The stability of social prominence and influence in a dynamic sow herd: A social network analysis approach

Social network analysis has developed as a valid technique in animal behaviour; despite an increase in the number of social network research in this field, comparatively few studies have focused on commercial animals. This study investigated a dry breeding sow herd (average = 69) as a model species for identifying socially prominent sows, individuals engaging in significantly higher levels of direct interactions than their conspecifics and socially influential sows, individuals with significantly higher levels of indirect interactions than their conspecifics. Animals were frequently added and removed from the herd during the study, creating a dynamic unstable social environment. In total, 63 h of video observations were obtained over three production cycles. Each production cycle covered 21 days, and behavioural observations occurred on days 1, 2, 3, 7, 14, 20, 21. The social network metrics of degree centrality (number of interactions), betweenness centrality (number of times an individual falls along the shortest path between two individuals) and centralisation (the extent to which a single individual dominates a network) were analysed to assess the stability of aggressive behaviour and preferential associations between production cycles. Preferential associations refer to lying behaviour <1 m from selected partners. Agonistic interactions included biting, thrusting, chasing and displacement behaviour. The results identified subgroup formations within the herd, based upon connectedness. Socially prominent and influential sows were identified within the subgroups, a result unaffected by low centralisation in the preferential association networks and substantial centralisation variance between the aggression networks. Sows quantified as socially prominent in the preferential association and aggression networks did not generally hold this position between production cycles, showing significant changes in degree centrality. Sows also demonstrated inconsistency in social influence in the preferential association networks, again with significant differences in betweenness centrality between production cycles. However, in the aggression networks, betweenness centrality was found to be a stable individual social network metric, with no significant changes in behaviour. Knowledge of the stability of social influence in the aggression networks could improve levels of agonistic interactions as it becomes possible to make predictions of future response to aid and inform intervention or management strategies. These findings highlight betweenness centrality as a powerful alternative tool that extends beyond the traditional methods of analysing dyadic interactions to identify critical animals.