Most models for coevolution of hosts and parasites are based on the assumption that resistance of hosts to parasites is an all–or–nothing effect. In many cases, for example where parasites require an appropriate receptor on host cells, this is a reasonable assumption. However, in many other cases, for example where hosts mount an immune response, this picture may be too simple. An immune system is expensive to maintain, which poses a question as to how much of its resources a host should allocate to resist parasites: if the risk of infection is low, natural selection may favour hosts with less effective immune systems. As optimal allocation to defence will depend on the force of infection, and the force of infection, in turn, depends on the level of defence in the rest of the host population, a game–theoretic approach is necessary. Here I analyse a simple model for the evolution of the ability to recover from infection. If parasites are not allowed to coevolve, the outcome is a single evolutionarily stable strategy (ESS). If the parasites coevolve, multiple evolutionary outcomes are possible, one in which the parasites are relatively avirulent and common and the hosts invest little in recovery ability, and another (the escalated arms race) where parasites are rare but virulent and the hosts invest heavily in defence.