Characterising nosing behaviours in pigs after mixing using social network analysis

Affiliation as indicated by proximity has, to date, been used as the principal measure of positive relationships in farm animals. However, intensive housing may present a caveat to using proximity as a representation of affiliation as animals may be forced into proximity by stocking densities. To investigate affiliation patterns, this study examined differences in the expression of proximity and contact behaviours following regrouping in pigs. Animals (61 males and 56 females) were observed across eight groups (14.6 ± 0.69 SD pigs/group). Each group comprised a mix of familiar and unfamiliar finishing pigs (12 weeks old). Video observations occurred for two consecutive days following regrouping. Individuals were continuously observed for social nosing behaviour for 2 h in total, balanced across days. Social network analysis of directed networks provided group-level information including weighted degree centrality and density. Exponential Random Graph Models fitted to these networks were used to consider the underpinning social processes including reciprocity, homophily, and the effect of individual attributes. Groups differed in the expression of behaviours, whilst, at the global level, density was significantly lower (P < 0.001) for the snout-snout (0.68 ± 0.15 SD) than snout-head (0.92 ± 0.04 SD) proximity networks. Statistically significant differences were also shown in the density across the contact networks (P < 0.001) with the lowest cohesion in the snout-snout (0.33 ± 0.14 SD), compared to the snout-head (0.52 ± 0.07 SD), and snout-body (0.66 ± 0.09 SD) contact networks. Familiarity was a predictor of interaction (P = 0.0001) across behaviours. Familiar pigs were nearly twice as likely to assort in the contact networks and three times more likely to assort in the snout-snout and snout-head proximity networks. Sex was not a predictor of snout proximity; however, females received significantly less behaviour than males in the snout-snout (odds ratio (OR): 0.78, P = 0.046), and snout-head (OR: 0.69, P = 0.001) contact networks. Snout proximity behaviours showed significant reciprocity (snout head: OR = 2.56; P = 0.008; snout-snout: OR = 2.80; P = 0.0001). Contact behaviours showed significant reciprocity in the snout-snout (OR = 2.40; P = 0.0001), and snout-head (OR = 1.55; P = 0.004) networks. Our study highlights behavioural nuances, with groups differing in snout proximity and contact patterns, in which reciprocation is normal behaviour, and snout-snout proximity and snout-snout contact are the least observed. Furthermore, it shows the influence of attributes on network structure to inform grouping strategies.