Evolutionary models predict that parasite virulence (parasite-induced host mortality) can evolve as a consequence of natural selection operating on between-host parasite transmission. Two major assumptions are that virulence and transmission are genetically related and that the relative virulence and transmission of parasite genotypes remain similar across host genotypes. We conducted a cross-infection experiment using monarch butterflies and their protozoan parasites from two populations in eastern and western North America. We tested each of 10 host family lines against each of 18 parasite genotypes and measured virulence (host life span) and parasite transmission potential (spore load). Consistent with virulence evolution theory, we found a positive relationship between virulence and transmission across parasite genotypes. However, the absolute values of virulence and transmission differed among host family lines, as did the rank order of parasite clones along the virulence-transmission relationship. Population-level analyses showed that parasites from western North America caused higher infection levels and virulence, but there was no evidence of local adaptation of parasites on sympatric hosts. Collectively, our results suggest that host genotypes can affect the strength and direction of selection on virulence in natural populations, and that predicting virulence evolution may require building genotype-specific interactions into simpler trade-off models.