It is widely believed that the dishonest behavior of an individual is influenced by his peers. The peer effects of dishonesty are hard to identify in empirical works because of self-selection. Indeed, individuals with common preferences or charac- teristics tend to associate together and this may explain that they behave similarly (homophily). In this paper, we use a laboratory experiment to disentangle peer effects on behavior from the influence of endogenous network formation with an ap- plication on lying behavior. We create two controlled environments: one in which peers are randomly assigned to participants and one in which participants can choose their peers. Our results show that participants tend to be homophilious, i.e. participants who lie to a larger extent choose peers that are more likely to be liars. In contrast, we find little evidence of pure peer effects.