Using a reciprocal cross-infection experiment, we detected clinal variation in the relative ability of trematodes to infect snails taken from the same habitat. Snails (Potamopyrgus antipodarum) collected from three habitats in a New Zealand lake were exposed to a digenetic trematode (Microphallus sp.) collected from infected snails from the same three locations. The habitats were arranged according to depth, from shallow (the shore bank habitat: less than 1 m deep), to intermediate (the Isoetes habitat: 1.5-3 m), to deep (the Elodea habitat: 4-6 m). Snails were assayed for infection at two points in the development of the parasite: 4-months post infection (the blastocercariae stage), in which the parasite was not yet infective to the final hosts (ducks), and 7-months post infection (the metacercariae stage), in which the worms were fully encysted and competent for transmission to the final host. At 7-months post infection, the results showed that shallow-water parasites were significantly more infective to shallow-water hosts, but deep-water parasites were about equally infective to snails from all three habitats. This result suggests that gene flow by the parasite is primarily from shallow to deep habitats, and that recycling of the parasite and host-parasite coevolution is primarily in shallow water. Such recycling would be consistent with our observations of foraging by the definitive host of the parasite, and may explain the maintenance of high frequencies of sexual individuals around the shallow-water margins of the lake.