Aim We evaluate, for the first time, the decay of similarity over distance in four species of marine fishes with different vagility: Trachurus murphyi (pelagic, highly vagile), Merluccius gayi (demersal, highly vagile), Sebastes capensis (demersal, low vagility) and Hippoglossina macrops (benthic, non-vagile).Location We use our own data of the species composition of parasite communities in four host species: T. murphyi (Perciformes), M. gayi (Gadiformes), H. macrops (Pleuronectiformes) and S. capensis (Scorpaeniformes), from the coast of Chile and southern Argentina. Trachurus murphyi, M. gayi and H. macrops in the area studied live closely associated with the cold waters of the Humboldt Current System, while S. capensis in southern Chile is associated with the fjord areas south of c. 43°S, and in Argentina this species lives in close association with the cold water of the Malvinas Current, which mixes with the warm water of the Brazilian Current in the sampled area. Thus, for all four host species there are no known geographical (oceanographic) barriers that interrupt their continuous distributions.Methods Fish were collected and frozen ()18°C) until examination in the laboratory. After thawing, each visceral organ was dissected separately, washed in running water, filtered using a mesh of 0.25 mm, and examined under a stereomicroscope (20·). For S. capensis, only ectoparasites were included in this study because data on endoparasites for some locations are not available. The approach of Poulin [Journal of Biogeography 30 (2003) 1609] was used. For each host species, we computed similarity in parasite composition using the BrayCurtis distance on a presence/absence matrix. Similarity between each pair of localities was correlated (linear regression on raw and transformed data) with the distance (in nautical miles) among those sites.Results Among conspecific populations of host fish, there was no correlation between sample size and parasite species richness. The same pattern was evident when all host populations were pooled and treated as a single sample. Except for H. macrops, an important proportion of the geographical ranges of the host species was covered. The decay in similarity of parasite communities over distance was recorded in three of the four hosts analysed. For H. macrops, no relationship between distances and decay of similarity was evident. A linear regression on nontransformed data provided a good fit, and the regression on log(n + 1) does not change significantly for both r 2 and probability, but slopes are smaller for logtransformed than for non-transformed data.Main conclusions Three of the four fish parasite systems studied fit well with the expected decay of similarity over distance. The slopes of the relationship between distance and similarity in the marine fishes studied are lower, but significant (three orders of magnitude) than those recorded in freshwater fishes by Poulin (2003). In our case, Hippoglossina macrops (benthic, non-vagile host)