Spatially-variable processes can be an important element of host–parasite interactions, but their longer term demographic and evolutionary effects depend on the magnitude of variation in space, the scale at which variation occurs and the degree to which such processes are temporally stable. Here, we use multiple years of data from a study of two closely related tit species (Paridae), infected with two congeneric species of avian malaria (Plasmodium), to evaluate the roles of extrinsic and intrinsic factors in driving spatial heterogeneity in infection risk, and to address questions of scale and temporal stability in these vector-driven host–parasite interactions. We show that the two malaria parasite species exhibit markedly different spatial epidemiology: P. relictum infections are effectively randomly distributed in space, with no temporal consistency, whereas P. circumflexum infections exhibit pronounced spatial structuring that is stable over the six years of this study and similar in both host species. We show that both conspecific and heterospecific host density contribute to elevated infection risk, but that the main determinants of elevated risk of P. circumflexum infection risk are habitat features probably associated with vector distribution and abundance. We discuss the implications of these findings, both for our understanding of the epidemiology of malaria in the wild, but also in terms of the longer-term evolutionary and demographic consequences that spatially variable parasite-mediated selection may have on host populations.