In laboratory studies, hamsters (Mesocricetus spp.) exhibit intense male-male aggression, thus making them an excellent model system for studies of the functional and mechanistic bases of aggression. In a field study of golden hamsters (M. auratus) in the wild, however, the few documented male-male interactions were not highly aggressive. Thus, we tested the hypothesis that familiarity modulates aggression in hamsters. Previous investigations of the effects of familiarity on aggression have mostly involved familiarization of unfamiliar individuals through agonistic interactions. Here we allowed male Turkish hamsters (M. brandti) to become familiar with each other by housing them together but separated by a wire-mesh partition (thus 'non-agonistic' familiarity). We found that although non-agonistic familiarity did not decrease investigation of the familiar male, it did decrease the occurrence of fights, the number of fights, and the percentage of time fighting; it also increased the latency to fight. These results are consistent with the 'dear enemy' hypothesis, which proposes that males are less aggressive toward familiar neighbors than to unfamiliar conspecifics because previous interactions have provided enough information about the other individual to render severe aggression unnecessary. Most importantly, our results suggest that information gained about other individuals through non-agonistic interactions decrease the frequency and intensity of fights with those individuals. We conclude that results from laboratory studies on aggression that do not consider the kind of social interactions that individuals have in nature should be interpreted with caution.