Within a host population, parasite infracommunities vary in both richness and species composition. If interspecific interactions among parasites are important in shaping infracommunities, the structure of these assemblages is expected to differ from the one predicted by null models, i.e. from the one that would result from chance alone. Using data from the literature, I tested for discrepancies between observed and random patterns in the richness and composition of gastrointestinal helminth infracommunities of birds and mammals. Both the Poisson distribution and a more sophisticated null model, derived from prevalence of the different parasite species in the host population, usually provided a good fit to the observed distributions of infracommunity richness among hosts. This suggests that parasite species do not co-occur more or less frequently than expected by chance. In mammals, the co-occurrence of all available parasite species in the same host individual, or maximum potential infracommunity richness, was less likely to be observed when several parasite species were available; this is also a phenomenon expected from the random assembly of parasite species. Finally, there was no evidence for a nested subset pattern among parasite species in a host population: rate species were distributed independently of common ones. The overall picture emerging from these results is one in which parasite assemblages are more likely to be the product of random events than of predictable and repeatable processes.