Coinfection of wild birds by influenza A viruses is thought to be an important mechanism for the diversification of viral phenotypes by generation of reassortants. However, it is not known whether coinfection is a random event or follows discernible patterns with biological significance. In the present study, conducted with viruses collected throughout 15 years from a wild-duck population in Alberta, Canada, we identified three discrete distributions of coinfections. In about one-third of the events, which involved subtypes of viruses that appear to be maintained in this duck reservoir, coinfection occurred at rates either close to or significantly lower than one would predict from rates of single-virus infection. Apparently, the better adapted an influenza A virus is to an avian population, the greater is its ability to prevent coinfections. Conversely, poorly adapted, nonmaintained viruses were significantly overrepresented as coinfectants. Rarely encountered subtypes appear to represent viruses whose chances of successfully infiltrating avian reservoirs are increased by coinfection. Mallards (Anas platyrhynchos) and pintails (A. acuta) were significantly more likely to be infected by a single influenza A virus than were the other species sampled, but no species was significantly more likely to be coinfected. These observations provide the first evidence of nonrandom coinfection of wild birds by influenza A viruses, suggesting that reassortment of these viruses in a natural population does not occur randomly. These results suggest that even though infections may occur in a species, all subtypes are not maintained by all avian species. They also suggest that specific influenza A virus subtypes are differentially adapted to different avian hosts and that the fact that a particular subtype is isolated from a particular avian species does not mean that the virus is maintained by that species.