Intolerant baboons avoid observer proximity, creating biased inter-individual association patterns.

Social network analysis is an increasingly popular tool for behavioural ecologists exploring the social organisation of animal populations. Such analyses require data on inter-individual association patterns, which in wild populations are often collected using direct observations of habituated animals. This assumes observers have no influence on animal behaviour; however, our previous work showed that individuals in a habituated group of chacma baboons (Papio ursinus griseipes) displayed consistent and individually distinct responses to observer approaches. We explored the implications of our previous findings by measuring the inter-individual association patterns of the same group of chacma baboons at different observer distances. We found a strong positive association between individual tolerance levels (towards observers) and how often an animal appeared as a neighbour to focal animals when observers were nearer, and a neutral relationship between the same variables when the observer was further away. Additionally, association matrices constructed from different observation distances were not comparable within any proximity buffer, and neither were the individual network metrics generated from these matrices. This appears to be the first empirical evidence that observer presence and behaviour can influence the association patterns of habituated animals and thus have potentially significant impacts on measured social networks.