Social networks created with different techniques are not comparable

Abstract

The recent application of social network analysis to animal populations has provided a tool to quantify group dynamics and individual social positions, which may enhance our understanding of the costs and benefits of sociality and the evolution of behavioural strategies within societies. Despite this, uncertainties remain about whether comparisons can be drawn between studies in which different sampling techniques have been used. We compared social networks constructed from two interaction and three proximity techniques that are frequently used in the literature, at both the ego network and global network levels, using data collected annually for two troops of chacma baboons, Papio ursinus, over 3 years. We obtained very different results at both the global and individual levels, demonstrating the clear distinction between networks built using different interaction and proximity techniques. While interaction techniques may be comparable at the whole global level, proximity techniques were not, and we found the opposite at the ego network level: proximity techniques could be compared whereas interaction techniques could not. As there was a clear distinction between the networks created, caution should be taken when using proximity as a proxy for social interactions (and vice versa) in social network studies. Further, our results showed high variation between troops and study seasons, reemphasizing the importance of incorporating temporal change in the analysis of social networks. Researchers should consider the effects of sampling technique on the networks produced when comparing networks created from different techniques.