Time is of the essence: an application of a relational event model for animal social networks

Abstract

Understanding how animal social relationships are created, maintained and severed has ecological and evolutionary significance. Animal social relationships are inferred from observations of interactions between animals; the pattern of interaction over time indicates the existence (or absence) of a social relationship. Autonomous behavioural recording technologies are increasingly being used to collect continuous interaction data on animal associations. However, continuous data sequences are typically aggregated to represent a relationship as part of one (or several) pictures of the network of relations among animals, in a way that parallels human social networks. This transformation entails loss of information about interaction timing and sequence, which are particularly important to understand the formation of relationships or their disruption. Here, we describe a new statistical model, termed the relational event model, that enables the analysis of fine-grained animal association data as a continuous time sequence without requiring aggregation of the data. We apply the model to a unique data set of interaction between familiar and unfamiliar steers during a series of 36 experiments to investigate the process of social disruption and relationship formation. We show how the model provides key insights into animal behaviour in terms of relationship building, the integration process of unfamiliar animals and group building dynamics. The relational event model is well suited to data structures that are common to animal behavioural studies and can therefore be applied to a range of social interaction data to understand animal social dynamics.Electronic supplementary materialThe online version of this article (doi:10.1007/s00265-015-1883-3) contains supplementary material, which is available to authorized users.

Pregnant cattle associations and links to maternal reciprocity

“…Repeated patterns of shorter, fleeting encounters can suggest coming together for reassurance or reinforcing familiar relationships, whereas short encounters that are not repeated can indicate agonistic encounters, such as fighting and pushing followed by avoidance. Such encounter patterns are decipherable from proximity logger data when combined with additional analyzes that take into account events as a time series (Patison et al, 2015).…”

Understanding Allogrooming Through a Dynamic Social Network Approach: An Example in a Group of Dairy Cows

“…To date, several studies have characterized social networks in cattle, using association networks based on spatial proximity data [ (12,(38)(39)(40)(41)(42)(43)(44)(45)(46)]. Although very valuable, association networks only provide information based on potential contacts or relations.…”

Social traits, social networks and evolutionary biology

“…Within SNA, the dynamic analysis of animal social networks has also started to receive more focus (Blonder et al ., ; Pinter‐Wollman et al ., ). There is an array of techniques developed in SNA for the dynamic analysis of networks, which incorporate individual‐ and network‐level influences on changes in social associates and other individual traits, such as temporal exponential random graph models (Hanneke et al ., ; Krivitsky & Handcock, ), stochastic actor‐orientated models (Steglich et al ., ; Fisher et al ., ) and relational event models (Patison et al ., ; Tranmer et al ., ), the latter two allowing the explicit modelling of network structure–trait codynamics. Integrating such explicitly dynamic models with models for evolutionary change in populations of socially interacting individuals (outlined above) may give us a better appreciation of how dynamic changes in individual behaviour, the social environment and ecological and evolutionary processes related to these such as selection are causally linked.…”