Networks never rest: An investigation of network evolution in three species of animals

Abstract Despite considerable advancement in the study of network evolution, three basic limitations are common to the data collected: (1) examining a small number of networks, (2) not observing networks from scratch, (3) not collecting time-stamped, continuous records of all interactions among all members of groups. Here, we avoid these limitations by observing all aggressive interactions leading to network formation from the moment of introduction among all members of 45 groups of four individuals each in three species of animals: chickens, cichlid fish, and mice. We apply several recently developed methods for the visualization and analysis of network evolution to these unique datasets. We discover, first, that network evolution is a remarkably dynamic process across all three species: networks do not evolve to specific structures and then remain in those configurations. Instead, we find dynamic stability in which many groups continually return to a general class of structures. Second, we find considerable similarity across species in the pathways that the groups take through different possible network configurations as they evolve. Third, we show that transitive component triads are more stable than intransitive ones. Fourth, we track the evolution of individual ranks within groups and discover that many individuals do not have stable positions. Finally, we discuss fundamental questions that our findings raise for the study of networks in both animals and humans.