Abstract Theory on the evolution of niche width argues that resource heterogeneity selects for niche breadth. For parasites, this theory predicts that parasite populations will evolve, or maintain, broader host ranges when selected in genetically diverse host populations relative to homogeneous host populations. To test this prediction, we selected the bacterial parasite Serratia marcescens to kill Caenorhabditis elegans in populations that were genetically heterogeneous (50% mix of two experimental genotypes) or homogeneous (100% of either genotype). After 20 rounds of selection, we compared the host range of selected parasites by measuring parasite fitness (i.e. virulence, the selected fitness trait) on the two focal host genotypes and on a novel host genotype. As predicted, heterogeneous host populations selected for parasites with a broader host range: these parasite populations gained or maintained virulence on all host genotypes. This result contrasted with selection in homogeneous populations of one host genotype. Here, host range contracted, with parasite populations gaining virulence on the focal host genotype and losing virulence on the novel host genotype. This pattern was not, however, repeated with selection in homogeneous populations of the second host genotype: these parasite populations did not gain virulence on the focal host genotype, nor did they lose virulence on the novel host genotype. Our results indicate that host heterogeneity can maintain broader host ranges in parasite populations. Individual host genotypes, however, vary in the degree to which they select for specialization in parasite populations.