Abstract. There is a large body of theoretical studies that investigate factors that affect the evolution of virulence, that is parasite-induced host mortality. In these studies the host population is assumed to be genetically homogeneous. However, many parasites have a broad range of host types they infect, and trade-offs between the parasite virulence in different host types may exist. The aim of this paper is to study the effect of host heterogeneity on the evolution of parasite virulence. By analyzing a simple model that describes the replication of different parasite strains in a population of two different host types, we determine the optimal level of virulence in both host types and find the conditions under which strains that specialize in one host type dominate the parasite population. Furthermore, we show that intrahost evolution of the parasite during an infection may lead to stable polymorphisms and could introduce evolutionary branching in the parasite population.Key words. Evolution of virulence, genetic trade-offs, host heterogeneity, host-parasite coevolution, intrahost evolution, serial passage.Received December 31, 1998. Accepted July 16, 1999 The relationship between a parasite and its host is a story of benefits and harms. The parasite benefits from the host by living in and on it and by using host resources to reproduce. The parasite's benefit gives rise to the host's harm: The life span of the host is usually shortened by infection, and important fitness traits of the host, such as fecundity, are often negatively affected by the parasite (Ewald 1994). However, by reducing the life span or the fitness of its host, the parasite may inflict harm upon itself. The question emerges: What evolutionary forces determine the level of harm inflicted on the host? In this paper, we focus on the evolution of parasite virulence, here defined as parasite-induced host mortality.If virulence were an independent trait, natural selection should favor parasites with low virulence. A parasite that does not kill its host has more time to exploit it and to be transmitted, thus increasing its own fitness. Therefore, in the long run the parasite should evolve to be avirulent. Nevertheless, there are many examples of intermediately or highly virulent host-parasite systems with a long coevolutionary history (Fenner and Ratcliffe 1965;Herre 1993). To account for virulent host-parasite interactions, trade-offs between virulence and other parasite traits such as infectivity, transmissibility, or reproduction rate were postulated , 1982, 1991May and Anderson 1979, 1983, 1990Levin and Pimentel 1981;Bremermann and Pickering 1983;Ewald 1983;Knolle 1989; Frank 1992;Antia et al. 1994). Furthermore, it was argued that in systems in which the parasite can super-or coinfect hosts or in which the parasites frequently generate new mutant strains within a host, a parasite strain with high virulence may have a competitive advantage over less virulent strains within the host (Hamilton 1972;Bremermann and Pickering 1983;Knolle 1989;Sas...