The results of a previous experiment we conducted suggested that agonistic interactions may, counterintuitively, attract organisms to food sites rather than drive them away. In this study we test whether interactions attract or deter organisms by manipulating the distribution of interactions in a natural population of mallard ducks (Anas platyrhynchos L.). We manipulated the distribution of interactions by changing the size of the bread fed to the ducks from two food sites. In the condition with large pieces of bread the distribution of interactions between the two food sites was more extreme than in the small bread condition. The attraction hypothesis predicted that the distribution of ducks should be more extreme in this condition, whereas the deterrence hypothesis predicted that the distribution of ducks should be less extreme. We found that the distribution of ducks was significantly more extreme in the large bread condition. This supports the hypothesis that agonistic interactions act as a net attraction to foraging organisms. The possible benefits organisms may gain from agonistic interactions are discussed.