Social networks predict selective observation and information spread in ravens

Animals are predicted to selectively observe and learn from the conspecifics with whom they share social connections. Yet, hardly anything is known about the role of different connections in observation and learning. To address the relationships between social connections, observation, and learning patterns, we investigated information transmission in two raven (Corvus corax) groups. First, we quantified social connections in each group by constructing networks on affiliative interactions, aggressive interactions, and proximity. We then seeded novel information by training one group member on a novel task and allowing others to observe. In each group, an observation network based on who observed whose task-solving behavior was strongly correlated with networks based on affiliative interactions and proximity. Ravens with high social centrality (strength, eigenvector, information centrality) in the affiliative interaction network were also central in the observation network, possibly as a result of solving the task sooner and being observed frequently. Network-Based Diffusion Analysis (NBDA) revealed that the order in which ravens first solved the task was best predicted by connections in the affiliative interaction network in a group of subadult ravens, and by social rank and kinship (which influenced affiliative interactions) in a group of juvenile ravens from three families. Our results demonstrate that not all social connections are equally effective at predicting the patterns of selective observation and information transmission.