Social networks are ubiquitous in animals and humans. Variation in social network structure can affect many biological processes, from the spread of pathogens to the evolution of cooperation. Testosterone is a key mediator of social behavior and studies have established that testosterone can often influence how individuals interact with their social partners. We investigated whether testosterone could also predict group-level social network dynamics in the wire-tailed manakin (Pipra filicauda), a lekking bird in which males perform complex coordinated displays with other males. We used an automated proximity system to longitudinally monitor male behavior several leks and we collected repeated samples of each male’s circulating testosterone. We found that social networks that were composed of high-testosterone dominant males were less specialized, less stable, and had more negative behavioral assortment. These results support our main hypothesis that hormone physiology can predict group-level dynamics. We also observed that larger leks with more interacting individuals had more positive behavioral assortment, suggesting that small groups may constrain the processes of homophily and behavior-matching. Overall, these results provide evidence that hormone-mediated behavior can shape the broader architecture of entire social groups, and that testosterone-mediated behavior can impede the evolution of stable cooperation.