theoretical models predict that balancing selection can maintain polymorphism within a population if the fitness effects of a resistance allele vary from a net cost to a net benefit, depending upon the extent of pathogen damage. However, there have been a few attempts to determine how commonly this mechanism operates in natural plant–pathogen interactions. Ipomoea purpurea populations are often polymorphic for resistance and susceptibility alleles at a locus that influences resistance to the fungal pathogen, Coleosporium ipomoeae. We measured the fitness effects of resistance over three consecutive years at natural and manipulated levels of damage to characterize the type of selection acting on this locus. Costs of resistance varied in magnitude from undetectable to 15.5%, whereas benefits of resistance sometimes equaled, but never exceeded, these costs. In the absence of net benefits of resistance at natural or elevated levels of disease, we conclude that selection within individual populations of I. purpurea probably does not account completely for maintenance of this polymorphism. Rather, the persistence of this polymorphism is probably best explained by a combination of variable selection and meta-population processes.