Multi-group analysis of grooming network position in a highly social primate

Abstract

Individual variation in complex social behavioral traits, like primate grooming, can be influenced by the characteristics of the individual and those of its social group. To better grasp this complexity, social network analysis can be used to quantify direct and indirect grooming relationships. However, multi-group social network studies remain rare, despite their importance to disentangle individual from group-level trait effects on grooming strategies. We applied social network analysis to grooming data of 22 groups of zoo-housed bonobos and investigated the impact of three individual (sex, age, and rearing-history) and two group-level traits (group size and sex ratio) on five social network measures (out-strength, in-strength, disparity, affinity, and eigenvector centrality). Our results showed age-effects on all investigated measures: for females, all measures except for affinity showed quadratic relationships with age, while in males, the effects of age were more variable depending on the network measure. Bonobos with atypical rearing histories showed lower out-strength and eigenvector centrality, while in-strength was only impacted by rearing history in males. Group size showed a negative association with disparity and eigenvector centrality, while sex ratio did not influence any of the investigated measures. Standardization for group size did not impact the effects of sex and age, indicating the robustness of these findings. Our study provides comprehensive insights into the complexity of grooming behavior in zoo-housed bonobos, and underlines the importance of multi-group analyses for the generalizability of social network analysis results for species as a whole.

Fig 1. The effect of age on the different grooming network measures.

(a) Out-strength, (b) in-strength, (c) disparity, (d) affinity, (e) eigenvector centrality, and on (f) agonistic dominance rank as measured by normalized David’s scores (nDS). Black dots represent individual datapoints. If an age-by-sex interaction was found, we indicated males in blue datapoints with a blue trendline, while females are indicated in red. Shaded area represents the 95% confidence interval.

Fig 2. The effect of rearing history on grooming network measures.

Rearing-history significantly affected (a) out-strength, (b) in-strength, and (c) eigenvector centrality. Boxplot figure with lower and upper box boundaries at 25th and 75th percentiles, respectively. Line inside box shows median, black dots show individual data points for both sexes, while red dots represent females and blue dots represent males in case of rearing-by-sex interactions. * P < 0.05, ** P < 0.01, *** P < 0.001.

Fig 3. The effect of group size on social network measures.

Group size significantly affected (a) disparity and (b) eigenvector centrality. Black dots show individual data points for both sexes, while red dots represent females and blue dots represent males in case of group size-by-sex interactions. Shaded area represents the 95% confidence interval.