According to current evolutionary dogma, multiple infections generally increase a parasite's virulence (i.e. reduce the host's reproductive success). The basic idea is that the competitive interactions among strains of parasites developing within a single host select individual parasites to exploit their host more rapidly than their competitors (thereby causing an increase in virulence) to ensure their transmission. Although experimental evidence is scarce, it often contradicts the theoretical expectation by suggesting that multiple infections lead to decreased virulence. Here, we present a theoretical model to explain this contradiction and show that the evolutionary outcome of multiple infections depends on the characteristics of the interaction between the host and its parasite. If we assume, as current models do, that parasites have only lethal effects on their host, multiple infections indeed increase virulence. By contrast, if parasites have sub–lethal effects on their host (such as reduced growth) and, in particular, if these effects feed back onto the parasites to reduce their rate of development, then multiplicity of infection generally leads to lower virulence.