Parasites relying on trophic transmission to complete their life cycles often induce modifications of their host's behavior in ways that may increase their susceptibility to predation by final hosts. These modifications have often been interpreted as parasite adaptations, but very few studies have demonstrated that host manipulation has fitness benefits for the parasite. The aim of the present study was to address the adaptive significance of parasite manipulation by coupling observations of behavioral manipulation to estimates of trophic transmission to the definitive host in the natural environment. We show that the acanthocephalan parasite Pomphorhynchus laevis manipulates the drifting behavior of one of its intermediate hosts, the amphipod Gammarus pulex, but not of a sympatric host, the introduced amphipod Gammarus roeseli. We found a 26.3-28.3 times higher proportion of infected G. pulex in the stomach content of one of the definitive hosts of P. laevis, the bullhead Cottus gobio, than in the benthos. No such trend was observed for G. roeseli. The bell-shaped curve of mean parasite abundance (MPA) relative to host size observed in G. pulex also supported an increased predation mortality of P. laevis-infected individuals compared to uninfected amphipods. Again, no such pattern was observed in G. roeseli. Furthermore, our results indicate that the modifications induced by P. laevis are specific to the definitive host and do not increase the risk of predation by inappropriate hosts, here the adult edible frog Rana esculenta. Overall, our study is original in that it establishes, under field conditions, a direct link between parasitic manipulation and increased transmission to the definitive host, and more importantly, identifies the specificity of the manipulation both in the intermediate host species and toward the definitive host.