Costs of resistance are often invoked to explain the maintenance of polymorphisms for resistance to fungal pathogens in natural plant populations. To investigate such costs, 27 half-sib families of Silene alba, collected from a single host population, were grown in experimental populations in the presence and absence of the anther-smut fungus Ustilago violacea, a host-sterilizing pathogen transmitted by insects that are both pollinators and vectors of the disease. Host families differed significantly in resistance to inoculation, indicating the presence of genetic variation for mechanisms that impede fungal growth once the disease is encountered ("biochemical" resistance) within the host population. In addition, host families differed significantly in onset of flowering and in flower production in the absence of the disease. Path analysis revealed that late onset of flowering in male host families made a direct contribution to high field resistance (P < 0.01), probably due to a reduced rate of contact between hosts and vectors carrying high spore loads (avoidance, or "phenological" resistance). The contribution of low flower production to field resistance only approached significance (P < 0.10). There was a significantly positive genetic association between biochemical and phenological resistance, suggesting that delayed flowering is either a pleiotropic effect of biochemical resistance, or that genes governing these traits are in linkage disequilibrium. Path analysis revealed that biochemical resistance made both a direct contribution to field resistance (P < 0.01) and a positive indirect contribution via its association with phenology and flower production (P < 0.05) in male hosts. Costs of resistance were sex specific. Male host families with high field resistance had significantly lower reproductive success in healthy populations, indicating a fitness cost of field resistance (P < 0.01), whereas no costs were detected for female hosts. Path analysis revealed that the biochemical component of field resistance made no direct contribution to the observed fitness cost in male hosts, whereas its indirect effect through phenology was only marginally significant (P < 0.10). This finding indicates that fitness costs were mainly due to the phenological component of field resistance. Because the host population had no known history of disease, it is not clear whether the fitness costs are responsible for maintenance of the resistance polymorphism or whether the polymorphism is present for reasons unrelated to pathogen infection. Interactions between host families and pathogen strains with respect to inoculation success were not significant. Hence, there was no evidence for indirect costs of biochemical resistance, that is, reduced resistance to alternative strains. Infection rates in experimental populations with an initially patchy distribution of the pathogen were lower than in populations with a uniform pathogen distribution, suggesting that the effective pathogen pressure and hence the relative success of susceptible and resistant individuals may, in addition to fitness costs of resistance, depend on the spatial population structure of the pathogen.