Ebert (1994) has proposed the rule that parasites are, with few exceptions, more infective to sympatric hosts than to allopatric hosts. We test this rule using field data for schistosome infections of planorbid snails and find that, although sympatric parasite-host combinations do tend to be more compatible, there are exceptions where particular allopatric parasite-host populations are significantly more compatible. We develop a mathematical model of the dynamics of the parasite-host interaction where parasite infectivity and host susceptibility are defined by the matching of genotypes in a diploid system. The model predicts dynamic polymorphisms where parasite allele frequencies track host allele frequencies but with lag. Because of this lag it is possible for allopatric combinations to be more compatible than sympatric combinations. Any `rule' that precludes this possibility is unlikely to prove robust.