Variation in infection rate arises from variation in host exposure and resistance to parasites both within and among populations. All things being equal, phenotypes that increase exposure risk should covary positively with infection among individuals. It might therefore be expected that populations with mean phenotypes that increase exposure might also have higher rates of infection. However, such positive covariance between exposure and infection at the population level might be undermined by other factors such as geographic variation in parasite abundance or host resistance, negating or reversing in between-population comparisons. We studied rates of infection of two parasites among 18 populations of threespine stickleback (Gasterosteus aculeatus). As predicted, within populations, trophic morphology covaries with infection of two trophically transmitted parasites: individuals with benthic (or limnetic) phenotypes were more likely to be infected with a benthic (or limnetic) parasite. However, across populations, the relationship between morphology and infection rate was absent (limnetic parasite) or reversed (benthic parasite). Our results confirm the importance of phenotype-dependent exposure, but stress different factors or processes, such as the evolution of reduced susceptibility, might shape variation in infection at larger spatial scales.