Explaining parasite virulence is a great challenge for evolutionary biology. Intuitively, parasites that depend on their hosts for their survival should be benign to their hosts, yet many parasites cause harm. One explanation for this is that within-host competition favors virulence, with more virulent strains having a competitive advantage in genetically diverse infections. This idea, which is well supported in theory, remains untested empirically. Here we provide evidence that within-host competition does indeed select for high parasite virulence. We examine the rodent malaria Plasmodium chabaudi in laboratory mice, a parasite-host system in which virulence can be easily monitored and competing strains quantified by using strain-specific real-time PCR. As predicted, we found a strong relationship between parasite virulence and competitive ability, so that more virulent strains have a competitive advantage in mixed-strain infections. In transmission experiments, we found that the strain composition of the parasite populations in mosquitoes was directly correlated with the composition of the blood-stage parasite population. Thus, the outcome of within-host competition determined relative transmission success. Our results imply that within-host competition is a major factor driving the evolution of virulence and can explain why many parasites harm their hosts.